Ph.D.

Thesis

Use of Geographic Information System in

Environmental Science

Ondřej Sáňka
2015

Masaryk University, Faculty of Science

Research Centre for Toxic Compounds in the
Environment
Brno, Czech Republic

Supervisor: RNDr. Pavel Čupr, Ph.D.

BIBLIOGRAPHIC IDENTIFICATION

Author: Ondřej Sáňka

Faculty of Science, Masaryk University,

Research Centre for Toxic Compounds in
the Environment

Title of Dissertation: Use of Geographic Information System in

Environmental Science

Degree Programme: Chemistry

Field of Study: Environmental Chemistry

Supervisor: RNDr. Pavel Čupr, Ph.D.

Academic Year: 2015

Number of Pages: 156

Keywords: Environmental science, geographic

information system, cartographic
visualization, spatial data, environmental
pollution, spatial analysis, environmental
modelling
BIBLIOGRAFICKÝ ZÁZNAM

Autor: Ondřej Sáňka

Přírodovědecká fakulta, Masarykova

Univerzita, Centrum pro výzkum
toxických látek v prostředí

Název práce: Využití geografických informačních

Systémů v environmentálních vědách

Studijní program: Chemie

Studijní obor: Chemie životního prostředí

Školitel: RNDr. Pavel Čupr, Ph.D.

Rok obhajoby: 2015

Počet stran: 156

Klíčová slova: Environmentální věda, geografický

informační systém, kartografická
vizualizace, znečištění životního
prostředí, prostorová analýza,
environmentální modelování
© ONDŘEJ SÁŇKA, MASARYK UNIVERSITY, 2015
ACKNOWLEDGEMENTS
First, I would like to thank my supervisor RNDr. Pavel Čupr, Ph.D. for the
guidance, support and valuable advice he provided to me during all my
postgraduate study. I am also very grateful to Luca Nizzetto, Ph.D. for his
support and advice, as well as to my colleague and partner Lisa Melymuk,
Ph.D. for moral and scientific support and unlimited patience in last years of
my study. I would also like to thank my work colleagues, friends and office
mates for creating a friendly and nice work environment. Finally, my
gratitude also belongs to my family, especially to my father, who gave me
the idea of doing postgraduate studies.
ABSTRACT
Environmental science as a young scientific discipline is undergoing rapid
development in recent decades. Together with population growth and the
introduction of new technologies the variety and number of environmental
issues is growing, resulting in diversification to more environmental
disciplines. One of these is environmental chemistry, dealing with chemical
pollution of the environment and its consequences. However, the challenges
of environmental chemistry, such as complex pollutant behaviour and fate
call for tools and techniques from scientific disciplines beyond only
chemistry, such as meteorology, geology, hydrology, sociology, geography,
and cartography and geographic information systems (GIS).

This thesis describes and demonstrates the importance of the spatial

component of information, which is a part of the absolute majority of
environmental chemistry data. These data have a large potential to be
processed, analysed and visualized by GIS, and GIS can help reveal possible
relations with other physical and socioeconomic environmental descriptors.
This thesis also describes the broad range of GIS approaches, tools and data
available for use in environmental chemistry, as well as some challenges
related to it. Several cases where GIS was successfully applied in
environmental chemistry studies are described more in detail in the
experimental part of the thesis, with particular emphasis on the use of GIS
in environmental modelling.

The first section of the experimental part of the thesis (Chapter 2) reports
on two case studies where GIS was used to process input data for modelling
polycyclic aromatic hydrocarbon (PAH) dispersion in air, as well as to
visualize results of the modelling. GIS was an irreplaceable tool in these
studies, through its ability to deal with large amounts of input data, and also
contributed to understanding the reasons for differences between the
dispersion model and passive air sampling results. The second section of
Chapter 2 introduces the application of a complex multi-compartmental
environmental model on a watershed of the Morava River to estimate the
transport paths and fate of p’p-DDT in soils and river water. GIS also played
a crucial role in this study in the preparation and processing of input data
on the watershed. The third section briefly describes three cases where GIS
analyses were used to produce and visualize information supporting the
results of scientific studies.
ABSTRAKT
Environmentální věda, jakožto mladá vědní disciplína, prochází
v posledních desetiletích rychlým vývojem. Společně s populačním růstem a
vznikem nových technologií roste rozmanitost a množství
environmentálních problémů, což vede k diverzifikaci na více
environmentálních disciplín. Jednou z nich je environmentální chemie,
zabývající se chemickým znečištěním životního prostředí a jeho následky.
Nicméně, problémy environmentální chemie, jako složitost chování
znečišťujících látek a jejich osud, vyžadují zapojení nástrojů a technik jiných
disciplín, než jen chemie, např. meteorologie, geologie, hydrologie,
sociologie, geografie, a kartografie a geografické informační systémy (GIS).

Tato disertační práce popisuje a demonstruje důležitost prostorové složky

informace, která je součástí naprosté většiny dat environmentální chemie.
Tato data disponují velkým potenciálem pro analýzu, a vizualizaci pomocí
GIS, přičemž GIS může pomoci objasnit vztahy s jinými fyzikálními a
socioekonomickými faktory. Tato práce zároveň popisuje širokou paletu
přístupů, nástrojů a dat GIS, vhodných k použití v environmentální chemii,
ale také problémy s tím spojené. Experimentální část této dizertační práce
detailněji popisuje několik případů úspěšné aplikace GIS v případech
environmentální chemie, s důrazem na využití GIS v environmentálním
modelování.

První oddíl experimentální části práce (kapitola 2) popisuje dvě případové

studie použití GIS k přípravě vstupních dat pro modelování rozptylu
polycyklických aromatických uhlovodíků (PAHs) v ovzduší, a pro vizualizaci
výsledků modelování. V těchto studiích se GIS ukázal jako nenahraditelný
díky své schopnosti zpracovat velké množství dat a jeho výsledky přispěly
k identifikaci možných příčin diferencí mezi rozptylovým modelem a
pasivním vzorkováním ovzduší. Druhý oddíl kapitoly 2 popisuje studii
použití komplexního modelu, zahrnujícího více složek životního prostředí
na povodí řeky Moravy k odhadu transportních procesů a osudu p’p-DDT v
půdě a říční vodě. Také v této studii představuje GIS významnou roli
v přípravě a zpracování vstupních dat o povodí pro model. Třetí oddíl
stručně popisuje tři případy použití GIS k analýze a vizualizaci dat,
podporující výsledky vědeckých studií.
LIST OF ABBREVIATIONS
BAP Benzo(a)pyrene
BBF Benzo(b)fluoranthene
BKF Benzo(k)fluoranthene
BFRs Brominated flame retardants
CHMI Czech Hydrometeorological Institute
DOC Dissolve organic carbon
DDT Dichlorodiphenyltrichloroethane
DIBAVOD Digital Base of Data on Water Management
EEA European Environment Agency
EMEP European Monitoring and Evaluation Air Quality Monitoring
programme
ESRI Environmental Systems Research Institute
FRs Flame retardants
FOSS Free and open source software
GCS Geographic coordinate system
GIS Geographic information system
GMP Global Monitoring Plan on Persistent Organic Pollutants of
the Stockholm Convention
GPS Global positioning system
ICIMOD The International Centre for Integrated Mountain
Development
IDW Inverse distance weighting
INCA Integrated Catchment Model
IPSI Index of potential source influence
IPY Indeno(123cd)pyrene
JRC Joint Research Centre
MONET Monitoring Network
NOAA National Oceanic and Atmospheric Administration
OCPs Organochlorine pesticides
PAHs Polycyclic aromatic hydrocarbons
PBDEs Polybrominated diphenyl ethers
PCBs Polychlorinated biphenyls
PCDDs/Fs Polychlorinated dibenzo-p-dioxins / furans
PCS Projected coordinate system
PERSiST Precipitation, Evapotranspiration and Runoff Simulator for
Solute Transport
PM Particulate matter
PM2.5 Particulate matter up to 2.5 µm diameter
PM10 Particulate matter up to 10 µm diameter
POPs Persistent organic pollutants
REZZO Registry of Sources of Air Pollution
RSD Road and Motorway Directorate of the Czech Republic
SOC Soil organic carbon
SVOCs Semi-volatile organic compounds
SYMOS’97 System of Modelling Stationary Sources
US EPA United States Environmental Protection Agency
USGS United States Geological Survey
WGS84 World Geodetic System 1984
TABLE OF CONTENTS

List of figures ....................................................................................................................................3

1 Introduction ............................................................................................................................6
1.1 GIS ......................................................................................................................................7
1.1.1 GIS structure.......................................................................................................7
1.1.2 GIS software........................................................................................................8
1.1.3 GIS Data structure and availability .........................................................9
1.2 Development of environmental science ...................................................... 13
1.2.1 Environmental Chemistry ........................................................................ 13
1.3 Use of GIS in environmental science ............................................................. 16
1.3.1 Overview of area of interest .................................................................... 16
1.3.2 Planning sampling campaigns................................................................ 17
1.3.3 Estimation techniques for unknown values ................................... 20
1.3.4 Spatial analyses and their use ................................................................ 25
1.3.5 The role of GIS in environmental models ......................................... 26
1.4 Challenges of GIS use in environmental science ..................................... 28
1.4.1 Environmental data availability and quality .................................. 28
1.4.2 Complexity of substance behaviour .................................................... 29
1.4.3 Linking GIS with models ........................................................................... 29
2 Experimental work........................................................................................................... 34
2.1 GIS-based assessment of PAHs concentrations in air using air
dispersion modelling ............................................................................................................ 36
2.2 Assessment of long term temporal trends of p,p’-DDT soil
concentration using GIS-aided mechanistic model .............................................. 41
2.3 Use of spatial analysis for environmental research support ............ 44
3 Highlights and Conclusions.......................................................................................... 52
4 References............................................................................................................................. 54
5 Short Curriculum Vitae .................................................................................................. 62

1
6 List of publications ........................................................................................................... 63
7 List of conference presentations and posters .................................................... 65
8 List of Annexes.................................................................................................................... 67

2
LIST OF FIGURES
Figure 1-1: Example of (a) raster and (b) vector data. ............................................ 11
Figure 1-2: Structure a five-ring PAH: benzo[a]pyrene. ......................................... 15
Figure 1-3: Structure of p,p’-DDT ....................................................................................... 16
Figure 1-4: Example of overview map of a study are ............................................... 17
Figure 1-5: Example of visualization of population count variability in the
area surrounding sampling sites ......................................................................................... 19
Figure 1-6: Comparison between results of spatial interpolation method of
kriging and land use regression, where kriging performs poorly due to lack
of measurement points.. .......................................................................................................... 21
Figure 1-7: Examples of four types of spatial interpolations: inverse
distance weighting, ordinary kriging, spline and radial basis functions, on
the same generic dataset, showing differences between these methods. ..... 23
Figure 2-1: concentrations of benzo(a)pyrene in Valašské Meziříčí,
calculated by SYMOS’97 dispersion model in winter and summer................... 40
Figure 2-2: Concentrations of benzo(a)pyrene in Liberec, calculated by
SYMOS’97 dispersion model in winter and summer. ............................................... 40
Figure 2-3: Land Use, subcatchment division and measurement sites in
study area of the Morava River watershed. ................................................................... 44
Figure 2-4: PCB concentrations at different elevations of mountain forest
soils of China. Calculated values of index of potential source influence (IPSI)
are also displayed........................................................................................................................ 46
Figure 2-5: Example of overlaying population data and backward
trajectories for IPSI calculation in GIS environment. ............................................... 47
Figure 2-6: Overlaying population data and PBDE sampling frequency. ....... 48
Figure 2-7: Visualization of active air sampling volumes in relation to
latitude .............................................................................................................................................. 48

3
4
 CHAPTER 1
INTRODUCTION AND BACKGROUND

5
1 INTRODUCTION
Many thousands of years ago, humans were creating rudimentary versions
of maps. These were simple two-dimensional depictions, such as 25000
year old drawing of a river valley etched on a mammoth tusk, found in the
Pavlov region of South Moravia (Klíma, 1988). Since that time, the art,
strategies, purposes and techniques of manufacturing maps has changed a
great deal, in parallel with the establishment of scientific discipline of
geography. Geography has always been considered as the discipline closest
to map-making, from its early beginnings in ancient times, through the
middle ages up to the present. Geography is a scientific discipline trying to
answer questions such as “where are things,” “how to get to them,” “what
are they like,” “what are the distances among them,” and “how can they be
influenced by things around them.” The most convenient and
comprehensive answers to these questions can be provided by a map.

As new parts of the world were being discovered in ancient, medieval, and
early modern times, new areas needed to be explored and documented and
maps were the best way of doing it. Beginning in the 1700s the
development of various scientific disciplines contributed to a massive
growth of phenomena and topics to be visualized in form of maps. So-called
“thematic maps” became an essential part of natural sciences such as
hydrology, geology, meteorology, geomorphology, climatology, glaciology
etc., as well as social sciences, such as sociology, history, economy, politics,
anthropology and many others (Buchroithner and Fernández, 2011; Clarke,
2003). As making maps for different fields of science could no longer be
considered as merely a part of geography, cartography was officially
acknowledged as an independent science in 1949 (Buchroithner and
Fernández, 2011). The development of technologies for map-making during
the 20th century enabled massive production of maps for both scientific and
also non-scientific purposes, such as tourism, decorative purposes etc.

The development and use of such a wide range of thematic maps led to the
demand for better tools for creation and manipulation of spatial data. This
need was answered by the development of geographic information systems
(GIS). The conceptual framework originated in the 1950s, when the now
common technique of map overlay was invented. Map overlay allowed
multiple topics to be combined, creating thematic maps that were useful for
many purposes (Clarke, 2003), although early use was much more

6
cumbersome, as it involved actual overlaying of transparent maps. The first
software packages appeared in the late 1960s and 70s, however the most
rapid development came with the boom of information technologies in the
1980s and 90s (Clarke, 2003; Goodchild, 2009). Today, there is a broad
range of software of dealing with spatial information, offering a multitude of
tools helping millions of users deal with data of many types, formats and
origins. Apart from professional licensed software there are also a many
free GIS software programs, together with much free spatial data available
online, which makes it possible for everyone with access to a PC to create
maps and/or spatial data. However, this can also have a negative aspect, as
it can increase the possibility of poor quality or erroneous spatial data and
maps.

The list of scientific disciplines taking advantage of cartography and GIS

continues to grow, as there are always new challenges, topics and
phenomena to be aided by spatial data and maps. As environmental science
arose in the 1960s and 1970s as a reaction to awareness of environmental
problems such as adverse effects of synthetic pesticides, industrial disasters
and other anthropogenic activities, spatial data and maps became important
tools in addressing these issues. For example, maps showing predicted
change of air temperature came with emergence of science on climate
change and maps of soil vulnerability to erosion were created due to the
extensive exploitation of soil for intensive agriculture. GIS was also
especially helpful in understanding chemical pollution, as maps displaying
and analysing the levels of pollution in environmental compartments were
used to assess newly identified pollutants. This thesis will explore in the
breadth of application of GIS in environmental sciences, present the current
state of the science, as well as suggest future directions and progress.

1.1 GIS
1.1.1 GIS STRUCTURE
As mentioned above, geographic information systems are a powerful tool of
use in many scientific disciplines. Definitions of GIS typically characterize it
as the whole of the techniques developed in organizing, using and
presenting spatial data, e.g., “Geographic Information Systems is a
computer-based tool that analyzes, stores, manipulates and visualizes
geographic information on a map” (GIS geography, 2015).

7
This suggests that the expression “GIS” encompasses much more than just
software for making maps. Firstly, GIS datasets are very often unusually
large amounts of data, depending on spatial coverage, spatial resolution,
amount of detail and format. These require sufficient hardware capacity to
be able to create, store or update these datasets. Secondly, GIS would be of
no use without data of proper format and quality, adequate to the intended
use. These can be either purchased from commercial data providers,
obtained for free from online databases or created by the user, which is
usually very time consuming work, however sometimes the only possible
way. Finally, there are also qualified, trained and experienced people
needed to create, store and update the data, as well as to perform analysis
that provides reliable results and create a correct and high-quality maps.

1.1.2 GIS SOFTWARE

Many GIS software packages are available, from commercial packages to
free and open source software (FOSS). Currently, probably the most widely
used one program is ArcGIS, developed by the Environmental Systems
Research Institute (ESRI) (ESRI, 2015), offering a user-friendly graphic
interface, ability to handle different types of geographic objects and a set of
powerful tools and extensions (Pistocchi, 2014). However, there is a
growing family of FOSS of different scripting languages and thoroughness.
Among the most comprehensive and popular are GRASS, QuantumGIS,
ILWIS, uDig, SAGA, OpenJump, MapWindow GIS and others (Shekhar and
Xiong, 2008; Steiniger and Bocher, 2008; Steiniger and Hunter, 2013).
These FOSSs may have wider use in the scientific community than in private
companies, as they are more suitable for developing environmental models
that need original equations and methods (Pistocchi, 2014).

In all the studies carried out for this thesis, the commercial ArcGIS software
was used, mostly due to its universality and broad functionality. ArcGIS is
also broadly used by scientific community and in scientific publications, e.g.
Gulliver et al. (2011), Hafner et al. (2005), Zou et al. (2011), Pistocchi et al.
(2010b), Kuboš ová et al. (2009) and VoPham et al. (2015). One of its
greatest advantages is also a very rich user support. This is provided
officially by ESRI thorough a Help section for the software, as well as
support discussions online (http://blogs.esri.com/esri/ or
https://geonet.esri.com/) and a broad variety of handbooks and tutorials.
However, often more important and more useful are the many online GIS

8
forums and online discussions, where both the correct use of tools and
analyses is discussed and solutions given to errors that the software
occasionally encounters. Examples are http://www.giscafe.com/ or
http://www.gislounge.com/). Finally, there are a number of free ArcGIS
tutorials available on university web pages as a support for their GIS
classes, e.g., https://www.trentu.ca/library/madgic/gisguides.

1.1.3 GIS DATA STRUCTURE AND AVAILABILITY

1.1.3.1 Coordinate system
A necessary part of any spatial data to be used in GIS is a coordinate system,
usually based on an ellipsoid (defined by shape, size and orientation) and
geoid, serving as an approximation of Earth’s surface. Many coordinate
systems exist, varying in dimensions of the area they are meant to be used
for. We distinguish between two basic types of coordinate systems:
geographic and projected.

In geographic coordinate system (GCS), coordinates are specified by

latitude, longitude and elevation above or below the surface of the geoid.
The most common GCS is the World Geodetic System, defined in 1984
(WGS84). Several geodetic institutions on the country level have also
developed their own GCS, based on different ellipsoids and geoids in order
to represent region-specific data with higher accuracy.

A projected coordinate system (PCS) is used when working with smaller

distances. Then it is beneficial to approximate locally the ellipsoid by a
plane, or another surface that can be flattened on a plane (cylinder, cone).
Coordinates can be represented by a pair of distances from a reference
point of an orthogonal system. A PCS is defined by a mathematical function
applied to transform the coordinates (a projection), the reference ellipsoid,
and the location of the reference point of the orthogonal coordinates
(Pistocchi, 2014). For example, a projected coordinate system S-JTSK was
developed for the area of former Czechoslovakia in 1920s, based on
Křovák’s conical projection, in order to minimize distortion of the specific
area covered by Czechoslovakia. This projection is still used today by Czech
cartographic agencies.

9
Projections vary by properties of the map being distorted (distances, angles
or areas). Depending on the shape of area being displayed, scale and
purpose of map being created, an appropriate projection should be chosen
so the distortion of the displayed area is as smallest as possible. Thus,
selection of a projection is one of the key factors in the process of map-
making, as inappropriate distortion can optically change proportions of
areas of different character or quality, potentially causing the reader to
misinterpret key elements of the map.

1.1.3.2 GIS data formats

Apart from the coordinate system, another feature common to all GIS and
cartographic data is the spatially-referenced information. In other words,
the factual data (e.g., number of inhabitants, air temperature, elevation or
measured pollutant concentration) always has to have some kind of spatial
information attached (e.g., name of administrative unit, coordinates,
address...). These vary for the specific type of the data. Two basic types of
data formats are used: vector and raster (Fig. 1-1).

Vector data describe objects by points, having a position defined by

coordinates and the relations among the points. There are three main types
of vector data: (a) points, represented by a single pair of coordinates, (b)
polylines, represented by a sequence of lines (each defined by two points, so
called vertices) and (c) polygons which are closed polylines. An example of
a phenomenon represented by a point feature is a sampling site, an example
of polyline is a road or a stream, and example of polygon is a preserved area
or a water body. The most popular form of vector GIS data is ESRI shape file
which is represented by three or more files of the same name and different
extensions, each containing different pieces of information. The most
important ones are *.shp, containing information on the location of the
objects, and *.dbf, containing a table of various attributes related to the
objects

10
a) b)
Figure 1-1: Example of (a) raster and (b) vector data.

In raster formats, data are associated to grid or cells, where the region of
interest is segmented into a regular network of squares, often called pixels
(picture elements). The location in the coordinate system is specified by
position of the corner (usually lower left), size of the grid cell, and number
of rows and columns of cells. The value of the phenomenon displayed is
represented by numeric values for each cell. Typically, continuous data are
presented in raster form, such as digital elevation models, remote sensing
data and results of spatial analyses such as spatial interpolations. The most
common raster formats are ESRI ASCII grid, GeoTIFF or NetCDF (The
network Common Data Form) (Pistocchi, 2014).

Typically, GIS software offers tools that allow data conversion from raster
to vector and vice versa, usually with certain level of information loss,
however for some types of analyses this conversion is necessary (Clarke,
2003).

1.1.3.3 GIS data availability

It is often extremely costly and time consuming for the user to individually
create necessary input data for a GIS analysis, whether by data collection
and survey in terrain using GPS, scanning and vectorising older hardcopy
maps, or by processing data captured by remote sensing techniques.
Moreover, advanced GIS analysis and modelling often requires use of
datasets describing very specific phenomena and variables, which cannot be
acquired by any of abovementioned ways.

11
An important method for getting needed data is by downloading from
available online data sources. There are two main options: (a) data to be
purchased and (b) free data. Aside the cost, there may be another important
difference governing the choice of data. All GIS data contain some level of
generalization, compared to the reality, which itself means a specific and
also acceptable type of error. The level of generalization, accuracy,
precision, resolution and completion in the data will be defined based on
the requirements of the project and type of analysis. Thus, choice of data is
governed by the availability of data and the project requirements.

One major source of GIS data is governmental or public organizations, such

as the United States Geological Survey (USGS, http://www.usgs.gov) and
the National Oceanic and Atmospheric Administration (NOAA,
www.noaa.gov) for the USA, and the National Geoportal (geoportal.gov.cz)
and Digital Base of Data on Water Management (DIBAVOD,
http://www.dibavod.cz) for the Czech Republic. Some basic data is also
available from local software providers, such as ArcDATA, Czech Republic
(www.arcdata.cz) in the case of ESRI.

For global or regional data related to environment sciences, more options

are available. There are several public web portals, such as DIVA-GIS
(http://www.diva-gis.org) and EDENext (http://www.edenextdata.com),
offering worldwide elevation and climate data as well as other human
geography data on the national or global level. More focused environmental
data, such as worldwide land use and vegetation coverage, is offered from
http://www.iscgm.org/. On the other hand, http://www.webgis.com/ and
http://srtm.csi.cgiar.org/ are more focused particularly on elevation data. If
not focusing strictly on global data, then webpages of European
Environment Agency (EEA) can be recommended, as well as Joint Research
Centre (JRC) European Soil Portal http://eusoils.jrc.ec.europa.eu/ can be an
excellent sources. The International Centre for Integrated Mountain
Development (ICIMOD, http://apps.geoportal.icimod.org/hkhglacier/#),
focusing on glaciers coverage in Southern Asia, is a non-European example
of a regional data source. Links for a large number of other GIS data sources
can be found at http://freegisdata.rtwilson.com/. An overview table of free
online GIS data sources can be found in Annex 7.

As concern for the environment and study of environmental sciences

continues to grow on a global scale, the amount of spatial data related to the

12
environment and its issues is also growing. This can make the process of
choosing the highest available quality and most up-to-date GIS data more
challenging than finding the data itself. However, the utility of GIS in
environmental science on a global scale will generally only be strengthened
with greater availability and accessibility of data.

1.2 DEVELOPMENT OF ENVIRONMENTAL SCIENCE

The scientific discipline of environmental science evolved in the 19th
century, as the development of the steam engine brought on the Industrial
Revolution, and it continued into the 20th century. With growing population
and development of industries dependent on coal and liquid fuels, adverse
effects on humans and the environment started to be observed. Around the
middle of the 20th century several events accelerated the already growing
concern about the environment. For example, in London in 1953, thousands
of deaths were caused by smog formed by a combination of the intensive
use of coal for heating and poor air dispersion conditions; many
photochemical smog events triggered by traffic emissions occurred in the
USA from the 1940s to 1960s (Finlayson-Pitts and Pitts, 2000); and several
accidents occurred in the chemical industry (Seveso 1976, Bhopal 1984,
etc.). As environmental pollution became a public concern, more research
was carried out and regulations on pollutant emissions were developed. As
pollutants were discovered to migrate among environmental compartments
and regions of the world, new questions arose. In addition to requiring
information on the risk of pollutants to humans and the environment,
suddenly important questions were also “where are they?” and “how and
why did they get there?” These questions cannot be answered solely by
chemistry and toxicology. Thus, environmental science became the
combination of many different scientific disciplines, such as meteorology,
oceanography, hydrology, biology, geography and cartography, and GIS
became a powerful tool to aid in understanding the fate and transport of
pollution in our environment.

1.2.1 ENVIRONMENTAL CHEMISTRY

In the 20th century a high number of new chemical substances came into use
for agriculture, industry and in daily use, such as organochlorine pesticides
(OCPs), polychlorinated biphenyls (PCBs) and brominated flame retardants
(BFRs). Other compounds such as polycyclic aromatic hydrocarbons (PAHs)
started to occur in much higher levels mostly as by-products of industrial

13
activities and traffic (Gabrieli et al., 2010). However, it was not long before
adverse effects on both human health and the environment were
discovered. Moreover, many of these substances were measured in the most
remote regions of the world and in organisms that were never expected to
be exposed to these chemicals, and were found to bioaccumulate in the food
web (Schwarzenbach et al., 2003). The study of these compounds, their
emissions and their interactions with the environment has become known
as the discipline of environmental chemistry.

The focus of this thesis is specifically related to the use of GIS in

environmental chemistry applications, focussing on semivolatile organic
compounds in the environment, a few examples of which are described in
greater detail below.

Through the study of these substances’ properties, their environmental

behaviour and transport was better understood. With this better
understanding of their environmental properties and toxic effects, the
production and use of many of these compounds (e.g., DDT, PCBs) was
eliminated or greatly reduced (Schwarzenbach et al., 2003). However, large
amounts of these persistent and toxic pollutants were emitted to the
environment during their lifetime of use, and now are found distributed
across all environmental compartments and regions of the world
(Macdonald et al., 2000; Pozo et al., 2006). Some compounds (e.g., PAHs,
polychlorinated dibenzo-p-dioxins/furans ̵ PCDD/Fs) are the unintended
by-products of fossil fuel combustion, and thus, while their emissions can be
reduced, they cannot be entirely eliminated. Moreover, new substances are
being constantly introduced to the marketplace, with poorly known
properties and effects (Howard and Muir, 2010), potentially creating
furthers risks for humans and the environment, and presenting another
challenge for environmental scientists and policy makers.

As certain pollutants are the focus of the studies described in this thesis
(Section 2.1 describes studies focussed on PAHs and Section 2.2 on DDT),
they will be briefly described in following paragraphs.

1.2.1.1 Polycyclic Aromatic Hydrocarbons (PAHs)

PAHs are group of air pollutants formed and released into ambient air
during the process of incomplete combustion of organic matter (Finlayson-
Pitts and Pitts, 2000). They were always present in the environment due to

14
forest fires and volcanic activity, however today the most significant
sources of PAHs are linked to human activities such as industry,
transportation and heating. Levels of PAHs are often highest in urban areas
(Vallero, 2008).

PAHs are formed by two or more benzene rings (Fig. 1-2). PAHs vary in
their physicochemical properties and behaviour depending on the number
of benzene rings and their arrangement. In the atmosphere, their transport
and behaviour also strongly depends on the meteorological conditions. Due
to their carcinogenicity and mutagenicity, levels of PAHs are monitored and
regulated by many environmental monitoring programmes and national
environmental agencies. In Europe, this is through the Convention on Long-
range Transboundary Air Pollution
(http://www.unece.org/env/lrtap/lrtap_h1.html) and its associated
monitoring program, the European Monitoring and Evaluation Programme.

Figure 1-2: Structure a five-ring PAH: benzo[a]pyrene.

1.2.1.2 Dichlorodiphenyltrichloroethane (DDT)

Unlike PAHs, DDT (Fig. 1-3) is a substance that was synthetized by humans
and from the 1940s to the 1980s it was heavily used as an agricultural
insecticide. In the 1960s side effects of exposure to DDT were discovered
and described, leading to a ban on use and manufacturing of DDT in most
countries. Today use DDT is also regulated by the Stockholm Convention on
Persistent Organic Pollutants, and only a limited use is permitted in some
countries to control the spread of malaria. Despite not being used for many
decades, due to its long environmental half-life DDT is still often detected at
high levels in soils, and not only in areas of past use. Moreover, DDT is a
substance with complex behaviour. It is subject to long range transport,
mostly by air and water, often accumulating in sediments, but also biota.
(Schwarzenbach et al., 2003)

15
 Cl
Cl Cl

Cl Cl

Figure 1-3: Structure of p,p’-DDT

1.3 USE OF GIS IN ENVIRONMENTAL SCIENCE

GIS is now an essential part of environmental science. A majority of
environmental data would be almost valueless without their spatial
reference component. Together with knowledge from other scientific
disciplines and GIS tools, spatial reference is a key piece of information,
helping to reveal relations between the measured concentrations and other
environmental parameters such as wind speed, wind direction, air
temperature, and human activity parameters such as population,
agriculture and industrial activity. In this section, several examples of
successful GIS use in environmental science will be described.

1.3.1 OVERVIEW OF AREA OF INTEREST

The first time GIS was used for purposes of environmental science is not
documented, but was likely for a simple visualisation of area of interest or a
sampling site. There is no major innovation or sophisticated analyses in
simply displaying a piece of land, possibly with some points of interest such
as sampling sites, sources of pollution or polluted areas. However, there is
still strong value in such a simple picture, giving a quick, clear and
comprehensive overview of the area of interest, which often eases and
improves a reader’s impression and understanding of the topic itself. It can
give a lot of valuable details and relationships that would need to be
described in text in a long and complicated way (Fig. 1-4).

16
Figure 1-4: Example of overview map of a study area (from Sáňka et al.,
2014)

1.3.2 PLANNING SAMPLING CAMPAIGNS

A well-designed sampling campaign is a key component for a successful
environmental science study. When planning a sampling campaign, factors
such as money, sampling equipment and well trained staff must be
considered in order to select the correct number and the location of
sampling sites that will be sufficient for purposes of the study. Often, a
realistic limiting factor is cost, as sampling equipment and sample analysis
can be some of the most expensive items on a project budget. Therefore,
sampling campaigns should be developed considering many potential uses
of the data, so that resources are used as efficiently as possible. There are
usually two main purposes of sampling campaigns; (a) monitoring of
pollutant levels in relation to temporal or/and spatial changes and (b)
targeted case studies to address particular environmental questions about
fate and transport of particular compounds in particular locations,
providing data often used in environmental models. There are hundreds of

17
established environmental monitoring programs worldwide, focussed on
many types of substances and environmental matrices, varying also in
spatial scale of their coverage, time resolution of data they provide and
other parameters. Some major environmental monitoring programs
relevant to environmental chemistry are the European Monitoring and
Evaluation air quality monitoring programme (EMEP, www.emep.int) and
the Global Monitoring Plan on Persistent Organic Pollutants of the
Stockholm Convention (GMP, http://www.pops-gmp.org/). Other
monitoring of matrices such as soil and water are collected by EEA,
(www.eea.europa.eu). Smaller scale environmental monitoring campaigns
are carried on by national environmental institutes, government agencies or
other research institutes (e.g., Borůvková et al., 2015; Poláková et al., 2010).
The usefulness of monitoring depends in part on the choice of sampling
sites, and GIS can be used to improve the effectiveness of data through
better sampling design. For example, Behera et al. (2015) used spatial data
on NO2 levels, population density and urban land use to identify sampling
areas representative of traffic-influenced, urban and suburban pollution
levels.

Despite environmental matrices such as air, soil and water having very
different characteristics, GIS is a valuable tool to describe variability of all of
these and therefore potentially useful when designing a sampling campaign.
GIS provides an easy and efficient way of assessing the potential suitability
of each site by quantifying variables that can significantly influence
sampling results. Typical tasks may be to calculate total population in a
defined surrounding of the site (Fig. 1-5), assess local land use or air back
trajectories (Westgate and Wania, 2011), or to calculate the distance of the
nearest source of pollution (factory, road…). Often also characteristics like
hill slope, orientation or the site elevation are useful when taking the
decision of which site to select. A few examples on the most common
environmental matrices are given further.

18
Figure 1-5: Example of visualization of population count variability in
the area surrounding sampling sites (adapted from Čupr et al. (2015))

Ambient air is a very complex matter, mostly because of its high temporal
variability, driven mainly by meteorological processes. Therefore,
meteorological conditions such as prevailing wind direction, wind speed
and temperature are very important when choosing air sampling sites. A
good example of use of GIS to assess the appropriateness of sampling sites
is that described by Righini et al. (2014), who assessed the spatial
representativeness of air quality monitoring stations in Italy using spatial
variability in particulate matter (PM2.5 and PM10) emission sources,
suggesting sites with lower spatial variability in emissions are more
representative for the purposes of air quality monitoring.

Compared to air, soil has lower temporal variability, but spatial variability
of soils is much higher, and key characteristics often change within meters.
Therefore different sampling strategies are required. While air requires
greater temporal resolution in sampling, results of soil sampling requires
greater spatial resolution, and the temporal aspect is less important. Soil
sampling campaigns often focus on soils of specific type (e.g., fluvisols,
contaminated soils), use (agriculture, forestry), or other properties
(floodplains, soils with geogenic contamination). These features or their
intersection can be reliably determined by using GIS, given access to data

19
sources of sufficient quality and spatial resolution. Additionally, other
relevant conditions can be calculated, such as minimum distance from
potential pollution sources, maximum distance from a river, and elevation
(Daly et al., 2007). Finally, sampling sites can be selected using a random
selection function within the area of interest to ensure unbiased sampling
site selection, as it was done for example by Borůvka et al., (2014).

In case of sampling of surface waters such as streams and rivers,

information describing upstream watershed characteristics (area, average
slope, land use, population and presence of pollution sources) may be
required in order to support the choice of representative sampling sites
selection. Bridges or other river features potentially easing access to a river
for sampling, can be also identified by GIS analyses. In case of sediment
sampling, stream flow speed, derived from river slope, can be useful for
estimating potential sediment accumulation zones. The United States
Environmental Protection Agency (US EPA) uses GIS-based site selection
techniques to select the >1000 stream sampling sites in their Wadeable
Streams Assessment Program (US Environmental Protection Agency, 2005).

1.3.3 ESTIMATION TECHNIQUES FOR UNKNOWN VALUES

As mentioned above, sampling is an extremely time-consuming and costly
process and the density of sampling points is often not enough to cover the
whole study area and fully address all the study goals. This becomes even
more critical with studies over larger areas. Thus, techniques for estimating
values at unknown locations/regions based on the known values and
conditions at measured points become a necessary tool. GIS offers a few
methods for estimating unknown values.

1.3.3.1 Spatial interpolation

The term “spatial interpolation” refers to a group of methods developed to
estimate unknown values from point measurements. The majority of these
methods are based on the principle of spatial autocorrelation, which
assumes that the values of the environmental properties are more similar to
each other at sites that are closer to each other. This is, to a certain extent, a
limiting factor in these methods, as they are not designed to deal with
sudden changes and/or sharp gradients in environmental properties.
Therefore, while these methods can be extremely useful, careful
consideration should be given to which variables these methods can be
used for. For example, ambient air at background sites is a variable with a

20
sufficiently homogenous distribution (Halse et al., 2011), while in densely
populated urban areas with high heterogeneity in air concentrations, spatial
interpolation techniques fail (see fig. 1-6). Assuming there is enough
measurement sites, in soils, interpolation can also be useful for certain
parameters, such as concentrations of heavy metals (British Geological
Survey, 2015). However the concentration of a pesticide in soil is an
example of a phenomenon with a potential for abrupt changes, e.g., one field
could be treated with the pesticide, but not the neighbouring one or the
nearby forest.

a) b)

Figure 1-6: Comparison between results of (a) spatial interpolation

method of kriging and (b) land use regression, where kriging
performs poorly due to lack of measurement points. From Melymuk
(2012).

Probably the simplest and most known spatial interpolation method is

inverse distance weighting (IDW), calculating unknown values for each
location based on weighting distances of measured values, where the only
parameters involved in the computation are the search radius from each
interpolated location and the number of points involved in the
interpolation.

21
Another spatial interpolation method is kriging, originally developed in the
1950s for mining and named after its inventor Danie Krige (Krige, 1951)
and which has since diversified into many types of kriging. The most
common types are simple kriging, ordinary kriging and universal kriging.
Instead of a distance power function, kriging uses a variogram, analyzing
the relationship between distance between measurement points and the
relative change of values. This method then allows estimation of the
maximum distance between points within which the values influence each
other. Based on the properties of the data kriging is applied to, a suitable
type of kriging using a slightly different type of variogram can be used. For
example, a 3-dimensional suface version of a variogram can reveal possible
differences in variance depending on direction and adjust the variogram
model accordingly. For example, this situation may occur when modelling
dispersion from a source of air pollution under conditions when the wind
has one strongly prevailing direction. Kriging was successfully applied by
Guo et al. (2007) and Shad et al. (2009), who used more advanced types of
kriging, fuzzy membership kriging and indicator kriging, to predict PM10
concentrations across study areas in California and Tehran, respectively.

There exist several more methods of spatial interpolation. A very simple

one is using Thiessen polygons, for each point delineating the area of the
surrounding surface which is closer to the point than to any other. The
triangulated irregular network method connects selected neighbouring
points with lines, resulting into a network of triangles with values derived
from known values of their corners. This method is often used for
interpolating variables that tend towards abrupt changes, such as flat
valleys bordered by hilly areas (Pistocchi, 2014). Spline methods use
polynomial curves, applied on parts of the interpolated surface,
guaranteeing continual changes of values. The application of radial basis
functions is analogous to fitting a rubber membrane through measured
sample values while minimizing the curvature of the surface. A specific
character of this method compared to other interpolation methods, such as
IDW, is the ability to model higher values than the highest measured or
lower than the lowest measured, e.g., to extrapolate values. This can be
useful in case of interpolating elevation of a mountain, but missing the
measurement of its top; in that case IDW would create a “fake crater”. The
comparison of results of four different types of spatial interpolation on the
same data is presented by Figure 1-7.

22
Figure 1-7: Examples of four types of spatial interpolations: (a)
inverse distance weighting, (b) ordinary kriging, (c) spline and (d)
radial basis functions, on the same generic dataset, showing
differences between these methods.

1.3.3.2 Regression analysis

Regression analysis is able to provide a mathematical equation which can
serve to calculate values of an unknown (dependent) variable, based on
values of known (explanatory) variables at the same locality. Each
independent variable is associated with a regression coefficient describing
the strength and the sign of that variable’s relationship to the dependent
variable. Then, values of the dependent variable can be calculated using the
equation, in this case for linear regression:

= + + +⋯⋯ + (Eq. 1.1)

23
Where Y is the dependent variable, are the regression coefficients, X are
values of explanatory variables and ε is the residual.

Applying a regression equation on a raster data (as raster datasets are the
common type of data used in regression analysis) will calculate the values
of unknown (dependent) variable for each cell of the raster, creating a new
dataset of estimated values of the unknown variable, covering continuously
the whole study area. The new dataset can serve as a valuable aid for
decision makers, as well as an input for further analysis.

Regression analysis has been successfully applied in several studies aimed

of environment quality. Hafner et al. (2005) proved a strong regression
relation between PAHs levels and population within a 25 km distance from
sampling sites. However, the results also revealed different relationships
between PAH levels and population density in different parts of the world,
suggesting the possible presence of another explanatory variable not
considered in the study. Melymuk et al. (2013) used linear regression to
identify potential sources of semivolatile organic compounds (SVOCs) and
their distribution in Toronto, Canada, based on a sampling campaign across
the city and using appropriate explanatory variables such as building
sealant use for PCBs and distribution of transportation infrastructure for
PAHs. Similarly, Briggs et al. (1997) applied regression modeling on NO2
concentrations in three European cities, and traffic and land cover proved
to be the most important explanatory variables. In another study, Beelen et
al. (2010) compared concentrations of NO2 in a Dutch urban area acquired
by land use regression and air dispersion modeling, though revealing the
dispersion model performs better if results of both approaches are
compared to measured values. A similar statistical stochastic modelling
technique with regression trees was used to estimate soil concentrations of
persistent organic pollutants (POPs), including OCPs in soils of Czech
Republic (Kuboš ová et al., 2009).

However like any other method, regression analysis also has its risks and
challenges. There are several common problems one could encounter when
doing spatial regression. Important explanatory variables can be rejected if
the presence of outliers decreases the correlation with measured data. Also
selected explanatory variables may have a nonlinear relationship; they may
be redundant (multicollinearity) or differently related in different regions
of study area (nonstationarity). Another problem is spatial autocorrelation,

24
meaning spatial clustering of overpredicted or underpredicted values,
suggesting the model is missing a key variable. ArcGIS software and other
GIS software offer a broad variety of spatial regression tools, including also
those for detection of above mentioned issues.

1.3.4 SPATIAL ANALYSES AND THEIR USE

Spatial analyses are any types of data analyses that extract and/or produce
spatial information based on operations on different layers of data. GIS
provides the organizational framework which allows numerical and
statistical analyses to be easily applied on real-world data (Clarke, 2003).
The simple, but essential and fundamental ability to overlay a basically
unlimited number of GIS layers together with hundreds of tools (GIS
functions performing computations on spatial data) to be used for solving
issues of spatial analysis and spatial statistics is what makes GIS so different
from using hardcopy maps. The possibility to combine those tools into
sequences further increases what they can be used for. Combining tools into
sequences allows unique and valuable information needed for a specific
type of research to be obtained in an easy and fast way. For example,
Mavroulidou et al. (2004) designed a simple screening model, composed of
a GIS functions, for estimating vulnerability to traffic-induced air pollution
based on weighting and combining raster layers such as topography,
buildings, and traffic, and this performed successfully compared a
dispersion model. In another study, population density together with
backward air trajectory density were used to calculate Index of potential
source influence (IPSI) for 30 mountain sites across China in studies of
(Zheng et al., 2015, 2014), in order to help to explain distribution of PCBs in
soils. Righini et al. (2014) used spatial analysis functions embedded in GIS
to combine different layers of emission sources of selected pollutants to
assess spatial variability of pollution as one of the key parameters for
monitoring station location representativeness. In this study, other GIS
layers such as road network, settlement and land cover were employed to
explain the acquired values of emissions spatial variability. In study of
Hafner et al. (2005) ArcGIS was used to calculate population within a 25 km
radius of sampling sites to serve as an input to a regression analysis.
Sharma et al. (2015) used spatial analysis for delineating watersheds and
calculating the percentage of glacier-covered area within the watersheds in
order to better estimate the meltwater contribution to the flow of the
Ganges River. Möller (2010) performed a GIS-based assessment of the

25
influence of the growing number of wind turbines on population in
northern Denmark, combining elevation, population density, and data from
the Danish national wind turbine registry, revealing that less than 5% of
population is located outside the viewshed of wind turbines. These are only
a few examples of studies where spatial analysis with GIS demonstrates the
breadth of its utility in the field of environmental sciences.

1.3.5 THE ROLE OF GIS IN ENVIRONMENTAL MODELS

The complexity of pollutant fate was already mentioned in previous
sections. There is a desire and necessity not only to understand the current
fate and transport of pollutants, but also to predict their future behaviour.
Currently a large spectrum of environmental models is available, dealing
with different types of pollutants, environmental matrices and time and
scale resolution. A good overview is offered by Holmes and Morawska
(2006). Further, environmental models vary in their availability. Some are
commercial and costly, while there are others that are free and used mostly
for research purposes. There is also a high variability in the complexity of
environmental models, ranging from simple fugacity-based fate models
with less than ten input parameters (Mackay et al., 1992) to complex
models involving hundreds of parameters describing the substance
properties and inputs, as well as the conditions of the environment in great
detail (Nizzetto et al., 2015). Finally, the level of user-friendliness is an
important criterion when selecting appropriate environmental models.

Regardless of which environmental model the user picks, the need of at

least a limited amount of spatial information together with the contaminant
measurements or other input data is necessary. Such information can be a
digital elevation model, population in a watershed, soil organic carbon
content, river slope, location of pollution sources, etc. Moreover, many
environmental models also require receptor points into which the results of
the model are stored. These often must also be created by GIS. Although
recently several mainly commercial models have been updated to contain
embedded GISextensions, enabling the user both input data pre-processing
and model output visualization, generally there is still a lack of
compatibility between GIS and environmental models. Therefore many
environmental models, particularly those developed by research
institutions, do not offer the user the benefits of input GIS data pre-
processing. This makes the work with those models more challenging, time-

26
consuming and requires the user to have expertise in GIS in additional to in
modelling. The advantage of GIS extensions is that they allow the model to
be accessible to much wider user community, and this is one of the main
reasons why GIS extensions are more frequently incorporated in recent
environmental models, e.g. AERMOD (U.S. Environmental Protection
Agency, 2013). When using a model without a GIS data input/output
extension the only way of acquiring the spatial data required by the model
is using GIS separately, as demonstrated in the models of Bou Kheir et al.
(2014), Matějíček et al. (2006, 2003), Pistocchi et al. (2009), Sáňka et al.
(2014; 2013) and Zou et al. (2011).

An alternative and also a popular way that developed from the spatial
analysis techniques described in section 1.3.4 is incorporation of a simple
model into GIS itself. In this case, a GIS-based model incorporated within a
GIS environment takes advantage of opportunity of all tools offered by the
GIS software, usually enhanced by a sequence of equations representing the
modelled reality in a more authentic way. The model must be programmed
in the programming language used by the particular GIS software. In most
reported studies, ArcGIS and its programming language Python are used.
This approach’s advantages are (a) the possibility to use all available GIS
tools in the model, (b) compatibility with most common GIS data formats
and (c) no need of GIS data pre-processing and transformations into
different formats for the model input, as well as no need of model outputs
processing and transformations for visualization. Considering these
advantages, the approach of incorporating simpler environmental models
into GIS appears to be a potentially successfully developing field in the
future (Pistocchi et al., 2010a). Several examples proving this approach is
meaningful are given. For example, Briggs et al. (1997), in developing his
regression-based model for estimation of NO2 air concentrations, did all
modelling within a GIS environment. Gulliver and Briggs (2011) developed
an emission sources grid-based model embedded in an ArcGIS environment,
using GIS functions and simple Gaussian plume model. Similarly, Pilla and
Broderick (2015) used the ArcGIS “Model builder” tool and Python scripting
language to integrate various air dispersion models into one ArcGIS tool,
successfully modelling PM10 concentrations for Dublin, Ireland, based on
road network, traffic, building geometry, PM10 measurements and
meteorology data. In another study, Wan (2015) implemented a simple
model using population, land use and pesticide usage data into GIS,

27
resulting in estimate of vulnerability of population exposure to pesticides.
Atmospheric concentrations and deposition fluxes of PCBs and PCDDs/Fs
were estimated by Pistocchi (2008), applying map algebra and mass
balance equations on raster datasets of emission, transport, landscape and
climate maps. Complex input data pre-processing using GIS was done in the
study of Vienneau et al. (2009), who used several types of NO2 source
databases from across Europe to build a model within the ArcGIS
environment, based on the moving window principle, combining all input
datasets within a defined surroundings. The model was successfully tested
on measured NO2 data and the approach can be applicable to various
substances.

1.4 CHALLENGES OF GIS USE IN ENVIRONMENTAL SCIENCE

As was already outlined in previous sections, there will always be
challenges and issues to emerge when two scientific disciplines intersect,
and this is just as much the case for environmental sciences and GIS.
Although there is now more overlap between these two fields of study than
ever before, there are challenges and questions related to data availability,
quality reliability and compatibility, as well as on the suitability of GIS
approaches and visualization methods used for handling environmental
data. Some of the most common challenges are discussed further.

1.4.1 ENVIRONMENTAL DATA AVAILABILITY AND QUALITY

The possible ways for obtaining spatial data for GIS analysis were
mentioned in section 1.1.3.3. As described earlier, the easiest way is to
obtain data from the large number of GIS datasets available online.
However, with this method, using datasets that someone else prepared, one
should always consider that the datasets are often the result of other
approximations. Thus, the values were produced using one of the
estimation methods (i.e., regression, extrapolation, interpolation or GIS-
based modeling) described in section 1.3.3, and

There is a large amount of both vector and raster spatial data available
online. An important considering with respect to available data is the level
of generalization applied, for example, socioeconomic data or pesticide use
data is often generalized at the level of administrative units; this can

28
present a challenge for the data user if that level of generalization is not
appropriate to the intended use.

To assess the suitability of GIS datasets for the intended purpose, metadata
are often helpful as they should contain information such as the data set
content, spatial resolution, coverage, format, recommended scale and year
of creation. Metadata of sufficient quality can ease and speed up the process
of selecting the most suitable dataset. This information is supposed to be
provided together with the data, however, it is often lacking. In reality, the
data user often has to rely on datasets of unsure origin, age and quality. Of
course there is a higher chance of encountering missing or incomplete
metadata in case of data that are available for free, rather than those which
the user must pay for.

Overall, it can be stated that there is a growing amount of environmental

data available for use in GIS analyses, enabling running more exact and
more reality representative analyses, however the background and the
origin of the data should be always kept in mind.

1.4.2 COMPLEXITY OF SUBSTANCE BEHAVIOUR

There are several challenges inherent in the sources and behaviour of
pollutants that make them challenging to characterize and model. Key
physicochemical parameters, namely solubility, vapor pressure, partition
coefficients and degradation rates, are determined based on results of
laboratory experiments (Pistocchi, 2014) or chemical structure modeling.
However, these values are often highly uncertain (Pontolillo and
Eganhouse, 2001). The conditions of the real environment are often much
more complex than in the laboratory, depending on many other
environmental properties, such as soil type, pH, organic matter content, air
temperature and humidity, biota etc., which are difficult to fully capture in
the necessary detail in spatial data or modelling.

1.4.3 LINKING GIS WITH MODELS

As already mentioned in section 1.3.5, the availability of GIS input data pre-
processing extensions in environmental models is still not common. More
often, the model requires the data as a text file of a very exact structure.
This very often means multiple conversions among data formats and time-

29
consuming formatting of the input text files. Still, this approach can be
considered a time saving alternative to preparing input data manually, also
increasing the accuracy and reducing probability of random errors
occurrence. Moreover, GIS tools can be useful for preparing a more complex
receptor point grid. Examples of studies where this approach was
successfully applied will be given in Chapter 2.

Furthermore, results of modelling are often delivered in a form which needs

manual processing in order to be imported into GIS for further analysis or
visualization.

There are several other minor issues to be considered, either when working
with a model that is able to deal with GIS data, or preparing the input files
outside the model environment. One of them is the coordinate system. It
was already mentioned in section 1.1.3 that there are many different
coordinate systems, often also using different units. Attention should be
given to ensuring that the coordinate system and units required by the
model are common among all input datasets. This avoids errors of incorrect
dataset placement and scaling. Another challenge is selection of the
appropriate level of spatial generalization of the input dataset. The
responsibility for this decision is mainly on the user, as he is the one to
select, e.g. which sources of pollution should be involved and which can be
neglected, or what is the required resolution of the input digital elevation
model. An overly high level of generalization increases the risk of skewing
the modelling results by omitting features influencing them. On the other
hand, using too detailed inputs can prolong the computational time. This
also depends on the performance of the hardware used, which is another
consideration when working with big datasets and models performing
complex computations.

30
31
32
 CHAPTER 2
EXPERIMENTAL WORK

33
2 EXPERIMENTAL WORK
The modelling of environmental pollution is a challenging task due to the
complexity of the environment, wide range of pollutant properties and
behaviours, and temporal and spatial heterogeneity. Therefore, there are
many different approaches to the modelling of environmental pollution.
Some are simple and fast, focusing only on one environmental
compartment, bringing results that are more likely suitable for screening
purposes; other ones can be complex, describing processes among more
matrices and giving a more reliable and more detailed elucidation, however
requiring long development, many types of input data and experienced
users.

This chapter gives details on case studies where GIS was employed in
environmental modelling. Six scientific publications will be described in
three sections according to the type of study, environmental matrix and the
way in which GIS contributed to the study. The variety of environmental
compartments and breadth of GIS techniques used demonstrate the
potential of GIS in the field of environmental science.

34
LIST OF ARTICLES DESCRIBED IN THIS CHAPTER AND THE AUTHOR’S
CONTRIBUTION TO THE ARTICLES

Sá ň ka, O., Dvorská , A., Klá nová , J. & Čupr, P. Modelování rozptylu
vybraných atmosférických PAU v ovzduší Valašského Meziříčí
pomocí modelu SYMOS’97 a GIS a jeho porovnání s výsledky
pasivních odběrů ovzduší. Ochr. Ovzduší 25, 13–21 (2013).
 Ondřej Sáňka did all GIS-related work, prepared input data
for the model and ran it, interpreted the results and wrote
Described
the manuscript
in Section
Sáňka, O., Melymuk, L., Čupr, P., Dvorská, A. & Klánová, J. Dispersion
2.1
modeling of selected PAHs in urban air: A new approach combining
dispersion model with GIS and passive air sampling. Atmos. Environ.
96, 88–95 (2014).
 Ondřej Sáňka did all GIS-related work, prepared input data
for the model and ran it, interpreted the results and wrote
the manuscript
Sanka, O., Kalina, J., Lin, Y., Brabec, K., Deutscher, J., Futter, M.,
Butterfield, D. & Nizzetto, L.,. Modeling hydrological and
biogeochemical controls of the dissipation of p,p’-DDT in soil.
Described
Submitted to Environ. Sci. Technol. (2015).
in Section
 Ondřej Sáňka did all GIS-related work, prepared input data
2.2
for the models, ran and calibrated one of the models
prepared all maps and most graphs, and contributed to
writing the manuscript
Zheng, Q., Nizzetto, L., Mulder, M., Sáňka, O., Lammel, G., Li, J., Bing, H.,
Liu, X., Jiang, Y Luo, Ch. & Zhang, G. Does an analysis of
polychlorinated biphenyl (PCB) distribution in mountain soils across
China reveal a latitudinal fractionation paradox? Environ. Pollut. 195,
115–122 (2014).
 Ondřej Sáňka did all GIS-related work and the associated
data processing and produced 3 figures
Zheng, Q., Nizzetto, L., Li, J., Mulder, M., Sáňka, O., Lammel, G., Bing, H.,
Liu, X., Jiang, Y., Luo, Ch. & Zhang, G. Spatial distribution of old and Described
emerging flame retardants in Chinese forest soils: Sources, trends in Section
and processes. Environ. Sci. Technol. 49, 2904–2911 (2015). 2.3
 Ondřej Sáňka did all GIS-related work and the associated
data processing and produced 3 figures
Melymuk, L., Bohlin, P., Sáňka, O., Pozo, K. & Klánová, J. Current
challenges in air sampling of semivolatile organic contaminants:
sampling artifacts and their influence on data comparability.
Environ. Sci. Technol. 48, 14077–91 (2014).
 Ondřej Sáňka did all GIS-related work and produced all the
figures

35
2.1 GIS-BASED ASSESSMENT OF PAHS CONCENTRATIONS IN AIR
USING AIR DISPERSION MODELLING
Despite decades of development, measuring SVOC concentrations in air still
remains challenging. Passive air sampling (PAS) is advantageous because it
is relatively cheap, not dependent on power supply and able to integrate air
concentrations over periods of weeks to months (Klá nová et al., 2006). PAS
are well suited to sampling gas-phase atmospheric compounds, and the
methods for quantifying air concentrations of these compounds are well-
validated (Bohlin et al., 2014; Klá nová et al., 2008). PAS are also able to
sample particle-phase atmospheric compounds, however there is greater
uncertainty in the uptake rates of particle-bound compounds, and thus this
is the topic of on-going research (Holt et al., in prep). PAHs are an important
compound for atmospheric monitoring because of the wide range of
emission sources and serious health effects (e.g., carcinogenic). However,
many PAHs can be challenging to quantify through PAS because their gas-
particle partitioning in air, determined by the volatility of the compound,
can shift between predominantly gas- or predominantly particle-bound at
ambient atmospheric temperatures. Thus, it is crucial to understand the
tools and techniques for proper quantification of these compounds.

In the Czech Republic, a national passive air sampling campaign called

MONET is on-going since 2003 (RECETOX Masaryk University, 2015). The
cities of Valašské Meziříčí and Liberec are characterized by the presence of
heavy industry and complex terrain conditions, making both of them
vulnerable to elevated concentrations of PAHs. Therefore, more attention
was given to these regions in the MONET sampling design, with four and
seven sites selected in Valašské Meziříčí and Liberec, respectively.Besides
direct air measurements, another method of air pollution assessment is air
dispersion modelling. SYMOS’97 (Stationary Sources Modeling System) is a
regional Gaussian dispersion model predicting atmospheric concentrations
of pollutants to a distance of up to 100 km from sources (Bubnik et al.,
1998). As Czech law requires the use of this model for air quality
assessment, it is used by state institutions, environmental agencies, and
private companies.

In two studies (Sáňka et al., 2014, 2013), SYMOS’97 was used to model
concentrations of four selected PAHs: benzo(a)pyrene,
benzo(b)fluoranthene, benzo(k)fluoranthene and indeno(123cd)pyrene

36
(BAP, BBF, BKF and IPY, respectively). These were selected based on their
known adverse effects on human health (Schwarzenbach et al., 2003) and
the availability of the emission inventory used for the purposes of
dispersion modelling. The pollutant emissions were adjusted to match the
13 passive air sampling intervals covering approximately the 2006 MONET
campaign. The SYMOS’97 model was run using the adjusted emission
inventory data, and the results from the dispersion modelling were
compared with measured values from the MONET sampling campaign.

Attention was given to precise preparation of four types of input data for
the dispersion modelling (elevation, sources, meteorology and receptors).
The model only allows loading input data in the form of a text file; therefore
several data format transformations were needed. Use of GIS was a crucial
part of preparation of all input data except for meteorology (obtained from
Czech Hydrometeorological Institute – CHMI), which is represented in a
limited way in the model, without considering any spatial aspects. GIS was
used to process the elevation data and prepare the elevation input file,
where a spatial resolution of 100 m was used for the study area of Valašské
Meziříčí and 150 m was used for the area of Liberec. Sources of pollution
required further division into three subcategories: (a) point sources, (b)
linear sources, and (c) areal sources. Point sources were mainly industry,
the records of which were taken from the Registry of Sources of Air
Pollution (REZZO) (CHMI, 1996), and located using their addresses. Linear
sources were road traffic; roads were digitized and the emissions were
calculated based on the major road census of the Road and Motorway
Directorate of the Czech Republic (RSD, 2006) and emission factors of
different fuel and vehicle types (Ntziachristos and Samaras, 2000). GIS was
used to convert linear features representing roads into points, each
representing emissions of a road segment of 50 m in the case of Valašské
Meziříčí and 100 m in the case of Liberec. Areal sources of pollution
described local heating, and were available in form of sums of fuels used per
settlement unit from REZZO. Annual fuel use was apportioned to individual
time periods based on daily mean air temperatures. Emission factors of
different fuels, which are part of the REZZO database, were used to convert
fuel into emissions. To locate the urbanized areas the CORINE Land Cover
database (European Environmental Agency, 2010) was used. As with the
linear sources, also these areas are required in the form of points.

37
Therefore, the polygons representing urban areas were converted into
regular grids of 200 m resolution.

Two grids of receptors were created to capture the outputs of the model
calculations. A regular grid was created to evenly cover both study areas,
and a extra irregular grid was added, increasing the density in urban areas,
along the roads and in the areas of sampling sites. The ArcGIS model builder
utility was used to automate the process of receptor creation along linear
pollution sources, placing receptors at the required distance from the road,
guaranteeing uniform distance between these receptors and linear
pollution sources to avoid predicting varying concentrations due to
different distances from the pollution source.

In both urban areas the dispersion model predicted lower concentrations of

PAHs compared to values measured by passive air sampling. Although both
approaches showed a seasonal trend with high concentrations in autumn
and winter months, the modelled trends were smoother than the measured
ones, showing also a significant variability between sequential samples.
According to the model, the strongest source of selected PAHs is local
heating, explaining the high concentration predictions in winter months,
and contrasting with concentrations close to zero in summer. An exception
is a site that is located close to a strong industrial source of pollution – a
bitumen mixing plant nearby Liberec. A better agreement between
modelled and measured data was observed at urban sites compared with
the suburban or rural sites. Suburban and rural sites are more likely to have
higher influence of emissions from unrecorded local biomass burning and
heating which are not captured in the model, thus accounting for some of
the disagreement between modelled and measured data. In comparison,
measured values at urban sites had lower variability between sequential
samples and a lower influence of sources not recorded in the emission
inventory, leading to better agreement between modelled and measured
data.

The discrepancies between measured and modelled data were generally

attributed to three factors: (a) weaknesses of the dispersion model, (b)
uncertainties in air concentration quantification and (c) the impact of
sources not included in the emission inventory, as mentioned above. The
model SYMOS’97 has insufficient representation of meteorology and
pollutant behaviour and does not consider processes such as scavenging or

38
pollutant transformations. Another source of discrepancies may be the fact
that passive air sampling is a semi-quantitative method, and the methods of
sampling rate estimation and its dependence on meteorological conditions
are still being developed (Holt et al., in prep; Klá nová et al., 2008; Petrich et
al., 2013). Finally, discrepancies between measured and modelled values
can also be attributed to the presence of pollution sources that are not
included in the emission inventory, such as combustion of wood for heating,
other biomass burning and summer forest fires, documented in nearby
regions of Liberec at the time that elevated concentrations were measured.
(JRC-IES, 2008, 2007).

The results of air dispersion modelling were processed using GIS. For
visualization, a method of ordinary kriging was used (Fig. 2-1 and 2-2).
Results of these studies showed that GIS is a useful tool, capable of
simplifying and speeding up pre-processing of input data for environmental
modelling, and of visualizing modelling results. Passive air sampling was
shown to be a useful source of data for air quality assessment; together with
air dispersion modelling it can be used to reveal undocumented sources of
pollution. Further research should be aimed at development of models that
integrate GIS with better abilities to describe meteorology, atmospheric
processes and pollutant behaviour in a more thorough way. Also more
attention should be put on GIS data pre-processing extensions for
environmental models.

The results of these studies were published in Ochrana Ovzduší -

Modelování rozptylu vybraných atmosférických PAU v ovzduší Valašského
Meziříčí pomocí modelu SYMOS'97 a GIS a jeho porovnání s výsledky
pasivních odběrů ovzduší (English title: Dispersion modelling of selected
atmospheric PAHs in the city of Valašské Meziříčí based on the SYMOS’97
dispersion model and GIS, and comparison of results using passive air
sampling) (Sáňka et al., 2013) and in Atmospheric Environment - Dispersion
modeling of selected PAHs in urban air: A new approach combining
dispersion model with GIS and passive air sampling (Sáňka et al., 2014) and
are attached as Annexes 1 and 2 of this dissertation thesis.

39
Figure 2-1: concentrations of benzo(a)pyrene in Valašské Meziříčí,
calculated by SYMOS’97 dispersion model in (a) winter and (b)
summer (Sáňka et al., 2013) .

Figure 2-2: Concentrations of benzo(a)pyrene in Liberec, calculated

by SYMOS’97 dispersion model in (a) winter and (b) summer (Sáňka et
al., 2014).

40
2.2 ASSESSMENT OF LONG TERM TEMPORAL TRENDS OF P,P’-DDT
SOIL CONCENTRATION USING GIS-AIDED MECHANISTIC MODEL
Very often, pollutant behaviour requires the use of models that consider
processes in multiple environmental compartments, as the pollutants
exchange among them, i.e., multimedia environmental models. However,
more complex pollutant behaviour and more environmental compartments
also inevitably leads to higher data requirements in order to predict
pollutant distribution in the environment.

This study (Sanka et al., 2015) reports on an investigation into the long
term trends of the pool of p,p’-DDT in Central Europe. Organochlorine
pesticides (OCPs) were widely used for agricultural application on a global
scale until successive prohibitions on their use from the 1970s to 1990s,
and their subsequent restriction through the Stockholm Convention. Today,
many agricultural soils have the legacy of the wide use of OCPs due to past
local use and long-range transport. For this study, the Morava River
catchment was selected due to its high agricultural land use, and the
availability of a unique dataset of several hundreds of records of OCP
concentrations in air, soil, river water, and river sediment. p,p’-DDT was
chosen for environmental modelling as a representative of the OCP
compound class. A stochastic model was used to determine temporal trends
of p,p’-DDT levels in soils using soil concentration measurements. These
trends were compared to the results of p,p’-DDT fate modelling performed
with the Integrated Catchment Model – Contaminants (INCA-
Contaminants). INCA-contaminants is a recently developed complex, high-
resolution integrated hydrobiogeochemical multimedia fate model,
designed for water quality assessment (Nizzetto et al., 2015).

The INCA-contaminants model required a broad set of spatial and

hydrological inputs, ranging from river parameters, meteorology, and
measured p,p’-DDT concentrations. For the description of the catchment
and river structure, data on stream orders, their length and width between
reach points, as well as characteristics of individual subcatchments,
including soil properties and land use information were required. Further,
the INCA-Contaminants model requires hydrologically effective rainfall and
soil moisture deficit time series to generate daily run-off and river flow
predictions; this was computed using the Precipitation, Evapotranspiration
and Runoff Simulator for Solute Transport model (PERSiST, Futter et al.,

41
2014). PERSiST outputs were calibrated using catchment-specific
observations of river flow, air temperature and rainfall that were obtained
from CHMI. The calibration of INCA–Contaminants for dissolved organic
carbon (DOC), size and properties of the sediment compartment and p,p’-
DDT concentrations was based on measured values from 2007 and 2008
(Czech Hydrometeorological Institute, 2015; Prokeš et al., 2012).

In this study GIS played a crucial role in acquiring spatial data for
representation of geographical and meteorological properties of the Morava
river catchment and its subcatchments. The river network was generalized
for the purposes of the study and the DIBAVOD database was used for
subcatchment delineation based on locations of discharge gauging stations
(Fig. 2-3). Attributes such as area and land use class coverage (from the
International Steering Committee for Global Mapping, 2015) for all
subcatchments were calculated, as well as length and average slope and
width of each river stretch. Daily rainfall and temperature data from nine
CHMI measurement stations were extrapolated to cover the whole
catchment using Thiessen polygons. From the polygons, the subcatchments
were allocated to measurement stations by the fraction coverage of the
polygons, and the fractional coverage was used to calculate a weighted
average for meteorological records. In this way, daily weighted mean
temperature and rainfall records for 2006-2008 were approximated for
each subcatchment. To verify the correctness of this approach, the total
amount of water discharge from each subcatchment for 2006-2008 was
calculated and compared to the total amount of precipitation each
subcatchment received within the same time period. Results ranged
between 91% of precipitation volume leaving the subcatchment as river
water in case of the subcatchments with the most forest, most precipitation
and lowest population density, and 8% in case of the catchments that were
driest, most densely populated and most heavily used for agriculture, which
is in agreement with expectations. GIS data on soil organic carbon (SOC)
and DDT concentration in soils of the Czech Republic at a spatial resolution
of 1 km, created by Kubošová et al. (2009) were averaged per land use class
and used as initial input values for the INCA model runs. GIS was further
used for visualization of model input data for the purpose of publication.

Based on these inputs and a successful calibration of water DOC, suspended

sediments and concentrations of p,p’-DDT in water and sediments, the
INCA–Contaminants model predicted temporal trends in p,p’-DDT soil

42
concentrations in the Morava catchment and overall p,p’-DDT fate in the
catchment for the period between 2006 and 2008. Results of the model
showed a satisfactory agreement with the observed values in terms of the
decline in concentrations of p,p’-DDT in soil compared to the trend
observed in hundreds of samples taken in the Czech Republic during last
decades (Borůvková et al., 2015). According to the model’s results the
catchment acts as a sink of p,p’-DDT as total losses are predicted to be more
than an order of magnitude higher than total atmospheric inputs to the
catchment. Degradation appeared to be the major dissipation pathway of
p,p’-DDT from the catchment, explaining approximately 90% of total losses.
The second most important dissipation pathway was runoff, exceeding
degradation only in cases of extreme floods. Additionally, the watershed
parameterization and calibrations of DOC and sediment can be useful for
modelling other pollutant fates for the same time period and area.

Results of this study were submitted to Environmental Science & Technology

- Modeling hydrological and biogeochemical controls of the dissipation of
p,p’-DDT in soil (Sanka et al., 2015) and are attached as Annex 3 of this
dissertation thesis.

43
Figure 2-3: Land Use, subcatchment division and measurement sites in
study area of the Morava River watershed (Sanka et al., 2015).

2.3 USE OF SPATIAL ANALYSIS FOR ENVIRONMENTAL RESEARCH

SUPPORT
As described in Chapter 1, there are a great number of techniques in GIS
which can be useful in environmental research. This section describes
additional studies in which GIS served as a complementary technique,
improving the development of studies, interpretation, and visualization and
presentation of results.

Spatial analysis was used in two studies focused on assessment of PCBs and
flame retardants (FRs), in mountain forest soils of China (Zheng et al., 2015,
2014). Thirty mountains were selected across China, from which 159 soil
samples were collected at 82 locations in 2012 and 2013. GIS was used to
plot measured concentrations of PCBs in relation to elevation (Fig. 2-4) as
well as to plot FRs in relation to population. Total annual precipitation and
annual mean temperature values were extracted for all mountain sites from
interpolated global climate data (Hijmans et al., 2005). In addition, Index of

44
potential source influence (IPSI) was created and calculated for each site.
IPSI combines data on population density distribution (expected to be a
proxy for PCB and FR source distribution) with backward trajectory
density. For each site, 360 ten-day backward trajectories were calculated,
covering one year preceding the sampling, for the purpose of IPSI
calculation, which was limited to radius of 6000 km for each site. An
example of IPSI calculation source data for one site is given in Fig. 2-5.

For PCBs it was not possible to separate the influence of IPSI due to a close
covariance between IPSI and total yearly precipitation. For concentrations
of the flame retardants, 13 of 16 measured FRs showed positive correlation
with IPSI; for seven substances this correlation was statistically significant
(p < 0.05). These are mostly brominated flame retardants (BFRs) that are
currently undergoing phase-out. This suggests that the primary sources of
these substances are mostly consumer products remaining in use, which are
expected to show high correlation with population density, from which IPSI
is derived. In contrast, FRs that are not regulated have significant
production in industrial and manufacturing areas in China, and thus show
lower correlation with population density-based IPSI.

Results of these studies were published in Environmental Pollution - Does

an analysis of polychlorinated biphenyl (PCB) distribution in mountain soils
across China reveal a latitudinal fractionation paradox? (Zheng et al., 2014)
and Environmental Science & Technology - Spatial distribution of old and
emerging flame retardants in Chinese forest soils: Sources, trends and
processes (Zheng et al., 2015) and are attached as annexes 4 and 5 of this
thesis.

45
Figure 2-4: PCB concentrations at different elevations of mountain
forest soils of China. Calculated values of index of potential source
influence (IPSI) are also displayed (Zheng et al., 2014).

46
Figure 2-5: Example of overlaying population data and backward
trajectories for IPSI calculation in GIS environment (Zheng et al.,
2014).

Spatial analysis was also used in a review paper describing the challenges in
air sampling of SVOCs. Polybrominated diphenyl ethers (PBDEs) were
selected as an example group of pollutants representing typical air
monitoring campaigns and case studies over the past ten years. All available
PBDEs sampling campaigns from 2003 - 2013 were reviewed and recorded,
noting variability in sampling methods, and spatial and temporal density of
air sampling. The sampling points corresponding with these studies were
imported into GIS, for a total of 779 records, to allow visualization of global
biases in the distribution of air sampling. Discrepancies between population
density and PBDEs sampling density were emphasized, as well as high
variability of PBDEs active air sampling volumes with a geographic bias
towards polar regions. The review suggests that due to the effects of air
sampling artifacts such as breakthrough, degradation, particle fractions and
sample deployment periods, different sampling methods may
underestimate SVOCs concentrations by up to 95%. The GIS allowed
visualization of how these air sampling artifacts may have geographic
biases (fig. 2-7) and a disconnect between population density and density of
air sampling for PBDEs (fig.2-6).

47
Results of these study were published in Environmental Science &
Technology - Current challenges in air sampling of semivolatile organic
contaminants: sampling artifacts and their influence on data comparability
(Melymuk et al., 2014) and are attached as annex 6.

Figure 2-6: Overlaying population data and PBDE sampling frequency

(Melymuk et al., 2014).

Figure 2-7: Visualization of active air sampling volumes in relation to

latitude (Melymuk et al., 2014).

48
49
50
 CHAPTER 3
HIGHLIGHTS AND CONCLUSIONS

51
3 HIGHLIGHTS AND CONCLUSIONS
Many environmental issues are related to chemical pollution and its effects,
which is the focus of environmental chemistry. The complex behaviour of
thousands of pollutants requires a multidisciplinary approach to
environmental pollution assessment, using knowledge and tools from other
scientific disciplines, such as meteorology, climatology, geology, hydrology,
demography, geography and cartography and geographic information
systems (GIS). This thesis described and demonstrated the potential for use
of cartography and GIS in environmental science, and particularly in
environmental chemistry.

The theoretical part of this thesis briefly introduced both environmental

chemistry and GIS, described the reasons for and benefits of use of GIS in
environmental chemistry and demonstrated the most useful GIS techniques
for solving the challenges of environmental chemistry. In particular, spatial
interpolation, regression analysis, and use of GIS in conjunction with
environmental models was discussed. The discussion also addressed the
most common challenges and complications related to the use of GIS in
environmental chemistry.

The experimental part of the thesis introduced six environmental studies

grouped in three sections, each focusing on a different use of GIS. The first
section (2.1) described the use of GIS to prepare input terrain, source, and
receptor data for modelling dispersion of PAHs in two case studies of cities
in Czech Republic. The results of the modelling were also processed and
visualized using GIS and further compared to results of passive air
sampling, identifying the presence of undocumented sources of pollution
such as local-scale biomass burning, heating and possibly also forest fires.

The second section of the experimental part of this thesis (2.2) described
the use of GIS for processing, preparation and visualization of input data for
complex, multi-compartmental modelling of p’p-DDT fate and transport in
the Morava River catchment in central Europe. The model was calibrated
with several types of p’p-DDT concentration measurements and predicted a
decline of p’p-DDT soil concentrations with a half-life of 3000 -3800 days, in
agreement with trends based on measurements. Degradation in soil was
identified as the major p’p-DDT dissipation pathway in the catchment. The
use of GIS played a crucial role in this study, creating the model of the

52
subcatchment network and relating data on meteorology, hydrology, land
use and soil properties to the subcatchment areas, resulting in a successful
model calibration. Moreover, the parameters of the watershed properties’
calibration can be used in further studies for modelling other pollutants’
fate.

Third section of the experimental part (2.3) introduced three studies where
GIS analysis was used as an important support to the outcomes of the
studies. Two studies were focused on analyzing and explaining the
distribution of PCBs and FRs in mountainous areas of China. Soil
concentrations of PCBs and FRs were evaluated in relation to temperature
and rainfall data and a site classification index, based on the combination of
population density and back trajectories. Finally, GIS analysis was used in a
review study on current air sampling techniques, revealing discrepancies
between population and air sampling network density and analyzing the
variability of air sampling methods across different parts of the world.

The broad variety of GIS techniques used in the studies presented in this
thesis for many areas of environmental chemistry research and on many
environmental compartments (air, soil, water) proves the potential for use
of GIS not only in environmental chemistry, but also broadly within the
environmental sciences. While there are still certain challenges to be
overcome when using GIS techniques in environmental sciences, the future
of data interpretation in environmental science will be greatly strengthened
by a multidisciplinary approach incorporating GIS techniques.

53
4 REFERENCES
Beelen, R., Voogt, M., Duyzer, J., Zandveld, P., Hoek, G., 2010. Comparison of
the performances of land use regression modelling and dispersion
modelling in estimating small-scale variations in long-term air
pollution concentrations in a Dutch urban area. Atmos. Environ. 44,
4614–4621.
Behera, S.N., Sharma, M., Mishra, P.K., Nayak, P., Damez-Fontaine, B., Tahon,
R., 2015. Passive measurement of NO2 and application of GIS to
generate spatially-distributed air monitoring network in urban
environment. Urban Clim. doi:10.1016/j.uclim.2014.12.003
Bohlin, P., Audy, O., Škrdlíková, L., Kukučka, P., Přibylová, P., Prokeš, R.,
Vojta, Š., Klánová, J., 2014. Outdoor passive air monitoring of semi
volatile organic compounds (SVOCs): a critical evaluation of
performance and limitations of polyurethane foam (PUF) disks.
Environ. Sci. Process. Impacts 16, 433–44. doi:10.1039/c3em00644a
Borůvka, L., Sáňka, M., Vácha, R., Šrámek, V., Drábek, O., Tejnecký, V.,
Antonín, N., Němeček, K., Houška, J., Fraňková, A., Ash, C., Galušková, I.,
Hubová, P., Fadrhonsová, V., Jurkovská, L., Čechmánková, J., Čermák, P.,
Horváthová, V., Skála, J., Čupr, P., Hofman, J., Sáňka, O., Rotter, P.,
Vašíčková, J., 2014. Metody hodnocení zátěže lesních půd rizikovými
látkami a identifikace ekologických rizik kontaminace lesních půd.
Prague.
Borůvková, J., Gregor, J., Šebková, K., Bednářová, Z., Kalina, J., Hůlek, R.,
Dušek, L., Holoubek, I., Klánová, J., 2015. GENASIS – Global
Environmental Assessment and Information System [WWW
Document]. GENASIS – Glob. Environ. Assess. Inf. Syst. [online].
Masaryk Univ. URL http://www.genasis.cz/ (accessed 6.17.15).
Bou Kheir, R., Shomar, B., Greve, M.B., Greve, M.H., 2014. On the quantitative
relationships between environmental parameters and heavy metals
pollution in Mediterranean soils using GIS regression-trees: The case
study of Lebanon. J. Geochemical Explor. 147, 250–259.
doi:10.1016/j.gexplo.2014.05.015
Briggs, D.J., Collins, S., Elliott, P., Fischer, P., Kingham, S., Lebret, E., Pryl, K.,
Van Reeuwijk, H., Smallbone, K., Van Der Veen, A., 1997. Mapping
urban air pollution using GIS: a regression-based approach. Int. J.
Geogr. Inf. Sci. 11, 699–718. doi:10.1080/136588197242158
British Geological Survey, 2015. Lead (Pb) in shallow soils [WWW
Document]. URL http://www.bgs.ac.uk/gbase/maps/soil/lead.html
(accessed 10.28.15).

54
Bubnik, J., Keder, J., Macoun, J., Manak, J., 1998. SYMOS’97 - Stationary
Sources Modeling System - a reference manual (in Czech). CHMI,
Prague.
Buchroithner, M.F., Fernández, P.A., 2011. Cartography in the Context of
Sciences: Theoretical and Technological Considerations. Cartogr. J. 48,
4–10. doi:10.1179/1743277411Y.0000000003
Clarke, K.C., 2003. Getting Started with Geographic Information Systems,
4th ed. Pearson Education, Inc., Upper Saddle River, NJ.
Czech Hydrometeorological Institute, 2015. ARROW - Assessment and
Reference Report of Water Monitoring [WWW Document]. URL
http://hydro.chmi.cz/isarrow/index.php (accessed 5.30.15).
Čupr, P., Prokeš, R., Sáňka, O., Přibylová, P., Bednářová, Z., Borůvková, J.,
Chropeňová, M., Šebková, K., Václavíková, J., Vaňková, L., Kalina, J.,
Audy, O., Kalábová, T., Holoubek, I., Klánová, J., 2015. Časové a
prostorové hodnocení koncentrací perzistentních organických
polutantů ve volném ovzduší ČR. Specializované mapy s odborným
obsahem. Brno, in press .
Daly, G.L., Lei, Y.D., Teixeira, C., Muir, D.C.G., Wania, F., 2007. Pesticides in
Western Canadian Mountain Air and Soil. Environ. Sci. Technol. 41,
6020–6025. doi:10.1021/es070848o
ESRI, 2015. Arc GIS Desktop.
European Environmental Agency, 2010. Corine Land Cover.
Finlayson-Pitts, B.J., Pitts, J.N., 2000. Chemistry of the Upper and Lower
Atmosphere, 1st ed. Elsevier, San Diego. doi:10.1016/B978-0-12-
257060-5.50027-7
Futter, M.N., Erlandsson, M.A., Butterfield, D., Whitehead, P.G., Oni, S.K.,
Wade, A.J., 2014. PERSiST: a flexible rainfall-runoff modelling toolkit
for use with the INCA family of models. Hydrol. Earth Syst. Sci. 18,
855–873. doi:10.5194/hess-18-855-2014
Gabrieli, J., Vallelonga, P., Cozzi, G., Gabrielli, P., Gambaro, A., Sigl, M., Decet,
F., Schwikowski, M., Gäggeler, H., Boutron, C., Cescon, P., Barbante, C.,
2010. Post 17th-century changes of European PAH emissions recorded
in high-altitude Alpine snow and ice. Environ. Sci. Technol. 44, 3260–6.
doi:10.1021/es903365s
GIS geography, 2015. GIS geography [WWW Document]. URL
http://gisgeography.com/what-gis-geographic-information-systems/
(accessed 11.26.15).

55
Goodchild, M.F., 2009. Geographic information systems and science: today
and tomorrow. Procedia Earth Planet. Sci. 1, 1037–1043.
doi:10.1016/j.proeps.2009.09.160
Gulliver, J., Briggs, D.J., 2011. STEMS-Air: A simple GIS-based air pollution
dispersion model for city-wide exposure assessment. Sci. Total
Environ. 409, 2419–2429. doi:10.1016/j.scitotenv.2011.03.004
Guo, D., Guo, R., Thiart, C., 2007. Predicting air pollution using fuzzy
membership grade Kriging. Comput. Environ. Urban Syst. Extr. Inf.
from Spat. Datasets 31, 33–51.
Hafner, W.D., Carlson, D.L., Hites, R.A., 2005. Influence of Local Human
Population on Atmospheric Polycyclic Aromatic Hydrocarbon
Concentrations. Environ. Sci. Technol. 39, 7374–7379.
doi:10.1021/es0508673
Halse, A.K., Schlabach, M., Eckhardt, S., Sweetman, A.J., Jones, K.C., Breivik, K.,
2011. Spatial variability of POPs in European background air. Atmos.
Chem. Phys. 11, 1549–1564. doi:10.5194/acp-11-1549-2011
Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high
resolution interpolated climate surfaces for global land areas. Int. J.
Climatol. 25, 1965–1978. doi:10.1002/joc.1276
Holmes, N.S., Morawska, L., 2006. A review of dispersion modelling and its
application to the dispersion of particles: An overview of different
dispersion models available. Atmos. Environ. 40, 5902–5928.
Holt, E., Borůvková, J., Kalina, J., Melymuk, L., Harner, T., Klánová, J., n.d.
Determining air concentrations of semi-volatile organic compounds
(SVOCs) from passive sampler measurements.
Howard, P.H., Muir, D.C.G., 2010. Identifying new persistent and
bioaccumulative organics among chemicals in commerce. Environ. Sci.
Technol. 44, 2277–85. doi:10.1021/es903383a
CHMI, 1996. REZZO, Registry of Sources of Air Pollution [WWW Document].
URL
http://portal.chmi.cz/files/portal/docs/uoco/oez/embil/oez99/99r1
4.html (accessed 8.6.09).
International Steering Committee for Global Mapping, 2015. Global Map
Data [WWW Document]. URL https://www.iscgm.org/gmd/ (accessed
1.24.15).
JRC-IES, 2008. Forest Fires in Europe 2007.
JRC-IES, 2007. Forest Fires in Europe 2006.

56
Klá nová , J., Cupr, P., Kohoutek, J., Harner, T., 2008. Assessing the Influence of
Meteorological Parameters on the Performance of Polyurethane Foam-
Based Passive Air Samplers. Environ. Sci. Technol. 42, 550–555.
doi:10.1021/es072098o
Klá nová , J., Kohoutek, J., Hamplova, L., Urbanova, P., Holoubek, I., 2006.
Passive air sampler as a tool for long-term air pollution monitoring:
Part 1. Performance assessment for seasonal and spatial variations.
Environ. Pollut. 144, 393–405.
Klíma, B., 1988. Nejstarší moravská mapa. Rod. země 110–121.
Krige, D.G., 1951. A statistical approach to some basic mine valuation
problems on the Witwatersrand. J. Chem. Metall. Min. Soc. South Africa
52, 119–139.
Kuboš ová , K., Komprda, J., Jarkovský , J., Sá ňka, M., Há jek, O., Duš ek, L.,
Holoubek, I., Klá nová , J., 2009. Spatially Resolved Distribution Models
of POP Concentrations in Soil: A Stochastic Approach Using Regression
Trees. Environ. Sci. Technol. 43, 9230–9236. doi:10.1021/es902076y
Macdonald, R.W., Barrie, L.A., Bidleman, T.F., Diamond, M.L., Gregor, D.J.,
Semkin, R.G., Strachan, W.M., Li, Y.F., Wania, F., Alaee, M., Alexeeva, L.B.,
Backus, S.M., Bailey, R., Bewers, J.M., Gobeil, C., Halsall, C.J., Harner, T.,
Hoff, J.T., Jantunen, L.M., Lockhart, W.L., Mackay, D., Muir, D.C.,
Pudykiewicz, J., Reimer, K.J., Smith, J.N., Stern, G.A., 2000. Contaminants
in the Canadian Arctic: 5 years of progress in understanding sources,
occurrence and pathways. Sci. Total Environ. 254, 93–234.
Mackay, D., Paterson, S., Shiu, W.Y., 1992. Generic models for evaluating the
regional fate of chemicals. Chemosphere 24, 695–717.
doi:10.1016/0045-6535(92)90531-U
Matejı́ček, L., Beneš ová , L., Tonika, J., Matě jı́ček, L., 2003. Ecological
modelling of nitrate pollution in small river basins by spreadsheets
and GIS. Ecol. Modell. 170, 245–263. doi:10.1016/S0304-
3800(03)00232-1
Matejíček, L., Engst, P., Janour, Z., 2006. A GIS-based approach to spatio-
temporal analysis of environmental pollution in urban areas: A case
study of Prague’s environment extended by LIDAR data. Ecol. Modell.
199, 261–277.
Mavroulidou, M., Hughes, S.J., Hellawell, E.E., 2004. A qualitative tool
combining an interaction matrix and a GIS to map vulnerability to
traffic induced air pollution. J. Environ. Manage. 70, 283–289.
doi:10.1016/j.jenvman.2003.12.002

57
Melymuk, L., 2012. Semi-volatile Organic Contaminants in the Urban
Atmosphere: Spatial and Seasonal Distributions and Implications for
Contaminant Transport. University of Toronto.
Melymuk, L., Bohlin, P., Sáňka, O., Pozo, K., Klánová, J., 2014. Current
challenges in air sampling of semivolatile organic contaminants:
sampling artifacts and their influence on data comparability. Environ.
Sci. Technol. 48, 14077–91. doi:10.1021/es502164r
Melymuk, L., Robson, M., Helm, P.A., Diamond, M.L., 2013. Application of
land use regression to identify sources and assess spatial variation in
urban SVOC concentrations. Environ. Sci. Technol. 47, 1887–1895.
doi:10.1021/es3043609
Möller, B., 2010. Spatial analyses of emerging and fading wind energy
landscapes in Denmark. Land use policy 27, 233–241.
doi:10.1016/j.landusepol.2009.06.001
Nizzetto, L., Butterfield, D., Futter, M., Lin, Y., Allan, I., Larssen, T., 2015.
Assessment of contaminant fate in catchments using a novel integrated
hydrobiogeochemical-multimedia fate model. Accept. to Sci. Total
Environ.
Ntziachristos, L., Samaras, Z., 2000. COPERT III Computer programme to
calculate emissions from road transport, Methodology and emission
factors (Version 2.1). EEA, Copenhagen.
Petrich, N.T., Spak, S.N., Carmichael, G.R., Hu, D., Martinez, A., Hornbuckle,
K.C., 2013. Simulating and explaining passive air sampling rates for
semivolatile compounds on polyurethane foam passive samplers.
Environ. Sci. Technol. 47, 8591–8. doi:10.1021/es401532q
Pilla, F., Broderick, B., 2015. A GIS model for personal exposure to PM10 for
Dublin commuters. Sustain. Cities Soc. 15, 1–10.
doi:10.1016/j.scs.2014.10.005
Pistocchi, A., 2014. GIS Based Chemical Fate Modeling: Principles and
Applications. John Wiley & Sons, Inc., Hoboken, NJ.
Pistocchi, A., 2008. A GIS-based approach for modeling the fate and
transport of pollutants in Europe. Environ. Sci. Technol. 42, 3640–7.
Pistocchi, A., Sarigiannis, D.A., Vizcaino, P., 2010a. Spatially explicit
multimedia fate models for pollutants in Europe: state of the art and
perspectives. Sci. Total Environ. 408, 3817–30.
doi:10.1016/j.scitotenv.2009.10.046
Pistocchi, A., Vizcaino, P., Hauck, M., 2009. A GIS model-based screening of
potential contamination of soil and water by pyrethroids in Europe. J.

58
 Environ. Manage. 90, 3410–21. doi:10.1016/j.jenvman.2009.05.020
Pistocchi, A., Zulian, G., Vizcaino, P., Marinov, D., 2010b. Multimedia
Assessment of Pollutant Pathways in the Environment, European Scale
Model (MAPPE-EUROPE). Publications Office of the European Union.
doi:10.2788/63765
Poláková, Š., Kubík, L., Němec, P., Malý, S., 2010. Bazální monitoring
zemědělských půd 1992 - 2007. Brno.
Pontolillo, J., Eganhouse, R.P., 2001. The Search for Reliable Aqueous
Solubility (Sw) and Octanol-Water Partition Coefficient (Kow) Data for
Hydrophobic Organic Compounds: DDT and DDE as a Case Study.
Reston, VA.
Pozo, K., Harner, T., Wania, F., Muir, D.C.G., Jones, K.C., Barrie, L.A., 2006.
Toward a Global Network for Persistent Organic Pollutants in Air:
Results from the GAPS Study. Environ. Sci. Technol. 40, 4867–4873.
doi:10.1021/es060447t
Prokeš, R., Vrana, B., Klánová, J., 2012. Levels and distribution of dissolved
hydrophobic organic contaminants in the Morava river in Zlín district,
Czech Republic as derived from their accumulation in silicone rubber
passive samplers. Environ. Pollut. 166, 157–66.
doi:10.1016/j.envpol.2012.02.022
RECETOX Masaryk University, 2015. MONET CZ [WWW Document]. Monet
CZ [online]. Masaryk Univ. URL http://monet.recetox.muni.cz/
(accessed 1.27.14).
Righini, G., Cappelletti, A., Ciucci, A., Cremona, G., Piersanti, A., Vitali, L.,
Ciancarella, L., 2014. GIS based assessment of the spatial
representativeness of air quality monitoring stations using pollutant
emissions data. Atmos. Environ. 97, 121–129.
doi:10.1016/j.atmosenv.2014.08.015
RSD, 2006. Road and Motorway Directorate of the Czech Republic [WWW
Document]. Road Traffic Census 2005. URL http://www.rsd.cz
Sá ňka, O., Dvorská , A., Klá nová , J., Čupr, P., 2013. Modelování rozptylu
vybraných atmosférických PAU v ovzduší Valašského Meziříčí pomocí
modelu SYMOS’97 a GIS a jeho porovnání s výsledky pasivních odběrů
ovzduší. Ochr. Ovzduší 25, 13–21.
Sanka, O., Kalina, J., Lin, Y., Brabec, K., Deutscher, J., Futter, M., Butterfield, D.,
Nizzetto, L., 2015. Modeling hydrological and biogeochemical controls
of the dissipation of p,p’-DDT in soil. Submitted to Environ. Sci.
Technol.

59
Sáňka, O., Melymuk, L., Čupr, P., Dvorská, A., Klánová, J., 2014. Dispersion
modeling of selected PAHs in urban air: A new approach combining
dispersion model with GIS and passive air sampling. Atmos. Environ.
96, 88–95. doi:10.1016/j.atmosenv.2014.07.002
Shad, R., Mesgari, M.S., Abkar, A., Shad, A., 2009. Predicting air pollution
using fuzzy genetic linear membership kriging in GIS. Comput. Environ.
Urban Syst. Spat. Data Mining-Methods Appl. 33, 472–481.
doi:10.1016/j.compenvurbsys.2009.10.004
Sharma, B.M., Nizzetto, L., Bharat, G.K., Tayal, S., Melymuk, L., Sáňka, O.,
Přibylová, P., Audy, O., Larssen, T., 2015. Melting Himalayan glaciers
contaminated by legacy atmospheric depositions are important
sources of PCBs and high-molecular-weight PAHs for the Ganges
floodplain during dry periods. Environ. Pollut. 206, 588–596.
doi:10.1016/j.envpol.2015.08.012
Shekhar, S., Xiong, H. (Eds.), 2008. Encyclopedia of GIS. Springer US, Boston,
MA. doi:10.1007/978-0-387-35973-1
Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental
Organic Chemistry, second. ed. John Wiley & Sons, Inc., Hoboken, NJ.
Steiniger, S., Bocher, E., 2008. An Overview on Current Free and Open
Source Desktop GIS Developments. Int. J. Geogr. Inf. Sci. 1–24.
Steiniger, S., Hunter, A.J.S., 2013. The 2012 free and open source GIS
software map – A guide to facilitate research, development, and
adoption. Comput. Environ. Urban Syst. 39, 136–150.
doi:10.1016/j.compenvurbsys.2012.10.003
U.S. Environmental Protection Agency, 2013. AERMOD.
US Environmental Protection Agency, 2005. Probability-based Design for
Selecting Sampling Sites.
Vallero, D.A., 2008. Fundamentals of Air Pollution, 4th ed. Elsevier,
Amsterdam. doi:10.1016/B978-0-12-373615-4.50041-8
Vienneau, D., de Hoogh, K., Briggs, D.J., 2009. A GIS-based method for
modelling air pollution exposures across Europe. Sci. Total Environ.
408, 255–266. doi:10.1016/j.scitotenv.2009.09.048
VoPham, T., Wilson, J.P., Ruddell, D., Rashed, T., Brooks, M.M., Yuan, J.-M.,
Talbott, E.O., Chang, C.-C.H., Weissfeld, J.L., 2015. Linking pesticides
and human health: A geographic information system (GIS) and Landsat
remote sensing method to estimate agricultural pesticide exposure.
Appl. Geogr. 62, 171–181. doi:10.1016/j.apgeog.2015.04.009

60
Wan, N., 2015. Pesticides exposure modeling based on GIS and remote
sensing land use data. Appl. Geogr. 56, 99–106.
doi:10.1016/j.apgeog.2014.11.012
Westgate, J.N., Wania, F., 2011. On the construction, comparison, and
variability of airsheds for interpreting semivolatile organic compounds
in passively sampled air. Environ. Sci. Technol. 45, 8850–7.
doi:10.1021/es202490b
Zheng, Q., Nizzetto, L., Li, J., Mulder, M.D., Sanka, O., Lammel, G., Bing, H., Liu,
X., Jiang, Y., Luo, C., Zhang, G., 2015. Spatial distribution of old and
emerging flame retardants in Chinese forest soils: Sources, trends and
processes. Environ. Sci. Technol. 49, 2904–2911.
doi:10.1021/es505876k
Zheng, Q., Nizzetto, L., Mulder, M.D., Sáňka, O., Lammel, G., Li, J., Bing, H., Liu,
X., Jiang, Y., Luo, C., Zhang, G., 2014. Does an analysis of polychlorinated
biphenyl (PCB) distribution in mountain soils across China reveal a
latitudinal fractionation paradox? Environ. Pollut. 195, 115–122.
doi:10.1016/j.envpol.2014.08.021
Zou, B., Wilson, J.G., Zhan, F.B., Zeng, Y., Wu, K., 2011. Spatial-temporal
variations in regional ambient sulfur dioxide concentration and
source-contribution analysis: A dispersion modeling approach. Atmos.
Environ. 45, 4977–4985.

61
5 SHORT CURRICULUM VITAE
Personal information

Name: Ondřej Sáňka

Address: Mošnova 21, 61500 Brno, Czech Republic
Date of birth: March 27th, 1984
Employer: Research Centre for Toxic Compounds in the
Environment (RECETOX), Masaryk University, Brno,
Czech Republic
E-mail: o.sanka@recetox.muni.cz
Phone: +420 549 495 956

Education

1999 – 2003 High School Elgartova 3, Brno

2003 – 2006 BSc. Cartography and Geoinformatics, Masaryk
University, Brno
2006 – 2008 MSc. Cartography and Geoinformatics, Masaryk
University, Brno
2008 onwards Ph.D. Environmental Chemistry, Masaryk University,
Brno

Work experience
2008 onwards Research Centre for Toxic Compounds in the
Environment (RECETOX), Masaryk University, Brno,
Czech Republic

Scientific stays

2010 (5 months) Lancaster Environment Centre, Lancaster University,

Lancaster, United Kingdom
2014 (1 month) The Norwegian Institute for Water Research, Oslo,
Norway

Language skills Fluent English

62
6 LIST OF PUBLICATIONS
Sáňka, O., Melymuk, L., Čupr, P., Dvorská, A., Klánová, J., 2014. Dispersion
modeling of selected PAHs in urban air: A new approach combining
dispersion model with GIS and passive air sampling. Atmos. Environ.
96, 88–95. doi:10.1016/j.atmosenv.2014.07.002
Sáňka, O., Dvorská , A., Klá nová , J., Čupr, P., 2013. Modelování rozptylu
vybraných atmosférických PAU v ovzduší Valašského Meziříčí pomocí
modelu SYMOS’97 a GIS a jeho porovnání s výsledky pasivních odběrů
ovzduší. Ochr. Ovzduší 25, 13–21.
Sanka, O., Kalina, J., Lin, Y., Brabec, K., Deutscher, J., Futter, M., Butterfield, D.,
Nizzetto, L., 2015. Modeling hydrological and biogeochemical controls
of the dissipation of p,p’-DDT in soil. Submitted to Environ. Sci.
Technol.
Melymuk, L., Bohlin, P., Sáňka, O., Pozo, K., Klánová, J., 2014. Current
challenges in air sampling of semivolatile organic contaminants:
sampling artifacts and their influence on data comparability. Environ.
Sci. Technol. 48, 14077–91. doi:10.1021/es502164r
Zheng, Q., Nizzetto, L., Li, J., Mulder, M.D., Sanka, O., Lammel, G., Bing, H., Liu,
X., Jiang, Y., Luo, C., Zhang, G., 2015. Spatial distribution of old and
emerging flame retardants in Chinese forest soils: Sources, trends and
processes. Environ. Sci. Technol. 49, 2904–2911.
doi:10.1021/es505876k
Zheng, Q., Nizzetto, L., Mulder, M.D., Sáňka, O., Lammel, G., Li, J., Bing, H., Liu,
X., Jiang, Y., Luo, C., Zhang, G., 2014. Does an analysis of polychlorinated
biphenyl (PCB) distribution in mountain soils across China reveal a
latitudinal fractionation paradox? Environ. Pollut. 195, 115–122.
doi:10.1016/j.envpol.2014.08.021
Sharma, B.M., Nizzetto, L., Bharat, G.K., Tayal, S., Melymuk, L., Sáňka, O.,
Přibylová, P., Audy, O., Larssen, T., 2015. Melting Himalayan glaciers
contaminated by legacy atmospheric depositions are important
sources of PCBs and high-molecular-weight PAHs for the Ganges
floodplain during dry periods. Environ. Pollut. 206, 588–596.
doi:10.1016/j.envpol.2015.08.012

63
Vacha, R., Sanka, M., Sanka, O., Skala, J., Cechmankova, J., 2013. The Fluvisol
and sediment trace element contamination level as related to their
geogenic and anthropogenic source. Plant, Soil Environ. 59, 136–142.
Borůvka, L., Sáňka, M., Vácha, R., Šrámek, V., Drábek, O., Tejnecký, V.,
Antonín, N., Němeček, K., Houška, J., Fraňková, A., Ash, C., Galušková, I.,
Hubová, P., Fadrhonsová, V., Jurkovská, L., Čechmánková, J., Čermák, P.,
Horváthová, V., Skála, J., Čupr, P., Hofman, J., Sáňka, O., Rotter, P.,
Vašíčková, J., 2014. Metody hodnocení zátěže lesních půd rizikovými
látkami a identifikace ekologických rizik kontaminace lesních půd.
Final project report (project NAZV QI112A201), Czech University of
Life Sciences, Prague.
Sáňka, M., Hofman, J., Vácha, R., Čupr, P., Čechmánková, J., Sáňka, O., Mikeš,
O., Skála, J., Horváthová, V., Šindelářová, L., Vašíčková, J., Nečasová, A.,
2015. Metodické postupy k omezení vstupu rizikových látek do
rostlinné produkce v oblastech postižených periodickými povodněmi,
1st ed. Masaryk University, Prague.
Skála, J., Čechmánková, J., Vácha, R., Horváthová, V., Sáňka, M., Sáňka, O.,
2013. Soubor map - Regionální struktura půdního pokryvu
zemědělsky využívaných půd ve vztahu k povodňové zonaci, 1st ed.
Research Institute for Soil and Water Conservation, Prague.
Skála, J., Čupr, P., Hofman, J., Sáňka, O., Čechmánková, J., Sáňka, M., Vácha, R.,
Horváthová, V., Mikeš, O., 2014. Obsahy rizikových prvků ve
fluvizemních půdách ČR – soubor map. Mapy s odborným obsahem,
1st ed. Research Institute for Soil and Water Conservation, Prague.
Rotter, P., Malý, S., Kalábová, T., Čižmár, D., Sáňka, O., Zbíral, J., Sáňka, M.,
Čechmánková, J., 2015. Is glomalin an appropriate indicator of forest
soil dynamics and reactive nitrogen status? Submitted to Eur. J. Soil Sci.
Čupr, P., Prokeš, R., Sáňka, O., Přibylová, P., Bednářová, Z., Borůvková, J.,
Chropeňová, M., Šebková, K., Václavíková, J., Vaňková, L., Kalina, J.,
Audy, O., Kalábová, T., Holoubek, I., Klánová, J., 2015. Časové a
prostorové hodnocení koncentrací perzistentních organických
polutantů ve volném ovzduší ČR. Specializované mapy s odborným
obsahem. Brno, in press.

64
7 LIST OF CONFERENCE PRESENTATIONS AND
POSTERS

Lammel, G.*, Nizzetto, L., Sáňka, O., Spitzy, A., Beudert, B., Kukučka, P., and
Přibylová., P. 2015. “Surface water pollution by PAHs, PCBs and
organochlorine pesticides in a central European subalpine mountain area
since 1985.” 15th EuCheMS International Conference on Chemistry and the
Environment. 20 - 24 September, Leipzig, Germany. Poster presentation.

Sáňka, O., Melymuk, L., Čupr, P.*, Dvorská, A. and Klánová, J. 2014.
“Dispersion modelling of PAHs in urban air: A new approach combining
dispersion model with GIS and passive air sampling” The 15th European
Meeting on Environmental Chemistry. 3 - 6 December, Brno, Czech Republic.
Platform Presentation.

Čupr, P.*, Prokeš, R., Moche, W., Weiss, P., Škrdlíková, L., Baráková, D., Sáňka,
O., Kukučka, P., Šebková, K., and Klánová, J. 2013. “Monairnet – Air-borne
POPs in the Czech-Austrian border region.” Sixth Meeting of the Conference
of the Parties of the Stockholm Convention. 28 April – 10 May, Geneva,
Switzerland. Platform presentation.

Prokeš, R.*, Čupr, P., Moche, W., Weiss, P., Hohenblum, P., Sáňka, O.,
Kukučka, P., Přibylová, P., and Klánová, J. 2013. “Results of passive air
sampling.” Final conference of Monairnet project. 3 December, Linz, Austria.
Platform presentation.

Prokeš, R.*, Baráková, D., Čupr, P., Moche, W., Weiss, P., Hohenblum, P.,
Sáňka, O., Kukučka, P., Přibylová, P. and Klánová, J. 2013. “Passive air
sampling and needles sampling – methodology, advantages and limitations.”
Final conference of Monairnet project. 5 December, Brno, Czech Republic.
Platform presentation.

Čupr, P.*, Prokeš, R., Vaňková, L., Baráková, D., Chropeňová, M., Václavíková,
I., Sáňka, O., Kukučka, P., Klánová, J., Holoubek, I., Přibylová, P., Šebková, K.,
Bányiová, K., Moche, W., Hohenblum, P. and Weiss, P. 2013. “Monitoring
network of ambient air in Czech Republic and Austria.” International

65
conference on Air quality, 27 – 29 November, Štrbské pleso, Slovakia,
Platform presentation.

Chropeňová, M., Čupr, P., Prokeš, P., Škrdlíková, L., Baráková, D.,
Václavíková, I., Sáňka, O., Kukučka, P., Klánová, J., Přibylová, P., Šebková, K.,
Moche, W. and Weiss, P. 2013. “MONAIRNET - Monitoring network for
persistent organic pollutants in ambient air of Czech Republic - Austria
region.” 14th EuCheMS International Conference on Chemistry and the
Environment. 25 - 28 June, Barcelona, Spain, Poster presentation.

Čupr, P.*, Václavíková, I., Prokeš, R., Škrdlíková, L., Kukučka, P., Sáňka, O.,
Klánová, J., Moche, W., and Weiss, P. 2013. “MONAIRNET – Monitoring
persistentních organických polutantů v ovzduší příhraničních regionů ČR a
Rakouska.” 11th conference Ovzduší 2013, 15–17 April, Brno, 2013. Platform
presentation.

Sanka, O., Sweetman, A., Kohut, L.*, and Cupr, P. 2012. “Modelling of
pesticides concentrations using air dispersion modelling and GIS” 3rd
SETAC CEE Annual Meeting. 17 – 19 September, Krakow, Poland. Poster
presentation.

Sáňka, O.,* Sweetman, A. and Čupr, P. 2011. “Modelování imisních

koncentrací pesticidů s využitím rozptylových modelů a GIS.” 10th
conference Ovzduší. 4 – 6 April, Brno, Czech Republic. Poster presentation.

Sáňka, O.,* Dvorská, A. and Čupr, P. 2009. “Aplikace rozptylového modelu

SYMOS´97 na město Valašské Meziříčí s využitím GIS, srovnání s výsledky
pasivního vzorkování.” 9th conference Ovzduší. 26 – 29 April, Brno, Czech
Republic. Poster presentation.

Sáňka, O.,* Alice Dvorská and Čupr, P. 2008. “Automatizace přípravy dat pro
modelování imisí některých PAHs z dopravy pomocí GIS v rámci rozptylové
studie provedené rozptylovým modelem SYMOS’97 pro město Valašské
Meziříčí.” 3rd conference Doprava, zdraví a životní prostředí. 4 – 5 November,
Litomyšl, Czech Republic. Poster presentation.

66
8 LIST OF ANNEXES

Annex 1:

Manuscript of Sá ňka, O., Dvorská , A., Klá nová , J. & Čupr, P. Modelování
rozptylu vybraných atmosférických PAU v ovzduší Valašského Meziříčí
pomocí modelu SYMOS’97 a GIS a jeho porovnání s výsledky pasivních
odběrů ovzduší. Ochr. Ovzduší 25, 13–21 (2013).

Annex 2:

Manuscript of Sáňka, O., Melymuk, L., Čupr, P., Dvorská, A. & Klánová, J.
Dispersion modeling of selected PAHs in urban air: A new approach
combining dispersion model with GIS and passive air sampling. Atmos.
Environ. 96, 88–95 (2014).

Annex 3:

Manuscript of Sanka, O., Kalina, J., Lin, Y., Brabec, K., Deutscher, J., Futter, M.,
Butterfield, D. & Nizzetto, L. Modeling hydrological and biogeochemical
controls of the dissipation of p,p’-DDT in soil. Submitted to Environ. Sci.
Technol. (2015).

Annex 4:

Manuscript of Zheng, Q., Nizzetto, L., Mulder, M., Sáňka, O., Lammel, G., Li, J.,
Bing, H., Liu, X., Jiang, Y Luo, Ch. & Zhang, G. Does an analysis of
polychlorinated biphenyl (PCB) distribution in mountain soils across China
reveal a latitudinal fractionation paradox? Environ. Pollut. 195, 115–122
(2014).

Annex 5:

Manuscript of Zheng, Q., Nizzetto, L., Li, J., Mulder, M., Sáňka, O., Lammel, G.,
Bing, H., Liu, X., Jiang, Y., Luo, Ch. & Zhang, G. Spatial distribution of old and
emerging flame retardants in Chinese forest soils: Sources, trends and
processes. Environ. Sci. Technol. 49, 2904–2911 (2015).

67
Annex 6:

Manuscript of Melymuk, L., Bohlin, P., Sáňka, O., Pozo, K. & Klánová, J.
Current challenges in air sampling of semivolatile organic contaminants:
sampling artifacts and their influence on data comparability. Environ. Sci.
Technol. 48, 14077–91 (2014).

Annex 7:

List of free online GIS data sources.

68
ANNEX 1

Manuscript of Sá ň ka, O., Dvorská , A., Klá nová , J. & Čupr, P. Modelování
rozptylu vybraných atmosférických PAU v ovzduší Valašského Meziříčí
pomocí modelu SYMOS’97 a GIS a jeho porovnání s výsledky pasivních
odběrů ovzduší. Ochr. Ovzduší 25, 13–21 (2013).
4/2013 OCHRANA OVZDUŠÍ

Imise

Modelování rozptylu vybraných atmosférických PAU

v ovzduší Valašského Meziříčí pomocí modelu
SYMOS'97 a GIS a jeho porovnání s výsledky
pasivních odběrů ovzduší
Ondřej Sáňka, Alice Dvorská, Jana Klánová, Pavel Čupr
RECETOX, Centrum pro výzkum toxických látek v prostředí, Masarykova Univerzita, Brno, cupr@recetox.muni.cz

ABSTRAKT Dispersion modelling of selected

atmospheric PAHs in the city of Valašské
Cílem studie bylo pomocí rozptylového modelu a geografického informač- Meziříčí based on the SYMOS'97 dispersion
ního systému (GIS) predikovat koncentrace vybraných polycyklických model and GIS, and comparison of results
aromatických uhlovodíků (PAU) ve  volném ovzduší a  získané hodnoty using passive air sampling
srovnat s výsledky pasivního vzorkování ovzduší. Na město Valašské Mezi-
říčí a jeho nejbližší okolí byl aplikován český rozptylový model gaussovské-
The aim of this study was to predict concentrations of selected polycyclic aro-
ho typu SYMOS'97. Při přípravě vstupních dat pro rozptylový model bylo
matic hydrocarbons (PAHs) in ambient air using a Geographic Information
ve  významné míře využito GIS, který se ukázal pro realizaci rozptylové
System (GIS) and to verify modelling results using the outcomes of the high-
studie podobného formátu jako klíčový. Zejména při zpracování liniových
density passive air sampling (PAS). SYMOS'97 (Czech Gaussian dispersion
a plošných zdrojů emisí, přípravě výškopisných dat a tvorbě referenčních
model) was applied on the area of the city of Valašské Meziříčí and its close
bodů bylo použití funkcí GIS nevyhnutelné. Výsledky modelování byly
surroundings. The application of the GIS turned out to be crucial when execut-
opět zpracovány a  vizualizovány pomocí GIS. Srovnání atmosférických
ing such a large dispersion study. It was used massively when processing input
koncentrací předikovaných modelem s experimentálními daty získanými
data for the dispersion model, but its application was inevitable especially
dlouhodobou kontinuální expozicí pasivních vzorkovačů ovzduší uká-
when processing the linear and areal emission sources and elevation data, or
zalo rozdíly, pro jejichž vysvětlení bylo navrženo několik hypotéz. Jako
when creating receptor points. The modelling results were also processed and
nejpravděpodobnější se jeví teorie, že na  kvalitu ovzduší mají velký vliv
visualized using the GIS. Several factors can contribute to the observed lack of
zdroje, které nejsou zahrnuty v použité emisní inventuře, například spa-
agreement between the modelled results and the outcomes of the passive air
lování biomasy. Dalším zdrojem nejistot mohou být některá zjednodušení,
sampling campaigns. An effect of significant emissions sources not included in
použitá při kvantifikaci emisních toků, ale také fakt, že pasivní vzorkování
the applied emission inventory (e.g. biomass combustion) can be one of them,
poskytuje pouze semikvantitativní výsledky. Jakkoli má tato jednoduchá
together with uncertainties stemming from simplifications in quantification
a  levná metoda nepřehlédnutelný potenciál pro dlouhodobé sledování
of emission flows. Semi-quantitative character of the passive air sampling
kvality ovzduší, je třeba mít na mysli, že v současné podobě je vhodná spíše
technique can also generate some uncertainties. While this simple method
pro relativní srovnání vzorkovacích lokalit nebo pro identifikaci dlouho-
certainly offers a significant potential for a long-term air quality monitoring,
dobých trendů, než pro přesnou kvantifikaci atmosférických koncentrací.
it is important to beare in mind that it is more suited for an assessment of the
Koeficienty umožňující přesnější přepočet koncentrací naměřených v pa-
long-term trends or relative comparison of the sampling sites than for precise
sivních vzorcích na objemové jednotky jsou předmětem neustálého výzku-
quantification of the atmospheric concentrations. Refining the coefficients
mu a zpřesňování. Tato omezení ovšem nemění nic na tom, že integrativní
enabling conversion of passive air sampling data to atmospheric concentra-
pasivní vzorkovače nabízejí finančně dostupnou možnost velkoplošného
tions is a subject of the on-going research. These drawbacks, however, do not
vzorkování ovzduší s  vysokým rozlišením potřebným pro verifikaci roz-
diminish a fact that integrative passive air samplers offer an opportunity for
ptylových modelů. Výsledky této studie ukazují zřejmé nedostatky rozpty-
cost-effective large-scale and high-density air monitoring which is needed for
lového modelu SYMOS'97. Přestože se stále těší velké oblibě uživatelů,
verification of dispersion models. Results of this study pointed out also several
v  mnoha ohledech se již jeví jako překonaný a  nevyhovující současným
drawbacks of the SYMOS'97 dispersion model. While it is still very popular
požadavkům a trendům. Realizace dalších podobných studií se zapojením
among the users, it does not seem to be fully meeting current trends and ex-
multiparametrických modelů pro interpretaci dat z pasivního vzorkování
pectations. The follow-up studies employing new multi-parametric model for
umožní lépe popsat slabá místa současných rozptylových modelů a pod-
interpretation of passive sampling data should further assess the weaknesses
pořit tak vývoj modelů nové generace.
of current dispersion models and support development of dispersion models of
Klíčová slova: polycyklické aromatické uhlovodíky (PAU), Valašské Me-
new generation.
ziříčí, modelování rozptylu, model SYMOS'97, Geografický informační
Key words: polycyclic aromatic hydrocarbons (PAHs), Valašské Meziříčí, dis-
systém, pasivní vzorkování ovzduší
persion modelling, SYMOS'97 model, Geographic Information System, pas-
sive air sampling

ÚVOD
S rozvojem průmyslu a automobilismu v 19. století byl zazna-
menán také nárůst znečištění všech složek životního prostředí
včetně ovzduší. Častý výskyt událostí s prudce zvýšenými kon-
centracemi polutantů v  ovzduší, nebo kontinuální vystavení
obyvatelstva jejich zvýšeným hladinám mělo za následek pro-
blémy především plic a dýchacího ústrojí. Absorpce polutantů
a jejich distribuce krví však často způsobuje i jiné zdravotní pro-
blémy [14]. Tento jev vyvolal zvýšený zájem o kvalitu životního
prostředí, především ovzduší. Během první poloviny 20. století
tak byly matematicky popsány základní děje v  atmosféře, což
v 60. letech umožnilo vznik gaussovských modelů, zaměřených
především na studium vlivu bodových zdrojů znečištění v lokál-
ním měřítku. Především rozvoj výpočetní techniky v posledních
desetiletích dal vzniknout mnoha rozptylovým modelům s nej-

– 13 –
OCHRANA OVZDUŠÍ
různějšími vlastnostmi. Samotný rozptylový model však nestačí
k  získání relevantních výsledků. Nedílnou součástí rozptylo-
vého modelování je také použití kvalitního a úplného souboru
vstupních dat, stejně jako jejich správné zpracování a příprava
pro vstup do rozptylového modelu. Zdaleka ne všechny rozpty-
lové modely však touto schopností disponují. Stejně tak běžnou
vlastností modelů nebývá schopnost generovat mapové výstu-
py, prezentující výsledky výpočtu. K  přípravě vstupních dat
a prezentaci výsledků modelování tedy musí být často použito
dalšího softwaru, ať už statistického, nebo geografického.
Při hodnocení znečištění ovzduší mají nepostradatelnou úlohu
také odběrové vzorkovací techniky. Důležitost pravidelných
měření koncentrací polutantů ve  vzduchu na  různých loka-
litách a  potřeba monitorovacích studií na  různých úrovních,
od bezprostředního okolí lokálních bodových zdrojů až po kon-
tinentální měřítko, inspirovala vývoj nových metod pro monito-
rování kvality ovzduší [19]. K nim patří zejména různé varianty
pasivních vzorkovačů [1] jako zařízení, která mohou být pou-
žita na celé řadě lokalit současně, což nabízí nové možnosti při
přípravě rozsáhlejších monitorovacích kampaní. Poskytují in-
formaci o dlouhodobé kontaminaci vybraného místa a mohou
být použity jako screeningová metoda pro semikvantitativní
srovnání různých lokalit s  výhodou malé citlivosti ke  krátko-
dobým náhodným změnám koncentrace polutantů. Stanovení
koncentrace ve vzduchu je pak odvozeno z kalibrace založené
na paralelním aktivním a pasivním vzorkování [7].
Na Centru pro výzkum toxických látek v prostředí byl k pro-
vedení této rozptylové studie pro město Valašské Meziříčí
použit rozptylový model SYMOS'97 [2](Idea – Envi s.r.o.).
Cílem této studie bylo nejen zjištění hlavních zdrojů znečištění
v ovzduší vybranými polycyklickými aromatickými uhlovodí-
ky (benzo[a]pyren, benzo[b]fluoranten, benzo[k]fluoranten
a indeno[123cd]pyren), ale také vzájemné porovnání výsledků
realizované odběrové kampaně s  pasivními vzorkovači a  vý-
sledků rozptylového modelu SYMOS'97. Výběr těchto látek
vycházel z průnikové shody informací jak vstupních emisních
dat, tak i výsledků pasivního monitoringu. K přípravě vstup-
ních dat a vizualizaci výsledků modelování bylo použito GIS
softwaru firmy ESRI, ArcGIS 9.2. Jako srovnávací data byly
použity koncentrace zjištěné metodou pasivního vzorkování
ovzduší na čtyřech lokalitách během roku 2006. Ten byl z dů-
vodu koncepce pasivního vzorkování rozdělen na 13 intervalů
po 28 dnech (viz tab. 3).
POUŽITÁ DATA
Realizace rozptylové studie není možná bez použití velkého
množství dat různého charakteru. Jedním z klíčových vstupů
jsou data o  emisích průmyslových zdrojů a  zdrojů vytápění.
Ta byla stejně jako větrné růžice poskytnuta Českým hydro-
meteorologickým ústavem (ČHMÚ). Potřebné informace
o mobilních zdrojích byly získány za pomoci Centra doprav-
ního výzkumu v.v.i.. Pro simulaci tvarů reliéfu v modelovaném
území a tvorbu sítě referenčních bodů byla použita GIS vrstva
vrstevnic s krokem 50 m, z databáze ArcČR 500 (firma ArcDa-
ta Praha, s. r. o.). Vzorkovací kampaň, součástí jejíchž výstupů
byla také srovnávací data, byla měřena Centrem pro výzkum
toxických látek v  prostředí. V  případě této studie se jednalo
o data ze čtyř lokalit v centru nebo v blízkém okolí Valašského
Meziříčí (Hvězdárna, Juřinka, Mštěnovice a  Příluky). Právě
– 14 –
4/2013
časové vymezení intervalů pasivního vzorkování bylo v  této
studii určující pro stanovení intervalů pro rozptylové modelo-
vání, a to z důvodu co nejvyšší relevance porovnání výsledků
modelování a vzorkování. Z důvodu zjištění možného příspěv-
ku vzdálenějších zdrojů ke  koncentracím sledovaných látek
ve Valašském Meziříčí byl výpočet rozptylového modelu pro-
veden pro dvě varianty zájmového území (viz obr. 2).
Autoři jsou si vědomi faktu, že polycyklické aromatické uhlo-
vodíky vybrané pro tuto studii nejsou svými vlastnostmi příliš
vhodné pro validaci rozptylového modelu SYMOS'97 pasivními
vzorkovači, jelikož ani jeden z přístupů není pro tyto látky opti-
mální. Rozptylový model SYMOS'97 nedokáže zcela zohlednit
fyzikálně chemické vlastnosti látek (a  tedy významné rozdíly
v chování různých PAU v atmosféře), pasivní vzorkovače jsou
primárně určeny pro měření atmosférických koncentrací polu-
tantů v plynném stavu, nikoli vázaných na částice, což je případ
PAU. Lepší možností by se mohla jevit například aplikace mo-
delu na skupinu polychlorovaných bifenylů (PCBs), které mají
mnohem uniformnější vlastnosti a v atmosféře se vyskytují pře-
vážně v plynném stavu. Pro tuto skupinu látek ovšem neexistuje
shoda mezi výběrem kongenerů, pro něž jsou dostupná data
v  databázi REZZO (Registr emisí a  zdrojů znečištění ovzdu-
ší) a kongenery, které jsou běžně sledovány v monitorovacích
programech. Z metodického hlediska a s ohledem na charakter
a rozsah použitých dat se však jedná o unikátní a zcela nový typ
přístupu hodnocení rozptylové studie.
METODIKA
Pasivní vzorkování
Pasivní vzorkování je jednoduchá metoda monitoringu kvality
ovzduší, při jejíž aplikaci sledované látky samovolně difundují
z prostředí do sběrného média. Množství látek akumulovaných
v odběrovém zařízení je funkcí délky expozice. Po skončení od-
běru se vzorkovací médium chemicky analyzuje a získaná data
se převádějí na skutečné atmosférické koncentrace za pomoci
různých modelů. K  výhodám pasivního vzorkování patří ze-
jména nízká cena zařízení a nízké provozní náklady (není nutná
přítomnost obsluhy a  zdroje elektrické energie), nehlučnost,
a malé nároky na instalaci a technickou údržbu. K jejich nevý-
hodám patří zejména nemožnost exaktního stanovení objemu
vzorkovaného vzduchu a  ovlivnění průběhu vzorkování me-
teorologickými podmínkami. Mezi hlavní faktory ovlivňující
sorpční kinetiku pasivních vzorkovačů patří teplota a rychlost
větru. Pasivní vzorkovač používaný v  této práci se skládá ze
dvou nerezových nádob miskovitého tvaru umístěných napro-
ti sobě na společné ose, která slouží i k upevnění vzorkovacího
média (obr. 1). Vnější průměr větší mísy je 30 cm, menší misky
24 cm. Mezi nimi je mezera o  velikosti 3 cm, která umožňuje
průchod vzduchu k vzorkovacímu povrchu. Všechny části vzor-
kovače jsou vyrobeny z nerezové oceli [8]. Vzorkovací médium
je umístěno uprostřed mezi miskami. Tento design umožňuje
tlumit vlivy rychlosti větru na  sorpci a  chrání vzorkovací me-
dium proti atmosférickým srážkám, usazování hrubých částic
a  ultrafialovému záření, které by mohlo rozkládat sledované
sloučeniny [19]. Pro pasivní vzorkování perzistentních orga-
nických polutantů (POPs) se jako sorpční médium používají
filtry z  bílé, nebarvené polyuretanové pěny (PUF). Polyureta-
nová pěna je vzorkovací médium běžně používané v tradičních
4/2013 OCHRANA OVZDUŠÍ

vysokoobjemových vzorkovacích zaří- Tab. 1: Kategorie zrojů REZZO

zeních a je velmi dobře popsané. Disky Kategorie REZZO REZZO1 REZZO2 REZZO3 REZZO4
jsou kruhového tvaru, tloušťky 15 mm Typ zdrojů bodové bodové, plošné plošné liniové
a  průměru 150 mm a  před umístěním Výkon > 5 MW 0,2 – 5 MW < 0,2 MW –
do  vzorkovače jsou čištěny Soxhleto- cementárna, kotelna, výtopna, souvislá městská silniční
Příklad
vou extrakcí 8 hodin v acetonu a 8 ho- spalovna, obalovna menší podnik zástavba doprava
din v dichlormethanu [7].
Tab. 2: Rozptylovým modelem požadované parametry pro jednotlivé druhy
Velkou výhodou použité metody pa- zdrojů znečištění
sivního vzorkování je její integrativní
Parametr Bodový zdroj Plošný zdroj Liniový zdroj
charakter umožňující kontinuální sledo-
Jednotné identifikační číslo ü ü ü
vání situace a  poskytování dlouhodobě
X souřadnice zdroje ü ü ü
reprezentativních dat. Na rozdíl od jed-
Y souřadnice zdroje ü ü ü
norázových krátkodobých měření ak-
Nadmořská výška zdroje ü ü ü
tivními metodami, kdy se dostáváme
Typ zdroje ü ü ü
sice k velmi přesným, ale v podstatě ná-
Indikace shluku ü ü ü
hodným hodnotám o znečištění ovzduší
Indikace skupiny ü ü ü
bez jakékoli informace o  tom, zda jsou
Množství látky odcházející komínem ü ü ü
to hodnoty minimální, maximální nebo
Výška koruny komína nad terénem ü ü
průměrné a  bez vazby na  dlouhodobou
Objem spalin odcházející komínem ü ü
situaci, v  případě pasivních vzorkovačů
Teplota spalin ü ü
s delší periodou měření dostáváme data
Vnitřní průměr komína ü ü
za dané období průměrná. Jejich dosta-
Výstupní rychlost spalin ü ü
tečné časové rozlišení nám přesto umož-
ňuje sledovat sezónní variabilitu dat. Relativní roční využití max. Výkonu ü ü ü
Jak bylo zmíněno výše, pasivní vzor- Počet hodin za den, kdy je zdroj v činnosti ü ü ü
kovače ovzduší jsou primárně určeny X souřadnice počátečního bodu elementu ü
ke  vzorkování plynné frakce, nikoli at- Y souřadnice počátečního bodu elementu ü
mosférických částic. K  přepočtu kon- X souřadnice koncového bodu elementu ü
centrací sledovaných látek v odběrovém Y souřadnice koncového bodu elementu ü
médiu (ng/PUF) na atmosférické kon- Šířka úseku ü
centrace (ng/m ) se v  hrubém přiblí-
3 Výška exhalací ü
žení používá pro dobu expozice 28 dní Vzdálenost zdrojů od sebe ü
koeficient 200 m3 [7]. Tento koeficient
představuje experimentálně stanovenou průměrnou hodnotu Práce s rozptylovým modelem SYMOS'97
objemu, z  něhož pasivní vzorkovač během doby své expozice
(objem je funkcí času) vyextrahoval sledované polutanty. Tento SYMOS'97 je český model gaussovského typu. Stejně jako
koeficient může být pro konkrétní lokality dále zpřesněn korekcí ostatní rozptylové modely také SYMOS'97 vyžaduje pro
na meteorologické podmínky (teplotu a rychlost větru). Pro tuto provedení studie různé druhy dat, která musejí být náležitě
korekci se běžně používá například obohacení vzorkovacího upravena podle specifických požadavků modelu. Vzhledem
média před expozicí izotopicky značenými standardy, přičemž k  velkému množství dat, která bylo třeba různými způsoby
rychlost jejich úniku je úměrná rychlosti záchytu nativních kon- zpracovat, bylo k  tomuto využito funkcí nejen GIS, ale také
taminantů z prostředí. Pokud jde o látky vázané na atmosféric-
ké částice, bylo v předchozích studiích prokázáno, že efektivita Tab. 3: Počet pracovních dní a denostupňů v jednotlivých
modelovaných intervalech, odpovídajících intervalům
jejich záchytu je ve srovnání s plynnými kontaminanty přibližně pasivního vzorkování
desetinová. Pro přibližný výpočet atmosférických koncentrací je
Číslo
tedy možné použít koeficient 20 m3. Je třeba ovšem vzít v úvahu, Rozsah intervalu Prac. dnů Denostupně
intervalu
že jednotlivé PAU jsou mezi plynnou a částicovou frakci distri- 1 30. 12. 2005 – 27. 1. 2006 20 657,8
buovány v různém poměru v závislosti na teplotě a jejich fyzikál- 2 27. 1. – 24. 2. 2006 20 614,6
ně-chemických vlastnostech. Je tedy žádoucí vyvinout komplex- 3 24. 2. – 24. 3. 2006 20 606,1
nější model, který pracuje se všemi potřebnými proměnnými. 4 24. 3. – 21. 4. 2006 19 354,4
Na vývoji tohoto modelu se intenzivně pracuje, zatímco pro tuto 5 21. 4. – 19. 5. 2006 18 150,3
práci byl pro všechny sledované PAU aplikován jednotně koefi- 6 19. 5. – 16. 6. 2006 20 83,6
cient 20 m3. Vzhledem k podobnosti vybraných 4 sloučenin ne- 7 16. 6. – 14. 7. 2006 18 0
bude toto zjednodušení zřejmě generovat významné chyby spe- 8 14. 7. – 11. 8. 2006 20 0
cifické pro jednotlivé látky. Dá se nicméně očekávat, že pro letní 9 11. 8. – 8. 9. 2006 20 0
období, kdy je distribuce PAU v ovzduší posunuta více směrem 10 8. 9. – 6. 10. 2006 19 22,9
k plynné fázi, budou odhadnuté vzorkovací objemy podhodno- 11 6. 10. – 3. 11. 2006 20 249,7
cené a atmosférické koncentrace tedy nadhodnocené. V zimním 12 3. 11. – 1. 12. 2006 19 351,1
období může toto zjednodušení naopak způsobovat nadhodno- 13 1. 12. – 29. 12. 2006 18 445,5
cení objemů a podhodnocení koncentrací. suma 251

– 15 –
OCHRANA OVZDUŠÍ
tabulkových a textových editorů. Tím byl proces přípravy dat
pro rozptylový model zkrácen na přijatelnou dobu a bylo také
možné se touto cestou vyhnout mnoha nahodilým chybám.
Požadovaná vstupní data lze rozdělit na  (a) výškopisná data
(b) větrné růžice, (c) údaje o zdrojích a (d) referenční body [2],
Výškopisná data
Tvar reliéfu je jedním z důležitých činitelů, ovlivňujících šíření
zplodin od  jejich zdroje. Pomocí meteorologických informací
dokáže model předpovědět, zda zplodiny ovlivní např. místo
ležící ve  vyšší nadmořské výšce nebo za  horským hřebenem.
Pokročilejší modely lagrangeovského a eulerovského typu, ma-
jící zabudovaný preprocesor meteorologických dat, navíc infor-
mace o tvarech reliéfu využívají k převodu izolovaných měření
směru a síly větru na plošnou informaci ve formě gridu [6].
Do  modelu SYMOS'97 je možné vložit pouze pravidelnou síť
výškopisu. V  případě potřeby simulovat rozptyl látek v  urči-
té části území ve větším detailu však lze vložit podrobnější síť
menšího rozměru, nadřazenou síti primární [3]. Než však data
dospějí do podoby na obr. 2, musejí projít několika kroky zpra-
cování. Původním zdrojem dat byla v případě zde prezentované
studie GIS vrstva obsahující vrstevnice s krokem 50 m pokrý-
vající zájmové území. Z té byl vytvořen rastrový digitální model
terénu s velikostí hrany pixelu 100 m (požadovaná vzdálenost
bodů výškopisu). Jednotlivé pixely pak byly převedeny na body
a těmto byly přiřazeny souřadnice v souřadném systému S-JT-
SK. Pro úpravu dat do formátu textového souboru, který lze vlo-
žit do modelu, bylo použito prostředí tabulkového editoru MS
Excel. Finální podobu získala tato data exportem do textového
formátu a následnými drobnými úpravami.
Větrné růžice
Příprava a  vložení větrných růžic do  modelu SYMOS'97 byla
poměrně jednoduchou záležitostí. Větrné růžice představují
především oproti zdrojům emisí nesrovnatelně menší objem
dat, proto bylo využito poměrně příjemného uživatelského roz-
hraní SYMOSu’97 a data o větrných růžicích byla do modelu
vložena manuální cestou. Zároveň jsou větrné růžice jediným
vstupem do modelu, který neobsahuje informaci o poloze, neby-
lo tedy potřeba jejich předzpracování pomocí GIS. Nutností je
však rozdělení větrných růžic na pět tříd stability ovzduší podle
klasifikace Bubníka a Koldovského pro definování zastoupení
výskytu různých podmínek pro rozptyl polutantů v atmosféře.
Zároveň je třeba větrné růžice rozdělit do  tří kategorií podle
rychlosti větrů v  rámci každé třídy stability [2]. Nejběžnějším
formátem pro SYMOS'97 je větrná růžice rozdělená do 8 směrů
(po 45°), avšak je možné vložit i větrnou růžici s počtem směrů
v rozmezí 4 (po 90°) až 36 směrů (po 10°).
Pro tuto studii byly větrné růžice dodány Českým hydrometeo-
rologickým ústavem. K jejich tvorbě bylo použito softwaru spe-
ciálně zkonstruovaného pro tento účel. Mezi hlavní výhody to-
hoto softwaru patří schopnost vygenerovat větrnou růžici pro
jakékoli místo a jakékoli časové rozmezí, čehož zde bylo využito.
Proto mohlo být v práci použito 13 větrných růžic pro 28den-
ní období totožná s  intervaly pasivního vzorkování ve  čtyřech
sledovaných lokalitách. Použitím tohoto přístupu je tato studie
do jisté míry ojedinělá, neboť většina rozptylových studií je pro-
váděna pro kratší (denní) nebo delší (roční) časové intervaly.
– 16 –
4/2013
Údaje o zdrojích
Zdroje emisí je nutné zpracovávat zvlášť s  ohledem na  je-
jich odlišný charakter. Na zdroje bodové (viz obr. 3), liniové
a plošné bylo proto nutné nahlížet odděleně. Rozptylový mo-
del SYMOS'97, podobně jako většina ostatních modelů, rozli-
šuje zdroje emisí právě na zmíněné tři kategorie.
Jako hlavní zdroj informací o činnosti zdrojů znečištění byl po-
užit Registr emisí a zdrojů znečišťování ovzduší – REZZO (viz
tab. 1) poskytnutý ČHMÚ, jehož Oddělení emisí a zdrojů je
od roku 1993 správcem této databáze.
Naprostá většina potřebných údajů o bodových zdrojích byla
získána z databáze REZZO1 a 2. Velké zdroje byly zpracová-
ny odděleně podle původního členění na spalovny, energetiku
a  technologie. Bylo nutné identifikovat provozovny, u  kte-
rých lze předpokládat činnost pouze v  pracovní dny. U  těch
pak byla celková suma emisí za  konkrétní 28denní interval
rozdělena pouze mezi pracovní dny (viz tab. 3). Důležitou
součástí bylo určení emisních faktorů velkých zdrojů katego-
rie REZZO1 závisejících na  výkonu zařízení a  druhu paliva.
S pomocí těchto emisních faktorů byl možný rozpočet sumy
4 PAUna jednotlivé látky. To již nebylo nutné v případě zdrojů
kategorie REZZO2, pro které byly emisní toky dodány odděle-
ně pro jednotlivé látky.
Hlavním zdrojem dat pro liniové zdroje emisí (silniční doprava)
byly informace ze sčítání dopravy, provedeného Ředitelstvím
silnic a  dálnic v  roce 2005 [13], doplněné o  expertní odhad
konzultovaný se zaměstnanci CDV, v.v.i.. Informace o  složení
dynamického vozového parku v ČR byly zjištěny ze studie Píši
et al. [12]. S přispěním informací o intenzitě dopravy pak byly
vypočteny hodnoty emisí vztažené k délce úseku komunikace.
K výpočtu bylo použito emisních faktorů z příručky k programu
COPERT [11]. (software k  výpočtu emisí z  dopravy) (http://
www.emisia.com/copert/). Kombinací dat z  výše zmíněných
zdrojů bylo při výpočtu emisí z  dopravy zajištěno zohlednění
parametrů, jako jsou složení a stáří vozového parku, podíl vo-
zidel spalujících olovnatý a bezolovnatý benzin nebo naftu nebo
jejich výskyt na různých typech komunikací.
Vzhledem k tomu, že rozptylový model SYMOS'97 neumožňu-
je vložení liniových zdrojů v podobě linií, bylo nutné tyto linie
nejprve převést na skupiny bodů v pravidelné vzdálenosti (v pří-
padě této studie 50 m) zastupující určitý úsek komunikace.
K těmto bodům je následně nutné přiřadit hodnotu emisí, vy-
počtených pro příslušný úsek liniového zdroje [2]. K provedení
výše popsaných kroků bylo ve značné míře využito funkcí GIS.
Informace o  plošných zdrojích (zdrojích lokálního vytápění)
byla stejně jako v  případě bodových zdrojů získána z  databá-
ze REZZO. Tento typ emisí byl dodán ve formě sum jednotli-
vých druhů paliv spotřebovaných za  dobu kalendářního roku
v  jednotlivých obcích. Takto rozdělená data však bylo třeba
rozpočítat s ohledem na provoz vytápění. K tomu bylo využito
informací o délce topného období [18], které je závislé na prů-
měrných denních teplotách, a počtu denostupňů [16] (viz tab.
3) vypočítaných pro každý 28denní interval. Součástí databáze
byly v případě zdrojů lokálního vytápění i emisní faktory, před-
stavující poměr hmotnosti do ovzduší přecházející znečišťující
látky ke hmotnosti nebo objemu paliva [17]. Pomocí nich pak
byly vypočteny celkové emise vztažené k obcím. Podobně jako
liniové zdroje, také zdroje plošného charakteru je nutné pro im-
port do  modelu SYMOS'97 transformovat na  body [2]. Toho
4/2013 OCHRANA OVZDUŠÍ

bylo dosaženo opět pomocí GIS. Zastavěné plochy obcí byly po- Obr. 1: Princip fungování pasivního vzorkovače. Schéma
kryty pravidelnou sítí bodů a celkové emise za každou obec byly naznačuje vertikální průřez zařízením, PUF disk je uprostřed.
Šipky naznačují proudění vzduchu vzorkovačem. (Upraveno
pravidelně rozděleny bodům tuto obec pokrývajícím. podle [7])
Referenční body
Referenční body (receptory) jsou body, do kterých jsou ulože-
ny výsledky výpočtů modelu. Z důvodu následného zpracová-
ní v  GIS pomocí prostorových interpolací je vhodné vytvořit
pravidelnou síť referenčních bodů. Na rozdíl od výškopisných
dat však lze vytvořit a  vložit také nepravidelnou síť, čehož
bylo využito pro zahuštění sítě referenčních bodů v oblastech,
kde je při vizualizaci výsledků požadována větší prostorová
přesnost. Takovými oblastmi jsou např. blízké okolí linio-
vých zdrojů znečištění, které působí těsně nad povrchem. Pro
tvorbu této skupiny referenčních bodů byl v GIS vytvořen spe-
ciální algoritmus, který vhodně kombinuje jednoduché funkce
GIS, jako např. vytvoření bufferu, Thiessenových polygonů,
odečítání vrstev nebo převedení linie na bod. Pouze v případě
velmi husté sítě komunikací algoritmus není schopen zajistit Obr. 2: Ukázka výškopisných dat
vzájemnou protilehlost bodů po  obou stranách komunika-
ce. Při správném použití naopak eliminuje nežádoucí efekt
doprovázející generování referenčních bodů pouze z  buf-
ferů. Tímto efektem je nahodilé rozmístění referenčních bodů
podél komunikace, a tedy jejich různá vzdálenost od bodů, re-
prezentujících liniový zdroj, která může mít za následek mo-
delování nepravých maximálních koncentrací v referenčních
bodech blíže k bodu liniového zdroje a nehomogenitu koncen-
trací, modelovaných podél komunikace.
Možnost vložení nepravidelné sítě byla využita také pro výpo-
čet koncentrací sledovaných škodlivin přesně v bodech pasiv-
ního vzorkování, což umožňuje lepší možnost porovnání vý-
sledků bez nutnosti spoléhání na výsledky interpolace.

Vizualizace výsledků
Výstupem rozptylového modelu SYMOS'97 je soubor uži-
vatelem předem definovaných referenčních bodů obsahující
predikované koncentrace. Aby měly výsledky srozumitelnou
a dobře čitelnou formu, je třeba referenční body dále zpraco-
vat, nejlépe pomocí GIS. Vhodným nástrojem k získání spo-
jitého pole koncentrací z  diskrétních bodů jsou prostorové
interpolace, kterých je však několik druhů. Obecně lze říci,
že čím hustší je síť interpolovaných bodů, tím méně rozdílné
výsledky jednotlivé metody podávají. V této práci byla použita
interpolační metoda Ordinary Kriging.

VÝSLEDKY
Pro všechny látky byly popsaným postupem vytvořeny mapy
koncentrací predikovaných rozptylovým modelem a  grafy
srovnávající časový průběh predikovaných a  pasivními vzor-
kovači měřených koncentrací na jednotlivých lokalitách. Zde
uvádíme pouze výstupy pro benzo[a]pyren (obr. 4 – 8), jakož-
to nejtoxičtější ze čtyř PAU sledovaných v této studii, Koncen-
trace ostatních PAU s koncentracemi benzo[a]pyrenu korelují
a výstupy jsou analogické. Nejvyšší koncentrace (kolem 2 ng/
m-3) byly pasivním vzorkováním zjištěny na  všech lokalitách
pro benzo[b]fluoranten, extrémní hodnota 7 ng/m-3 byla na-
měřena pouze na lokalitě Mštěnovice v předposledním vzor-
kovacím období. Koncentrace ostatních látek se pohybovaly
na přibližně polovičních hodnotách, do 1 ng/m-3.
Koncentrace sledovaných látek predikované pro lokality pa-
sivního vzorkování modelem SYMOS'97 se oproti tomu nej-
častěji pohybují okolo 0,2 ng/m-3, tedy na  hladinách přibliž-
ně 5× nižších. Pouze v případě benzo[b]fluorantenu dosáhla
predikovaná zimní maxima až 0,7 ng/m-3. V letních měsících
se predikované koncentrace sledovaných PAU na všech sledo-
vaných lokalitách blíží nule.
DISKUZE
Atmosférické koncentrace sledovaných látek měřené pasivními
vzorkovači vykazovaly na všech vzorkovacích lokalitách sezónní
variabilitu s nejvyššími hladinami v zimním období. Měsíční va-
riabilita koncentrací na lokalitě Juřinka byla přitom výrazně vyšší

– 17 –
OCHRANA OVZDUŠÍ 4/2013

Obr. 3: Lokalizace zdrojů REZZO1 a 2 v okolí Obr. 4: Koncentrace benzo[a]pyrenu zjištěné na sledovaných
Valašského Meziříčí lokalitách metodou pasivního vzorkování (ng.m-3) v průběhu
jednoho roku

devším na lokalitách Juřinka a Příluky. Mez tolerance koncentra-

než na ostatních lokalitách. V chování čtyř sledovaných látek ne-
byly pozorovány významné rozdíly. Je třeba si uvědomit, že pro
výpočet atmosférických koncentrací bylo v  tomto případě pou-
žito zjednodušeného modelu, kdy se pro celý rok používá stejná
hodnota teoretického vzorkovacího objemu. Ve skutečnosti může
být ale rozdíl ve vzorkovaném objemu mezi těmito obdobími až
řádový. Skutečné letní hladiny by pak byly až několikrát nižší, než
vyplývá z této práce, zatímco zimní hladiny naopak vyšší.
V mnoha případech jsou naměřené hodnoty vyšší, než je průměr-
ná roční zákonem povolená koncentrace 1 ng/m-3 [10], a to pře-
Obr. 5: Predikované a měřené koncentrace benzo[a]pyrenu na lokalitách Hvězdárna (a), Juřinka (b), Mštěnovice (c) a Příluky
(d). Levá osa y se vztahuje k výsledkům modelování, pravá k výsledkům pasivního vzorkování
cí však v roce 2006 byla 4 ng/m-3, přičemž do začátku roku 2010
postupně klesala na 0 % povolené koncentrace [9]. Nejnižší kon-
centrace byly pasivním vzorkováním zjištěny v městské lokalitě
Hvězdárna, která je pravděpodobně méně ovlivněna náhodnými
zdroji emisí, které nejsou obsaženy v použité emisní inventuře.
Proto i srovnání výsledků modelování a vzorkování vychází nej-
lépe pro tuto lokalitu. V časových řadách pasivního vzorkování
pro ostatní lokality lze pozorovat zvýšené koncentrace především
v  podzimním období, kdy lze ve  venkovských sídlech očekávat
činnosti jako pálení biomasy na  zahradách, které by tyto jevy
do značné míry vysvětlovalo.
Predikované časové řady se vyznačují výraznější sezonalitou s té-
měř nulovými koncentracemi v  letních vzorkovacích obdobích
(obr. 7). Ty jsou způsobeny absencí emisí ze zdrojů lokálního vy-
tápění ve výpočtech modelu pro letní období a podstatně lepšími
rozptylovými podmínkami [5]. Nejvyšší modelované koncentra-
ce benzo[a]pyrenu dosahují přibližně poloviny uvedeného cílové-
ho limitu.
Zjištěné rozdíly ve výsledcích pasivního vzorkování ovzduší a mo-
delovaných koncentrací pro benzo[a]pyren (obr. 5) lze vysvětlovat

– 18 –
4/2013 OCHRANA OVZDUŠÍ

několika způsoby. Ze strany rozptylového Obr. 6: Srovnání průměrných koncentrací benzo[a]pyrenu ve Valašském Meziříčí
modelu mohlo mít vliv na  výsledky mo- v prvním (27. 12. 2005 – 24. 1. 2006) vzorkovacím období (vlevo) a sedmém
(13. 6. – 11. 7. 2006) vzorkovacím období (vpravo) vypočtených rozptylovým
delování jak několik zobecnění vyplý- modelem SYMOS'97
vajících z  podstaty modelu, tak některé
generalizace provedené při práci s  mo-
delem. Rozptylový model SYMOS'97
především není určen pro modelování
emisí v městské zástavbě a uličních kaňo-
nech [2]. V této studii může toto omezení
ovlivňovat především městskou lokalitu
Hvězdárna. V metodice je také uvedeno,
že model nelze použít pro modelování
rozptylu pod inverzní vrstvou a za bezvět-
ří. Ve větrných růžicích speciálně vygene-
rovaných pro použití v této studii je však
bezvětří zastoupeno nezanedbatelným
podílem. Stejně tak výskyt inverzí je bě-
hem celého roku pravděpodobný. Menší
vliv může mít na  výsledky modelování
zobecnění tvaru liniových zdrojů, kdy byl
segment komunikace v modelu nahrazen přímou spojnicí počát- ho pracovat s těmito typy látek (představujícími potenciální riziko
ku a konce segmentu. Tato délka je brána jako délka komunikace, zdravotních efektů). K tomu je ovšem třeba zlepšit úroveň zna-
a  proto došlo ke  zkrácení některých klikatějších segmentů ko- losti o vlastnostech sledovaných látek (reakční konstanty) a jejich
munikace. V neposlední řadě bylo po odborné osobní konzultaci chování v různých environmentálních matricích [4].
zanedbáno pravidlo stanovující maximální délku elementu linio- Správnou cestou by tak v  budoucnu mohlo být zavádění
vého zdroje a maximální délku strany elementu plošného zdroje lagrangeovských a eulerovských modelů, které dokáží věrněji
ve vztahu ke vzájemné vzdálenosti nejbližších referenčních bodů. vystihnout rozptyl látek v atmosféře a poskytují často nástroje,
Pro dodržení tohoto pravidla by bylo třeba vytvořit síť referenčních které umožňují zohlednit specifické vlastnosti různých látek
bodů s přibližně pětkrát větší vzdáleností sousedních bodů, nebo a procesy, kterým tyto látky v ovzduší podléhají. Zde je třeba
naopak výrazně zmenšit délky elementů liniových zdrojů a stran dodat, že tyto modely jsou podstatně náročnější na data, pře-
elementů plošných zdrojů. Oba tyto postupy by znamenaly kom- devším data popisující meteorologii.
plikace v podobě ztráty prostorového rozlišení výsledku nebo neú-
měrného prodloužení výpočtu. ZÁVĚR
Významné zkreslení výsledků může být způsobeno také neú-
plností emisních dat či jejich nepřesnostmi. Jak již bylo uve- Popsaná studie demonstruje užitečnost propojení rozptylo-
deno výše, použité emisní inventury nezahrnovaly náhodné vého modelu SYMOS'97 s Geografickým informačním systé-
zdroje emisí. mem. Toto propojení podstatně urychlilo a  zautomatizovalo
Další nejistoty generuje technika pasivního vzorkování ovzduší. přípravu dat pro vstup do modelu a vyloučilo náhodné chyby
V této studii nebylo využito obohacení odběrového média izoto- při jejich zpracování. Funkcí GIS bylo následně využito také
picky značenými standardy a teoretické vzorkovací objemy byly pro přehlednou vizualizaci výsledků modelování.
odvozené z databáze dlouhodobého paralelního měření aktivní Dalším přínosem projektu je ověření metodicky nové mož-
a pasivní technikou na stanici ČHMÚ v Košeticích. Tato expe- nosti použít pasivní vzorkovač ovzduší k ověření atmosféric-
rimentální data naznačují, že dlouhodobá mediánová hodnota kých koncentrací sledovaných látek predikovaných modelem
vzorkovacího objemu se pro čtyři sledované PAU pohybuje při- SYMOS'97. Metoda pasivního vzorkování má pro verifikaci
bližně na 10 % vzorkovacích objemů atmosférických polutantů rozptylových modelů velký potenciál především díky své ceně,
v plynné fázi, tedy kolem 0,4 – 0,7 m3 za den (12 – 20 m3 za jed- jednoduché logistice a nízkonákladovému provozu. Umožňují
no vzorkovací období). Při tomto zjednodušení se ovšem nezo- tak paralelní vzorkování v husté síti lokalit nejen v krátkodo-
hledňuje vliv sezónních meteorologických podmínek. S rostou- bém, ale i  dlouhodobém horizontu. Právě schopnost postih-
cí teplotou se distribuce vysokomolekulárních PAU posunuje nout i sezónní variabilitu koncentrací je pro srovnávání mode-
od atmosférických částic k plynné fázi a s tím roste i vzorkovací lových výstupů s reálně naměřenými hodnotami klíčová.
rychlost (teoretický vzorkovací objem), v některých případech Ačkoli srovnání měřených a  modelovaných koncentrací vyka-
až řádově. Je proto nezbytné v dalších studiích použít zpřesněný zuje rozdíly různé míry závažnosti, studie ověřila, že kombinace
multiparametrický model zahrnující i meteorologické vlivy. rozptylového modelu s  geografickým informačním systémem
Největší slabinou rozptylového modelu SYMOS'97 je ovšem zřej- a  reprezentativními daty z  terénních měření je do  budoucna
mě to, že neumožňuje modelování látek vyskytujících se v obou správnou cestu. Dalším krokem by mělo být vytvoření multipa-
fázích. PAU byly proto v této studii modelovány pouze v plynné rametrického modelu pro interpretaci výsledků pasivního vzor-
fázi, což zcela neodpovídá realitě. Zejména v zimním období bude kování ovzduší v závislosti na meteorologických parametrech.
podstatná část vybraných PAU vázaná na atmosférické částice, Toto zpřesnění koncentračních dat s vysokým prostorovým i ča-
což ovlivní jejich chování i  perzistenci v  atmosféře. Logickým sovým rozlišením ukáže jasněji skutečné nedostatky rozptylo-
krokem by proto měl být vývoj rozptylového modelu umožňující- vých modelů a umožní jejich další cílený vývoj.

– 19 –
OCHRANA OVZDUŠÍ
Obr. 7: Srovnání koncentrací benzo[a]pyrenu ve Valašském
Meziříčí v jednotlivých vzorkovaných intervalech,
vypočtených rozptylovým modelem SYMOS'97
Poděkování
Autoři děkují za spolupráci při realizaci studie Ing. P. Machál-
kovi a RNDr. J. Kederovi, CSc. z ČHMÚ, dále Idea – Envi s. r. o.,
CDV v.v.i. a kolegům z Centra pro výzkum toxických látek v pro-
středí. Studie vznikla za  podpory projektu OP VaVpI CETO-
COEN (CZ.1.05/2.1.00/01.0001) a Výzkumné infrastruktury
RECETOX (projekt MŠMT LM2011028).
LITERATURA
[1] Bartkow, M. E., Huckins, J. N., Müller, J. F. (2004): Field-ba-
sed evaluation of semipermeable membrane devices (SPMDs)
as passive air samplers of polyaromatic hydrocarbons (PAHs).
Atmospheric environment 38, 5983 – 5900
[2] Bubník, J., Keder, J., Macoun, J.: SYMOS'97 – Systém modelo-
vání stacionárních zdrojů, Metodická příručka. Nakladatelství
ČHMÚ, Praha, 1998. 60 s. ISBN 80-85813-55-6
[3] Bubník, J., Keder, J., Macoun, J., Maňák, J. (2006): SYMOS'97,
Výpočet znečištění ovzduší, Nakladatelství ČHMÚ, Praha. Dostup-
né z: <http://www.idea-envi.cz/download/symos97v2006/Symos97
-master.pdf>. Citováno v lednu 2011. ISBN 80-85813-55-6, 46 s.
[4] Dvorská, A., Lammel, G., Klánová, J. (2011): Use of diagnostic
ratios for studying source apportionment and reactivity of am-
bient polycyclic aromatic hydrocarbons over Central Europe.
Atmospheric Environment 45, 420 – 427.
[5] Ekologické centrum Most: Rozptylové podmínky [online].
©2011. Dostupné z: <http://www.ecmost.cz/ovzdusi.php?page
=rozptylove_podminky>. Citováno v lednu 2011.
– 20 –
4/2013
Obr. 8: Srovnání příspěvků jednotlivých kategorií REZZO
k celkovému znečištění ve Valašském Meziříčí v prvním
vzorkovacím období (27. 12. 2005 – 24. 1. 2006),
vypočtených rozptylovým modelem SYMOS'97
[6] Hasel, M., Kost, W.-J., Nielinger, J. (2010): Atmospheric Disper-
sion Modelling – Prognostic Wind Field Libraries: 13th Confe-
rence on Harmonisation within Atmospheric Dispersion Mode-
lling for Regulatory Purposes, Sborník konference, s 600 – 604.
[7] Klánová, J., Čupr, P., Kohoutek, J., Harner, T. (2008): Asse-
ssing the influence of meteorological parameters on the per-
formance of polyurethane foam-based passive air samplers,
Environmental Science & Technology 42, 550 – 555.
[8] Kohoutek, J. (2005): SOP 4 – Metodika pasivního vzorkování
ovzduší. Recetox, MU Brno, 4 s.
[9] Nařízení vlády č. 350/2002, kterým se stanoví imisní limity
a podmínky a způsob sledování, posuzování, hodnocení a říze-
ní kvality ovzduší, Příloha č. 1.
[10] Nařízení vlády č. 597/2006 o sledování a vyhodnocování kvali-
ty ovzduší, Příloha č. 1.
[11] Ntziachrsitos, L., Samaras, Z.: COPERT III – Computer pro-
gramme to calculate emissions from road transport. Methodo-
logy and Emission Factors. Copenhagen, EEA, 2000, 86 s.
[12] Píša, V., Jareš, R., Karel, J., Říha, M., Štych, P.: Zjištění ak-
tuální dynamické skladby vozového parku a  jeho emisních
parametrů. Praha, ŘSD, 2001. 86 s. [online]. Dostupné z:
<http://www.rsd.cz/rsd/rsd.nsf/a3eda25d005dc6bec125737e-
0045602e/8285cc8dddd79828c1256dbf002ccf09/$FILE/vo-
zovy_park.pdf>.
[13] Ředitelství silnic a dálnic: Sčítání dopravy v roce 2005 [online].
©2005 [citováno 2008-04-07]. Dostupné z: <http://www.scita-
ni2005.rsd.cz/html/1_zl.htm>.
4/2013
[14] Samakovlis, E., Huhtala, A., Bellander, T., Svartengren, M.: (2005)
”Valuing health effects of air pollution-Focus on concentration-re-
sponse functions“, Journal of Urban Economics 58, 230 – 249.
[15] Sáňka, O.: Modelování a  vizualizace environmentálních dat
v prostoru a čase s využitím GIS. [diplomová práce], Masary-
kova univerzita, Brno, 2008. 90 s.
[16] TZB-info: Denostupně – teorie k  výpočetní pomůcce [online].
©2008. Dostupné z: <http://www.tzb-info.cz/t.py?t=2&i=2592>.
Citováno v říjnu 2010. ISSN 1801 – 4399.
[17] TZB-info: Emise z kotelen a ochrana ovzduší (III) [online]. ©2005.
Dostupné z: <http://www.tzb-info.cz/2309-emise-z-kotelen-a-ochra-
na-ovzdusi-iii>. Citováno v lednu 2011. ISSN 1801 – 4399.
spalovny
EVIDENCE ZAŘÍZENÍ PRO TEPELNÉ ZPRACOVÁNÍ ODPADU V ČR
Ilona Dvořáková
Český hydrometeorologický ústav Praha, oddělení emisí a zdrojů, ilona.dvorakova@chmi.cz
ABSTRAKT
Článek je zaměřen na problematiku zařízení pro tepelné zpracování od-
padu v ČR. Evidence těchto zařízení je vedena od roku 2002 na základě
legislativních povinností. Od této doby jsou na stránkách ČHMÚ zveřej-
ňovány údaje z registru spaloven a zařízení ke spoluspalování odpadu,
provádí se rovněž reporting k implementaci směrnice o spalování od-
padu. Důraz je kladen na inventarizaci emisí, ze které vyplývá, že podíl
emisí všech znečišťujících látek na  celkovém množství je velmi nízký.
Tepelné zpracování odpadu má proto velkou perspektivu z hlediska eko-
nomického i ekologického.
Klíčová slova: tepelné zpracování odpadu, evidence, legislativa, zne-
čišťující látky
SOUVISEJÍCÍ LEGISLATIVA
Evidence zařízení tepelně zpracovávajících odpad v  ČR je vede-
na od  roku 2002, kdy vstoupil v  platnost zákon č. 86/2002 Sb.,
o  ochraně ovzduší a  o  změně některých dalších zákonů (zákon
o ochraně ovzduší) a jeho prováděcí předpisy. Jedním z nich bylo
nařízení vlády č. 354/2002 Sb., kterým se stanovily emisní limity
a další podmínky pro spalování odpadu, které byly v souladu s ev-
ropskou legislativou (směrnice 2000/76/EC, o spalování odpadu).
Podle § 13 zákona vznikla povinnost vést registr emisí a zdrojů
znečišťování ovzduší (REZZO), jehož součástí je i evidence spalo-
ven odpadu; nařízení vlády č. 354/2002 Sb. dále uzákonilo povin-
nost vést evidenci povolení včetně závazných podmínek provozu.
Výše uvedené předpisy byly zrušeny v  roce 2012, kdy nabyl
účinnost nový zákon o  ochraně ovzduší (zákon č. 201/2012
Sb.) a vyhláška č. 415/2012 Sb., o přípustné úrovni znečišťo-
vání a jejím zjišťování a o provedení některých dalších ustano-
vení zákona o ochraně ovzduší (tzv. emisní vyhláška).
Ke změně došlo na základě implementace směrnice č. 2010/75/
EU, o  průmyslových emisích (integrované prevenci a  omezo-
vání znečištění), která nahrazuje několik doposud platných
právních předpisů včetně směrnice 2000/76/EC, o  spalování
– 21 –
OCHRANA OVZDUŠÍ
[18] Vyhláška č. 152/2001 Sb., kterou se stanoví pravidla pro vytápě-
ní a dodávku teplé užitkové vody, měrné ukazatele spotřeby tep-
la pro vytápění a pro přípravu teplé užitkové vody a požadavky
na vybavení vnitřních tepelných zařízení budov přístroji regulu-
jícími dodávku tepelné energie konečným spotřebitelům
[19] Wilford, B. H., Harner, T., Zhu J., Shoeib, M., Jones, K. C.
(2004): Passive Sampling Survey of Polybrominated Diphenyl
Ether Flame Retardants in Indoor and Outdoor Air in Ottawa,
Canada: Implications for Sources and Exposure. Environmen-
tal Science & Technology 38, 5312 – 5318.
[20] Zákon č. 86/2002 Sb. o ochraně ovzduší a o změně některých
dalších zákonů (zákon o ochraně ovzduší)
EVIDENCE OF INSTALLATIONS FOR THERMAL
TREATMENT OF WASTE IN CR
The article is focused on the issue of installations for thermal treatment
of waste in CR. Evidence of these installations is led since 2002 according
to requirements of legislation. Information from register of waste incin-
erators and co-incinerators are made available to the public on website
of Czech Hydrometeorological Institute. The article is mainly focused on
emission inventory, which shows that the share of emissions of all pol-
lutants in the total number is very low. Therefore, thermal treatment of
waste has great potential both economically and environmentally.
Key words: thermal treatment of waste, evidence, legislation, pollutants
odpadů. V případě zařízení pro tepelné zpracování odpadu však
došlo pouze k  nepatrným změnám oproti původní legislativě
včetně zachování emisních limitů.
POSKYTOVÁNÍ INFORMACÍ VEŘEJNOSTI
Český hydrometeorologický ústav zpracovává a  průběžně
aktualizuje databázi zařízení pro tepelné zpracování odpadu.
Na  webových stránkách ČHMÚ zpřístupňujeme tyto úda-
je z  registru spaloven komunálních a  nebezpečných odpadů
a zdrojů znečištění ovzduší spoluspalujících odpad:
– Měsíčně aktualizovaný přehled spaloven a zařízení spoluspa-
lujících odpad (http://portal.chmi.cz/files/portal/docs/uoco/
oez/emise/spalovny/index.html)
Potřebné údaje jsou získávány z pravidelných měsíčních hlá-
šení ČIŽP na základě informací oblastních inspektorátů, které
oznamují aktuální změny v provozu zařízení − změny názvu
nebo provozovatele, změny technologií, stavební úpravy, změ-
ny složení spalovaného odpadu, odstavení zdroje z  provozu
nebo uvedení do provozu, poskytují informace o provedených
měřeních a plnění emisních limitů.
ANNEX 2

Manuscript of Sáňka, O., Melymuk, L., Čupr, P., Dvorská, A. & Klánová, J.
Dispersion modeling of selected PAHs in urban air: A new approach
combining dispersion model with GIS and passive air sampling. Atmos.
Environ. 96, 88–95 (2014).
 Atmospheric Environment 96 (2014) 88e95

Contents lists available at ScienceDirect

Atmospheric Environment
journal homepage: www.elsevier.com/locate/atmosenv

Dispersion modeling of selected PAHs in urban air: A new approach

combining dispersion model with GIS and passive air sampling
Ond
rej Sa
n


ka, Lisa Melymuk, Pavel Cupr *  1, Jana Kla
, Alice Dvorska nova

RECETOX e Research Centre for Toxic Compounds in the Environment, Masaryk University, Kamenice 5/753, 625 00 Brno, Czech Republic

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Passive air sampling data and

dispersion modeling gives insight
into PAH sources.

The study identified probable pres-
ence of unquantified sources of PAHs.

Combined GIS and dispersion
modeling is promising for future air
quality assessment.

a r t i c l e i n f o a b s t r a c t

Article history: This study introduces a new combined air concentration measurement and modeling approach that we
Received 14 April 2014 propose can be useful in medium and long term air quality assessment. A dispersion study was carried
Received in revised form out for four high molecular weight polycyclic aromatic hydrocarbons (PAHs) in an urban area with in-
30 June 2014
dustrial, traffic and domestic heating sources. A geographic information system (GIS) was used both for
Accepted 2 July 2014
Available online 11 July 2014
processing of input data as well as visualization of the modeling results. The outcomes of the dispersion
model were compared to the results of passive air sampling (PAS). Despite discrepancies between
measured and modeled concentrations, an approach combining the two techniques is promising for
Keywords:
Passive air sampling
future air quality assessment. Differences between measured and modeled concentrations, in particular
Air dispersion modeling when measured values exceed the modeled concentrations, are indicative of undocumented, sporadic
GIS pollutant sources. Thus, these differences can also be useful for assessing and refining emission
Polycyclic aromatic hydrocarbons inventories.
Emission inventories © 2014 Elsevier Ltd. All rights reserved.

1. Introduction techniques, and have specific advantages and disadvantages

Methods for obtaining spatially-distributed air concentration
information for persistent organic pollutants (POPs)/semi-volatile
organic compounds (SVOCs) remain a key challenge within the
field of environmental science, despite their ongoing development.
However, knowledge of air concentrations of these compounds is
important both for regulatory treaties and human health risk
assessment. Both measurement and modeling are commonly used
* Corresponding author.

et al., 2008; Melymuk et al., 2011). For the gas phase-associated
E-mail address: cupr@recetox.muni.cz (P. Cupr).
1
Current address: Global Change Research Centre AS CR, v.v.i., Belidla 4a, 603 00
Brno, Czech Republic.
depending on sources, compounds and study application.
Various passive air sampling (PAS) devices have been intro-
duced and applied since late 1990s (Choi et al., 2007; Klanova et al.,
2006; Santiago and Cayetano, 2007) as they are capable of sampling
various SVOCs in a cost-effective and logistically easy manner
allowing for a dense spatial sampling coverage. Interpretation of
PAS results continuous to be, however, a major challenge. Theo-
retical sampling volumes enabling derivation of atmospheric con-
centrations of selected chemicals from their mass in PAS are usually
based on the field intercalibration of active and passive air sampling
devices (Harner et al., 2013; Hazrati and Harrad, 2007; Chaemfa
chemicals, they can be further refined using depuration com-
pounds (Persoon and Hornbuckle, 2009), which also account for

http://dx.doi.org/10.1016/j.atmosenv.2014.07.002
1352-2310/© 2014 Elsevier Ltd. All rights reserved.
 n
O. Sa
ka et al. / Atmospheric Environment 96 (2014) 88e95
Fig. 1. Location of the PAS sites and major industrial sources in the study area.
the effects of meteorological conditions. However, results for
chemical compounds associated with atmospheric particles are
more difficult to interpret. Their gas-particle partitioning is a
temperature-dependent process affecting their distribution among
the atmospheric phases and thus, also the efficiency of their sam-
pling (Klanova et al., 2008). Sampling of particulate matter cannot
be described easily nor quantified by depuration compounds, and
thus we rely on experimental data to estimate the efficiency of
particle sampling.
Modeling tools are also frequently used in characterizing con-
centrations of SVOCs, particularly fugacity-based multimedia
models (Csiszar et al., 2013; Mackay, 2001) and atmospheric
dispersion models (Hsu et al., 2003; Morselli et al., 2012; Tao et al.,
2006). A recent development is the use of PAS data in the evalua-
tion of modeling results (Csiszar et al., 2014, 2013; Estellano et al.,
2014; Halse et al., 2011), as in this study. However, the majority of
current models do not include a spatial analysis component or
ability for adequate visualization and spatial interpretation of re-
sults. Employing spatial tools can improve confidence and clarity of
modeling results. This can be either in the form of built-in spatial
functions or through incorporation with geographic information
systems (GIS).
The use of GIS in air pollution modeling and assessment has
seen significant growth in last decade. GIS alone can be useful in air
quality assessment as studies by e.g. (Guo et al., 2007; Shad et al.,
2009) show. It has also been used to make dispersion modeling
easier and more precise (Briggs et al., 2010; Gulliver and Briggs,
2011). Nevertheless, despite its utility, its application is still
seeing limited use because of lack of compatibility between these
89
two types of software. However, use of GIS functions is becoming
more common in the last decade, as popular dispersion models
incorporate GIS extensions to ease both data input and results
visualization, or offer additional software tools enabling users to
take advantage of GIS functions.
This paper reports on a novel application of a dispersion model
supported by GIS analysis to atmospheric polycyclic aromatic hydro-
carbons (PAHs). Four PAHs, benzo(a)pyrene, benzo(b)fluoranthene,
benzo(k)fluoranthene and indeno(123cd)pyrene (BAP, BBF, BKF and
IPY, respectively) were selected as the compounds of interest as they
are associated with significant negative health effects, particularly
carcinogenicity; BAP is identified as a known carcinogen, while BBF,
BKF, and IPY are classified as possible carcinogens (IARC, 2013), and all
are largely particle-bound at ambient temperatures, which can lead to
more complex environmental behavior. Additionally, these four
compounds had both measured concentrations and emission in-
ventories available, allowing for comparison of model results with
measured data. The dispersion model was used to predict the atmo-
spheric concentrations of these PAHs, and GIS were used to process
source and elevation input data, create a receptor grid, and carto-
graphically visualize model results. These results were then verified
using data from a passive air sampling network (Du
sek et al., 2010).
2. Methods
2.1. Study area
This case study was carried out in the area of Liberec (in
northern Czech Republic, close to the German and Polish borders)
90 n
O. Sa
ka et al. / Atmospheric Environment 96 (2014) 88e95
(Fig. 1). The city of approximately 100,000 inhabitants is situated in
a basin elevated between 330 and 500 m ASL, protected by two
mountain ranges reaching 1000 m from southwest and northeast.
Lower natural barriers enclose the city from southeast and north-
west. These orographic conditions suggest that local pollution
sources (including a waste incineration plant) are expected to have
a significant impact in this area.
2.2. Dispersion model
The applied SYMOS0 97 (Stationary Sources Modeling System)
model is a regional Gaussian dispersion model predicting atmo-
spheric concentrations of pollutants to a distance of up to 100 km
from sources (Bubnik et al., 1998). As Czech law requires the use of
this model for air quality assessment, it is used by state institutions,
environmental agencies, and private companies. However, the
model has some drawbacks, such as an inability to predict disper-
sion under inversions and calm wind conditions, and poor incor-
poration of the physicochemical properties of modeled substances.
On the other hand, the model does not limit the number of input
emission sources or receptors. Four types of input data are needed
in order to perform a model run: (a) terrain elevation, (b) meteo-
rological data, (c) emission data, and (d) receptors; all are discussed
further below.
2.2.1. Terrain elevation data
Terrain elevation data are important for modeling the influence
of natural topography and modifying the emission dispersion
direction.
The SYMOS0 97 dispersion model only allows for insertion of a
regular rectangular grid of terrain elevation data. In this study the
spatial resolution of the terrain elevation grid was 150 m, which
was sufficient to prevent smoothing of the terrain; a denser grid
would have negatively affected the computation duration.
2.2.2. Meteorological data
The only meteorological data required for SYMOS0 97 are wind
speed and direction. These have to be entered in a specific form,
classified according to five stability classes according to the classi-
fication of Bubnik and Koldovsky (1974). Each stability class is
divided into three wind speed classes with average wind speeds of
1.7, 5.0 and 11 m s1. For this study, meteorological data were
provided by the Czech Hydrometeorological Institute (CHMI) in a
form of 13 wind roses for an elevation of 10 m above ground and
covering the sampling periods of the passive air samplers
(Table S1).
2.2.3. Emission data
Emission data (year 2006) for industrial and local heating
sources were provided by CHMI, Department of Air Quality Pro-
tection. Their Registry of Emissions and Air Pollution Sources (CHMI,
1996) provides emission data of selected pollutants for the
following air pollution source categories in the Czech Republic:
small, medium and large stationary sources and mobile sources. For
medium (e.g. small enterprises) and large (industry) sources,
additional parameters such as smokestack height, diameter and
emission speed are included. All pollution sources were localized
using geographic coordinates (Fig. S1). Emission data were further
processed to obtain emission totals coinciding with the PAS sam-
pling intervals (Table S1). Emission data from sources outside of the
Czech Republic were not available.
Emissions from domestic heating sources (i.e. small sources in
the Registry) were provided in the form of individual fuel type use
per village/city per year. Emission factors (CHMI, 1996) were then
used to convert the amounts of fuel to emitted pollutant quantities.
The annual fuel use totals were then apportioned to individual time
periods based on the daily mean air temperatures (using temper-
ature records obtained from the CHMI) thus accounting for the
heating season. The Corine Land cover database (European
Environmental Agency, 2010) was used to map urban areas, and
small villages were added by manual vectorization. As the model
requires data on non-point sources, such as settlement areas, in the
form of a grid, all polygons were converted to grids with a resolu-
tion of 250 m (Fig. S2). The total emissions for each settlement were
divided by the number of corresponding grid cells in order to
spread the emissions evenly over the entire area of each city/
village.
As the emission data for mobile (linear) sources from the Czech
Registry of Emissions and Air Pollution Sources are not of sufficient
resolution, an alternative approach was developed using various
other data sources. As a first step, all roads were vectorized using the
GIS (Fig. S3). Traffic volumes were obtained from the 2005 major
road traffic census of the Road and Motorway Directorate of the
Czech Republic (RSD, 2006). Using these figures, fuel consumption
was calculated for each type of road and vehicle (Pisa et al., 2001).
Fuel-specific emission factors (Ntziachristos and Samaras, 2000)
were used to convert the amounts of fuel consumed to emissions
per length of each road. Similar to the non-point pollution sources,
the linear/mobile sources were input into the model in the form of
points placed in a line (Fig. S4). The distance between two adjacent
points was set to 100 m, where each point represented a specific
emission flux equivalent to the amount of fuel burned on that road
segment. Use of GIS was crucial for all these steps.
2.2.4. Receptors
The last crucial input for the SYMOS0 97 dispersion model was a
receptor point grid. These points did not affect the results of
calculation, but rather served to store the results. Sufficient density
of the grid and appropriate point placement was essential for
proper visualization of the model results. Unlike the terrain
elevation grid, receptor points can also be embedded into the
model as an irregular grid. This option was used in cases where
higher spatial resolution was required, such as in heavily populated
areas, near major PAH sources, or in zones parallel to roads (Fig. S5).
Also, to avoid relying on spatial interpolation when model results
were compared to field monitoring data, receptors were also placed
directly on the seven passive air sampling sites located in the study
area (Fig. 1).
As traffic emissions are released close to ground level, reduced
dispersion conditions and greater harmful effects are to be ex-
pected (Tomlin et al., 2009). Thus, a simple model enabling gen-
eration of receptor points along roads and at a constant distance
from roads was created using the ArcGIS Model builder. This is in
contrast to simply generating the receptor points by converting the
boundaries of buffers created along linear features into points. This
approach eliminates variations in distances between receptors and
linear sources. A similar approach to modeling atmospheric con-
centrations along linear sources has been incorporated in some
other models (Banerjee et al., 2011; Ganguly et al., 2008). A model
schematic is shown in Fig. S6 and examples of both approaches are
given in Fig. S7.
2.3. Modeled data visualization
The output of each dispersion model run was an MS Excel file
containing almost 12,000 receptors, which were then processed
using ArcGIS software. Ordinary kriging (Krige, 1951) spatial
interpolation was used to map modeled values, using different
color shades to represent predicted pollutant concentrations
(Fig. 3).
 n
O. Sa
ka et al. / Atmospheric Environment 96 (2014) 88e95
2.4. Passive air sampling data
Passive air samplers were deployed for 13 consecutive 28-day
periods at seven sampling sites in the centre and vicinity of Lib-
erec, Czech Republic. Samplers were deployed from December 2005
to December 2006. Three of them were situated in the inner city (1U,
2U, 3U), two in rural areas (4R, 5R), and two in the neighboring
mountains (6M, 7M) (Fig. 1 and Fig. S1). All sampling sites belong to
the RECETOX long-term MONET air quality monitoring network
(Du
sek et al., 2010). The atmospheric concentrations of the selected
PAHs (BAP, BBF, BKF and IPY) were calculated by adapting the
experimental values of the sampling rates of gas-phase compounds.
For this type of PAS, the gas-phase sampling rate is 7 m3/day over a
28 day sampling periods, and the sampling efficiency of particle-
associated chemicals is estimated to be an order of magnitude
lower (Klanova et al., 2008). Thus, assuming that the selected high-
molecular weight PAHs were mostly sorbed to particles during the
whole experimental period, we used a PAS sampling rate of 0.7 m3/
day to estimate their atmospheric concentrations. However, it
should be noted that gas-particle partitioning is strongly tempera-
ture dependent, and the possible implications on results of the study
are further discussed in section 3.2.
3. Results and discussion
3.1. Modeled concentrations
The results generated by the SYMOS0 97 dispersion model show
smooth seasonal trends, with higher values during winter and low
Fig. 2. Measured, smoothed measured (smoothed by weighted moving means of medians of adjacent values) and modeled concentrations of selected PAHs at one urban sampling
site (1U).
91
values during summer (Fig. 2, Figs. S8 and S9) caused by the
absence of domestic heating sources in summer and better condi-
tions for pollutant dispersion (Vecchi et al., 2004). The highest
concentrations of all PAHs were predicted at the urban sites near
the Liberec city centre (U1, U2, U3), likely as a result of higher
density of local heating as well as traffic sources. High concentra-
tions of BAP, BBF and BKF were also predicted at the 5R sampling
site, near the largest regional industrial source of PAHs (a bitumen
mixing plant near Liberec, at 1850 m from the 5R site, Fig. S10). IPY
did not have elevated concentrations at 5R because the plant is not
a significant source of this compound.
For BAP, the only PAH addressed in air quality regulations, the
modeled concentrations were always less than 0.25 ng m3, and
well below the Czech and European regulatory limit of 1 ng m3
(European Union, 2014; Chamber of Deputies of the Parliament of
The Czech Republic (2012)) at all sampling sites and over the
whole study duration.
The modeled concentrations had a strong negative relationship
with both elevation and settlement size, as can be observed at the
rural (4R) and mountain (6M and 7M) sampling sites (7M is at the
peak of Jested Mountain, 1000 m ASL). The lowest elevation site
(4R) is at 520 m ASL and is situated in a large village, 6M is at 740 m
ASL and in a small village, and there is almost no settlement on the
peak (7M). All three sites have almost the same distance to the
Liberec city centre (6 km) and a similar share of the winds blowing
from the city centre (5e7%) during the winter season. However,
concentrations of BAP at 7M are on average 1.67 times lower than at
6M and 2.32 times lower than at 4R, and distributions are similar
for the other three PAHs. As there are no industrial sources that
92 n
O. Sa
ka et al. / Atmospheric Environment 96 (2014) 88e95
Fig. 3. Spatial distribution development of BAP concentrations during 1st (winter) and 7th (summer) sampling period, obtained by air dispersion modeling. Note the emissions
from the bitumen mixing plant in northwest corner of the study area.
could be responsible for such differences in concentrations, it
suggests that the differences are attributed to the combined effects
of population density and elevation. Higher population density in
an area is associated with increased local emissions (Hafner et al.,
2005), and higher elevation is associated with lower concentra-
tions of atmospheric particulates and higher dispersion
(Whiteman, 2000), and thus, by extension, lower concentrations of
particulate-bound PAHs. However, due to the correlation of set-
tlement size and elevation we are not able to identify the relative
significance of each contribution.
Based on the model results, more than 95% of the winter BAP
emissions at all sampling sites except 5R originated from domestic
heating. The remainder was from traffic and industrial emissions.
However, at the site closest to the bitumen mixing plant (site 5R),
25% of winter emissions are attributed to the bitumen plant
(Fig. S11), and an even higher fraction of the summer emissions.
This is supported by the fact that at the 5R site almost 30% of winds
are blowing from the direction of the plant both in winter and
summer.
Modeled concentrations of BAP were very low during summer
due to the lack of domestic heating sources. The highest summer
levels were predicted at the 5R site due to the presence of the
bitumen mixing plant. A stronger contribution of traffic emissions
was predicted at the urban sites (1U, 2U and 3U). At the rural (4R)
and the mountain sites (6M and 7M), predicted BAP levels were
close to zero. This also confirms the hypothesis that local heating
sources are the major factor controlling the atmospheric levels of
BAP.
While none of the individual sampling sites had modeled con-
centrations exceeding 1 ng m3, the model did predict a concen-
tration hot spot at the bitumen mixing plant where BAP levels
exceeded the Czech regulatory air quality limit. The dispersion
model predicted BAP concentrations exceeding this regulatory limit
at this site in all time periods, with an annual average of
1.38 ng m3 and a maximum predicted concentration of
2.05 ng m3. As demonstrated in Fig. 3 and S12, the impact of the
bitumen mixing plant was constant throughout the year while local
heating sources only contributed in winter. The estimated contri-
bution of traffic sources was surprisingly small, between 0.6 and
2.8% in winter season and 4.0e84% in summer season due to
absence of local heating sources. Very low summer BAP levels were
predicted at the urban sites where traffic was the only summer
source (Figs. S11, S13).
3.2. Measured data
Measured PAHs air concentrations showed expected seasonal
trends; winter concentrations were highest, while summer con-
centrations were often below detection limits (Fig. 2 and S9, S14).
However, overlying the general seasonal trend, there was also
significant variability between sequential samples. For example, in
the eighth sampling interval (July 14eAugust 11) at the 4R site
concentrations of all studied pollutants were 18 times higher than
in seventh and ninth sampling periods, while in the eleventh
sampling interval (October 6eNovember 3) at the 2U site concen-
trations were three times higher than in tenth and twelfth sam-
pling periods.
The causes of the elevated concentrations in individual months
are uncertain. Concentrations in the eighth sampling period were
elevated at all sites, but the highest levels were measured at the
rural site (4R), situated in the southeastern part of the region. The
wind conditions at this site and in this sampling period were
similar to other sampling periods, and thus do not provide an
adequate explanation for the elevated concentrations; therefore
the most probable causes are incidental sources of pollution (e.g.
biomass burning, barbecuing, forest fires) (Lemieux et al., 2004)
which are not included in the emission inventory. In cases such as
the eighth sampling period, when elevated concentrations were
observed across the whole study region (Fig. S14), this is suggestive
of a larger regional unquantified source, such as forest fire; in fact,
Liberec and surrounding regions of Czech Republic, Poland and
Germany had 440 ha burned by forest fire in 2006 (JRC, 2012).
Furthermore, in Czech Republic, Poland and Germany, the peak of
the 2006 fire season was in July (JRC-IES, 2008, 2007). Elevated
levels of PAHs in individual months could also be related to tem-
perature and precipitation conditions. The elevated PAH concen-
trations measured during the eighth sampling period coincided
with very hot and dry weather compared to the long-term summer
 n
O. Sa
ka et al. / Atmospheric Environment 96 (2014) 88e95 93

average, increasing volatilization and limiting wash-out/rain-out,
while the following sampling period was relatively cold and very
rainy. Additionally, the PAS themselves can be biased by meteoro-
logical conditions: PAHs shift towards the gas phase at higher
temperatures, and thereby experience higher PAS sampling rates
during the hot seasons (Klanova et al., 2008). Therefore, derivation
of PAS air concentrations from a simple flat rate model could result
in overestimation of the atmospheric concentrations of these
compounds in warm periods. The likely explanation for the
anomalous high concentrations in the eighth sampling period is a
combination of sampling artifacts, meteorology and unquantified
regional sources.
The autumn extreme was most significant at the Liberec city
centre site (2U), close to the waste incineration plant (Fig. S14).
Interestingly, elevated PAH concentrations were not observed at
the other two urban sites (1U and 3U), despite these sites being in
close proximity (940 m north and 590 m south of 2U, respectively),
suggesting a relatively localized source. Although a change in pre-
vailing wind direction observed during this sampling interval (34%
coming from south and only 6% coming from northwest) can
partially explain some of the concentration variability, it cannot
explain very different levels measured at sites in close proximity.
The spatial variability is most likely the result of unregistered local
sources of PAHs, such as domestic biomass burning.
3.3. Comparison between modeled and measured data
Measured PAH concentrations were generally higher and more
variable than those predicted by dispersion model (Fig. 2 and S9).
Correlations between measured and modeled data (Table 1)
demonstrated that for the full yearly data, the relationships were
weak, with a maximum of R ¼ 0.326 (p ¼ 0.278). This is attributed
to presence of too many values below the detection limit during
winter and spring. To exclude this influence, we examined the
correlation for only summer and autumn. This gave much higher
values, up to R ¼ 0.980 at maximum. However, the lowest corre-
lations were for sites 2U and mainly 4R, due to occurrence of
elevated concentrations in individual months (as discussed in
Section 3.2). If excluding these two sites, the mean correlation co-
efficient between measured and modeled data for all studied sub-
stances in summer and autumn season is 0.772, p value is 0.104.
Discrepancies between measured and modeled data are attrib-
uted to (1) weaknesses in the dispersion model, (2) uncertainties in
quantification of the air concentrations, and (3) the impact of local
sources not included in the emission inventory.
SYMOS0 97, as a simple Gaussian model, cannot reflect the
impact of complex meteorological conditions that affect the
dispersion of pollutants. In addition, it is not very sensitive to the
various physicochemical properties of chemical compounds. This
model is also not recommended for modeling concentrations in
street canyons and in complex terrain during inversions and calm
periods (Bubnik et al., 1998). More advanced Gaussian models have
been improved with tools enabling prediction of chemical trans-
formations of pollutants and their atmospheric scavenging, as well
as modeling the air flow over complex terrains. Nevertheless, the
use of Lagrangian and Eulerian models appears to be better. These
models often enable modeling of multiple substances with
different physicochemical properties and perform more success-
fully with respect to simulation of atmospheric conditions and air
flow affecting pollutant dispersion. On the other hand, the use of
more complex models usually requires higher computing capacity
as well as detailed meteorological data, which is not always avail-
able. Thus, appropriate choice of a model requires consideration of
available data, properties of the compounds to be modeled, and the
requirements of the intended output in terms of temporal and
spatial resolution. In this study, the large discrepancy between the
measured and modeled data suggests SYMOS0 97 may not be
appropriate to model particle-bound PAHs. While SYMOS0 97 has
been shown to effectively model contaminant plumes and volatile
air contaminants on small scales (Keder, 2008; Keder et al., 2005),
the discrepancies observed here suggest that more information is
needed to determine to model contaminants with multiple diffuse
non-point sources and complex partitioning behavior.
Therefore further research is recommended to apply more so-
phisticated air quality models on this type of data.
The second potential source of discrepancies between modeled
and measured values is the estimation of air concentrations using
passive air samplers, as described in Section 3.2. In recent years,
several methodologies for estimation of the sampling rates have
been used, and work on improving the characterization of PUF-PAS
(especially for particle-bound compounds) is continuing. Incorpo-
ration of the influence of meteorological parameters (temperature
and wind speed) is crucial when developing such models as gas-
particle partitioning is the most important process driving the
sampling efficiency: for example, BAP is entirely particle bound
at 6  C but shifts to 10e20% gas phase at 22  C, leading to higher
PAS sampling rates (Klanova et al., 2008).
When interpreted with care, passive air sampling data can be
extremely valuable in identifying sources not included in emission
inventories (biomass burning, open-fire cooking), and assessing

Table 1
Correlation coefficients and p-values for comparison of measured and modeled data for whole year and summer þ autumn (S þ A) only for each site. Correlations that are
statistically significant (p < 0.05) are in bold italics.

Site Period BAP IPY BBF BKF

R p R p R p R p

1U Whole year 0.276 0.361 0.026 0.932 0.247 0.415 0.230 0.450
SþA 0.889 0.018 0.670 0.146 0.665 0.150 0.769 0.074
2U Whole year 0.026 0.933 0.065 0.833 0.125 0.685 0.013 0.965
SþA 0.357 0.487 0.368 0.473 0.356 0.488 0.403 0.428
3U Whole year 0.111 0.718 0.018 0.953 0.236 0.438 0.236 0.437
SþA 0.582 0.226 0.400 0.432 0.398 0.434 0.604 0.204
4R Whole year 0.322 0.283 0.288 0.340 0.329 0.272 0.331 0.269
SþA 0.220 0.675 0.156 0.768 0.260 0.619 0.155 0.769
5R Whole year 0.021 0.946 0.007 0.981 0.082 0.789 0.247 0.415
SþA 0.836 0.038 0.898 0.015 0.931 0.007 0.661 0.153
6M Whole year 0.161 0.599 0.284 0.347 0.326 0.278 0.111 0.717
SþA 0.970 0.001 0.938 0.006 0.910 0.012 0.908 0.012
7M Whole year 0.075 0.807 0.110 0.720 0.062 0.842 0.014 0.964
SþA 0.769 0.074 0.913 0.011 0.944 0.005 0.785 0.065
94 n
O. Sa
ka et al. / Atmospheric Environment 96 (2014) 88e95
their impact on local and regional air quality. These sources include
not only incidental combustion sources (other than domestic
heating sources) but also various secondary pollution sources,
including releases from historically contaminated areas. It is,
however, very important to select an appropriate time resolution of
passive sampling for the specific source types. Passive air sampling
periods vary between sampling networks. XAD samplers have a
high capacity and can be deployed over a period of one year
(Barthel et al., 2012; Hayward et al., 2010), providing annually-
averaged data which are useful for assessment of long-term time
trends. Conversely, PUF-based PAS are typically deployed in 3-
month intervals, which can provide information on the season-
ality of data, and winter and summer extremes (Gouin et al., 2005;
Melymuk et al., 2012; Motelay-Massei et al., 2005) It is still rather
difficult to interpret seasonally-resolved data as a function of
meteorological parameters. Monthly deployment is an option
when assessing chemical compounds with variable but sufficiently
high levels in the atmosphere. Additionally, combinations of pas-
sive sampling devices can be used for short-term surveys. Passive
samplers provide a strong advantage in areas with high spatial
heterogeneity in air concentrations, e.g., urban and industrial areas,
because they allow dense spatial coverage which can strengthen
the application of air quality models in such challenging locations.
The use of atmospheric modeling techniques can provide addi-
tional insight into the specific fate and transport processes gov-
erning atmospheric release and distribution of pollutants. Well-
designed case studies in combination with advanced dispersion
models can generate new knowledge, filling the gaps in our current
understanding of the fate and impact of chemicals released from
various primary and secondary sources. Modeling techniques can
also provide greater insight into emission inventories, by identifying
locations influenced by undocumented sources, e.g., forest fires in
the current study. Additionally, the incorporation of GIS in data pre-
processing and characterization of spatial source distributions im-
proves the accuracy of model inputs and speeds analysis and
interpretation, as was used in this study. GIS tools can also be helpful
in sampling network design. When incorporating measured data
into modeling frameworks, appropriate sampling design is espe-
cially important; for example, radial sampler placement is well
suited to modeling large point sources and dispersion plumes, while
regional assessment of non-point sources is better treated with
gridded sampling. The use of GIS tools also brings greater informa-
tion on factors such as orography, wind and land use, and provides
opportunity to assess model performance relative to measured data
under the influence of these conditions. This can be further useful in
identifying regionally representative sampling locations.
4. Conclusions
A new approach to air quality monitoring and assessment was
introduced in this study. A combination of air dispersion modeling,
GIS and passive air sampling was applied on a one-year-period,
where high molecular PAHs (benzo(a)pyrene, benzo(b)fluo-
ranthene, benzo(k)fluoranthene and indeno(123cd)pyrene con-
centrations were addressed in the area of Liberec, Czech Republic.
GIS was used to pre-process dispersion model input data, to process
modeling outcomes and visualize them in a map. GIS was shown to
be a useful tool, capable of improving the quality of pollution data
by connecting them to the external data files such as the CORINE
Land Cover database.
A combination of passive air sampling techniques with air
dispersion modeling was shown to be a useful tool for addressing
spatial and seasonal variability of atmospheric concentrations.
Certain discrepancies exist between the measured and modeled
data. Gaps between modeled and measured concentrations are
attributed to undocumented sources and model limitations. This,
however, requires further investigation in follow-up studies
focused on reduction of both measurement and modeling un-
certainties. However, overall the results suggest that the combi-
nation of air dispersion modeling, GIS and passive air sampling has
a great potential in the future of air pollution modeling and
monitoring, and further, can be useful in human risk assessment.
Acknowledgments
This research was supported by the Ministry of Education of the
Czech Republic (LM2011028 and LO1214), by the Technology
Agency of the Czech Republic (TB010MZP057) and the project
“Employment of Best Young Scientists for International Coopera-
tion Empowerment” (CZ.1.07/2.3.00/30.0037) co-financed from
European Social Fund and the state budget of the Czech Republic.
We thank the Czech Hydrometeorological Institute for providing
meteorological and emission inventory data.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.atmosenv.2014.07.002.
References
Banerjee, T., Barman, S.C., Srivastava, R.K., 2011. Application of air pollution
dispersion modeling for source-contribution assessment and model perfor-
mance evaluation at integrated industrial estate e Pantnagar. Environ. Pollut.
159, 865e875.
Barthel, P., Thuens, S., Shunthirasingham, C., Westgate, J.N., Wania, F., Radke, M.,
2012. Application of XAD-resin based passive air samplers to assess local
(roadside) and regional patterns of persistent organic pollutants. Environ.
Pollut. 166, 218e225.
Briggs, D.J., Collins, S., Elliott, P., Fischer, P., Kingham, S., Lebret, E., Pryl, K., Van
Reeuwijk, H., Smallbone, K., Van Der Veen, A., 2010. Mapping urban air pollu-
tion using GIS: a regression-based approach. Int. J. Geogr. Inf. Sci. 11, 699e718.
Bubnik, J., Keder, J., Macoun, J., Manak, J., 1998. SYMOS’97-Stationary Sources
Modeling System e a Reference Manual (In Czech). CHMI, Prague.
Bubnik, J., Koldovsky, M., 1974. Weather classification focused in air pollution. In:
Bo€ hm, B., et al. (Eds.), Air Pollution in Krusne Mts. HMÚ, Prague.
Chaemfa, C., Barber, J.L., Gocht, T., Harner, T., Holoubek, I., Klanova, J., Jones, K.C.,
2008. Field calibration of polyurethane foam (PUF) disk passive air samplers for
PCBs and OC pesticides. Environ. Pollut. 156, 1290e1297.
Chamber of Deputies of the Parliament of The Czech Republic, 2012. Law 201/2012
of Air Quality Protection.
CHMI, 1996. REZZO, Registry of Sources of Air Pollution [WWW Document]. URL.
http://portal.chmi.cz/files/portal/docs/uoco/oez/embil/oez99/99r14.html
(accessed 08.06.09.).
Choi, S.-D., Baek, S.-Y., Chang, Y.-S., 2007. Influence of a large steel complex on the
spatial distribution of volatile polycyclic aromatic hydrocarbons (PAHs) deter-
mined by passive air sampling using membrane-enclosed copolymer (MECOP).
Atmos. Environ. 41, 6255e6264.
Csiszar, S.A., Daggupaty, S.M., Verkoeyen, S., Giang, A., Diamond, M.L., 2013. SO-MUM: a
coupled atmospheric transport and multimedia model used to predict intraurban-
scale PCB and PBDE emissions and fate. Environ. Sci. Technol. 47, 436e445.
Csiszar, S.A., Diamond, M.L., Daggupaty, S.M., 2014. The magnitude and spatial range
of current-use urban PCB and PBDE emissions estimated using a coupled
multimedia and air transport model. Environ. Sci. Technol. 48, 1075e1083.
Du
sek, L., Klanova , J., Bor
uvkov a, J., H sek, M., Gregor, J.,
ulek, R., Jarkovský, J., Kuba


Sebkov 
a, K., H
rebí
cek, J., Holoubek, I., 2010. GENASIS: Passive Ambient Air and
Soil Sampling Czech Republic (MONET-cz) [WWW Document]. GENASIS e Glob.
Environ. Assess. Inf. Syst [Online]. Masaryk Univ. URL. http://www.genasis.cz/
index-en.php?pg¼data-sourcesemonet-cz (accessed 27.01.14.).
Estellano, V.H., Pozo, K., Silibello, C., Mulder, M.D., Efstathiou, C., Tomassino, M.P.,
Funaro, F., Donadio, I., Focardi, S., 2014. Characterization of urban pollution in
two cities of the Puglia region in Southern Italy using field measurements and
air quality (AQ) model approach. Atmos. Pollut. Res. 5, 34e41.
European Environmental Agency, 2010. Corine Land Cover.
European Union, 2014. Air Quality Standards [WWW Document]. URL. http://ec.
europa.eu/environment/air/quality/standards.htm (accessed 02.04.14.).
Ganguly, R., Broderick, B.M., O’Donoghue, R., 2008. Assessment of a general finite
line source model and CALINE4 for vehicular pollution prediction in Ireland.
Environ. Model. Assess. 14, 113e125.
Gouin, T., Harner, T., Blanchard, P., Mackay, D., 2005. Passive and active air samplers
as complementary methods for investigating persistent organic pollutants in
the Great Lakes Basin. Environ. Sci. Technol. 39, 9115e9122.
 n
O. Sa
ka et al. / Atmospheric Environment 96 (2014) 88e95
Gulliver, J., Briggs, D., 2011. STEMS-Air: a simple GIS-based air pollution dispersion
model for city-wide exposure assessment. Sci. Total Environ. 409, 2419e2429.
Guo, D., Guo, R., Thiart, C., 2007. Predicting air pollution using fuzzy membership
grade Kriging. Comput. Environ. Urban Syst. Inf. Spat. Datasets 31, 33e51.
Hafner, W.D., Carlson, D.L., Hites, R.A., 2005. Influence of local human population on
atmospheric polycyclic aromatic hydrocarbon concentrations. Environ. Sci.
Technol. 39, 7374e7379.
Halse, A.K., Schlabach, M., Eckhardt, S., Sweetman, A., Jones, K.C., Breivik, K., 2011.
Spatial variability of POPs in European background air. Atmos. Chem. Phys. 11,
1549e1564.
Harner, T., Su, K., Genualdi, S., Karpowicz, J., Ahrens, L., Mihele, C., Schuster, J.,
Charland, J.-P., Narayan, J., 2013. Calibration and application of PUF disk passive
air samplers for tracking polycyclic aromatic compounds (PACs). Atmos. Envi-
ron. 75, 123e128.
Hayward, S.J., Gouin, T., Wania, F., 2010. Comparison of four active and passive
sampling techniques for pesticides in air. Environ. Sci. Technol. 44, 3410e3416.
Hazrati, S., Harrad, S., 2007. Calibration of polyurethane foam (PUF) disk passive air
samplers for quantitative measurement of polychlorinated biphenyls (PCBs)
and polybrominated diphenyl ethers (PBDEs): factors influencing sampling
rates. Chemosphere 67, 448e455.
Hsu, Y.-K., Holsen, T.M., Hopke, P.K., 2003. Locating and quantifying PCB sources in
Chicago: receptor modeling and field sampling. Environ. Sci. Technol. 37,
681e690.
IARC, 2013. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans:
List of Classifications [WWW Document]. URL. http://monographs.iarc.fr/ENG/
Classification/index.php (accessed 29.01.14.).
JRC, 2012. EFFIS e European Forest Fire Information System [WWW Document]. Jt. Res.
Cent. URL. http://forest.jrc.ec.europa.eu/effis/applications/fire-history/ (accessed
29.01.14.).
JRC-IES, 2007. Forest Fires in Europe 2006.
JRC-IES, 2008. Forest Fires in Europe 2007.
Keder, J., 2008. Experiments with inverse-dispersion modelling. In: 12th Conference
on Harmonisation with the Atmospheric Dispersion Modelling for Regulatory
Purposes (Croatia).
Keder, J., Bubnik, J., Macoun, J., 2005. Validation of Czech Reference Model
SYMOS’97, adapted for odour dispersion modeling. In: 10th Conference on
Harmonisation with the Atmospheric Dispersion Modelling for Regulatory
Purposes, pp. 139e142. Crete.
Klanova, J., Cupr, P., Kohoutek, J., Harner, T., 2008. Assessing the influence of
meteorological parameters on the performance of polyurethane foam-based
passive air samplers. Environ. Sci. Technol. 42, 550e555.
Klanova, J., Kohoutek, J., Hamplova, L., Urbanova, P., Holoubek, I., 2006. Passive air
sampler as a tool for long-term air pollution monitoring: part 1. Performance
assessment for seasonal and spatial variations. Environ. Pollut. 144, 393e405.
Krige, D.G., 1951. A statistical approach to some basic mine valuation problems on
the Witwatersrand. J. Chem. Metall. Min. Soc. S. Afr. 52, 119e139.
Lemieux, P.M., Lutes, C.C., Santoianni, D.A., 2004. Emissions of organic air toxics from
open burning: a comprehensive review. Prog. Energy Combust. Sci. 30, 1e32.
95
Mackay, D., 2001. Multimedia Environmental Models: the Fugacity Approach, sec-
ond ed. CLC Press.
Melymuk, L., Robson, M., Helm, P.A., Diamond, M.L., 2011. Evaluation of passive air
sampler calibrations: selection of sampling rates and implications for the
measurement of persistent organic pollutants in air. Atmos. Environ. 45,
1867e1875.
Melymuk, L., Robson, M., Helm, P.A., Diamond, M.L., 2012. PCBs, PBDEs, and PAHs in
Toronto air: spatial and seasonal trends and implications for contaminant
transport. Sci. Total Environ. 429, 272e280.
Morselli, M., Ghirardello, D., Semplice, M., Raspa, G., Di Guardo, A., 2012. Integration
of an atmospheric dispersion model with a dynamic multimedia fate model:
development and illustration. Environ. Pollut. 164, 182e187.
Motelay-Massei, A., Harner, T., Shoeib, M., Diamond, M.L., Stern, G., Rosenberg, B.,
2005. Using passive air samplers to assess urban-rural trends for persistent
organic pollutants and polycyclic aromatic hydrocarbons. 2. Seasonal trends for
PAHs, PCBs, and organochlorine pesticides. Environ. Sci. Technol. 39,
5763e5773.
Ntziachristos, L., Samaras, Z., 2000. COPERT III Computer Programme to Calculate
Emissions from Road Transport, Methodology and Emission Factors (Version
2.1). EEA, Copenhagen.
Persoon, C., Hornbuckle, K.C., 2009. Calculation of passive sampling rates from both
native PCBs and depuration compounds in indoor and outdoor environments.
Chemosphere 74, 917e923.
Pisa, V., Jares, R., Karel, J., Riha, M., Stych, P., 2001. Survey on Current Dynamic Car
Park Composition and its Emission Parameters. Road and Motorway Directorate
of the Czech Republic, Prague.
RSD, 2006. Road Traffic Census 2005 [WWW Document]. Road and Motorway
Directorate of the Czech Republic. URL. http://www.rsd.cz (accessed 17.10.10.).
Santiago, E.C., Cayetano, M.G., 2007. Polycyclic aromatic hydrocarbons in ambient
air in the Philippines derived from passive sampler with polyurethane foam
disk. Atmos. Environ. 41, 4138e4147.
Shad, R., Mesgari, M.S., Abkar, A., Shad, A., 2009. Predicting air pollution using fuzzy
genetic linear membership Kriging in GIS. Comput. Environ. Urban Syst. Spat.
Data Mining-Meth. Appl. 33, 472e481.
Tao, S., Li, X., Yang, Y., Coveney, R.M., Lu, X., Chen, H., Shen, W., 2006. Dispersion
modeling of polycyclic aromatic hydrocarbons from combustion of biomass and
fossil fuels and production of coke in Tianjin, China. Environ. Sci. Technol. 40,
4586e4591.
Tomlin, A.S., Smalley, R.J., Tate, J.E., Barlow, J.F., Belcher, S.E., Arnold, S.J., Dobre, A.,
Robins, A., 2009. A field study of factors influencing the concentrations of a
traffic-related pollutant in the vicinity of a complex urban junction. Atmos.
Environ. 43, 5027e5037.
Vecchi, R., Marcazzan, G., Valli, G., Ceriani, M., Antoniazzi, C., 2004. The role of at-
mospheric dispersion in the seasonal variation of PM1 and PM2.5 concentration
and composition in the urban area of Milan (Italy). Atmos. Environ. 38,
4437e4446.
Whiteman, D.C., 2000. Mountain Meteorology: Fundamentals and Applications,
first ed. Oxford University Press, New York.
ANNEX 3

Manuscript of Sanka, O., Kalina, J., Lin, Y., Brabec, K., Deutscher, J., Futter, M.,
Butterfield, D. & Nizzetto, L. Modeling hydrological and biogeochemical
controls of the dissipation of p,p’-DDT in soil. Submitted to Environ. Sci.
Technol. (2015).
Modeling hydrological and biogeochemical controls of the dissipation
of p,p’-DDT in soil
Ondej Sanka†, Jiri Kalina†, Yan Lin§, Jan Deutscher ¶, Martyn Futter‡, Dan Butterfield⊥, Karel
Brabec†, and Luca Nizzetto*†§

† Research Centre for Toxic Compounds in the Environment (RECETOX), Faculty of Science, Masaryk University, Kamenice
5/753, Pavilion A29, Brno 62500, Czech Republic
§ Norwegian Institute for Water Research, Oslo NO-0349, Norway
¶ Department of Landscape Management, Faculty of Forestry and Wood Technology, Mendel University in Brno,
Zemědělská 3, Brno 61300, Czech Republic
‡ Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences, Uppsala, Sweden
⊥ Enmosys, Glen Ellyn, IL 60138, USA

Abstract
Despite not being used for decades in most countries, DDT is still one of the most abundant
pollutants in the environment, mostly present in soils. This study assesses p,p’-DDT dissipation in a
large Central European river catchment, comparing dissipation rates acquired by statistical analysis of
a large dataset of soil concentration measurements and mechanistic modelling. INCA-Contaminants,
a recently developed multi-compartment model was calibrated against hydrological records, organic
carbon, sediment and water concentration data, covering the period of 2006-2008. The INCA-
Contaminants model results were in agreement with the statistical model, suggesting p,p’-DDT half-
life in soil of 3000 < τ < 3800 days. Degradation appeared to be the main driver of dissipation, almost
an order of magnitude higher than the sum of runoff, river discharge, and erosion. In this study an
integrative model framework linking hydrological, biogeochemical and physical processes with a very
high level of detail in parameterization was used for the first time, delivering results with a high level
of detail and reliability, giving deeper insight in the fate of p,p’-DDT.

1. Introduction
Dichlorodiphenyltrichloroethane (DDT) and its metabolites are arguably among the most
environmentally-relevant endocrine disruptors ever studied.1 DDT was mainly used as pesticide for
crop protection and after international bans on its application; current use is essentially limited to
malaria disease vector control in several tropical countries. In Europe, DDT was the most used
pesticide until the late 1970s (27,000 T y-1). Since then its use in Europe dropped to about 300 T y-1 in
the early 1990s, following evidence of bioaccumulation. After 1995 no legal use is thought to have
occurred in European countries.2

p,p’-DDT and its homologues are highly hydrophobic, semivolatile3 and environmentally persistent4
compounds. The combination of these properties explains DDT’s ubiquitous occurrence in the
environment and its effective storage in soils and sediments, both in areas where the substance was
used and in background areas receiving DDT loads from atmospheric deposition.5 The largest
environmental pool of DDT is stored in global soils. This is estimated to represent 95% of the total
environmental burden.6 Changing environmental conditions or different types of disturbances can
enhance remobilization from soils to the atmosphere or to the aquatic ecosystem.7 In countries that
have completely banned DDT, re-volatilization from contaminated agricultural soils represents the
principal atmospheric source. Monitoring data from these areas generally display a steady decline of
DDT in both biota and abiotic compartments.6,8–11 It is important to track such a decline and address
the mechanisms of the underlying processes to enable meaningful prediction of the future
environmental and human exposure to this harmful substance.

Large scale multimedia fate models have been successfully used to reproduce the long term trend of
DDT environmental contamination.6 Nevertheless these approaches miss the level of detail necessary
to mechanistically assess hydrological and biogeochemical controls on its dissipation.

We focused on investigating the processes controlling the long-term trends of the pool of p,p’-DDT in
soil in a region representative of a temperate environment (i.e. a large Central European river
catchment) where DDT use was banned several decades ago. The study area (the Morava River
catchment in the Czech Republic, Slovakia and Austria) was selected based on the availability of a
unique dataset encompassing several hundred historical records of air, soil, river water, and river
sediment concentration data spanning over one decade.

The analysis followed three steps: i) analysis of long-term p,p’-DDT concentration trends in central
European soils using stochastic models; ii) set-up and calibration of a state-of-the-art integrated
hydrobiogeochemical/contaminant fate model (INCA-Contaminants) to mechanistically simulate the
observed p,p’-DDT trends in soil, water and sediment of the catchment; and iii) analysis of p,p’-DDT
fluxes modeled through the mechanistic framework to assess the relevance of different hydrological
and biogeochemical controls influencing p,p’-DDT dissipation from soils.

Figure 1: Morava River catchment land use, sampling sites and river and subcatchment structure
used for INCA-Contaminants parametrization.
2. Materials and methods
2.1. Case study and contamination datasets
The area selected for this study is the Morava River catchment. The Morava is a left tributary of the
Danube River with a confluence located at the border between Austria and Slovakia, draining an area
of 26658 km2, mainly in the eastern part of the Czech Republic. The river is 354 km long, originating
in northeast of Czech Republic at an elevation of 1371 m. The average discharge is 120 m3/s. Land
use types include croplands (58.7 %), forests (39.4 %), grasslands (1.4 %) and urban (0.5 %) (Table
S2).The landscape is typical of central Europe. DDT usage peaked in Czechoslovakia during the early
1970s, and was banned in 1974.12

A large dataset of DDT soil concentration is available for the Morava catchment, spanning 2 decades
(1995-2015). Data are categorized by soil type (e.g. agricultural, forested, industrial, urban, etc.), soil
horizon, level of impact (i.e. impacted or not directly impacted). The time series of atmospheric
concentrations for DDT are also available for this region from the Košetice monitoring station in
central Czech Republic. Air measurements for DDT started in 1996 and are currently on-going with a
weekly resolution. Similarly, DDT concentration data in river water and sediments are available from
a case study conducted in 2007 and 2008 in two central reaches of the Morava, covering a full year of
continuous monitoring.13,14 All the DDT data used in this study were generated over the years by the
Research Centre for Toxic Compounds in the Environment (RECETOX) and the Central Institute for
Supervising and Testing in Agriculture (ÚKZÚZ) laboratories using consistent methods and quality
assurance/control standards. The entire dataset is included in the RECETOX GENASIS database, an
open-access information system reporting geo-time referenced contamination data.15

In summary, this study incorporated a total of 1269 records of soil concentrations from 221 different
sites across Czech Republic,15 15 records of river water concentrations (truly dissolved phase)
collected using passive samplers, 29 records of river bed sediment concentration, 60 records of liquid
phase and particulate phase concentrations13,14 and 884 records of air concentrations.15

2.2. Statistical analysis of p,p’-DDT time trend in soils

Statistical regression analysis was conducted to resolve the temporal trends of p,p’-DDT in soil. Of
the 1269 available records, about 10% were below limits of detection or limits of quantification.
During the statistical analysis, these values were replaced by half of these limits in order to
semiquantitatively preserve information on low concentrations. Only records from areas labeled in
the GENASIS dataset as “non-impacted” were considered (in total 816 records). Five apparent outlier
values were also removed (threshold for exclusion was set at >100 ng g-1). The final dataset
encompassed 811 soil concentration records from 137 sites, including locations with detailed
information on four soil horizons (thus, 182 individual combinations of sites and soil horizons).

The stochastic analysis of DDT temporal trends in soils was conducted using two approaches: a
“naïve” approach (ignoring labels of location and horizon), and ii) an “informed” approach
(computing a globally-averaged trend from reiterated trend analyses of individual combinations of
locations and soil horizons). For the “informed” approach all combinations of site and soil horizon,
having less than three records were excluded. In total, 69 site-horizon combinations of 664 records
were considered.
An exponential model was chosen to regress soil concentration data over time in both the naïve and
informed approaches. This model was assumed to be a good descriptor of the decline of an
environmental pollutant with no current use. This is consistent with the general assumptions of first-
order kinetics controlling multicompartment exchange and degradation of contaminants in the
environment (e.g. MacLeod et al.16 and Wania et al.17). Trend analysis was therefore conducted after
linearization through the logarithm function. The non-parametric Theil-Sen estimator was chosen to
estimate the trends.18,19 In addition, the non-parametric Mann-Kendall test was run to calculate p-
value of the trend and tau statistics. Resulting trends from the two different approaches were
compared to assess the robustness of the analysis.

2.3. Mechanistic analysis of p.p’DDT time trend in soil

To quantitatively infer the processes controlling the long-term fate of p,p’-DDT in soil, a state-of-the-
art high-resolution integrated hydrobiogeochemical-multimedia fate model – INCA-Contaminants –
was employed. INCA (Integrated Catchment) models are a family of complex, one-dimensional
dynamic flow catchment models.20,21 INCA-Contaminants22 was recently developed by integrating a
multimedia contaminant fate model, a hydrological model, a biogeochemical model detailing organic
carbon budget in soil and river water (INCA-C23 simulating organic matter mass balance in multi-
branched, multi-land use catchments), and a sediment dynamics model (INCA-Sed24 simulating soil
erosion and in-stream transport, deposition and entrainment of sediments). Briefly, the
discretization of catchment and river structure is provided by detailing the order of streams, their
length and width between reach points, and by compiling dimension and characteristics of individual
subcatchments (including soil properties and land use information).20 An underlining rainfall-runoff
model (PERSiST)25 computes hydrologically-effective rainfall and soil moisture deficit time series that
are used by INCA models to generate daily run-off and river flow predictions. PERSiST outputs are
calibrated using catchment-specific observations of river flow. INCA-C computes carbon (C) budgets
in soil, sediment and stream water by linking different pools (solid organic C (SOC), dissolved organic
C (DOC) and inorganic C) using calibrated first-order transformation kinetics in individual
compartments. Exchange of C across the different pools, surface and diffuse run-off from soil,
vertical water advection and riverine transport of mineral and organic C are computed as function of
hydrology and temperature. The same representation of terrestrial soil structure used by INCA-C is
employed in INCA-Contaminants

Within INCA-Contaminant all submodules are automatically coupled. For example SOC fluxes from
land phase run-off are distributed across different sediment size classes. This was done to facilitate
particle-size dependent contaminant partitioning and in-stream sediment transport. The
contaminant fate module is effectively a multimedia box model overlying the INCA-C/INCA-Sed
coupled structure. It computes mass budgets of organic contaminants as a function of compound-
specific, user-definable physical-chemical properties. Once informed with atmospheric contaminant
concentration data, atmospheric deposition data and meteorological data (i.e. information on
hydrologically-effective precipitation and air temperature) INCA-Contaminants tracks, on a daily
basis, the mass budget of water, C, sediments and contaminants as a function of land use, soil
properties, river network structure, river bed morphology and contaminant-specific physical-
chemical properties, in soils and in selected locations of the river network.
2.3.1. Morava catchment representation in INCA-Contaminants
The watershed was divided into 16 sub-catchments (Fig. 1) based on a geographic information
system (GIS) layer with information of streams and relative sub-catchment boundaries.26 For each
sub-catchment and reach, parameters such as area, reach length, width and slope were extracted
(see table S3). Detailed information on daily discharge data were available for 10 sites in the
catchment from the Czech Hydrometeorological Institute (CHMI), as well as average daily air
temperature and daily precipitation records for nine stations across the watershed (Fig. 1).
Temperature and precipitation data were extrapolated from the available measuring stations to
cover the whole watershed. The values for each sub-catchment were calculated as a function of
proximity to nearest measuring stations using an ArcGIS function creating Thiessen polygons. Daily
meteorological and hydrological data covered a three-year period (2006 – 2008). This time frame
coincided with the availability of data of river water and sediment concentrations.

In INCA models, each sub-catchment is divided in land use units. Land use data were obtained from
International Steering Committee for Global Mapping webpage.27 An initial 16 classes were clustered
into six major classes encompassing broadleaf forest, needle leaf forest, mixed forest, arable land,
grassland and urban (Fig. 1, S1 and table S2). Daily data of hydrologically-effective rainfall and soil
moisture deficit for each land use type and sub-catchment were calculated starting from measured
precipitation and river discharge data using the rainfall-run-off model (PERSiST).25 The initial range-
finding of PERSiST parameters was performed manually, and final optimization was performed using
Monte Carlo analysis.

2.3.2 Organic carbon and suspended sediment: data structure and model calibration
To provide a realistic simulation of organic C fluxes and suspended sediment dynamics the model
was calibrated against data collected in different sections of the Morava and its tributaries.28 The
carbon cycle, as described in INCA-Contaminants include inputs to soils from litterfall and/or
amendments (for agricultural soils). From these inputs OC is distributed by first-order functions
across the different pools, including degradation to DOC and full mineralization. C can reach the
stream in different forms by superficial or diffuse run-off. In the stream OC transformation can also
occur, following dynamics that are independent from those of the land phase. OC can also be
incorporated to the sediment bed through deposition of suspended sediment (for SOC) or by
diffusion for DOC. Sediments classified by size and density are delivered to the stream by erosion
from soil or entrainment from the sediment bed. Both these functions are regulated by climate
through soil hydrology and stream conditions. Calibration of the C and sediment modules consisted
of manually setting data on litterfall inputs for individual land use classes, tuning first-order rate
constants for the transformation of C among different forms, tuning the parameters defining the
physical characteristics of the sediment classes and the thresholds for their dynamic entrainment
from soils and sediment bed, as described elsewhere.23,24 To initialize the model, simulation was set
to start six years before the periods of calibration data availability.

2.3.3. Inputs and parameterization of p,p’-DDT in INCA-Contaminants

The initial set of p,p’-DDT properties (including KOW, Henry’s law constant, the enthalpies of phase
transfer between octanol and water and air and water, molecular weight and volume, and
degradation rates in soil and sediments were taken from the literature.3,4

The initial soil concentrations were set according to the historical time series of p,p’-DDT in soils from
the GENASIS database.15 Data were clustered as averaged concentrations for individual land use
classes using a regression-tree based approach, as described elsewhere.29 Finally, using the time-
regression models (described in Section 3.1), data were scaled according to their expected values on
1 January 2000 (initial date of simulations).

Air concentration and atmospheric deposition data were used as inputs assuming that feedbacks
from volatilization from soil and water do not significantly affect atmospheric loadings. This approach
is appropriate for models that describe the sub-regional scale in which the average residence time of
air is of the order of few hours (as for the Morava catchment). Daily atmospheric concentrations of
p,p’-DDT were estimated by regressing measured concentrations from Košetice monitoring station
against time. After screening for different types of decay models, satisfactory results were obtained
simply using an exponential decay curve. This model was previously assumed to describe declining
atmospheric concentrations of persistent organic pollutants (e.g. Gioia et al.30).

Other inputs of DDT to the soil and water include atmospheric wet and dry deposition and litterfall-
associated deposition. Litterfall-associated deposition was estimated by considering yearly averaged
deposition velocities from Horstmann and McLachlan31 combined with the modeled air
concentrations. Daily wet deposition inputs were generated by assuming equilibrium between air
and raindrops/snow combined with rainfall and snowfall measurements (and their extrapolation
through the sub-catchments). Particle-bound deposition was calculated starting from particle-bound
concentrations estimated using a KOA-based model,32 assuming a molecular weight of aerosol organic
matter of 500 g·mol-1 as recommended by Götz et al.33 Historical records of background atmospheric
particulate organic carbon for this region were taken from the Košetice monitoring station (data for
PM10 generated through low-flow air sampler) and available from the EBAS database
(http://ebas.nilu.no). The average value of 5 ± 3 μg m3 was assumed throughout the period of the
simulations. The gas-particle partitioning model yielded a particle-bound fraction of p,p’-DDT in the
range of 10%-20%. A generic dry particle deposition velocity of 1 cm·s-1 was assumed which is in the
range of data measured in other locations in the world.34–38 This deposition velocity and the modeled
daily p,p’-DDT gaseous concentrations were used to calculate daily dry deposition data as an input
for INCA-Contaminants.

3. Results and discussion

3.1. Long term trends of DDT in soils
The concentration of p,p’-DDT in background soils of Czech Republic displayed a highly significant
(p<0.001) decline. Similar result was achieved both through the naïve and informed statistical models
(Table 1). Results of the trend analysis for individual locations and soil horizons used to calculate
overall trends in the informed method are reported in Table S1.

Table 1. Statistical model of p,p’-DDT trends.

Percentage
Model method a b Half-life (y)
yearly decline
Naïve 56.3 -0.0904 8.64 7.69
𝐶𝐷𝐷𝐷 = 𝑎 ∙ 𝑒 𝑏(𝑦𝑦𝑦𝑦−1970)
Informed 19.89 -0.0490 4.78 14.146
Figure 2: Estimate of p,p’-DDT soil concentrations in Morava watershed ( INCA-Contaminants,
“informed” approach, naïve approach, ◦ p,p’-DDT soil concentration measurement)

The two methods delivered slightly different trends with a slower (factor of 2) decline resulting from
the informed method. These rates are consistent with previous observations from other regions of
Europe. Schuster et al.39 observed a decline in organic matter normalized soil concentrations of p,p’-
DDT along a UK-Norwegian transect with an average half-life of 10 years. DDT dissipation in
European soils described by Schenker et al.6 is consistent with half-lives in the range of 15 years.
Finally, Meijer et al.40, by analyzing archived agricultural soil samples collected between 1968 and
1990 reported dissipation half-lives of 10.5-11.8 y.

The decline obtained in the present study are in agreement with decline of p,p’-DDT observed in air
samples from at the Košetice monitoring station (percentage yearly decline: 7.48, half-life: 9.26 y)
suggesting that the decline of atmospheric inputs may represent the limiting factor controlling p,p’-
DDT clearance from soil.

3.2. Simulation results

The simulated daily runoff of the river was successfully calibrated against measured river discharges
(Figure 3a and 3b). The Nash-Sutcliffe (N-S) coefficient reached 0.667.41 Due to an insufficient
number of observations and low variability of measured data of dissolved OC and suspended
sediments, a N-S evaluation of the calibration process is not useful, however the model captured
with an acceptable level of accuracy both trends and levels of DOC and suspended sediment
simultaneously measured in different reaches of the river (Fig. 3c-3f). Mean percentage error
between modeled and observed data was 20% and -14% for DOC and suspended sediment
concentrations, respectively (Fig. 3), demonstrating that INCA-Contaminants can typically predict
these parameters with an error of less than a factor of 1.2. The values of the parameters controlling C
mass budgets obtained from the calibration exercise were in the range of those obtained from a
previous application of INCA-C in which the model was thoroughly calibrated against a detailed
dataset of observed DOC concentration in rivers of boreal catchments.42

Concentrations of p,p’-DDT in water and sediments were simulated with an acceptable level of
accuracy (Figure 4). Considering the uncertainty of passive sampler-derived water concentrations,
conservatively assumed to range 2x above and below the measured value, there was no significant
difference between mean dissolved phase concentrations predicted by the model and those
measured continuously over two years in two central reaches of the Morava. Assumptions about the
accuracy of passive sampler performance were defined by conservatively considering the errors in
silicon rubber/water equilibrium partitioning, its temperature dependence, and performance
reference compound release rate.43 The 50th and 75th percentiles of the ratio between predicted and
observed values were 0.7 and 1.1, respectively. Predicted dissolved concentrations were significantly
higher than those measured through the passive samplers in only three cases in Reach 345 and one
case in Reach 350. These experimental data points were outliers displaying levels of p,p’-DDT at least
one order of magnitude lower than the mean of other samples at the same sites.

The magnitude of bed sediment concentrations was also successfully simulated in the same two sub-
catchments. The 50th and 75th percentiles of the ratio between predicted and observed values were
3.8 and 11.6 showing that the model typically successfully captured the order of magnitude of
sediment concentrations. For bed sediment, the model failed in predicting three experimental
outliers.

The bulk of DDT (over 90%) in the water column was predicted to be associated to suspended
particles and DOC. This was confirmed by experimental data from grab sampling and sediments traps
sampling performed simultaneously with the passive sampling campaign (Fig. 4).13,14 The simulations
revealed highly dynamic trends in water. Temporal variations of water concentrations had both short
term variation (with typical periods of 4-15 days), determined by fluctuations in river hydrology and
temperature, and seasonal fluctuation, with summer bulk water concentration higher than winter.

The model simulated the effect of an extreme flood event during 2-3 weeks in 2006, in which river
discharge was over one order of magnitude higher than its yearly average (Fig. 3). As a result p,p’-
DDT discharge was also higher during this high flow event, however bulk water concentration
displayed minima in this period (Fig S3) suggesting dilution had a pivotal role in controlling
concentration in the suspended and DOC-bound phases.
measured
800

700 a Reach 345 - Zlin district

Flow INCA
600
Flow measured
500
m 3/s

400

300

200

100

0
2006 2007 2008

b
800

700
Reach 350 - Uherske Hradiste district
Flow INCA
600
Flow measured
500
m 3/s

400

300

200

100

0
2006 2007 2008

10
9
8
c Reach 345 - Zlin district
Open Water DOC INCA

7 Open Water DOC measured

6
m g/L

5
4
3
2
1
0
2006 2007 2008

d
10
9 Reach 350 - Uherske Hradiste district
Open Water DOC INCA
8
7 Open Water DOC measured

6
m g/L

5
4
3
2
1
0
2006 2007 2008

10
9
8
e Reach 339 - Olomouc district
Open Water DOC INCA

7 Open Water DOC measured

6
m g/L

5
4
3
2
1
0
2006 2007 2008

f
1000
900 Reach 350 - Uherske Hradiste district Suspended Sediment INCA
Suspended Sediment measured
800
700
600
ng/kg

500
400
300
200
100
0
2006 2007 2008

Figure 3: Hydrobiogeochemistry calibration results of INCA-Contaminants for selected sub-

catchments: a) Morava discharge in Zlin district, b) Morava discharge in Uherske Hradiste district, c)
open water DOC in Zlin district, d) open water DOC in Uherske Hradiste district, e) open water DOC in
Olomouc district, f) suspended sediments in Uherske Hradiste district.
Figure 4: p,p’-DDT concentration calibration results of INCA-Contaminants for selected sub-
catchments: a) truly dissolved concentrations in Zlin district, b) truly dissolved concentrations in
Uherske Hradiste district, c) upper sediment bulk concentrations in Zlin district, d) upper sediment
bulk concentrations in Uherske Hradiste district.

3.3. Simulated long term trends of p,p’-DDT in soil and analysis of dissipation fluxes
The calibrated model predicted a similar long-term temporal trend of p,p’-DDT in soil to what was
observed from experimental data (Fig. 2). The dissipation trend was primarily sensitive to
degradation half-life in soil, suggesting degradation as the controlling process. The best agreement
between predicted and observed dissipation trends was obtained by using a degradation half-life in
soil (τ) in the range 3000 < τ < 3800 days. There is a large uncertainty in DDT degradation half-life in
soil. The τ obtained from INCA-Contaminant calibration was 3x longer than the upper τ range
reported by Mackay et al.4 but within the range of 187 to 4331 days reported by Hornsby et al.44
Recent observations identified a τ for p,p’-DDT >7000 days in Canadian soils.45 To our knowledge this
is the first landscape-based assessment of p,p’-DDT half-life that considered a large set of soil
concentration data repeatedly measured over one decade.
That degradation of p,p’-DDT is the controlling process for the dissipation in soil is confirmed by the
flux analysis (Figure 5). When INCA-Contaminants used the optimal set of parameters, degradation
represented 88% of the total p,p’-DDT loss (on a yearly base) from the soil, while the remaining loss
was due to diffuse and superficial run-off (Figures 5a and 5d). Previous applications of INCA-
Contaminants investigated the sensitivity of model output to individual parameters.20,21,23,42 The most
influential parameters for water and sediment concentrations outputs were organic KOW, τ and the
SOC mineralization rate. To acknowledge the uncertainty around these sensitive parameters, flux
analysis was conducted considering combinations of these parameters values ranging one order of
magnitude around their “optimal” value (Table S7). When this uncertainty is considered, the
influence of diffuse run-off on p,p’-DDT dissipation increased to 40 %. During periods of high
precipitation and high discharge the run-off flux can temporarily exceed total degradation in soil
(Figure 5a), demonstrating that hydrologically- and biogeochemically-driven remobilization from soil
to the water is non-negligible when assessing the fate of p,p’-DDT, even over long temporal scales
and large spatial scales.

Figure 5: p,p’-DDT fate and processes in the Morava watershed, calculated by INCA-Contaminants:
a) two main p,p’-DDT removal processes from soil – run-off and degradation, b) comparison of air-
water and air-soil exchange rates, c) rate of total atmospheric input to soil and total losses form soil,
d) all processes contribution to p,p’-DDT removal from soil.
Total atmospheric inputs had a very small influence on the overall budget of p,p’-DDT in the
landscape. The sum of wet and dry deposition, forest litter-associated deposition, and net dry
gaseous exchange was over one order of magnitude lower than the total p,p’-DDT losses from the
catchment (i.e. sum of degradation, run-off, erosion and volatilization) (Fig. 5c). Total atmospheric
input was also lower than the remobilization (i.e. sum of run-off and erosion) by about one order of
magnitude. Similarly, the dry gaseous exchange between soil surface and the atmosphere is strongly
oriented toward volatilization throughout the catchment. The volatilization flux was however a
negligible loss process for the overall budget of DDT in soil if compared to the sum of degradation
and run-off (i.e. more than one order of magnitude smaller) (Figures 5b and 5d). The catchment is
predicted therefore to behave as a strong net secondary source of p,p’-DDT to the water and air, the
bulk source being the historical burden accumulated in soils during past decades.

Interestingly, contrary to what predicted on land, the dry gaseous exchange over the river surface
was oriented toward deposition (Figure 5b), suggesting that the biogeochemical processing within
the water column and sediment bed act in maintaining low fugacity in the truly dissolved phase. The
sorption on DOC and suspended particles in water and burial of contaminated sediments in river
depositional areas likely represent the major drivers behind this behavior. Net deposition over the
water was however one to two orders of magnitude smaller compared to the total net volatilization
from the land when these fluxes are spatially integrated over the landscape (Figures 5b and 5d).

The processes controlling the long term dissipation of p,p’-DDT from soils were elucidated in a
catchment representative of central European conditions. For the first time this assessment was
obtained from an integrative model frame linking hydrological, biogeochemical and physical
processes with a very high level of detail in scenario description. The model was also constrained to
simultaneously reproduce a large number of historical time series of experimental observations. In
agreement with previous model assessment, degradation was the main dissipation driver; however
the hydrological- and biogeochemical-driven remobilization (i.e. the sum of run-off and erosion) is
not negligible. This process is often under-parameterized in many multimedia fate models. The
assessment also provided a new estimate of p,p’-DDT degradation half-life that exceed by a factor of
1.5-3 the values commonly adopted to parameterize multimedia fate models.

Acknowledgments
The research leading to these results has acquired a financial grant from the People Program (Marie
Curie action) of the Seventh Framework Program of EU according to the REA Grant Agreement No.
291782. The research is further co-financed by the South-Moravian Region. This paper reflects solely
the authors’ views and the European Union is not liable for any use that may be made of the
information contained therein. This work was supported by the GENASIS core facility of the Research
Centre for Toxic Compounds in the Environment, with financial support from the Czech Ministry of
Education (LM2011028 and LO1214) within the framework of “Major infrastructures for research,
development and innovations” and the National Sustainability Programme. This study was supported
by the Norwegian Institute for Water Research (NIVA) through its Strategic Program on contaminants
of emerging concern (CECs).

References
(1) Vos, J. G.; Dybing, E.; Greim, H. A.; Ladefoged, O.; Lambré, C.; Tarazona, J. V; Brandt, I.;
Vethaak, A. D. Health effects of endocrine-disrupting chemicals on wildlife, with special
 reference to the European situation. Crit. Rev. Toxicol. 2000, 30 (1), 71–133.

(2) Pacyna, J. M.; Breivik, K.; Münch, J.; Fudala, J. European atmospheric emissions of selected
persistent organic pollutants, 1970–1995. Atmos. Environ. 2003, 37, 119–131.

(3) Shen, L.; Wania, F. Compilation, Evaluation, and Selection of Physical−Chemical Property Data
for Organochlorine Pesticides. J. Chem. Eng. Data 2005, 50 (3), 742–768.

(4) Mackay, D.; Shiu, W. Y.; Ma, K.-C. Illustrated Handbook of Physical-Chemical Properties of
Environmental Fate for Organic Chemicals, Volume 5; CRC Press, 1997.

(5) Skrbić, B.; Durisić-Mladenović, N. Distribution of chlorinated organic pollutants in a wide

variety of soils from Europe and Asia: a multivariate statistical approach. Arch. Environ.
Contam. Toxicol. 2007, 52 (4), 466–474.

(6) Schenker, U.; Scheringer, M.; Hungerbühler, K. Investigating the Global Fate of DDT: Model
Evaluation and Estimation of Future Trends. Environ. Sci. Technol. 2008, 42 (4), 1178–1184.

(7) Nizzetto, L.; Macleod, M.; Borgå, K.; Cabrerizo, A.; Dachs, J.; Di Guardo, A.; Ghirardello, D.;
Hansen, K. M.; Jarvis, A.; Lindroth, A.; et al. Past, present, and future controls on levels of
persistent organic pollutants in the global environment. Environ. Sci. Technol. 2010, 44 (17),
6526–6531.

(8) Bignert, A.; Olsson, M.; Persson, W.; Jensen, S.; Zakrisson, S.; Litzén, K.; Eriksson, U.; Häggberg,
L.; Alsberg, T. Temporal trends of organochlorines in Northern Europe, 1967-1995. Relation to
global fractionation, leakage from sediments and international measures. Environ. Pollut.
1998, 99 (2), 177–198.

(9) Bogdal, C.; Schmid, P.; Kohler, M.; Müller, C. E.; Iozza, S.; Bucheli, T. D.; Scheringer, M.;
Hungerbühler, K. Sediment record and atmospheric deposition of brominated flame
retardants and organochlorine compounds in Lake Thun, Switzerland: lessons from the past
and evaluation of the present. Environ. Sci. Technol. 2008, 42 (18), 6817–6822.

(10) Bossi, R.; Larsen, B.; Premazzi, G. Polychlorinated biphenyl congeners and other chlorinated
hydrocarbons in bottom sediment cores of Lake Garda (Italy). Sci. Total Environ. 1992, 121,
77–93.

(11) Villeneuve, J.-P.; Carvalho, F. P.; Fowler, S. W.; Cattini, C. Levels and trends of PCBs,
chlorinated pesticides and petroleum hydrocarbons in mussels from the NW Mediterranean
coast: comparison of concentrations in 1973/1974 and 1988/1989. Sci. Total Environ. 1999,
237-238, 57–65.

(12) Holoubek, I.; Adamec, V.; Bartoš, M.; Bláha, K.; Bláha, L.; Budňáková, M.; Černa, M.; Čupr, P.;
Demnerová, K.; Drápal, J.; et al. The National Implementation Plan for Implementation of the
Stockholm Convention in the Czech Republic; Brno, 2006.

(13) Prokeš, R.; Vrana, B.; Komprdová, K.; Klánová, J. Annual dynamics of persistent organic
pollutants in various aquatic matrices: a case study in the Morava River in Zlín district, Czech
Republic. J. Soils Sediments 2014, 14 (10), 1738–1752.

(14) Prokeš, R.; Vrana, B.; Klánová, J. Levels and distribution of dissolved hydrophobic organic
contaminants in the Morava river in Zlín district, Czech Republic as derived from their
accumulation in silicone rubber passive samplers. Environ. Pollut. 2012, 166, 157–166.

(15) Borůvková, J.; Gregor, J.; Šebková, K.; Bednářová, Z.; Kalina, J.; Hůlek, R.; Dušek, L.; Holoubek,
 I.; Klánová, J. GENASIS – Global Environmental Assessment and Information System
http://www.genasis.cz/ (accessed Jun 17, 2015).

(16) MacLeod, M.; von Waldow, H.; Tay, P.; Armitage, J. M.; Wöhrnschimmel, H.; Riley, W. J.;
McKone, T. E.; Hungerbuhler, K. BETR global--a geographically-explicit global-scale multimedia
contaminant fate model. Environ. Pollut. 2011, 159 (5), 1442–1445.

(17) Wania, F.; Breivik, K.; Persson, N. J.; McLachlan, M. S. CoZMo-POP 2 – A fugacity-based
dynamic multi-compartmental mass balance model of the fate of persistent organic
pollutants. Environ. Model. Softw. 2006, 21 (6), 868–884.

(18) Sen, P. K. Estimates of the Regression Coefficient Based on Kendall’s Tau. J. Am. Stat. Assoc.
1968, 63 (324), 1379–1389.

(19) Theil, H. A Rank-Invariant Method of Linear and Polynomial Regression Analysis. Proc. R.
Netherlands Acad. Sci. 1950.

(20) Whitehead, P.; Wilson, E. J.; Butterfield, D. A semi-distributed Integrated Nitrogen model for
multiple source assessment in Catchments (INCA): Part I — model structure and process
equations. Sci. Total Environ. 1998, 210-211, 547–558.

(21) Whitehead, P.; Wilson, E. J.; Butterfield, D.; Seed, K. A semi-distributed integrated flow and
nitrogen model for multiple source assessment in catchments (INCA): Part II — application to
large river basins in south Wales and eastern England. Sci. Total Environ. 1998, 210-211, 559–
583.

(22) Nizzetto, L.; Butterfield, D.; Futter, M.; Lin, Y.; Allan, I.; Larssen, T. Assessment of contaminant
fate in catchments using a novel integrated hydrobiogeochemical-multimedia fate model.
Accept. to Sci. Total Environ. 2015.

(23) Futter, M. N.; Butterfield, D.; Cosby, B. J.; Dillon, P. J.; Wade, A. J.; Whitehead, P. Modeling the
mechanisms that control in-stream dissolved organic carbon dynamics in upland and forested
catchments. Water Resources Research. American Geophysical Union February 3, 2007.

(24) Lazar, A. N.; Butterfield, D.; Futter, M. N.; Rankinen, K.; Thouvenot-Korppoo, M.; Jarritt, N.;
Lawrence, D. S. L.; Wade, A. J.; Whitehead, P. G. An assessment of the fine sediment dynamics
in an upland river system: INCA-Sed modifications and implications for fisheries. Sci. Total
Environ. 2010, 408 (12), 2555–2566.

(25) Futter, M. N.; Erlandsson, M. A.; Butterfield, D.; Whitehead, P. G.; Oni, S. K.; Wade, A. J.
PERSiST: a flexible rainfall-runoff modelling toolkit for use with the INCA family of models.
Hydrol. Earth Syst. Sci. 2014, 18 (2), 855–873.

(26) Research Institute for Water Management. DIBAVOD - Digital Database of Water
Management Data http://www.dibavod.cz/ (accessed Sep 28, 2014).

(27) International Steering Committee for Global Mapping. Global Map Data
https://www.iscgm.org/gmd/ (accessed Jan 24, 2015).

(28) Czech Hydrometeorological Institute. ARROW - Assessment and Reference Report of Water
Monitoring http://hydro.chmi.cz/isarrow/index.php (accessed May 30, 2015).

(29) Kubošová, K.; Komprda, J.; Jarkovský, J.; Sáňka, M.; Hájek, O.; Dušek, L.; Holoubek, I.; Klánová,
J. Spatially Resolved Distribution Models of POP Concentrations in Soil: A Stochastic Approach
Using Regression Trees. Environ. Sci. Technol. 2009, 43 (24), 9230–9236.
(30) Gioia, R.; Steinnes, E.; Thomas, G. O.; Mejier, S. N.; Jones, K. C. Persistent organic pollutants in
European background air: derivation of temporal and latitudinal trends. J. Environ. Monit.
2006, 8 (7), 700–710.

(31) Horstmann, M.; Mclachlan, M. S. Atmospheric deposition of semivolatile organic compounds

to two forest canopies. Atmos. Environ. 1998, 32 (10), 1799–1809.

(32) Harner, T.; Bidleman, T. F. Octanol−Air Partition Coefficient for Describing Particle/Gas
Partitioning of Aromatic Compounds in Urban Air. Environ. Sci. Technol. 1998, 32 (10), 1494–
1502.

(33) Götz, C. W.; Scheringer, M.; MacLeod, M.; Roth, C. M.; Hungerbühler, K. Alternative
Approaches for Modeling Gas−Particle Partitioning of Semivolatile Organic Chemicals: Model
Development and Comparison. Environ. Sci. Technol. 2007, 41 (4), 1272–1278.

(34) Demircioglu, E.; Sofuoglu, A.; Odabasi, M. Particle-phase dry deposition and air-soil gas
exchange of polycyclic aromatic hydrocarbons (PAHs) in Izmir, Turkey. J. Hazard. Mater. 2011,
186 (1), 328–335.

(35) González-Gaya, B.; Zúñiga-Rival, J.; Ojeda, M.-J.; Jiménez, B.; Dachs, J. Field measurements of
the atmospheric dry deposition fluxes and velocities of polycyclic aromatic hydrocarbons to
the global oceans. Environ. Sci. Technol. 2014, 48 (10), 5583–5592.

(36) Liu, F.; Xu, Y.; Liu, J.; Liu, D.; Li, J.; Zhang, G.; Li, X.; Zou, S.; Lai, S. Atmospheric deposition of
polycyclic aromatic hydrocarbons (PAHs) to a coastal site of Hong Kong, South China. Atmos.
Environ. 2013, 69, 265–272.

(37) López-García, P.; Gelado-Caballero, M. D.; Santana-Castellano, D.; Suárez de Tangil, M.;
Collado-Sánchez, C.; Hernández-Brito, J. J. A three-year time-series of dust deposition flux
measurements in Gran Canaria, Spain: A comparison of wet and dry surface deposition
samplers. Atmos. Environ. 2013, 79, 689–694.

(38) Zhang, L.; Cheng, I.; Wu, Z.; Harner, T.; Schuster, J.; Charland, J.-P.; Muir, D.; Parnis, J. M. Dry
deposition of polycyclic aromatic compounds to various land covers in the Athabasca oil sands
region. J. Adv. Model. Earth Syst. 2015, 7 (3), 1339–1350.

(39) Schuster, J. K.; Gioia, R.; Moeckel, C.; Agarwal, T.; Bucheli, T. D.; Breivik, K.; Steinnes, E.; Jones,
K. C. Has the burden and distribution of PCBs and PBDEs changed in European background
soils between 1998 and 2008? Implications for sources and processes. Environ. Sci. Technol.
2011, 45 (17), 7291–7297.

(40) Meijer, S. N.; Halsall, C. J.; Harner, T.; Peters, A. J.; Ockenden, W. A.; Johnston, A. E.; Jones, K.
C. Organochlorine Pesticide Residues in Archived UK Soil. Environ. Sci. Technol. 2001, 35 (10),
1989–1995.

(41) McCuen, R. H.; Knight, Z.; Cutter, A. G. Evaluation of the Nash–Sutcliffe Efficiency Index. J.
Hydrol. Eng. 2006, 11 (6), 597–602.

(42) Futter, M. N.; Löfgren, S.; Köhler, S. J.; Lundin, L.; Moldan, F.; Bringmark, L. Simulating
dissolved organic carbon dynamics at the Swedish integrated monitoring sites with the
integrated catchments model for carbon, INCA-C. Ambio 2011, 40 (8), 906–919.

(43) Allan, I. J.; Booij, K.; Paschke, A.; Vrana, B.; Mills, G. A.; Greenwood, R. Field Performance of
Seven Passive Sampling Devices for Monitoring of Hydrophobic Substances. Environ. Sci.
Technol. 2009, 43 (14), 5383–5390.
(44) Hornsby, A. G.; Wauchope, R. D.; Herner, A. Pesticide Properties in the Environment; Springer,
USA: New York, USA, 1996.

(45) Crowe, A. S.; Smith, J. E. Distribution and persistence of DDT in soil at a sand dune-marsh
environment: Point Pelee, Ontario, Canada. Can. J. Soil Sci. 2011.
ANNEX 4

Manuscript of Zheng, Q., Nizzetto, L., Mulder, M., Sáňka, O., Lammel, G., Li, J.,
Bing, H., Liu, X., Jiang, Y Luo, Ch. & Zhang, G. Does an analysis of
polychlorinated biphenyl (PCB) distribution in mountain soils across China
reveal a latitudinal fractionation paradox? Environ. Pollut. 195, 115–122
(2014).
 Environmental Pollution 195 (2014) 115e122

Contents lists available at ScienceDirect

Environmental Pollution
journal homepage: www.elsevier.com/locate/envpol

Does an analysis of polychlorinated biphenyl (PCB) distribution in

mountain soils across China reveal a latitudinal fractionation paradox?
Qian Zheng a, f, Luca Nizzetto b, c, *, Marie D. Mulder b, Ondrej Sa
n ka b,
Gerhard Lammel b, d, Jun Li a, Haijian Bing e, Xin Liu a, Yishan Jiang a, Chunlin Luo a,
Gan Zhang a, *
a
State Key Laboratory of Organic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou 510640, China
b
Masaryk University, Research Centre for Toxic Compounds in the Environment, Brno, Czech Republic
c
Norwegian Institute for Water Research, Oslo, Norway
d
Max Planck Institute for Chemistry, Mainz, Germany
e
The Key Laboratory of Mountain Surface Processes and Ecological Regulation, Institute of Mountain Hazards and Environment,
Chinese Academy of Sciences, Chengdu 610041, China
f
Graduate University of the Chinese Academy of Sciences, Beijing 100039, China

a r t i c l e i n f o a b s t r a c t

Article history: Organic and mineral soil horizons from forests in 30 mountains across China were analysed for poly-
Received 9 July 2014 chlorinated biphenyl (PCB). Soil total organic carbon (TOC) content was a key determinant of PCB dis-
Received in revised form tribution explaining over 90% of the differences between organic and mineral soils, and between 30% and
17 August 2014
60% of the variance along altitudinal and regional transects. The residual variance (after normalization by
Accepted 20 August 2014
Available online
TOC) was small. Tri- to tetra-CB levels were higher in the South in relation to high source density and
precipitation. Heavier congeners were instead more abundant at mid/high-latitudes where the advection
pattern was mainly from long range transport. This resulted in a latitudinal fractionation opposite to
Keywords:
POP
theoretical expectations. The study showed that exposure to sources with different characteristics, and
PCB possibly accumulation/degradation trends of different congeners in soils being out-of-phase at different
Fractionation latitudes, can lead to an unsteady large scale distribution scenario conflicting with the thermodynamic
Cold trapping equilibrium perception.
Mountain © 2014 Elsevier Ltd. All rights reserved.
Soil
Latitude
Forest

1. Introduction waste management (Breivik et al., 2002; Li et al., 2009). Secondary

Persistent organic pollutants (POPs) have been the focus of
research and regulation due to their persistence, ubiquitous dis-
tribution and bioaccumulative and toxic properties (Aichner et al.,
2013; Jones and de Voogt, 1999; Klanova et al., 2011; Zhang et al.,
2008b). Polychlorinated biphenyls (PCBs) represent a class of
POPs with a broad range of physicalechemical properties (Bozlaker
et al., 2008), ideal for investigating the mechanisms involved in the
control of fate and global distribution of POPs in general (Aichner
et al., 2013).
Although PCBs are banned in most parts of the world, primary
emissions continue mostly from old equipment and inadequate
* Corresponding authors.
E-mail addresses: Luca.nizzetto@niva.no, nizzetto@recetox.muni.cz (L. Nizzetto),
Ghanzhang@gig.ac.cn (G. Zhang).
emissions from environmental reservoirs such as soils, vegetation
and oceans hosting PCBs burdens emitted during the past, also
significantly contribute in feeding long-range atmospheric trans-
port (e.g. (Lammel and Stemmler, 2012)).
Predicting the future of POP exposure requires understanding
the dynamic balance between atmospheric sources, atmospheric
depositions, environmental degradation and remobilization from
environmental reservoirs. Analyzing the current regional and
global distribution of the environmental burden of POPs in the
major reservoir compartments (such as soils) represent an effective
approach to gather information useful for predicting possible
future exposure conditions (Meijer et al., 2003).
China has 134 million hectares of forested land including a range
of different biomes including subtropical forests, boreal forests, and
semi-arid environments (Fang et al., 2001). Large geographic scale,
climate variability, complex topography, ecosystem diversity, and

http://dx.doi.org/10.1016/j.envpol.2014.08.021
0269-7491/© 2014 Elsevier Ltd. All rights reserved.
116 Q. Zheng et al. / Environmental Pollution 195 (2014) 115e122
densely populated conglomerates in well-defined areas make
China a very interesting case study (Jiang et al., 1999).
In order to provide an assessment of anthropological and
environmental controls on POP distribution, the patterns of PCBs in
forest soils of several mountain sites in China were investigated in
this study, considering gradients such as soil depth, soil type, alti-
tude, latitude and distance from urban/industrial clusters.
2. Materials and methods
2.1. Monitoring design
Thirty mountain sites across China were chosen (shown in Fig. 1). In each
mountain site, up to four sampling locations were selected along the same aspect on
the altitudinal transect. In total 159 forest-soil samples were collected in 82 loca-
tions from the 16th of May 2012 to the 15th of March 2013. Major gradients
included: latitude ranging between 21
and 53
; altitude ranging between 200 m
and 3800 m; yearly mean temperature (ymt) and averaged total yearly precipitation
(typ) ranging 6
Ce21
C and 245 mme2129 mm, respectively; distance from the
major urbanized areas (assumed here as proxy of major primary sources locations)
ranging between a few tenths to several hundred km.
The monitoring plan was defined taking into consideration the need of sampling
in all major accessible mountain areas of China, achieving homogeneous
geographical distribution, and collecting multiple samples along slopes at locations
differing at least 200 m in elevation.
2.2. Sampling of soil
Three small trenches located at about 5 m distance from each other were
excavated at the depth of 30 cm in each individual altitude location. Vegetation litter
Fig. 1. Altitudinal and regional distribution of PCBs in organic (red bars) and mineral (blue bars) soil horizons, and the index of potential source influence (IPSI) (defined in Section
2.6). Different climate zones are represented by points with different colors: green: Climate zone 1: humid continental climate; orange: Climate zone 2: humid subtropical climate;
purple: Climate zone 3: semiarid continental climate. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
was carefully removed and a preliminary classification of the soil layers was per-
formed in-situ based on colour and structure of the material present in each horizon.
Samples from the O- (organic) and A- (mineral) horizons were collected separately
from each trench using a metal spoon, folded in aluminium foil, placed in poly-
ethylene zip-bags, cooled and transported to the laboratory where their mass, water
content, bulk density and total organic carbon (TOC) content were determined. The
samples were then freeze dried and stored at 20
C for maximum 3 months before
chemical analysis began.
2.3. Chemical analysis
Dry samples were sieved in order to remove stones and aggregates larger than
2 mm. The samples from each trench were homogeneously mixed to create aggre-
gated samples of the O-horizon and A-horizon reflecting average conditions of the
sampling location at the selected altitude. Sample extraction and preparation,
instrumental analysis and adopted quality assurance and control methods were
consistent with those described in a previous study (Huang et al., 2013). Details on
analytical methods are also reported in the Supplementary information (SI) avail-
able on line.
2.4. Geographic information
A set of spatially explicit datasets were used in this study to feed the exploratory
analysis performed in order to assess physical and anthropogenic influences on
contaminant distribution. Geo-referenced datasets of elevation, ymt, typ (0.50
resolved) and human population counts (2.50 resolved) were obtained from
WorldClim (Jurado et al., 2004; WORLDCLIM) and CIESIN (CIESIN; Gioia et al., 2006).
Geographic information was elaborated using ArcGIS Desktop software. Multiannual
averaged ymt of individual sites was calculated using monthly averages of time
series (1950e2000). The information on human population distribution was used as
 Q. Zheng et al. / Environmental Pollution 195 (2014) 115e122
proxy for the distribution of PCB primary atmospheric sources in China, as further
discussed in Section 2.6 of this paper.
2.5. Atmospheric back trajectories analysis
Atmospheric back trajectories were analysed using Hysplit (Draxler and Hess,
1998) to evaluate, on an annual base, the main origin of air masses reaching indi-
vidual sampling sites. For each sampling site, a total of 360 ten-days backward
trajectories were computed covering the full year preceding the sampling. Back
trajectories were released from an altitude of 100 m above ground level every 2 days
at 6 and 18 UTC (2 and 14 Beijing time (CST)) in order to take into account the day/
night variation of the depth of the (mixed) planetary boundary layer (BL). The in-
formation on air mass trajectory was elaborated to obtain maps of back trajectories
density. The back trajectory density fx,i value for a given point in space i (identified by
the area delimited by grid cells with 2.5' resolution) was calculated in relation to
each sampling location x as:
BTnx;i
fx;i ¼
360
where BTnx,i is the number of times a back trajectory originated from site x was
passing over location i during the year preceding the sampling, and 360 is the total
number of computed back trajectories. The rationale of such an elaboration was to
obtain aggregated information showing the areas surrounding the sampling loca-
tions with the highest potential to serve as source of airborne PCBs for the selected
sampling areas. For example, emissions of contaminants from locations with low fx,i
value are likely to have little influence on the contamination profile of site x.
2.6. Index of potential source influence
In order to further parameterize the potential influence of primary urban
sources on the regional distribution of PCBs in China, a numerical index of potential
source influence (IPSI) was defined combining the information on back trajectories
density and human population distribution. More specifically IPSI was calculated as:
X  Unfortunately typ and IPSI geographical distribution co-varied
IPSIx ¼ 108 fx;i $pi
i
where pi is the human population residing within a generic point in space i, and the
numerical parameter is simply a scaling factor. IPSI for a given sampling site x is
proportional to the product of back trajectory density and human population counts
calculated for a circular geographical area (6000 km diameter) centred in x.
In the definition of IPSI it is implicitly assumed that the geographical distribution
of human population reflects the distribution of PCBs primary sources. PCBs were
used in China mainly in electric equipment and painting (SEPA, 2003). Similarly to
other parts of the world (Motelay-Massei et al., 2005; Yeo et al., 2004; Zhang et al.,
2008a), previous assessment of PCBs distribution in soils and other environmental
compartments in China evidenced the existence of an urban-rural gradient with
highly populated urban and industrial districts representing the areas with highest
density of potential primary atmospheric sources (Cui et al., 2013; Ren et al., 2007;
Xing et al., 2005; Zheng et al., 2010). In China industrial and major urban clusters
closely overlap (Xing et al., 2005). For this reason population count is an adequate
proxy for describing potential PCB usage and primary source location while the
associated back-trajectory density value weights different locations influence based
on how often they were upwind the sampling point. The IPSI values defined here
represent therefore a strategy to rank the influence of potential primary source areas
on the contamination pattern in the receiving mountain soils. In addition, IPSI can be
considered as an index of potential influence from sources in the subregional/local
domain rather than from long-range transport. This derives from the fact that fx,i
values, which obviously tend to be higher in proximity of the back trajectory origin,
were not weighted for the effect of distance. The potential influence of long-range
transport was instead assessed by computing the fraction of time air masses were
traveling above the BL using the Hysplit outputs for BL height and back trajectory
elevations.
2.7. Statistical analysis
Multivariate statistical analysis was carried out using principal component
analysis (PCA). Principal components were calculated based on correlation matrixes
using normalized, standardized and centred data. Raw data of contamination and
typ data were normalized by natural logarithm transformation. All variables were
standardized by dividing by one standard deviation and centred by subtracting the
mean. Non-parametric rank correlation (Spearman) analysis was performed on raw
or transformed data, to elucidate significance of observed co-linearity.
3. Results and discussion
3.1. Air mass patterns and index of potential source influence (IPSI)
The dominant large-scale air mass advection pathway into the
BL of China at locations above 30
N is from northwest. In winter, air
117
from Europe and Central Asia moves eastwards and turns south
under the influence of a large semi-permanent high pressure sys-
tem. Advection from west is prominent for upper tropospheric
levels eventually intersecting high altitude sites in western China.
In summer, there is a semi-permanent low pressure system above
Central Asia, with air masses flowing mainly from the North-
Northwest and less frequently from the Northeast toward the low
pressure system.
IPSI values for different sampling sites ranged between 1 and 37
(Fig. 1) following a negatively skewed distribution with most
frequently observed values around 25. The highest IPSI values are
observed in the South-East reflecting the distribution of highly
populated areas. The trajectories arriving at sampling locations in
this part of China (in particular sites 21e23) are estimated to have
spent a considerable fraction of the last 5 days before arrival to the
sampling site within the BL. These were the locations which likely
received the highest contribution from local sources. In contrast,
sites located in the South-West (Yunnan province), despite
frequently receiving air masses from highly populated areas in
Northern India, scored with intermediate IPSI values, reflecting
relatively low population density in the local-regional domain.
The lowest IPSI values were observed in the Northern part of
China and along the Greater Khingan range in Inner Mongolia
(NorthEast), due to relatively low population in this region of China.
These sites too, received often air masses from large scale advection
carrying contamination signals from long range transport.
(p < 0.05). This was essentially due to the fact that human popu-
lation in China is concentrated in the Southern and Western
provinces experiencing higher precipitations, while sampling sites
in populated region of northern and eastern China are located
mostly upwind of conurbations. Because of this covariance, it was
not possible to separate IPSI and typ during exploratory and
regression statistical analysis to assess their individual influence on
PCB distribution.
3.2. Summary of PCB concentration results
Measured total PCB concentrations are displayed in Fig. 1 indi-
vidually for O-horizon and A-horizon. The full dataset including
individual congener results is reported in Table S3.
P
The total concentrations of PCBs ( 29PCBs) averaged over all O-
and A-horizon samples were 510 ng kg1 (57e1320 ng kg1) and
227 ng kg1 (36e679 ng kg1), respectively. These concentration
ranges are consistent with those reported by a previous study on
background soil contamination in Asia (Li et al., 2009). PCB con-
taminations in Chinese background soil appeared to be consider-
ably lower than in North America (mean 4300 ng kg1) South
America (mean 1400e4300 ng kg1) (Li et al., 2009), and Europe
(e.g. UK (mean 4500 ng kg1) (Meijer et al., 2002) and forest soils in
Germany (mean 24,700 ng kg)(Aichner et al., 2013)). This reflects
lower historical usage of PCBs in China compared to the western
countries (Breivik et al., 2002). Seven indicator congeners (PCB 28,
PCB 52, PCB 101, PCB 118, PCB 138, PCB 153 and PCB 180) accounted
P
for approximately 49.8% of 29PCBs, while PCB -28, -138 and -153
were the main contributors.
The differences between the highest and lowest concentrations
were within a factor of 20. When considering dry weight based
concentration data, no outliers were found. TOC normalized con-
centration data showed instead a considerably lower variance (i.e. a
factor of 10 between the highest and the lowest measured values).
In this case only a single outlier was found. This was the Dinghu
Mountain in the south of Guangdong province (Site 23) with TOC
normalized concentrations about 2e3 times higher compared to
the maximum of any other site (consistently for all sampled
118 Q. Zheng et al. / Environmental Pollution 195 (2014) 115e122
elevations). Site 23 was among the locations with the highest IPSI
value, suggesting significant impact from local sources, while its
soil had the lowest TOC content observed in this study.
3.3. Distribution across soil layers
Concentrations in the O-horizon were generally higher (average
factor of 2.2) compared to the A-horizon. This difference nearly
completely disappeared when TOC normalized data were
compared. After TOC normalization, the residual difference be-
tween O-horizon and A-horizon concentrations scattered following
a normal distribution with mean approaching 0 (Fig. S3). Such a
residual variance could not be explained by any environmental
factor considered in this study. TOC normalized concentrations
were consistent and significantly correlated (p < 0.01) between the
two soil horizons (Fig. S2) confirming the expected pivotal control
of soil layer specific TOC content on the vertical distribution of POPs
(Moeckel et al., 2008), and showing lack of significant influence of
other potential drivers (e.g. climate, altitude, soil type, distance
from sources) over such a control.
3.4. Distribution along the altitudinal gradient
The analysis of the altitudinal trends was based on selected 20
sites (86% of the entire data set), as, obviously, mountain sites with
only one location sampled along the altitudinal gradient and sites
with two sampling locations less than 500 m apart in altitude were
excluded.
An increasing trend of TOC normalized concentrations of penta-
CBs or higher chlorinated congeners was observed at increasing
altitude in 16 sites out of 20. In order to assess the relevance of this
result from a statistical point of view, the significance of the
observed trend was estimated from the p-value of the correlation
analysis between the scaled values of concentration and altitude of
all the 20 considered sites. Scaling was performed by dividing in-
dividual data points of concentration and altitude by their median
value at each respective mountain transect. The occurrence of a
significant (p < 0.05) positive correlation between TOC normalized
concentrations in soil and altitude was confirmed for all penta-CBs
and higher chlorinated congeners (Fig. 2).
The present result highlights the tendency of enrichment of
heavier PCBs in high altitude soils. This trend was consistent across
sites with different climates and conditions and confirms some
previous evidences of orographic cold trapping collected in Chinese
high mountains (Chen et al., 2008; Wang et al., 2009) and other
regions of the world.
Many physical, ecological and anthropological factors can
contribute to determine the observed altitudinal distribution
pattern. Although the present dataset does not allow a detailed
analysis of the influence of individual drivers due to co-linearity
(e.g. temperature, precipitation, vegetation type, and potential
exposure to local sources co-varied, as expected, with altitude),
temperature and wet deposition are suggested as pivotal variables.
These were in fact the only variables significantly (p < 0.05)
correlated with PCBs enrichment trends along the slopes. Such a
result is consistent with theoretical expectations of altitudinal cold
trapping. Orographic cold trapping is expected to mainly derive
from different magnitude of wet depositions along altitudes (Wania
and Westgate, 2008). A number of other factors can also contribute
to the observed pattern, including higher capacity of fugacity of
soils and slower bio-degradation processes at lower temperature/
high altitude soils. Differential wet deposition along altitude,
however, specifically results in a more effective entrapment of
compounds with values of the scavenging ratio (at 25
C) between
103.5 and 105.5 represented here by penta- and heavier-CB
Fig. 2. Correlation between scaled TOC normalized PCB concentrations of tetra- and
higher congeners and scaled altitude. r values are Spearman's rank correlation
coefficients.
congeners (Wania and Westgate, 2008). Coincidently these were
also the congeners for which significant altitudinal trends were
observed in the present dataset.
Unlike previous observations from other few mountain sites in
China (Chen et al., 2008; Wang et al., 2009), the influence from local
PCB sources at lower altitudes was visible here only in few cases
(sites 9, 14, 25 and 26). Here in fact, TOC normalized concentrations
of heavier PCBs had maxima at lower altitudes. Coincidently these
sites had IPSI values in the medium-to-high range (13.5, 28.1, 25.6
and 27.4, respectively); they were exposed to air masses with
relatively high residence time in the BL; and were located in
proximity (e.g. less than 100 km apart) of major conurbations.
3.5. Soil TOC influence on regional distribution
The exploratory analysis of factors controlling regional distri-
bution of PCBs was performed here considering the following set of
uncorrelated (when considering the entire dataset) environmental
variables: ymt, altitude, TOC and typ. It is recalled that typ has to be
regarded here also as a proxy of IPSI owing to the significant pos-
itive correlation (p < 0.05) between these two variables (Table S4).
TOC content explained large part of the variance of PCB concen-
trations (expressed on soil dry weight) for both the O-horizon and
A-horizon as illustrated by the biplot in Fig. S4. Covariance is
showed here by the similar orientation of the vectors representing
P
29PCBs and TOC.
Several previous studies had provided parameterizations for the
relationship between PCB concentrations and TOC using data from
global and regional scale surveys (Meijer et al., 2003; Salihoglu
et al., 2011). TOC in the O-horizon and A-horizon ranged
3.9e44.6% with mean 26.8% and 2.0e26.4% with mean 10.9%,
respectively. The relationship between PCB concentrations and TOC
was analyzed individually here for O-horizon and A-horizon. In
order to perform such an analysis, PCB concentrations ([PCB]) and
TOC were log-transformed. As expected, a significant (p < 0.05)
positive relationship between log TOC and log [PCB] was observed.
Table 1 reports correlation coefficients and slope values of the
regression curve for selected representative congeners. Regression
parameters are generally consistent with previous observations
(Meijer et al., 2003). PCBs high thermodynamic affinity for organic
matter, as described by the very high values of the octanoleair and
octanolewater equilibrium partitioning (Li et al., 2003), and
elevated persistence in soil solids bounded phase, underpin such a
strong relationship.
 Q. Zheng et al. / Environmental Pollution 195 (2014) 115e122 119

Table 1
Spearman correlation coefficient (r) and regression line slopes describing the relationship between PCB concentrations and soil TOC. Data are reported for the analysis
including the full dataset and grouped climate type. Bold numbers highlight correlation coefficients of significant (p < 0.05) relationships.
P
PCBs PCB 8 PCB 28 PCB 52 PCB 101 PCB 118 PCB 153 PCB 138 PCB 180

r Slope r Slope r Slope r Slope r Slope r Slope r Slope r Slope r Slope

O-horizon
Global 0.47 0.48 0.18 0.26 0.35 0.33 0.42 0.54 0.45 0.71 0.47 0.48 0.18 0.26 0.35 0.33 0.42 0.54
Group 1 (Temperate) 0.41 0.81 0.12 0.28 0.37 0.68 0.41 0.95 0.34 1.02 0.41 0.81 0.12 0.28 0.37 0.68 0.41 0.95
Group 2 (Subtropical) 0.66 0.37 0.41 0.26 0.42 0.37 0.62 0.52 0.59 0.45 0.66 0.37 0.41 0.26 0.42 0.37 0.62 0.52
A-horizon
Global 0.50 0.48 0.17 0.18 0.28 0.34 0.28 0.29 0.44 0.60 0.50 0.48 0.17 0.18 0.28 0.34 0.28 0.29
Group 1 (Temperate) 0.52 0.55 0.32 0.32 0.44 0.56 0.21 0.26 0.37 0.59 0.52 0.55 0.32 0.32 0.44 0.56 0.21 0.26
Group 2 (Subtropical) 0.56 0.35 0.06 0.14 0.14 0.04 0.39 0.32 0.63 0.48 0.56 0.35 0.06 0.14 0.14 0.04 0.39 0.32

The analysis of the present dataset, including observations precipitation and IPSI which both had the highest values in the
performed across the boundaries of different climate zones, can southern part of China (e.g. PCB 28: p < 0.01, r ¼ 0.41, and p ¼ 0.06,
contribute to establish a relationship on how the relationship be- r ¼ 0.22, for typ and IPSI respectively). In contrast heavier conge-
tween soil organic matter and POP distribution is modulated in ners did not display such dependence but when only the temperate
different environmental and ecological conditions. To this end, humid climate zone dataset was considered they tended to increase
sampling locations were divided into 3 climate zones individuated their abundance at northern latitudes. These trends for the heavier
based on typ and ymt values. These zones are: Climate type 1 with congeners were poorly significant or non-significant (e.g. PCB 180:
ymt and typ of 4.8
C, and 736 mm, respectively, identified as the p ¼ 0.056, r ¼ 0.27).
humid continental climate zone; type 2 with ymt ¼ 13.7
C and
typ ¼ 1700 mm identified as the subtropical climate zone; and type 3.7. Latitudinal fractionation
3 (cold semi-arid climate zone) with ymt ¼ 4.2
C and
typ ¼ 255 mm which only included a single site (#30). Sampling Latitudinal cold trapping is a process expected to influence
sites grouped based on their climate zone are shown in Fig. 1 and phase distribution equilibrium of semivolatile chemicals between
Table S2. PCB concentrations were correlated to TOC (p < 0.05) for the atmosphere and earth's surface resulting in varying contami-
most of the congeners (with the exception of PCB 8) also when nant patterns along the latitudinal temperature gradient (Wania
climate specific sub-datasets of data were considered. Slope values and MacKay, 1996). Experimental evidences of cold trapping
for the regression between OC and PCB concentrations (in case of driven latitudinal fractionation of POPs were reported by previous
O-horizons) were generally lower by a factor of 2e10 in the sub- studies in a range of geographical scales and locations and in
tropical climate (zone 2) compared to the temperate climate relation to different environmental matrixes, including vegetation
(Table 1). This behavior is consistent with previous observations of (Calamari et al., 1991) and soils (Meijer et al., 2002). In agreement
lower dependence of PCB concentration on soil organic matter with thermodynamics-based predictions these results show that
content at lower latitudes (Meijer et al., 2003). Interestingly, the
same behavior was not observed for the A-horizon dataset. This
possibly reflects the influence of different characteristics of soil
organic matter or even a higher influence of precipitation and
temperature, which had consistently higher values in the sub-
tropical China. Less acidic constituents in vegetation litter and
higher soil organic matter turn-over are important distinguishing
characteristics of subtropical soils (compared to soils in temperate
areas). These properties are both dependent on temperature and
precipitation and might have an influence in determining the lower
slope values observed here in climate zone 2. Recent experimental
evidences seem in fact to suggest a relationship between fast turn-
over of labile organic matter and rapid remobilization of hydro-
phobic contaminants from soils (Liu et al., 2013; Wong and
Bidleman, 2011). The detailed study of the dynamic coupling be-
tween soil organic matter turn-over and remobilization of POPs
from soils is however still in its early stages.

3.6. Residual variance of TOC normalized concentrations

After normalization by TOC content, the residual variance of the

dataset considerably dropped (factor of 2) making difficult to find
significant relationships with possible environmental or anthro-
pological drivers. However some spatial distribution trend could be
observed. For example (Fig. 3), TOC normalized concentrations of
more volatile congeners tended to be higher at lower latitudes,
especially in relation to the more populated areas in the subtropical
humid climate zone. Altogether up to 20e40% of the residual
variance of lighter congeners could potentially be associated to
Fig. 3. Biplots from the Principal Component Analysis depicting relationships between
TOC normalized PCB 28 and PCB 180 (taken as examples) concentration patterns and
environmental descriptors. Light blue vectors in the biplots represent auxiliary variable
which were correlated with one of the variables included in the analysis and for this
reason not directly included in the computation of principal components. Vectors
having similar orientation indicate covariance between the associated variable. (For
interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)
120 Q. Zheng et al. / Environmental Pollution 195 (2014) 115e122

the lighter congeners tend to increase their relative abundance at
high latitudes (low myt) in the condensed phase (e.g. soil) while
heavier congeners are more abundant in the south (high myt)
(Gouin and Wania, 2007; Wang et al., 2012). Latitudinal fraction-
ation of PCBs was investigated here by calculating the ratio be-
tween the concentrations of selected congeners representative of
P
different homologue groups and 29PCBs. Results are plotted in
Fig. 4. When the full dataset is included in the analysis a progressive
decline of relative abundance of tri- and tetra-congeners could be
observed at increasing latitude (p < 0.05), while penta- and hexa-
congeners were more abundant in the north (both in relative and
absolute terms (p < 0.05)). Fractionation trends were not significant
for PCB 138 and PCB 180 (Fig. 4). In order to reduce the possible
confounding factor associated to source proximity to the sites
located in the southern part of China, the analysis was repeated
considering only high altitude sites (e.g. >1500 m). These sites were
more prone to receive air masses with a shorter residence time in
BL and therefore less influenced by local sources. Fractionation
trends of tri- to hexa-CBs for this subset of data were consistent
with those obtained considering the entire dataset. In this case,
however dependence on latitude was higher with correlation co-
efficients higher of a factor of 2) (Table S5). Increasing trends of
heavier congeners (PCB 138 and 180) relative abundance were
observed too, however they were still not significant from a sta-
tistical point of view (Table S5).
The PCB fractionation displayed in Fig. 4 is opposite to that ex-
pected from the cold trapping theory (Wania and MacKay, 1996)
and previously observed from other global scale surveys (Meijer
et al., 2002). As highlighted in Section 3.6 considerably higher
precipitations and proximity to PCBs sources have likely deter-
mined the higher levels of di- to tetra-chlorinated PCBs in Southern
China forested soils. The latitudinal pattern displayed by lighter
congeners had a strong influence in determine the general frac-
tionation results. The relative importance of heavier congeners (or,
in other terms, the relatively lower abundance of lighter PCBs) in
Northern China could not be ascribed to the influence of different

Fig. 4. “Inverse” latitudinal fractionation of PCB congeners in Chinese soils. Blue data points and black trend line and regression parameters refer to the full dataset. Data points
highlighted in red and red trend lines and regression parameters refer to the high altitude (i.e. locations above 1500 m) subset of data (see Section 3.6). (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)
 Q. Zheng et al. / Environmental Pollution 195 (2014) 115e122
types of local sources since all the sites in this region scored with
low values of IPSI (Fig. 1). Sites in the Northern China (especially
those located at the higher altitudes) were instead more prone to
receive contamination signals associated to long range atmospheric
transport from North-West (as described in Section 3.1). It is likely
these signals carried a different PCB profile compared to that
expressed by local sources in Southern China. Higher abundance of
low chlorinated congeners in commercial mixtures are in fact a
peculiarity of PCB usage in China (Ren et al., 2007; Xing et al., 2005).
Some considerations on soil exposure dynamics may also
contribute to explain the apparent inverse fractionation pattern
displayed by the present data. Spatially resolved multicompart-
ment modeling (Lammel and Stemmler, 2012) highlighted that
temporal trends of different PCB congeners in soils are out of phase,
with declining burdens of the less chlorinated, peaking burdens of
mid chlorinated, and still increasing burdens of high chlorinated
congeners in soils of east Asia in 2001e2010. Hereby, peaking of
heavier congeners in north China soils is expected to lag 5e15 years
behind south China soils, while lighter congeners might start
declining earlier by means of their considerably shorter degrada-
tion half-lives in soil. The combined effect of different degradation
rates, accumulation from the atmosphere, and exposure to sources
could have therefore contributed, in this decade and region, to the
definition of an unsteady scenario which resulted in fractionation
trends inverse to the thermodynamic global-scale perception.
3.8. Conclusions
This regional-scale analysis of POP distribution in eastern Asia
highlighted that soil TOC plays a controlling role in the vertical
(across soil layers), altitudinal and regional distribution of PCBs.
Between 37% and 60% of the total variance of concentrations data
was explained by the variance of TOC. TOC also explained nearly
90% of the differences of PCBs levels between O- and A-horizons.
The residual limited variance could not be clearly related to any
of the considered environmental and anthropological drivers. PCB
contaminations of mountain soil of China reflect exposure to
background atmospheric levels which are quite uniform
throughout China. Under these conditions a strong direct influence
of primary sources could not be clearly detected in most of the
selected sites. Conversely, the influence of environmental factors
(especially temperature and rainfall) emerged from the observed
trends of orographic enrichment in most of the sampled sites.
Enrichment of lighter congeners at lower latitudes determined a
general latitudinal fractionation pattern opposite to that expected
from merely thermodynamic considerations. Such a behavior was
interpreted as the combined results of different factors, including in
particular: exposure to sources with different characteristics in
different regions (e.g. local sources vs remote sources) and accu-
mulation/degradation trends of PCBs in soils being out of phase at
different latitudes in this region. In summary, the study experi-
mentally showed that different driving factors acting in different
regions can produce PCB latitudinal distribution patterns opposite
to expectations from the fractionation theory, even at relatively
large geographical scales.
Acknowledgements
This study was supported by the National Natural Science
Foundation of China (Nos. 41173082 and 21307133), and the Joint
Funds of the National Natural Science Foundation of China and the
Natural Science Foundation of Guangdong Province, China (Nos.
U1133004). Part of the work was carried out with the support of
core facilities of the Research Centre for Toxic Compounds in the
Environment (RECETOX) e National Infrastructure for Research of
121
Toxic Compounds in the Environment, project number LM2011028,
funded by the Ministry of Education, Youth and Sports of the Czech
Republic under the activity “Projects of major infrastructures for
research, development and innovations”.
The authors wish to thank Dusan Lago for running Hysplit and
post-processing of the output and Klara Komprdova  for providing
advice on statistical analysis methodology.
Appendix A. Supporting information
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.envpol.2014.08.021.
References
Aichner, B., Bussian, B., Lehnik-Habrink, P., Hein, S., 2013. Levels and spatial dis-
tribution of persistent organic pollutants in the environment: a case study of
German forest soils. Environ. Sci. Technol. 47, 12703e12714.
Bozlaker, A., Odabasi, M., Muezzinoglu, A., 2008. Dry deposition and soil-air gas
exchange of polychlorinated biphenyls (PCBs) in an industrial area. Environ.
Pollut. 156, 784e793.
Breivik, K., Sweetman, A., Pacyna, J.M., Jones, K.C., 2002. Towards a global historical
emission inventory for selected PCB congeners d a mass balance approach: 1.
Global production and consumption. Sci. Total Environ. 290, 181e198.
Calamari, D., Bacci, E., Focardi, S., Gaggi, C., Morosini, M., Vighi, M., 1991. Role of
plant biomass in the global environmental partitioning of chlorinated hydro-
carbons. Environ. Sci. Technol. 25, 1489e1495.
Chen, D., Liu, W., Liu, X., Westgate, J.N., Wania, F., 2008. Cold-trapping of persistent
organic pollutants in the mountain soils of Western Sichuan, China. Environ.
Sci. Technol. 42, 9086e9091.
CIESIN, http://sedac.ciesin.columbia.edu.
Cui, S., Qi, H., Liu, L.Y., Song, W.W., Ma, W.L., Jia, H.L., Ding, Y.S., Li, Y.F., 2013.
Emission of unintentionally produced polychlorinated biphenyls (UP-PCBs) in
China: has this become the major source of PCBs in Chinese air? Atmos. Envi-
ron. 67, 73e79.
Draxler, R.R., Hess, G.D., 1998. An overview of the HYSPLIT_4 modelling system for
trajectories, dispersion, and deposition. Aust. Meteorol. Mag. 47, 295e308.
Fang, J.Y., Chen, A.P., Peng, C.H., Zhao, S.Q., Ci, L., 2001. Changes in forest biomass
carbon storage in China between 1949 and 1998. Science 292, 2320e2322.
Gioia, R., Steinnes, E., Thomas, G.O., Mejier, S.N., Jones, K.C., 2006. Persistent organic
pollutants in European background air: derivation of temporal and latitudinal
trends. J. Environ. Monit. 8, 700e710.
Gouin, T., Wania, F., 2007. Time trends of arctic contamination in relation to
emission history and chemical persistence and partitioning properties. Environ.
Sci. Technol. 41, 5986e5992.
Huang, Y., Xu, Y., Li, J., Xu, W., Zhang, G., Cheng, Z., Liu, J., Wang, Y., Tian, C., 2013.
Organochlorine pesticides in the atmosphere and surface water from the
equatorial Indian Ocean: enantiomeric signatures, sources and fate. Environ.
Sci. Technol. 47, 13395e13403.
Jiang, H., Apps, M.J., Zhang, Y.L., Peng, C.H., Woodard, P.M., 1999. Modelling the
spatial pattern of net primary productivity in Chinese forests. Ecol. Model. 122,
275e288.
Jones, K.C., de Voogt, P., 1999. Persistent organic pollutants (POPs): state of the
science. Environ. Pollut. 100, 209e221.
Jurado, E., Lohmann, R., Meijer, S., Jones, K.C., Dachs, J., 2004. Latitudinal and sea-
sonal capacity of the surface oceans as a reservoir of polychlorinated biphenyls.
Environ. Pollut. 128, 149e162.
Klanova, J., Diamond, M., Jones, K., Lammel, G., Lohmann, R., Pirrone, N.,
Scheringer, M., Balducci, C., Bidleman, T., Blaha, K., Blaha, L., Booij, K.,
Bouwman, H., Breivik, K., Eckhardt, S., Fiedler, H., Garrigues, P., Harner, T.,
Holoubek, I., Hung, H., MacLeod, M., Magulova, K., Mosca, S., Pistocchi, A.,
Simonich, S., Smedes, F., Stephanou, E., Sweetman, A., Sebkova, K., Venier, M.,
Vighi, M., Vrana, B., Wania, F., Weber, R., Weiss, P., 2011. Identifying the research
and infrastructure needs for the global assessment of hazardous chemicals ten
years after establishing the Stockholm convention. Environ. Sci. Technol. 45,
7617e7619.
Lammel, G., Stemmler, I., 2012. Fractionation and current time trends of PCB con-
geners: evolvement of distributions 1950e2010 studied using a global
atmosphere-ocean general circulation model. Atmos. Chem. Phys. 12,
7199e7213.
Li, N.Q., Wania, F., Lei, Y.D., Daly, G.L., 2003. A comprehensive and critical compi-
lation, evaluation, and selection of physical-chemical property data for selected
polychlorinated biphenyls. J. Phys. Chem. Ref. Data 32, 1545e1590.
Li, Y.-F., Harner, T., Liu, L., Zhang, Z., Ren, N.-Q., Jia, H., Ma, J., Sverko, E., 2009. Pol-
ychlorinated biphenyls in global air and surface soil: distributions, airesoil
exchange, and fractionation effect. Environ. Sci. Technol. 44, 2784e2790.
Liu, X., Ming, L.-L., Nizzetto, L., Borga, K., Larssen, T., Zheng, Q., Li, J., Zhang, G., 2013.
Critical evaluation of a new passive exchange-meter for assessing multimedia
fate of persistent organic pollutants at the air-soil interface. Environ. Pollut. 181,
144e150.
122 Q. Zheng et al. / Environmental Pollution 195 (2014) 115e122
Meijer, S.N., Ockenden, W.A., Sweetman, A., Breivik, K., Grimalt, J.O., Jones, K.C.,
2003. Global distribution and budget of PCBs and HCB in background surface
soils: Implications or sources and environmental processes. Environ. Sci.
Technol. 37, 667e672.
Meijer, S.N., Steinnes, E., Ockenden, W.A., Jones, K.C., 2002. Influence of envi-
ronmental variables on the spatial distribution of PCBs in Norwegian and
U.K. soils: implications for global cycling. Environ. Sci. Technol. 36,
2146e2153.
Moeckel, C., Nizzetto, L., Di Guardo, A., Steinnes, E., Freppaz, M., Filippa, G.,
Camporini, P., Benner, J., Jones, K.C., 2008. Persistent organic pollutants in boreal
and montane soil profiles: distribution, evidence of processes and implications
for global cycling. Environ. Sci. Technol. 42, 8374e8380.
Motelay-Massei, A., Harner, T., Shoeib, M., Diamond, M., Stern, G., Rosenberg, B.,
2005. Using passive air samplers to assess urban-rural trends for persistent
organic pollutants and polycyclic aromatic hydrocarbons. 2. Seasonal trends for
PAHs, PCBs, and organochlorine pesticides. Environ. Sci. Technol. 39,
5763e5773.
Ren, Que, Li, Y.-F., Liu, Wan, Xu, Sverko, E., Ma, J., 2007. Polychlorinated biphenyls in
Chinese surface soils. Environ. Sci. Technol. 41, 3871e3876.
Salihoglu, G., Salihoglu, N.K., Aksoy, E., Tasdemir, Y., 2011. Spatial and temporal
distribution of polychlorinated biphenyl (PCB) concentrations in soils of an
industrialized city in Turkey. J. Environ. Manag. 92, 724e732.
SEPA, C., 2003. Building the Capacity of the People's Republic of China to Implement
the Stockholm Convention on POPs and Develop a National Implementation
Plan. GEF Project Brief (GF/CPR/02/010).
Wang, P., Zhang, Q., Wang, Y., Wang, T., Li, X., Li, Y., Ding, L., Jiang, G., 2009. Altitude
dependence of polychlorinated biphenyls (PCBs) and polybrominated diphenyl
ethers (PBDEs) in surface soil from Tibetan Plateau, China. Chemosphere 76,
1498e1504.
Wang, X.-P., Sheng, J.-J., Gong, P., Xue, Y.-G., Yao, T.-D., Jones, K.C., 2012. Persistent
organic pollutants in the Tibetan surface soil: spatial distribution, airesoil ex-
change and implications for global cycling. Environ. Pollut. 170, 145e151.
Wania, F., MacKay, D., 1996. Peer reviewed: tracking the distribution of persistent
organic pollutants. Environ. Sci. Technol. 30, 390Ae396A.
Wania, F., Westgate, J.N., 2008. On the mechanism of mountain cold-trapping of
organic chemicals. Environ. Sci. Technol. 42, 9092e9098.
Wong, F., Bidleman, T.F., 2011. Aging of organochlorine pesticides and poly-
chlorinated biphenyls in muck soil: volatilization, bioaccessibility, and degra-
dation. Environ. Sci. Technol. 45, 958e963.
WORLDCLIM, http://www.worldclim.org.
Xing, X., Lu, Y.L., Dawson, R.W., Shi, Y.J., Zhang, H., Wang, T.Y., Liu, W.B., Ren, H.C.,
2005. A spatial temporal assessment of pollution from PCBs in China. Chemo-
sphere 60, 731e739.
Yeo, H.G., Choi, M., Chun, M.Y., Kim, T.W., Cho, K.C., Young, S.W., 2004. Concentra-
tion characteristics of atmospheric PCBs for urban and rural area, Korea. Sci.
Total Environ. 324, 261e270.
Zhang, G., Chakraborty, P., Li, J., Sampathkumar, P., Balasubramanian, T.,
Kathiresan, K., Takahashi, S., Subramanian, A., Tanabe, S., Jones, K.C., 2008a.
Passive atmospheric sampling of organochlorine pesticides, polychlorinated
biphenyls, and polybrominated diphenyl ethers in urban, rural, and wetland
sites along the coastal length of India. Environ. Sci. Technol. 42, 8218e8223.
Zhang, Z., Liu, L., Li, Y.-F., Wang, D., Jia, H., Harner, T., Sverko, E., Wan, X., Xu, D.,
Ren, N., Ma, J., Pozo, K., 2008b. Analysis of polychlorinated biphenyls in
concurrently sampled Chinese air and surface soil. Environ. Sci. Technol. 42,
6514e6518.
Zheng, X., Chen, D., Liu, X., Zhou, Q., Liu, Y., Yang, W., Jiang, G., 2010. Spatial and
seasonal variations of organochlorine compounds in air on an urban-rural
transect across Tianjin, China. Chemosphere 78, 92e98.
ANNEX 5

Manuscript of Zheng, Q., Nizzetto, L., Li, J., Mulder, M., Sáňka, O., Lammel, G.,
Bing, H., Liu, X., Jiang, Y., Luo, Ch. & Zhang, G. Spatial distribution of old and
emerging flame retardants in Chinese forest soils: Sources, trends and
processes. Environ. Sci. Technol. 49, 2904–2911 (2015).
 Article

pubs.acs.org/est

Spatial Distribution of Old and Emerging Flame Retardants in

Chinese Forest Soils: Sources, Trends and Processes.
Qian Zheng,† Luca Nizzetto,*,‡,§ Jun Li,† Marie D. Mulder,‡ Ondřej Sáňka,‡ Gerhard Lammel,‡,∥
Haijian Bing,⊥ Xin Liu,† Yishan Jiang,† Chunling Luo,† and Gan Zhang*,†
†
State Key Laboratory of Organic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou
510640, China
‡
Masaryk University, Research Centre for Toxic Compounds in the Environment, Brno 601 77, Czech Republic
§
Norwegian Institute for Water Research, Oslo NO-0349, Norway
∥
Max Planck Institute for Chemistry, Mainz 55128, Germany
⊥
The Key Laboratory of Mountain Surface Processes and Ecological Regulation, Institute of Mountain Hazards and Environment,
Chinese Academy of Sciences, Chengdu 610041, China
*
S Supporting Information

ABSTRACT: The levels and distribution of polybrominated diphenylethers (PBDEs),

novel brominated flame retardants (NBFRs) and Dechlorane Plus (DP) in soils and their
dependence on environmental and anthropological factors were investigated in 159 soil
samples from 30 background forested mountain sites across China. Decabromodiphenyl-
ethane (DBDPE) was the most abundant flame retardant (25−18 000 pg g−1 and 5−13 000
pg g−1 in O-horizon and A-horizon, respectively), followed by BDE 209 (nd-5900 pg g−1
and nd-2400 pg g−1 in O-horizon and A-horizon, respectively). FRs distributions were
primarily controlled by source distribution. The distributions of most phasing-out PBDEs,
DP isomers and TBPH were in fact correlated to a population density-based index used as
proxy of areas with elevated usage and waste of FR containing products. High
concentrations of some NBFRs were however observed in industrialized regions and FR
manufacturing plants. Strongly positive correlations were observed between PBDEs and
their replacement products suggesting similar emission pattern and environmental behavior.
Exposure of mineral subsoils depended on precipitations driving leaching of FRs into the
soil core. This was especially evident for some emerging BFRs (TBE, TBPH, and TBB etc.) possibly indicating potential for
diffuse groundwater contamination.

1. INTRODUCTION avoiding compounds with POP characteristics, some NBFRs

Flame retardants (FRs) are a class of chemicals used to reduce
flammability in different kind of products. They are commonly
added in construction materials, furniture, electronics, etc.1
Brominated flame retardants (BFRs) and chlorinated flame
retardants (CFRs) are among the most widely used.
BFRs are structurally heterogeneous.2,3 Among them,
polybrominated diphenylethers (PBDEs) were historically
marketed as three different commercial mixtures commonly
referred to as penta-BDE, octa-BDE, and deca-BDE.1 However,
due to their persistence and bioaccumulative properties some
tetra-BDE, penta-BDE, hexa-BDE, and octa-BDE were added
to the list of banned Persistent Organic Pollutants (POPs)
under the Stockholm Convention.4 Restrictions on the use of
PBDE congeners have paved the way for the use of “novel”
BFRs (NBFRs).5,6 Most commonly used NBFRs are 1,2-
bis(pentabromodiphenyl)ethane (DBDPE), 1,2-bis(2,4,6-
tribromophenoxy)ethane (TBE), 2-ethylhexyl-2,3,4,5-tetrabro-
mobenzoate (TBB), bis(2-ethylhexyl)-3,4,5,6-tetrabromo-
phthalate (TBPH), hexabromobenzene (HBB), and pentabro-
moethylbenzene (PBEB).5,7 Despite the impelling demand for
were shown to have potential for long-range atmospheric
transport, persistence, and bioaccumulation.4,8,9
Dechlorane plus (DP) belongs to the CFRs group and is
common in electrical and electronic items such as wires and
cables, as well as plastic roofing materials.10,11 DP has been
shown to be bioavailable, bioaccumulative and potentially
toxic.12
Soils represent a principal receptor and environmental
reservoir for many persistent semivolatile organic contami-
nants.13 Forest soils, in particular have been pointed out as
effective accumulators of several hydrophobic and persistent
substances due to the forest canopy effect in enhancing
atmospheric depositions, elevated soil organic carbon (OC)
content (representing the bulk capacity for storing these
contaminants), and relatively low level of disturbance, which
Received: December 3, 2014
Revised: February 7, 2015
Accepted: February 8, 2015
Published: February 8, 2015

© 2015 American Chemical Society 2904 DOI: 10.1021/es505876k

Environ. Sci. Technol. 2015, 49, 2904−2911
Environmental Science & Technology
could prevent remobilization.14−17 Research on soils as
environmental reservoirs of semivolatile contaminants have
mainly focused on legacy POPs.18,19 Forest soils may also be
important reservoirs of airborne old and emerging FRs,20,21 and
high persistence in soils has been described for some NBFRs
(e.g., the half-lives of TBE and HBB in soil are reported to be
8600 h).22
China has 134 million hectares of forested land including
tropical, subtropical, temperate, and boreal biomes.23 Main
forested areas in China are prevalently far away from major
conurbations and therefore possible primary sources. Con-
taminant levels and distribution in these soils can therefore be
regarded as indicators of background environmental contam-
ination expectedly reflecting the influence of major environ-
mental drivers of distribution and distance from sources.24
Meteorological conditions (temperature and precipitation
patterns),25 influence distribution of semivolatile contaminants
over large geographical scales.26 The large land area, climate
variability, topographic complexity, ecosystem diversity, and
expected elevated production, use and emissions of FRs make
China an interesting case study and possibly an important
region influencing global contamination and distribution
pattern of these contaminants.27
Research on NBFRs and DP have traditionally mainly
focused on air, sediment, water, sewage sludge, and dust with
very limited information on accumulation and levels in
soils.5,8,28−31 The scope of this study was to (i) identify the
levels and spatial distribution of target compounds in
background forest soils along a range of environmental and
anthropological gradients, (ii) explore colinearity between the
distribution of the target compounds, and (iii) assess possible
influences of environmental and geographical variables
(namely: precipitation, altitude, soil OC content, and exposure
to/distance from primary sources) on contaminant distribution.
2. EXPERIMENTAL SECTION
2.1. Monitoring Design. Thirty mountain sites across
major accessible mountainous areas of China were chosen
(shown in Supporting Information (SI) Figure S1). In each site
1−4 sampling locations were chosen along the same aspect on
the altitudinal transect. In total, 159 forest-soil samples
including O-horizon (77) and A-horizon (82) were collected
at 82 locations between 16 May 2012 and 15 March 2013.
Major transects included: latitude ranging from 21° to 53°;
altitude ranging from 200 to 3800 m; the yearly mean
temperature ranging from 6 to 21 °C and mean yearly
precipitation ranging from 245 mm to 2129 mm. Geo-
referenced data sets of elevation, yearly mean temperature
and precipitation (0.5′ resolved) were obtained from
WorldClim. Details on sampling location characteristics are
presented in SI Table S2.
2.2. Sampling of Soil. In each sampling location, three
small trenches located at about 5 m apart from each other were
excavated. Vegetation litter was carefully removed and the soil
layers were classified based on color and structure of the
material present in each horizon. Samples from the O- (humus)
and A- (topsoil) horizons were collected separately from each
trench, folded in aluminum foil, placed in polyethylene zip-
bags, cooled and transported to the laboratory. The samples
were then freeze-dried and stored at −20 °C until chemical
analysis began.
2.3. Chemical Analysis. In order to remove stones, each
dry sample was sieved to exclude structures larger than 2 mm.
2905
Article
The samples were mixed by pooling together equivalent
amounts of sample from the three trenches in order to create
aggregated samples of the O-horizon and A-horizon (sepa-
rately), reflecting average conditions of the sampling location.
Sample extraction and preparation were consistent with those
described in a previous study.32 Details are reported in the
Supporting Information. Total organic carbon (TOC) in soil
samples was determined with an elemental analyzer (CHNS
Vario Ei III, Elementar) after removal of carbonates with HCl
as described somewhere else.33
2.4. QA/QC. Ten procedural blanks and 20 repeated analysis
of an individual sample were included in the running list to
assess potential laboratory derived contamination and ensure
repeatability of analysis. The Instrument Detection Limit (IDL)
ranged from 0.09 to 0.36 pg (injected) and Method Detection
Limit (MDL) 1.1−28.6 pg g−1, depending on substance
(further details are reported in SI Text S1.2 ). 20 ng PCB
198 and PCB 209 recovery standards were added to each
sample prior to extraction to monitor quantitative analysis
performance. Obtained recoveries were respectively 82% ±
4.9% and 85% ± 3.7% (variance is expressed here as standard
error). Results reported in the study are expressed on a dry
weight basis (pg g−1 dry wt) and not corrected for recovery
results.
2.5. Atmospheric Back Trajectories Analysis and
Index of Potential Source influence. In a primary sources
controlled scenario, FRs concentrations will be highest close to
emission sources. Similarly to other parts of the world,34
previous assessment of FRs (such as PBDEs) distribution in
soils and other environmental compartments in China
evidenced the existence of an urban-rural gradient with highly
populated urban and industrial districts representing the areas
with highest density atmospheric sources.35,36 In China,
industrial and major urban clusters closely overlap.37 For this
reason population count is an adequate proxy for describing
potential sources from FRs usage, wasting and manufacturing.
A few important industrial plants manufacturing or processing
NBFR can however be also found in areas with intermediate
population density. In order to track the possible influence of
use/related primary sources from high density urban areas on
background soil contamination, air mass back trajectories (BT)
and a numerical index of potential source influence (IPSI) in
the regional domain were computed. We elaborated the
concept of IPSI in a previous paper38 (details are also reported
in SI Text S2 of the present paper). In short, IPSI combines
data on population density distribution (as proxy of source
distribution) and BT density calculated for each sampling
location considering a full year of simulations (twice a day). BT
density is used by IPSI to weigh the influence of potential
sources in a given area for an individual sampling location based
on how often this area was upwind of the sampling point.
3. RESULTS AND DISCUSSION
3.1. Summary of FRs Concentration Results. All 16
selected compounds were analyzed in all the samples. SI Table
S3 reports data of individual compounds in the O-horizon and
A-horizon. The concentrations of target compounds in O-
horizon were generally higher than those in A-horizon by an
average factor of 5.4, possibly reflecting the more direct
exposure to atmospheric deposition of the superficial O-
horizon. (SI Table S3 reports details of soil characteristics).
The mean concentrations of the total PBDEs (∑8PBDEs) in
the O- and A-horizon samples were 920 pg g−1 (3−6300 pg
DOI: 10.1021/es505876k
Environ. Sci. Technol. 2015, 49, 2904−2911
Environmental Science & Technology Article

g−1) and 220 pg g−1 (not detected-2500 pg g−1), respectively. PBEB and the major components of PBDE congeners (i.e.,
As expected from previous results, BDE 209 was the most BDE 28, BDE 153, and BDE 183) showed a considerably lower
abundant PBDE congener in these samples,33,39 accounting for variability, due to the lower usage volume or possibly current
more than 60% of ∑8PBDEs, followed by BDE 47 and BDE regulation. Figure 1 shows the spatial variations of target
99. The concentrations of BDE 209 were in the range of nd-
5900 pg g−1 and nd-2400 pg g−1 in O and A-horizon,
respectively. These values are much lower than those reported
for agricultural soils in southern China (39.8−95.2 ng g−1)40
and soils in proximity of e-waste recycling areas (28.8−468 ng
g−1).41 They are nevertheless much higher than forest soil
concentrations measured in Sweden (15−750 pg g−1).42 Such
elevated values can be linked to the higher consumption of
PBDEs in China (30 000 t in 2005) which experienced a sharp
increase during last several years.43 The concentrations of other
PBDEs (∑7PBDEs excluding deca-BDE) in O- and A-horizon
were 3−800 pg g−1 and nd-320 pg g−1, respectively. ∑7PBDEs
in A-horizon was comparable to that observed in other
locations in background soils, including Kuwait (24.7−296 pg
g−1),34 UK (73.5−285 pg g−1),44 and Brazil (434 pg g−1),39
somehow lower compared to that found in mountain soils in
Italy (710 ± 830 pg g−1),45 however higher than that in the
Tibetan Plateau soils.20,21
NBFRs and DP data sets of the soil compartment are very
limited. DBDPE was the FR with the highest concentrations
measured in this study. It ranged 25−18 000 pg g−1 (mean
2643 pg g−1) and 5−13 000 pg g−1 (mean 660 pg g−1) in O-
and A-horizon, respectively. The present data are lower than
data reported for agricultural soils in southern China (28.1 ng
g−1).40 DBDPE predominance in environmental samples was
also previously shown in air samples at a rural site of China.36
Such a behavior can be related to the large success of this
product in the market. DBDPE market is currently increasing at
a yearly rate of 80% in China.5 Some studies however have
reported DBDPE concentrations lower than BDE 209 in
sewage sludge samples in Europe2,16 and air samples in Chinese
e-waste area.36 This discrepancy might be attributed to several
factors including regional and temporal variability in FR usage
and emissions and type of samples.
The remaining set of NBFRs was present at generally lower
concentrations. TBB: nd-1400 pg g−1 (mean 250.5 pg g−1) and
nd-1600 pg g−1 (mean 184.0 pg g−1), TBPH: 4.0−643.7 pg g−1
(mean 131.7 pg g−1) and 5.5−526.4 pg g−1(mean 68.9 pg g−1),
TBE: nd-330 pg g−1 (mean 48.5 pg g−1) and nd-240 pg g−1
(mean 22.9 pg g−1), HBB: nd-340 pg g−1 (mean 46.0 pg g−1)
and nd-42 pg g−1(mean 7.2 pg g−1), and PBEB showed the
lowest concentrations in the range of nd-92 pg g−1 (mean 6.8
pg g−1) and nd-70 pg g−1 (mean 9.0 pg g−1) in O-horizon and
A-horizon, respectively. The detection frequencies varied
among different compounds (details are given in SI Table S3).
Both anti-isomer and syn-isomer of DP were detected in
100% of the samples, demonstrating the currently broad
Figure 1. Spatial distribution of selected FRs in organic (O-horizon)
distribution of this FRs in the Chinese markets. The soil samples. (a:BDE47, BDE99, anti-DP, syn-DP, TBPH, TBB, HBB
concentrations of anti-isomer and syn-isomer were 5−680 pg and TBE; b:BDE 209;c:DBDPE).
g−1 and 4−390 pg g−1 in O-horizon and 3−220 pg g−1 and 2−
170 pg g−1 in A-horizon, respectively. Concentrations measured
in the selected forest soils were lower than those previously compounds in O-horizons. Overall, high levels of many FRs
measured in urban, rural, and remote surface soils of Northern were mainly found in the Southeast China. For instance, the
China (nd-8.45 ng g−1 and nd-3.76 ng g−1 for anti-DP and syn- highest average concentrations of BDE 47, BDE 99, and TBPH
DP, respectively).28 were measured in site 21; whereas, anti-DP, syn-DP, BDE 209,
3.2. Spatial Trends in Superficial Soils. High spatial and TBE showed peaking concentrations in site 23. Sites 21
variability of FR concentrations was observed during this and 23 are both in the Guangdong province, already reported
survey. For DBDPE, the differences between the highest and as one of the important e-waste recycling and the most
lowest concentrations were up to a factor of 2300. In contrast, urbanized/industrialized regions in South China.36,41,46 Some
2906 DOI: 10.1021/es505876k
Environ. Sci. Technol. 2015, 49, 2904−2911
Environmental Science & Technology Article

interesting and informative outliers were found too. DBDPE

displayed the highest average concentration in site 10 (Figure
1c) in Shandong Province, due to important DBDPE
production facilities are located in this area,5,8 while two
outliers for TBB and HBB were located in site 6 and 7 (Figure
1a) in Hebei province, also an important industrial area. In
China, 600 tons/year of HBB are produced in Shandong
province,5 and as expected high concentration could be found
in site 10. Unfortunately, there is no production or usage
information publically disclosed for TBB.5,8 As far as we know,
TBB combined with TBPH was used in Firemaster BZ-54,
Firemaster 550 (in a 4:1 ratio) and DP-45 (91% TBPH
together with 2−6% TBB by weight).47 The ratios of TBB to
TBPH in samples in Northern China averaged 5.2 (in contrast
with diagnostic ratios observed in the Southern part averaging
0.55), indicating different sources or fate in these subregions as
well as the northern samples probably been more affected by
proximity to manufacture related sources. The high con-
sumption of Firemaster 550 in Northern China, especially in
Hebei province appears to be related to the peculiar type of
manufacturing district present here which includes oil, coal,
metallurgical, mechanical, and electronics industry. The
considerably lower ratio value found in Southern China
suggests that manufacturers in this area may have used different
FR formulations (such as DP-45, neoprene and polyvinyl
chloride (PVC)31) with a relatively higher level of TBPH.
Moreover, TBB has been reported to be more vulnerable to
photodegradation with a half-life twice shorter than TBPH.47,48
Warmer climate and higher solar radiation in Southern China
might accelerate degradation and result in lower ratios observed
in the soils of this region. In general, the presence of these
outliers highlights the strong influence of local production
point sources for the exposure of background mountain soils.
These data indicate that the current emission control measures
applied during industrial production and processing in China
may be insufficient to prevent off-site exposure to FRs at the
subregional level.
In summary, spatial distribution of flame retardants in
background soils in China appears to be dependent on source
distribution and sensitive to the typology of sources (e.g., e-
waste facilities and manufacturing facilities as well as highly
populated areas where the majority of FR containing products
are used).49
In order to provide a more quantitative assessment of this
hypothesis, we statistically explored the interaction between
concentrations of FRs in the O-horizon and the results for IPSI.
IPSI values for different sampling sites ranged between 1 and
37 following a slightly negatively skewed distribution, with most
frequently observed values around 25 (Figure 1). Consistently, Figure 2. Correlation between selected FR concentrations and IPSI
the highest IPSI values are observed in the Southeast (in values.
particular sites 21−23). Most of the target compounds in the
O-horizon showed high levels in locations with a high IPSI.
Nearly all the FRs (with the exception of PBEB, HBB, and point of view, penta-BDE and octa-BDE technical mixtures are
TBB) showed positive correlation with IPSI with correlation currently undergoing phasing out, therefore primary emissions
coefficients (Spearman) typically ranging between 0.1 and 0.5 are continuing mostly from consumer products still in use or
(with exception of BDE 28: r = 0.08 and BDE 209: r = 0.06) inadequate waste management. These types of sources are
(SI Table S6). These relationships were statistically significant obviously associated with the use of FR rather than
(p < 0.05) for most of the PBDE congeners (BDE 47, BDE manufacturing. Hence use-related sources are a function of
100, BDE 153, BDE 154, and BDE 183), DP-isomers and population density which in turn is a key for IPSI calculation. In
TBPH (Figure 2 and SI Table S6a). We can only speculate on contrast, BDE 209 is still produced and largely used in China.
the reasons behind the higher sensitivity of these PBDEs on Similarly to emerging FRs, emissions from production and
use-related source distribution, compared to other FR which processing plants is expected to lower the dependence on IPSI.
did not display a clear dependence on IPSI. From a regulatory As an example, the extremely high level of BDE 209 observed at
2907 DOI: 10.1021/es505876k
Environ. Sci. Technol. 2015, 49, 2904−2911
Environmental Science & Technology
site 10 (Shangdong province, in spite of the relatively low IPSI
score) is likely associated with the documented presence of
production plants in this area.50 As for DPs and TBPH
(displaying dependence on IPSI), may already follow a
consolidated distribution in the market and their usage- and
disposal-related sources. In contrast, the distribution of some
more abundant NBFRs in soils appeared to be influenced by
production plant/processing location and did not display
dependence on IPSI. This trend is in general agreement with
previous studies that had shown that the signal of FR sources
from manufacturing locations can be detected in air and other
media on the local scale range.8,51 The present results expand
this paradigm showing that emission point sources can possibly
affect soil concentrations even at subregional level.
3.3. Co-Linearity between Compound Distributions.
The large-scale data set of both discontinued and current use
flame retardants presented here provides a unique opportunity
to explore how market shifts may have affected environmental
levels and distribution patterns. For example a mixture of
TBPH and TBB, TBE and DBDPE are used as alternatives for
penta-BDE, octa-BDE, and deca-BDE formulation, respec-
tively.5,7,31,52,53 Hence, a spatial distribution pattern with
PBDEs levels being negatively correlated to NBFR levels
could be hypothesized as a result of growing emissions of
NBFR and simultaneously declining emissions and ongoing
degradation of the environmental burden of PBDEs. In contrast
to this hypothesis, levels of PBDEs and NBFRs were strongly
positively correlated, suggesting similar drivers controlling their
spatial distribution. A summary of the correlation analysis
results are reported in SI Table S4. Significantly positive
relationships between BDE 209 and DBDPE were observed in
both O- and A-horizon, such a behavior was previously
observed by Ma et al. in the Great Lakes atmosphere.7
Similarly, BDE 183 and TBE (TBE being a replacement for
octa-BDE) were strongly positively correlated in both O- and
A- soil horizons (SI Table S4). TBPH was significantly
correlated to the penta-BDE mixtures (sum of BDE 47, 99,
100, 153, and 154) in both O- and A-horizons, despite this
novel FR is used as a substitute of banned PBDEs. This result
indicates that emission patterns and environmental behavior of
old and emerging FRs are, to a certain extent, similar. In
addition, there is evidence here that the reduced emissions of
PBDEs following the international ban are not yet reflected in
background soil samples over the large spatial scale. Kawai et al.
proposed half-lives of brominated organic compounds in soil
(e.g., BDE 35 and BDE 99 in the order of 17 000 and 55 000 h,
respectively).54 These values reflect the elevated persistence of
PBDEs in soil. Hence it is unlikely that management actions to
reduce emissions can be reflected in changes of soil
contamination in the short term. In addition, the observed
correlation between the behavior of old and emerging flame
retardants shed (at least from the environmental exposure point
of view) a critical light on the effectiveness of the substitution
strategy adopted in recent years by the FRs industry.
TBB and TBPH distribution were negatively correlated in
soil samples despite these substances are used together (yet
with varying proportions) in FR commercial formulations. The
ratio of cTBB/c(TBPH+TBB) in these soil samples ranged from 0 to
0.97 (mean 0.48) and 0 to 0.99 (mean 0.45) in O-horizon and
A-horizon respectively (shown in SI Table S5). A wide range of
cTBB/cTBPH (0.05−50) was also previously reported in another
study focusing on the analysis of dust samples.31 Differences in
sources (e.g., commercial formulations) and physical chemical
2908
Article
properties controlling environmental fate of the different
ingredients have to be considered among the most likely
causes, as also discussed in section 3.2 of this paper.
In spite of the great variation in DP levels across different
forest sites, anti- and syn-isomers showed extremely strong
relationship (R2 = 0.96, p < 0.01 and R2 = 0.92, p < 0.01 in O-
horizon and A-horizon, respectively) (SI Figure S2). This is
consistent with other reports focusing on different matrices
(e.g., soil of Huai’an City (R2 = 0.98)55 and ambient air around
Shanghai (R2 = 0.93).56 The ratio between anti- or syn- isomer
has widely been used as a proxy of remoteness (in space and
time) of the contamination following their different environ-
mental behavior.55,56 In this study, fanti is defined as the ratio of
canti‑DP/c(anti‑DP+syn‑DP) (SI Table S5). fanti ranged 0.43 to 0.72
(mean 0.56) in O-horizon and 0.41 to 0.79 (mean 0.58) in A-
horizon. Possibly because of the large spatial scale covered in
our survey, these ranges are larger and their mean values are
lower than those reported by Wang et al. (0.67−0.85),55 Hoh
et al.57and Qiu et al.(0.75−0.80),57,58 but similar to those
reported in a case study in Lake Ontario (0.53).59 Lower fanti
values (<0.5) were also observed in Great Lakes air and
sediment samples, which were collected in distal sites from
production facilities.57 The forested sites selected in the present
study are classified as semiremote areas. fanti values lower than
0.5 measured here can also be discussed as a result of different
environmental behavior of the two isomers (including stereo-
selective degradation) and remoteness from sources.
3.4. Influence of Environmental Parameters. Biogeo-
chemical Control. The spatial trends of FRs in the A-horizons
were often poorly consistent with those observed in O-
horizons. The O-horizon is directly exposed to atmospheric
depositions, while the presence of flame retardants in the
“aged” A-horizon has to be related to the effects of a range of
processes driving to their downward vertical transport in the
soil column. These processes, in particular, are water driven
leaching, vertical diffusion and, up to a minor extent,
bioturbation. The generally higher measured levels and
detection frequency of TBE (66%) in A-horizon compared to
its “predecessor” BDE 183 (29%) suggested that mobility in
soils might be significantly higher for TBE. Levels of TBE and
other NBFRs (TBPH and TBB, etc. (SI Table S3)) in the A-
horizon were in some cases similar to those observed in the O-
horizon, confirming rapid vertical translocation of these
contaminants in soils. In soil ecotoxicology, the concepts of
mobility and “availability” (including bioavailability) are strictly
related.60 High leaching rates are also associated with higher
potential for groundwater contamination. The fact that many
NBFRs are found in mineral soils at similar levels as observed in
superficial soil, after relatively short time from their
introduction is of some concern. Interestingly, NBFRs were
more frequently detected in the A-horizon of forested sites of
the subtropical part of China, where precipitation is more
intense and turnover of organic matter in O-horizon is faster. In
conclusion, the climatic and ecological conditions of this region
may configure an increased risk for leaching.61
Total Organic Carbon (TOC) in Soil. Soil OC is a key
parameter controlling distribution of hydrophobic persistent
contaminants (including POPs) at the global scale. The
relationship between concentration of POPs and soil OC is
underpinned by the large affinity of these compounds to the
organic matter and their persistence.24,38,62 In contrast, the
present assessment shows that among the PBDE congeners
only BDE 47 had a significantly positive correlation with soil
DOI: 10.1021/es505876k
Environ. Sci. Technol. 2015, 49, 2904−2911
Environmental Science & Technology
OC in both O- and A-horizon (SI Table S6). Only PBEB was
significantly and positively correlated with OC in A-horizon.
Some compounds were negatively related to OC either in O-
horizon or A-horizon (TBE in O-horizon (r = −0.14; p > 0.05);
BDE 183 in A-horizon (r = −0.05; p > 0.05)). These negative
correlations have to be regarded as spurious and essentially due
to the contingent inverse (nonsignificant) relationship between
source distribution (as described by IPSI) and soil TOC.
The majority of the target compounds did not display any
dependence on TOC. This is an important result supporting
the hypothesis that source distribution (rather than environ-
mental factors) is pivotal for controlling distribution of both
“old” and novel FRs in background soils.
Precipitation, Temperature and Latitude. Precipitation and
temperature influence deposition, degradation and reemission
processes of many POPs,25 possibly resulting in a general
dependence of concentration (or pattern) on latitude.63 In the
O-horizon, more than 80% of the FRs did not display any
dependence on precipitation and latitude (SI Table S6a). The
situation was totally different in the A-horizon in which 75% of
contaminants had significant positive correlations with
precipitation and negative correlations with latitude (SI Table
S6b). Heavy precipitation in Southern China accelerates
leaching, enhancing exposure of mineral soil layers in this
region. Negative correlations between environmental levels and
latitude in A-horizon (SI Table S6b) have therefore simply to
be interpreted as the result of the influence of areas with higher
source density in the rainy Southern China, as described above.
Similarly, spurious correlations between the levels of many
compounds and temperature were observed in both O- and A-
horizon owing to colinearity between source distribution and
temperature.
The present study highlighted the importance of source
location in controlling distribution of both old and emerging
FRs in soils of background mountain environments across a
large regional scale. The contribution of local sources
dominated over the possible influence of environmental factors
and soil biogeochemical processes affecting contaminants fate.
The only exception is the exposure of mineral subsoils that
appeared to be controlled by the amount of precipitation. Co-
linearity between the distribution of PBDEs and NBFRs (used
as their substitute) suggests still high emission of phasing-out
substances and similar environmental behavior of phasing-in
ones.
■
ASSOCIATED CONTENT
*
S Supporting Information
Details on chemical analysis methodology, the full experimental
data set of concentrations and environmental/anthropogenic
variables and results of statistical analysis are reported. This
material is available free of charge via the Internet at http://
pubs.acs.org/.
■ AUTHOR INFORMATION
Corresponding Authors
*(L.N.) Phone: +47 98215393; fax: +47 22185200; e-mail:
nizzetto@recetox.muni.cz and luca.nizzetto@niva.no.
*(G.Z.) Phone: +86 20 8529 0805; fax: +86 20 8529 0130; e-
mail: zhanggan@gig.ac.cn.
Notes
The authors declare no competing financial interest.
2909
Article
■ ACKNOWLEDGMENTS
This work is supported by the National Scientific Foundation
of China (Nos. 41125014, 41390240), the Joint Funds of the
National Natural Science Foundation of China and the Natural
Science Foundation of Guangdong Province, China (Nos.
U1133004). Part of the work was carried out with the support
of core facilities of the Research Centre for Toxic Compounds
in the Environment (RECETOX) − National Infrastructure for
Research of Toxic Compounds in the Environment, project
number LM2011028, funded by the Ministry of Education,
Youth and Sports of the Czech Republic under the activity
“Projects of major infrastructures for research, development
and innovations”. OS participation to the study was supported
by the RECETOX GENASIS infrastructure.
■
REFERENCES
(1) Stapleton, H. M.; Misenheimer, J.; Hoffman, K.; Webster, T. F.
Flame retardant associations between children’s handwipes and house
dust. Chemosphere 2014, 116, 54−60.
(2) Gorga, M.; Martinez, E.; Ginebreda, A.; Eljarrat, E.; Barcelo, D.
Determination of PBDEs, HBB, PBEB, DBDPE, HBCD, TBBPA and
related compounds in sewage sludge from Catalonia (Spain). Sci. Total
Environ. 2013, 444, 51−59.
(3) WHO (1997) Environmental Health Criteria; Flame Retardants: A
General Introduction; World Health Organization: Geneva, Switzerland,
http://www.inchem.org/documents/ehc/ehc/ehc192.htm.
(4) Ma, Y.; Venier, M.; Hites, R. A. 2-Ethylhexyl tetrabromobenzoate
and bis(2-ethylhexyl) tetrabromophthalate flame retardants in the
Great Lakes atmosphere. Environ. Sci. Technol. 2012, 46 (1), 204−208.
(5) Covaci, A.; Harrad, S.; Abdallah, M. A. E.; Ali, N.; Law, R. J.;
Herzke, D.; de Wit, C. A. Novel brominated flame retardants: A review
of their analysis, environmental fate and behaviour. Environ. Int. 2011,
37 (2), 532−556.
(6) Betts, K. New flame retardants detected in indoor and outdoor
environments. Environ. Sci. Technol. 2008, 42 (18), 6778−6778.
(7) Ma, Y.; Salamova, A.; Venier, M.; Hites, R. A. Has the phase-out
of PBDEs affected their atmospheric levels? Trends of PBDEs and
their replacements in the Great Lakes atmosphere. Environ. Sci.
Technol. 2013, 47 (20), 11457−11464.
(8) de Wit, C.; Kierkegaard, A.; Ricklund, N.; Sellström, U.Emerging
brominated flame retardants in the environment. In Brominated Flame
Retardants; Springer: Berlin Heidelberg, 2010; pp 241−286.
(9) de Wit, C. A. An overview of brominated flame retardants in the
environment. Chemosphere 2002, 46 (5), 583−624.
(10) Xiang, N.; Chen, L.; Meng, X.-Z.; Li, Y.-L.; Liu, Z.; Wu, B.; Dai,
L.; Dai, X. Polybrominated diphenyl ethers (PBDEs) and dechlorane
plus (DP) in a conventional wastewater treatment plant (WWTP) in
Shanghai: Seasonal variations and potential sources. Sci. Total Environ.
2014, 487, 342−349.
(11) Sverko, E.; Tomy, G. T.; Reiner, E. J.; Li, Y.-F.; McCarry, B. E.;
Arnot, J. A.; Law, R. J.; Hites, R. A. Dechlorane Plus and related
compounds in the environment: A review. Environ. Sci. Technol. 2011,
45 (12), 5088−5098.
(12) Yu, Z.; Lu, S.; Gao, S.; Wang, J.; Li, H.; Zeng, X.; Sheng, G.; Fu,
J. Levels and isomer profiles of Dechlorane Plus in the surface soils
from e-waste recycling areas and industrial areas in South China.
Environ. Pollut. 2010, 158 (9), 2920−2925.
(13) Wang, Y.; Cheng, Z.; Li, J.; Luo, C.; Xu, Y.; Li, Q.; Liu, X.;
Zhang, G. Polychlorinated naphthalenes (PCNs) in the surface soils of
the Pearl River Delta, South China: Distribution, sources, and air-soil
exchange. Environ. Pollut. 2012, 170, 1−7.
(14) Moeckel, C.; Nizzetto, L.; Strandberg, B.; Lindroth, A.; Jones, K.
C. Air-boreal forest transfer and processing of polychlorinated
biphenyls. Environ. Sci. Technol. 2009, 43 (14), 5282−5289.
(15) Nizzetto, L.; Cassani, C.; Di Guardo, A. Deposition of PCBs in
mountains: The forest filter effect of different forest ecosystem types.
Ecotoxicol. Environ. Saf. 2006, 63 (1), 75−83.
DOI: 10.1021/es505876k
Environ. Sci. Technol. 2015, 49, 2904−2911
Environmental Science & Technology
(16) Su, Y. S.; Wania, F. Does the forest filter effect prevent
semivolatile organic compounds from reaching the Arctic? Environ. Sci.
Technol. 2005, 39 (18), 7185−7193.
(17) Wania, F.; Mackay, D. Global fractionation and cold
condensation of low volatility organochlorine compounds in polar-
regions. Ambio 1993, 22 (1), 10−18.
(18) Bergknut, M.; Wiberg, K.; Klaminder, J. Vertical and lateral
redistribution of POPs in soils developed along a hydrological
gradient. Environ. Sci. Technol. 2011, 45 (24), 10378−10384.
(19) Aichner, B.; Bussian, B.; Lehnik-Habrink, P.; Hein, S. Levels and
spatial distribution of persistent organic pollutants in the environment:
A case study of German forest soils. Environ. Sci. Technol. 2013, 47
(22), 12703−12714.
(20) Yuan, G.-L.; Xie, W.; Che, X.-C.; Han, P.; Liu, C.; Wang, G.-H.
The fractional patterns of polybrominated diphenyl ethers in the soil
of the central Tibetan Plateau, China: The influence of soil
components. Environ. Pollut. 2012, 170, 183−189.
(21) Zheng, X.; Liu, X.; Jiang, G.; Wang, Y.; Zhang, Q.; Cai, Y.;
Cong, Z. Distribution of PCBs and PBDEs in soils along the altitudinal
gradients of Balang Mountain, the east edge of the Tibetan Plateau.
Environ. Pollut. 2012, 161, 101−106.
(22) Xiao, H.; Shen, L.; Su, Y.; Barresi, E.; DeJong, M.; Hung, H.;
Lei, Y.-D.; Wania, F.; Reiner, E. J.; Sverko, E.; Kang, S.-C. Atmospheric
concentrations of halogenated flame retardants at two remote
locations: The Canadian High Arctic and the Tibetan Plateau. Environ.
Pollut. 2012, 161, 154−161.
(23) Fang, J. Y.; Chen, A. P.; Peng, C. H.; Zhao, S. Q.; Ci, L. Changes
in forest biomass carbon storage in China between 1949 and 1998.
Science 2001, 292 (5525), 2320−2322.
(24) Meijer, S. N.; Ockenden, W. A.; Sweetman, A.; Breivik, K.;
Grimalt, J. O.; Jones, K. C. Global distribution and budget of PCBs
and HCB in background surface soils: Implications or sources and
environmental processes. Environ. Sci. Technol. 2003, 37 (4), 667−672.
(25) Xu, Y.; Tian, C.; Zhang, G.; Ming, L.; Wang, Y.; Chen, Y.; Tang,
J.; Li, J.; Luo, C. Influence of monsoon system on α-HCH fate in Asia:
A model study from 1948 to 2008. Journal of Geophys. Res.: Atmos.
2013, 118 (12), 6764−6770.
(26) Kirchner, M.; Faus-Kessler, T.; Jakobi, G.; Levy, W.;
Henkelmann, B.; Bernhoeft, S.; Kotalik, J.; Zsolnay, A.; Bassan, R.;
Belis, C.; Kraeuchi, N.; Moche, W.; Simoncic, P.; Uhl, M.; Weiss, P.;
Schramm, K. W. Vertical distribution of organochlorine pesticides in
humus along Alpine altitudinal profiles in relation to ambiental
parameters. Environ. Pollut. 2009, 157 (12), 3238−3247.
(27) Jiang, H.; Apps, M. J.; Zhang, Y. L.; Peng, C. H.; Woodard, P.
M. Modelling the spatial pattern of net primary productivity in
Chinese forests. Ecol. Modell. 1999, 122 (3), 275−288.
(28) Ma, J.; Qiu, X.; Liu, D.; Zhao, Y.; Yang, Q.; Fang, D. Dechlorane
Plus in surface soil of North China: Levels, isomer profiles, and spatial
distribution. Environ. Sci. Pollut. Res. Int. 2014, 21 (14), 8870−7.
(29) Hoh, E.; Hites, R. A. Brominated flame retardants in the
atmosphere of the east-central United States. Environ. Sci. Technol.
2005, 39 (20), 7794−7802.
(30) Gouteux, B.; Alaee, M.; Mabury, S. A.; Pacepavicius, G.; Muir,
D. C. G. Polymeric brominated flame Retardants: Are they a relevant
source of emerging brominated aromatic compounds in the environ-
ment? Environ. Sci. Technol. 2008, 42 (24), 9039−9044.
(31) Stapleton, H. M.; Allen, J. G.; Kelly, S. M.; Konstantinov, A.;
Klosterhaus, S.; Watkins, D.; McClean, M. D.; Webster, T. F. Alternate
and new brominated flame retardants detected in U.S. house dust.
Environ. Sci. Technol. 2008, 42 (18), 6910−6916.
(32) Wang, Y.; Luo, C.; Li, J.; Yin, H.; Zhang, G. Influence of plants
on the distribution and composition of PBDEs in soils of an e-waste
dismantling area: Evidence of the effect of the rhizosphere and
selective bioaccumulation. Environ. Pollut. 2014, 186, 104−109.
(33) Mai, B. X.; Chen, S. J.; Luo, X. J.; Chen, L. G.; Yang, Q. S.;
Sheng, G. Y.; Peng, P. G.; Fu, J. M.; Zeng, E. Y. Distribution of
polybrominated diphenyl ethers in sediments of the Pearl River Delta
and adjacent South China Sea. Environ. Sci. Technol. 2005, 39 (10),
3521−3527.
2910
Article
(34) Gevao, B.; Ghadban, A. N.; Uddin, S.; Jaward, F. M.; Bahloul,
M.; Zafar, J. Polybrominated diphenyl ethers (PBDEs) in soils along a
rural-urban-rural transect: Sources, concentration gradients, and
profiles. Environ. Pollut. 2011, 159 (12), 3666−3672.
(35) Jaward, T. M.; Zhang, G.; Nam, J. J.; Sweetman, A. J.; Obbard, J.
P.; Kobara, Y.; Jones, K. C. Passive air sampling of polychlorinated
biphenyls, organochlorine compounds, and polybrominated diphenyl
ethers across Asia. Environ. Sci. Technol. 2005, 39 (22), 8638−8645.
(36) Tian, M.; Chen, S.-J.; Wang, J.; Zheng, X.-B.; Luo, X.-J.; Mai, B.-
X. Brominated flame retardants in the atmosphere of e-waste and rural
sites in southern China: Seasonal variation, temperature dependence,
and gas-particle partitioning. Environ. Sci. Technol. 2011, 45 (20),
8819−8825.
(37) Xing, X.; Lu, Y. L.; Dawson, R. W.; Shi, Y. J.; Zhang, H.; Wang,
T. Y.; Liu, W. B.; Ren, H. C. A spatial temporal assessment of pollution
from PCBs in China. Chemosphere 2005, 60 (6), 731−739.
(38) Zheng, Q.; Nizzetto, L.; Mulder, M. D.; Ondřej, Sáňka; Lammel,
G.; Li, J.; Bing, H.; Liu, X.; Jiang, Y.; Luo, C.; Zhang, G. Does an
analysis of polychlorinated biphenyl (PCB) distribution in mountain
soils across China reveal a latitudinal fractionation paradox? Environ.
Pollut. 2014, 195, 115−122.
(39) Thorenz, U. R.; Bandowe, B. A. M.; Sobocka, J.; Wilcke, W.
Method optimization to measure polybrominated diphenyl ether
(PBDE) concentrations in soils of Bratislava, Slovakia. Environ. Pollut.
2010, 158 (6), 2208−2217.
(40) Shi, T.; Chen, S.-J.; Luo, X.-J.; Zhang, X.-L.; Tang, C.-M.; Luo,
Y.; Ma, Y.-J.; Wu, J.-P.; Peng, X.-Z.; Mai, B.-X. Occurrence of
brominated flame retardants other than polybrominated diphenyl
ethers in environmental and biota samples from southern China.
Chemosphere 2009, 74 (7), 910−916.
(41) Cheng, Z.; Wang, Y.; Wang, S.; Luo, C.; Li, J.; Chaemfa, C.;
Jiang, H.; Zhang, G. The influence of land use on the concentration
and vertical distribution of PBDEs in soils of an e-waste recycling
region of South China. Environ. Pollut. 2014, 191, 126−31.
(42) Ejarrat, E.; Marsh, G.; Labandeira, A.; Barcelo, D. Effect of
sewage sludges contaminated with polybrominated diphenylethers on
agricultural soils. Chemosphere 2008, 71 (6), 1079−1086.
(43) Zou, M.-Y.; Ran, Y.; Gong, J.; Maw, B.-X.; Zeng, E. Y.
Polybrominated diphenyl ethers in watershed soils of the Pearl River
Delta, China: Occurrence, inventory, and fate. Environ. Sci. Technol.
2007, 41 (24), 8262−8267.
(44) Harrad, S.; Hunter, S. Concentrations of polybrominated
diphenyl ethers in air and soil on a rural-urban transect across a major
UK conurbation. Environ. Sci. Technol. 2006, 40 (15), 4548−4553.
(45) Parolini, M.; Guazzoni, N.; Binelli, A.; Tremolada, P.
Polybrominated diphenyl ether contamination in soil, vegetation,
and cow milk from a high-mountain pasture in the Italian Alps. Arch.
Environ. Contam. Toxicol. 2012, 63 (1), 29−44.
(46) Wang, Y.; Jiang, G.; Lam, P. K. S.; Li, A. Polybrominated
diphenyl ether in the East Asian environment: A critical review.
Environ. Int. 2007, 33 (7), 963−973.
(47) La Guardia, M. J.; Hale, R. C.; Harvey, E.; Mainor, T. M.;
Ciparis, S. In situ accumulation of HBCD, PBDEs, and several
alternative flame-retardants in the bivalve (Corbicula f luminea) and
gastropod (Elimia proxima). Environ. Sci. Technol. 2012, 46 (11),
5798−5805.
(48) Davis, E. F.; Stapleton, H. M. Photodegradation pathways of
nonabrominated diphenyl ethers, 2-ethylhexyltetrabromobenzoate and
di(2-ethylhexyl)tetrabromophthalate: Identifying potential markers of
photodegradation. Environ. Sci. Technol. 2009, 43 (15), 5739−5746.
(49) She, J.; She, Y.; Song, W. Comparison of PBDE congener
profiles and concentration levels in human specimens from China and
the US and identification of human exposure sources. Sci. China-Chem.
2010, 53 (5), 995−1002.
(50) XUE Zheng-ran, L. H.-j. The productive mechanism and
technology research of new highly efficiency flame retardant of
decabromodiphehyl ether. Shandong Chem. Ind. 2002, 04, 31−32.
DOI: 10.1021/es505876k
Environ. Sci. Technol. 2015, 49, 2904−2911
Environmental Science & Technology Article

(51) Qiu, X.; Hites, R. A. Dechlorane plus and other flame retardants
in tree bark from the Northeastern United States. Environ. Sci. Technol.
2008, 42 (1), 31−36.
(52) Renner, R. In U.S., flame retardants will be voluntarily phased
out. Environ. Sci. Technol. 2004, 38 (1), 14A−14A.
(53) Hoh, E.; Zhu, L. Y.; Hites, R. A. Novel flame retardants, 1,2-
bis(2,4,6-tribromophenoxy)ethane and 2,3,4,5,6-pentabromoethylben-
zene, in United States’ environmental samples. Environ. Sci. Technol.
2005, 39 (8), 2472−2477.
(54) Kawai, T.; Jagiello, K.; Sosnowska, A.; Odziomek, K.; Gajewicz,
A.; Handoh, I. C.; Puzyn, T.; Suzuki, N. A new metric for long-range
transport potential of chemicals. Environ. Sci. Technol. 2014, 48 (6),
3245−3252.
(55) Wang, B.; Iino, F.; Huang, J.; Lu, Y.; Yu, G.; Morita, M.
Dechlorane Plus pollution and inventory in soil of Huai’an City,
China. Chemosphere 2010, 80 (11), 1285−1290.
(56) Yu, Z.; Liao, R. e.; Li, H.; Mo, L.; Zeng, X.; Sheng, G.; Fu, J.
Particle-bound Dechlorane Plus and polybrominated diphenyl ethers
in ambient air around Shanghai, China. Environ. Pollut. 2011, 159 (10),
2982−2988.
(57) Hoh, E.; Zhu, L. Y.; Hites, R. A. Dechlorane plus, a chlorinated
flame retardant, in the Great Lakes. Environ. Sci. Technol. 2006, 40 (4),
1184−1189.
(58) Qiu, X.; Marvin, C. H.; Hites, R. A. Dechlorane plus and other
flame retardants in a sediment core from Lake Ontario. Environ. Sci.
Technol. 2007, 41 (17), 6014−6019.
(59) Tomy, G. T.; Pleskach, K.; Ismail, N.; Whittle, D. M.; Helm, P.
A.; Sverko, E.; Zaruk, D.; Marvin, C. H. Isomers of dechlorane plus in
Lake Winnipeg and Lake Ontario food webs. Environ. Sci. Technol.
2007, 41 (7), 2249−2254.
(60) Riding, M. J.; Doick, K. J.; Martin, F. L.; Jones, K. C.; Semple, K.
T. Chemical measures of bioavailability/bioaccessibility of PAHs in
soil: Fundamentals to application. J. Hazard. Mater. 2013, 261, 687−
700.
(61) Bearr, J.; Stapleton, H.; Mitchelmore, C. Accumulation and
DNA damage in fat-head minnows (Pimephales promelas) exposed to
2 brominated flame-retardant mixtures, firemaster 550 and firemaster
BZ-54. Environ. Toxicol. Chem. 2012, 29 (3), 722−729.
(62) Hassanin, A.; Breivik, K.; Meijer, S. N.; Steinnes, E.; Thomas, G.
O.; Jones, K. C. PBDEs in European background soils: Levels and
factors controlling their distribution. Environ. Sci. Technol. 2004, 38
(3), 738−745.
(63) Grimalt, J. O.; Borghini, F.; Sanchez-Hernandez, J. C.; Barra, R.;
Garcia, C. J. T.; Focardi, S. Temperature dependence of the
distribution of organochlorine compounds in the mosses of the
Andean mountains. Environ. Sci. Technol. 2004, 38 (20), 5386−5392.

2911 DOI: 10.1021/es505876k

Environ. Sci. Technol. 2015, 49, 2904−2911
ANNEX 6

Manuscript of Melymuk, L., Bohlin, P., Sáňka, O., Pozo, K. & Klánová, J.
Current challenges in air sampling of semivolatile organic contaminants:
sampling artifacts and their influence on data comparability. Environ. Sci.
Technol. 48, 14077–91 (2014).
 Critical Review

pubs.acs.org/est

Current Challenges in Air Sampling of Semivolatile Organic

Contaminants: Sampling Artifacts and Their Influence on Data
Comparability
Lisa Melymuk,*,† Pernilla Bohlin,*,† Ondřej Sáňka,† Karla Pozo,† and Jana Klánová†
†
Research Centre for Toxic Compounds in the Environment (RECETOX), Faculty of Science, Masaryk University, Kamenice 5/753,
Pavilon A29, Brno 62500, Czech Republic
*
S Supporting Information

ABSTRACT: With current science and policy needs, more attention

is being given to expanding and improving air sampling of semivolatile
organic contaminants (SVOCs). However, a wide range of techniques
and configurations are currently used (active and passive samplers,
different deployment times, different sorbents, etc.) and as the SVOC
community looks to assess air measurements on a global scale,
questions of comparability arise. We review current air sampling
techniques, with a focus on sampling artifacts that can lead to
uncertainties or biases in reported concentrations, in particular
breakthrough, degradation, meteorological influences, and assump-
tions regarding passive sampling. From this assessment, we estimate
the bias introduced for SVOC concentrations from all factors. Due to
the effects of breakthrough, degradation, particle fractions and sampler
uptake periods, some current passive and active sampler config-
urations may underestimate certain SVOCs by 30−95%. We then recommend future study design, appropriateness of sampler
types for different study goals, and finally, how the SVOC community should move forward in both research and monitoring to
best achieve comparability and consistency in air measurements.

■ INTRODUCTION AND BACKGROUND

Recent scientific and policy discussion has addressed air
sampling of semivolatile organic compounds (SVOCs)/
persistent organic pollutants (POPs), particularly regarding
long-term monitoring networks and the data needs of the
SVOC community.1−3 SVOCs are organic compounds with
vapor pressures typically between ∼1 and 10−10 Pa. Many of the
SVOCs are classified as POPs, and are subject to international
agreements and conventions such as the Aarhus protocol on
POPs (1998) under the Convention on Long-Range Trans-
boundary Air Pollution (LRTAP, 1979) (http://www.unece.
org/env/lrtap/) and the Stockholm Convention (SC) on POPs
(2001/2004) (www.pops.int). These conventions aim to
eliminate or restrict the production and emission of a range
of SVOCs, notably organochlorine pesticides (OCPs),
polychlorinated biphenyls (PCBs), polychlorinated dibenzo-p-
dioxins and furans (PCDD/Fs), polybrominated diphenyl
ethers (PBDEs), polycyclic aromatic hydrocarbons (PAHs),
per- and polyfluorinated compounds (PFCs). To provide
information on long-term trends and effectiveness of regulatory
actions, air monitoring strategies have been established,1
including the European Monitoring and Evaluation Programme
(EMEP) for LRTAP compounds (1999) and the Global
Monitoring Plan (GMP) for SC compounds (2008). 1
Monitoring networks require appropriate sampling strategies
to provide reliable and broadly comparable air data and current
regional monitoring networks, with differences in sampling
methodology and procedures, cannot adequately provide the
data needed for global assessments, such as the GMP.1
Furthermore, scientific concern is also growing regarding
currently unregulated SVOCs such as endocrine disrupting
compounds (EDCs).4
Beyond the policy demands of international conventions, the
SVOC scientific community also requires air sampling to
understand the fate and transport of SVOCs and to provide
necessary information for scientific models and human
exposure assessments.2 This type of data is often produced
through smaller-scale case studies or local- or regional-scale
monitoring networks. While less standardized than long-term
monitoring networks, some very critical knowledge of SVOCs
has come from case studies, as they allow specific questions to
be addressed in a focused manner. In particular, they provide a
platform for testing samplers (e.g., polyurethane foam (PUF)
passive air samplers,5 XAD-passive air samplers6) and provide
data in vulnerable areas that are not covered by long-term
monitoring networks (e.g., e-waste sites in China7 or Africa,
intensive agricultural regions in India or South America8−10).
Received: May 2, 2014
Revised: September 23, 2014
Accepted: October 20, 2014
Published: October 20, 2014

© 2014 American Chemical Society 14077 dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
Figure 1. Global distribution of PBDE measurements in air (active and passive) from 2003 to 2013, as an example of the unbalanced spatial
distribution of SVOC sampling. Shading is used to aid in visualization of aggregated sampling distribution and frequency. Darker orange indicates a
high density of sampling (e.g., regular sampling over many years), while the lightest orange indicates collection of a single event sample (e.g., ship
cruises). Dots reflect global population density as of 2000 (http://sedac.ciesin.columbia.edu/gpw).
Together, monitoring networks and case studies provide our
information on the global distribution of SVOCs, their
atmospheric transport, fate, and relationship with sources.
However, there are still crucial gaps and biases in this
information which may lead to uncertainties in comparability
of data, namely spatial distribution and temporal coverage of
measurements, and lack of sampling standardization within and
between networks/case-studies. Problems related to spatial
distributions of established networks are discussed by Hung et
al.2 and Klánová and Harner.1 The uneven spatial distribution is
somewhat improved when one considers the contribution of
case studies (Supporting Information (SI) Figure S1b) and
new/planned monitoring networks in Australia, Southeast Asia,
and South America,2 but the issue of data comparability
remains. Additionally, there are discrepancies in the spatial
distribution of measurements and the spatial scale of
concentration variations; SVOCs vary on small scales11−13
and current sampling distributions infrequently address this.
There are also discrepancies in the frequency of sampling at
each site, with some regions having repeated continuous
monitoring capturing seasonal trends and short-term variability,
and others with only one or two reported measurements in the
past ten years (Figure 1). Although individual monitoring
networks are effective at accomplishing their respective goals
(e.g., providing data for modeling exercises, or broad spatial
coverage) the different network structures may lead to
problems with the intercomparability of scientific results.
With increasing international collaboration and data sharing
(e.g., the GMP) the question of inter-network comparability is
brought to the forefront.1 There are a wide range of techniques
used across different networks and case studies, including active
and passive air samplers, and differences in sampling volumes,
sorbents, sampling length and sampling frequencies. There has
never been an agreed mandate to establish standardized
procedures within and between SVOC networks and there is
a lack of quality standard operating procedures for SVOC
14078
Critical Review
sampling. While some studies have assessed comparability
between samplers or sampling networks,14−21 efforts never
went beyond these to address the potential implications. Most
current discussion focuses on the problems related to spatial
distributions, as well as analytical quality assurance and quality
control (e.g., interlaboratory or inter-network compari-
sons14,16,17,22,23), but little attention is given to sampling
artifacts or spatial intercomparability. Herein, we review and
discuss influences on the most frequently used SVOC air
sampling methods/techniques in both established networks and
case studies and analyze how they impact measurements and
thus data comparability. We then make recommendations on
what sampling techniques are appropriate for particular study
goals, how to enhance data comparability, and what is needed
to improve SVOC air measurements.
■ SAMPLING TECHNIQUES
Active Air Sampling Methods. Active air samplers (AAS)
are currently perceived as the most accurate method of
obtaining SVOC air concentrations as they accumulate both gas
and particle phase compounds under a controlled flow. Despite
identified sampling artifacts,24 this technique is almost
unchanged over the past 40 years.25−27
AAS are commonly classified as either high volume or low
volume samplers, with the main differences being the flow rate
and sample collection time. In high volume samplers, flow rates
are typically 15−80 m3/hour16,28,29 yielding total sample
volumes of >400 m3, although there is a large variation in
what is considered a high volume sample. For example, the
Integrated Atmospheric Deposition Network (IADN) typically
collects ∼820 m3 of air at American sites, and ∼350 m3 at
Canadian sites,16 while some Arctic sampling networks collect
up to 13000 m3.30 Low volume air samplers typically have flow
rates of <3 m3/hour.15,31−34 This often results in small sample
volumes (e.g., <200 m3), but low volume samplers can also be
used for longer deployment times, for example, 7−14 days of
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology Critical Review

Figure 2. Distribution of active air sampling volumes used in measurements of PBDEs from 2003 to 2013. Studies that did not report a sample
volume are excluded. There is a general trend of larger volumes collected in remote areas, particularly the Arctic, however there is also significant
variability in air volumes that is not directly related to the expected concentrations at a given site and this variability may affect what is reported as
concentrations in a given location (e.g., occurrence of either breakthrough or lack of detection).

Table 1. Common Active and Passive Sampling Techniques for SVOCs

sampler high volume sampler low volume sampler double-bowl PUF PAS XAD PAS SIP PAS
type of sampler active active passive passive passive
typical sampling 600 m3 100 m3 80−400 m3 100−200 m3 100−500 m3
volume (m3)
typical sampling 15−80 m3/hour <3 m3/hour 0.3−20 m3/day 0.4−0.8 m3/ 4−12 m3/day
rate/flow rate day
sampling time hours to month days to weeks weeks to months months to weeks to months
year
gas-phase sorbent PUF, XAD or combination PUF, XAD or PUF XAD XAD-impregnated PUF
of PUF, XAD combination of
PUF, XAD
particle-phase GFF or QFF GFF or QFF debatable. inconsistent. with lower no debatable. inconsistent. with lower
sampling sampling rate than gas-phase (10%). sampling rate than gas-phase (10%).
use in long-term yes (inter alia EMEP, IADN, rarely (selected yes (GAPS, MONET) yes (GAPS) yes (GAPS)
monitoring AMAP, MONAIRNET) OSPAR stations)
networks?

continuous sampling resulting in air volumes of 500−1000
m3.31−33 Low volume sampling is rarely used in long-term
sampling networks; only the Coordinated Atmospheric
Monitoring Programme of OSPAR and EMEP use low volume
samplers at selected stations.31,35
Environmental parameters and SVOC concentrations vary
widely, for example, very low concentrations in remote/polar
regions or high temperatures in tropical regions, and thus it can
be problematic to use identical sampling configurations in all
locations. The use of different sampling configurations results
in different sampling artifacts. Sampling events can only be
compared if a sufficient measure of accuracy has been reached.
Current sampling procedures/protocols have wide variations in
sampled volumes that may not always reflect the best choice for
the temperature and concentrations in a region, and rather may
simply rely on established sampling protocols not tailored to
specific sampling conditions (Figure 2). While these choices
allow individual networks to have long-term temporal data
continuity, they may reduce the accuracy of results and
comparability between networks or studies. A greater analysis
14079
of the influences of sample volume and sampling media is given
in the following sections.
Passive Air Sampling Methods. Passive air sampling
techniques for SVOCs were introduced in the end of 1990s to
simplify and reduce the costs for air monitoring of SVOCs.
Unlike AAS, passive air samplers (PAS) do not need electricity;
instead chemicals are trapped by diffusive uptake to a sorbent
material. PAS have enabled broader monitoring network
distribution and, as a result, increased the spatial breadth of
SVOC measurements.1,36 Several types of PAS have been
evaluated and used during the last 10−15 years, and a few of
them have received greater attention and been implemented in
monitoring networks (Table 1). Currently, the disk-shaped
PUF−PAS,5 and the XAD-resin based PAS6 are most used in
monitoring networks. Recently, an XAD sorbent-impregnated
PUF (SIP) PAS has also been implemented in the Global
Atmospheric Passive Sampling network (GAPS).37 Other PAS
samplers have been used (e.g., POGs,38 SPMDs39), and new
techniques are under development,40 however these are largely
used in case studies. The available PAS differ in characteristics
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
and applicability, which affects data comparability, as described
below. For an in-depth review of PAS theory and application
we recommend dedicated articles on theoretical framework,41
calibration,42 indoor use,43 new developments and techni-
ques.40,44
■ SAMPLING INFLUENCES
Below, we describe and discuss some of the major sampling
influences for both AAS and PAS. All air sampling techniques
have biases that may affect measurements and thereby also the
comparability of results between different sampling events. We
consider: (i) is the sampling technique accurate−does it
provide the real picture of atmospheric levels of the target
contaminants, (ii) is the sampling technique robust−how
sensitive is it to environmental conditions, and (iii) are data
from different techniques comparable−do we get the same
number when using different sampling techniques and different
sampling approaches? A focus is often on analytical procedures
and intercomparisons of chemical analyses;17,23 comparisons of
sampling procedures/methodologies/techniques are less com-
mon. Yet, given the range of sampling techniques, the
parameters of sample collection (e.g., sorbent type, sample
volumes) should be carefully considered, as they have the
potential to introduce significant biases. A summary of the
advantages and disadvantages of individual sampler config-
urations is given in SI Table S1.
Sampling InfluencesActive Air Sampling. Sampling
Head. The most commonly used sampling heads in AAS are
the TSP (total suspended particulates) and PM10 (particles
with aerodynamic diameter <10 μm). TSP theoretically
includes all airborne particulate matter, although in practice,
the upper particle size cutoff of a TSP sampling head is between
50 and 100 μm, dependent on flow rate and ambient
conditions.45 As a result of new legislations within the
European Union (EU), recently PM2.5 (particles <2.5 μm)
has also been suggested as a sampling threshold for monitoring
networks. Different sampling heads and a specific particle cut-
point can be useful, as particles <10 μm are most important for
atmospheric chemistry and physics46 while particles <2.5 μm
are most relevant for human health and risk assessment
estimates,47 but may result in lower data comparability. The
choice of sampling head should not result in large differences in
reported SVOC concentrations, as the majority of SVOCs are
typically associated with the finest particles (e.g., <1.5 μm),
which are sampled by the TSP, PM10 and PM2.5 sampling
heads.48−52 On average PM2.5 represents >75% of TSP by mass
and most of the SVOCs (>80%) are found in this fraction.51,52
However, there may be circumstances when the use of different
sampling heads leads to very different results, such as
environments with high TSP, or industrial settings where
SVOCs are found on coarser particles.
Sorbent and Filter Type. A range of different sorbent types
are used for sampling gas-phase SVOCs in AAS (e.g., PUF,
XAD, Tenax TA, polydimethylsiloxane), but published data on
SVOCs is dominated by sampling with PUF and XAD, and we
focus our discussion on these. PUF is a low density foam
consisting of a polymer bound by urethane links that has been
used in SVOC air sampling for over 40 years.25 XAD is a
styrene-divinylbenzene copolymer more recently used in AAS,
either independently or in conjunction with PUF. Two types of
XAD are commonly used in SVOC air sampling:53 (i)
Amberlite XAD-2, commonly used in AAS and XAD-
PAS,6,54,55 and (ii) Amberlite XAD-4, in active denuder
14080
Critical Review
samplers and SIP-PAS.56,57 The main difference between
XAD and PUF is sorptive capacity; for equal amounts of
sorbent material, PUF has a much lower capacity than XAD,
largely due to differences in surface area. PUF has a specific
surface area of 0.007−0.035 m2/g while the surface area of
XAD-2 is 300−600 m2/g.53 On a practical basis, this means that
samplers using XAD as the gas-phase sorbent can collect larger
volumes of air before breakthrough occurs (for AAS) or be
deployed for longer periods of time before equilibrium is
reached (for PAS). However, PUF does have the advantage of
lower cost and ease-of-use compared to XAD, as manipulation
of the PUF is simple and blanks are often lower.
There are also differences in the surface bonding of particular
compounds to PUF and XAD.58,59 XAD is a stronger sorbent
for PFCs, current-use pesticides (CUPs), and other emerging
chemicals,59,60 yet the differences are not fully understood, and
research is ongoing. Both sorbents are influenced by temper-
ature and humidity, but the amount of variability introduced by
these parameters is unclear and needs further research.53,59
Generally, they are linked to changes in sorptive capacities,59
and thus effects can be significant when large seasonal shifts are
expected (e.g., monsoon-impacted areas). The choice of PUF
vs XAD should be based on the choice of compound to be
sampled and the sampling parameters, but is often due to
previous regional practices (e.g., PUF is more commonly used
in AAS in Canada and Europe, while XAD is more commonly
used in the U.S.). PUF/XAD combinations are seeing
increasing global use in recent years (SI Figure S2).
Three types of filters are used in AAS for SVOCs: quartz
fiber filters (QFF), glass fiber filters (GFF), and Teflon/Teflon-
impregnated filters. GFF and QFF are most common, and are
perceived to give comparable results,61 although specific studies
assessing differences between them are limited. The recom-
mendations on filter type instead suggest that choice of filter
should consider what, if any, additional analyses will be
completed in conjunction with SVOCs (e.g., PM, SO2, organic/
elemental carbon, or metals).45,62
While the choice of sampling head determines the upper end
of the particle sizes collected on the filter, the lower size cutoff
is related to filter choice, but is more ambiguous. Typical GFFs
or QFFs have a collection efficiency of >99%, but do not often
have a specified fine particle cutoff. Particles <50 nm may pass
through the filter and subsequently be trapped by the gas-phase
sorbent material.63 This might be an issue for ultrafine particles
(typically <100 nm) which are of concern for human
health.64,65 Although numbers of ultrafine particles are high,
their total mass is small and therefore they are challenging to
quantify and require adaptations of current instrumentation or
new sampler types. There is limited information on how
ultrafine particles are distributed within an AAS, and this may
contribute to uncertainty in gas-particle partitioning.
Breakthrough. An important consideration in AAS is what
volume of air can be collected before the sampling medium
experiences “breakthrough”, that is, the loss of compounds
downstream of the sampling medium. This can be a result of a
saturated sampling medium or desorption of compounds from
the sampling medium. Sampling design must achieve the right
balance between collecting sufficient sample for analysis of
lower concentration SVOCs, while not collecting so much
volume as to have breakthrough of more volatile SVOCs. In
breakthrough estimates or controlled experiments <10%
breakthrough is considered an acceptable threshold.66−68
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
The key factors affecting breakthrough are (1) type and
geometry of sampling medium, (e.g., PUF vs XAD), (2) target
compound properties and concentration, (3) competition for
the sampling medium (i.e., from other gas-phase compounds),
(4) sample volume, (5) sampler flow rate, (6) temperature, and
(7) humidity. Of these, sampling medium, sample volume and
sampler flow rate are the most easily controlled, and therefore
breakthrough is most often discussed in terms of what sampling
medium and sample volume are appropriate for a given class of
compounds.
Basic chromatographic relationships can be used to estimate
breakthrough,24,67 but given the number of variables influencing
breakthrough, it can be more reliable to conduct small
breakthrough experiments with the exact study configurations.
Results from early breakthrough studies67,68 tend to be used as
an accepted framework for present studies and little
consideration is given to breakthrough in typical sampling
set-ups. Despite this, measurements and calculations suggest
that in a typical high-volume AAS breakthrough of more
volatile compounds (e.g., PCB-28, HCB, HCHs, fluorene) can
occur already at <600 m3.24,34,59,66,68−70 For example, at
tropical temperatures (e.g., around 30°C) a typical high volume
AAS (2 PUF plugs, 600 m3 of air, 24 h) can lose 10−15% of the
PCB-28 due to breakthrough (calculations in SI). However, it
should be noted that this is also highly temperature dependent:
with the same sampling configuration at 5°C the breakthrough
volume is >3000 m3. Many current AAS configurations may
experience breakthrough for lower molecular weight SVOCs,
leading to systematic underestimation of these compounds;
consideration of this is warranted in interpretation of results.
Filter Artifacts. The filter within the AAS (GFF or QFF) can
be the site of two sampling artifacts which influence measured
concentrations: blowoff and filter adsorption. Blowoff is the
volatilization loss of SVOCs from the filter, thereby
disproportionately reducing SVOC filter masses, while filter
adsorption is the adsorptive partitioning of gaseous compounds
onto the filter, increasing the mass of SVOCs on the filter.
This is affected by sampler configurations (e.g., flow rate,
sampling time, filter type) and occurs largely under three
conditions: (i) when there are large variations in the SVOC
concentration in incoming air, (ii) when there are large
temperature variations, (iii) when there is very high particle
loading to filters. Under these conditions, the gas-particle
distribution of SVOCs in the AAS may no longer be in
equilibrium with the influent air, and filter blow-off can occur in
the case of higher temperatures/lower concentrations, or filter
adsorption in the case of lower temperatures/higher concen-
trations. By using a higher flow rate or longer sampling time,
surface-sorbed SVOCs can more easily be stripped from
particles than those that are less available for exchange with the
atmosphere. In large sample volumes or in areas of high PM,
overloading of filters can cause further artifacts. If the filter
becomes physically blocked by PM, sampler flow rates can be
reduced and thus true sample volumes could be lower, resulting
in an underestimate of concentrations.
Filter adsorption and blowoff further vary by compound,
depending on partitioning coefficients, ionic interactions, and
hydrophilicity.71,72 For example, significant filter adsorption was
identified for perfluorooctanesulfonate (PFOS), and certain
perfluoroalkyl carboxylic acids (PFCAs) in a high-volume AAS,
resulting in overestimation of particle-associated fractions by up
to 80%, while other PFCs did not have significant filter
adsorption.24
14081
Critical Review
If filter blowoff/adsorption occurs, bulk air concentrations
may be correctly characterized, but gas-particle partitioning
estimates may be incorrect. This is expected to be most
significant for compounds with short-term temporal variations
in concentrations71 (such as CUPs) which may not be in
equilibrium between the gas and particle phases, and those with
intermediate gas-particle partitioning that vary between largely
gas- and largely particle-associated at ambient air temperatures.
Degradation. Degradation of atmospheric SVOCs occurs as
a natural process, caused by reaction with atmospheric reactive
species, in particular hydroxyl radicals, ozone, and NO3, or by
photolysis (SI Figure S3). Degradation has been found to
significantly affect ambient atmospheric concentrations, notably
for PAHs through reaction with OH, ozone, and NO3,73,74 for
PCBs and pesticides through reaction with OH,75−78 and for
PBDEs and novel flame retardants (NFRs) through photol-
ysis.79−81 While a sampler should provide a snapshot of the
atmospheric SVOC concentrations which may be already
affected by degradation, there is also the possibility for
degradation to continue within the sampler, as atmospheric
reactive species are drawn into the sampling medium along with
the SVOCs themselves. Within-sampler degradation is an
unwanted sampling artifact that can result in underestimation
of ambient concentrations.
UV radiation has a minimal direct influence on within-
sampler degradation due to the protection provided by the
sampler housings, but there is the potential for degradation due
to ozone, hydroxyl radicals and other atmospheric reactive
species to continue within the sampler. Significant within-
sampler degradation has been observed for PAHs both on
filters82,83 and in gas-phase sorbents.84 Particle-phase PAH
concentrations can typically be underestimated by 20−40% due
to degradative losses on AAS filters, with underestimates >70%
for particularly reactive compounds (e.g., benzo[a]-
pyrene).82,83,85 Gas-phase PAHs may also experience degrada-
tion within the sorbent, with losses of up to 50% due to
reaction with both OH and ozone.84,86
The amount of potential within-sampler degradation
depends on sampling conditions, site conditions, and
compound, but the range of variability due to these factors is
highly uncertain. For example, there are large spatial and
temporal ranges in levels of atmospheric reactive species. On a
global scale, ozone varies by a factor of 2−3, with higher levels
in populated areas,87 while OH varies by a factor of 6, with
higher levels in tropical regions,75,88 and there is potential for
even higher local-scale variability in both ozone and OH.89
Ozone can vary seasonally by a factor of 2 and diurnally to a
lesser extent87 while OH can vary up 10× both seasonally and
diurnally.75,88,90 Given the direct relationship between the
amount of degradative losses and levels of atmospheric reactive
species, within-sampler degradation may be significantly higher
(up to 100%) if the concentrations of atmospheric reactant
species are very high.84,85 This can then affect the temporal
(e.g., summer vs winter) and spatial comparability (urban vs
remote) of measured concentrations. For example, estimates
based on reaction with ozone74,91 suggest that measured
concentrations at an urban site in summer with 60 ppb ozone
would be underreported by 30% due to particle phase reaction
with ozone, while at a remote site in winter with 20 ppb ozone,
there would be minimal within-sampler degradation. While site
and season are very important for degradation, filter and
sorbent type, sampling duration, amount and composition of
PM, temperature and humidity also affect reactive losses.74 The
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
length of sampler deployment may have a minor influence on
degradation,74,85 but more characterization of these processes is
needed.
Experimental evidence only exists for PAHs, but within-
sampler degradation should also be of importance for other
reactive SVOCs, according to their susceptibility to degradation
in the atmosphere. Gas-phase compound degradation is largely
via reaction with the OH radical,73,86,92 and thus we expect
other gas-phase compounds with OH reaction rate constants in
the same range as PAHs to also have potential within-sampler
degradative losses (SI Table S2). In contrast, on-filter
degradation is largely through reaction with ozone,82,86 and
thus degradative losses of particle-associated compounds are
important mainly for PAHs (SI Table S3). Based on rate
constants and gas-particle partitioning, we estimate that PAHs,
CUPs and some NFRs may experience significant within-
sampler degradative losses.
Sampling InfluencesPassive Air Sampling. Despite
the benefits and widespread use of PAS1 they are still associated
with challenges and limitations affecting the comparability of
measurements. These are important to recognize and control,
but few of them can be fully quantified in a field deployment
setting.
Calculation of Air Concentrations. The most relevant
limitation of PAS is their lower accuracy compared to AAS.5
PAS have a semiquantitative nature, since air concentrations
can only be derived if an estimated sampling rate (m3/day) is
applied. The sampling rate is specific to the type of sampler
(e.g., sorbent, sampler housing) and the environmental
conditions (e.g., indoor, outdoor, meteorology), and can vary
by orders of magnitude due to these differences.6,93−96 In
theory sampling rates should not vary by compound, but in
practice large variations are found due to compound-specific
behaviors or uncertainties in the methods of determining
sampling rates. For example, measured sampling rates for
PUF−PAS span 3 orders of magnitude (e.g., 0.02−20 m3/
day),32,33,94,95,97−99 and some studies have observed systematic
variations in sampling rates according to compound physical-
chemical properties.32,33,93,100 The choice of using one general
sampling rate or nongeneric sampling rates is an important
current discussion point within the PAS community. The use of
nongeneric sampling rates (such as homologue-specific) can
“correct” for some of the potential errors a general sampling
rate may introduce under certain circumstances. For example,
using a general sampling rate of ∼4 m3/day36,101 vs a
nongeneric sampling rate of 0.3 m3/day for benzo[a]pyrene94
results in 13× concentration differences due to the choice of
sampling rate alone. However, a general sampling rate for all
SVOCs is often used in monitoring networks, as it is unclear
whether the corrective potential of nongeneric sampling rates is
significant considering the semiquantitative nature of PAS. It
may not be a problem when the goal is to study the sum of
compounds but may lead to errors if compound-specific
concentrations or the SVOC fingerprint is of interest. The PAS
community has not yet reached consensus regarding selection
of appropriate sampling rates.
The methods for obtaining sampling rates add an additional
level of uncertainty to their validity. Sampling rates are obtained
from three methods: (1) use of depuration compounds (DCs)/
performance reference compounds (PRCs),102,103 (2) calibra-
tion studies with comparison to (i) continuous low volume
AAS,31,33,104 or (ii) intermittent high volume AAS,18,19,59,105 or
(3) modeling exercises106,107 (e.g., applying a PUF-air
14082
Critical Review
partitioning coefficient). Method (i) is considered most
accurate, and is particularly advantageous when calibrating
compounds with high short-term variability, which could be
biased with intermittent high volume AAS depending on the
timing of sampling.31 DCs account for site-specific environ-
mental conditions but do not cover the full-range of
compounds, as they are not applicable for particle-associated
compounds, and cannot be used for XAD samplers due to their
high sorptive capacity.
PAS for SVOCs are used as time-integrated samplers, and
sampling rates are only applicable when the sampler is in the
linear uptake phase.41,95 The length of the linear uptake phase
varies for compounds within a SVOC class as well as between
classes and it is important to keep deployment times within the
reported linear time frames for each compound. While the
conventional PAS model assumes uniform distribution of a
compound within the sorbent,41,95 recent studies have
identified that kinetic resistance within the sorbent material
may be limiting and as such, model-based calculations may
overestimate the length of the linear uptake phase.59,100,108 This
contention is supported by experimental work.33,93 For
example, penta+hexachlorobenzene and low molecular weight
PAHs and PCBs enter a curvilinear uptake after 6−9 weeks,93
and thus air concentrations from a three-month PUF−PAS
deployment may be underestimates. Furthermore, the length of
the linear uptake phase may be affected by environmental
conditions such as temperature, wind speed and air
concentrations, but this is not yet fully characterized.
Sorbent Material. Two sorbent types are commonly used in
PAS: PUF and XAD. As with AAS, other sorbents have been
used but not applied on a large scale. Their specific
characteristics result in different sampling potentials. The
PUF sorbent has a lower capacity and higher sampling rate than
the XAD sorbent (Table 1).31 Typical deployment times in
established air monitoring networks are months (∼1−4
months) for PUF−PAS36,101,109 and up to one year for XAD-
PAS.6 As a consequence, data from the two samplers are only
comparable if sampling is conducted within the linear uptake
phase of each sampler, and time weighted average concen-
trations from multiple PUF−PAS cover the same sampling
period as XAD-PAS.
Another difference between the two sorbents is the type of
compounds they can sample. While XAD is solely a gas-phase
sampler, PUF is able to accumulate both gas-phase and particle-
associated SVOCs, although particles are sampled with a lower
accuracy and more variable sampling rates.93,97 On the other
hand, PUF−PAS have shown poorer or inconsistent perform-
ance for more volatile and polar SVOCs (e.g., CUPs, PFOS/
PFOA), for which XAD perform well.31,56,59,110 The low
sorptive capacity of PUF−PAS for most PFCs60,110 suggests
different partitioning mechanisms for polar and nonpolar
compounds. The SIP-PAS has been shown to be more effective
for PFCs,56 CUPs,37 methyl siloxanes,111 and phthalates.112
However, consistency with SIP-PAS may be challenging, as
XAD resin can be lost from the SIP disk during sampler
deployment, particularly in windy, outdoor conditions,113
thereby introducing added uncertainties.
PUF disks of different densities (0.021, 0.030, and 0.035 g
cm − 3 ) are currently used by diff erent research
groups.36,105,114,115 Although density is included in the
theoretical calculation of PUF-air partition coefficients, the
comparability of results from different densities is not fully
understood. A higher density PUF disk has higher capacity for
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
SVOCs and thus a longer linear uptake phase. However, kinetic
resistance within the PUF also differs with density. This can
result in different uptake scenarios, with lower density PUFs
having enhanced rates of SVOC transfer into the interior of the
disk, whereas higher density PUF has a rapidly filling “surface
compartment” and slower filling of the interior of the disk (e.g.,
a two-phase uptake mechanism).114 Additionally, particle
uptake may be enhanced in a lower-density PUF, which
could be a contributing factor in the discrepancies found for
particle associated PAHs (0.1 vs 5.0 m3/day) in different
studies.93,96
Sampler Housing. PAS sorbents are deployed inside a
sampler housing (chamber) to reduce the influence of
environmental factors that affect the performance of the PAS,
such as wind speed and UV radiation. Studies have indicated
that the PAS configuration, including the position of the
sampling media (both PUF and XAD) inside the housing as
well as the alignment of the housing (tight or open), affects
uptake/accumulation and sampling rates.108,116,117
The housings used for PUF−PAS consist of two stainless
steel bowls fixed on a common vertical axis around the PUF
disk. This type of housing may vary slightly in geometry and
deployment. Differences in the bowl geometries do not
significantly affect uptake114 but the way the housing is fixed
results in different uptake scenarios, especially under windy
conditions. A freely hanging sampler housing dampens the
variation in external wind speed and results in small variations
in sampling rates, while a fixed sampler results in up to 3×
higher sampling rates under windy conditions.115,118 The XAD-
PAS housing consists of two parts: an inner stainless steel mesh
tube in which the XAD is placed and an outer larger stainless
steel cylinder acting as a protective shelter.6 Differences in
housing configuration for XAD-PAS and air movement within
the sampler housing also cause significant variation in sampling
rates (up to 50%).108
Degradation. No study has investigated degradation in PAS
in detail. However, it is expected to affect the same compounds
and be related to the same environmental variables as in AAS.
Kennedy et al.119 identified photodegradation when PUF−PAS
were not shielded by a sampler housing and that this is reduced
by the double bowl chamber; typical PAS housings should limit
UV exposure. The long deployment times of PAS and thereby
long exposures to atmospheric reactive species may exacerbate
degradation, both of the SVOCs and the sorbent material itself.
The effect on the structure and uptake capacity of the sorbent is
unknown. Compound-specific degradation may also be a source
of variability in sampler calibrations.
Environmental Factors/Location. The main causes of
variability in PAS are environmental variables (e.g., wind
speed, temperature, air concentrations) and most refinements
to PAS aim to quantify/control these, through calibration of
sampling rates, sampler housing design, use of DCs, etc.
Despite these efforts, environmental factors continue to lead to
bias/errors in estimated air concentrations.
Sampler housings are effective up to wind speeds of 4 m/s as
they maintain the air flow within the chamber at less than ∼1
m/s, which has a minor effect on sampling rates.115 At wind
speeds >5 m/s the sampler housing cannot sufficiently dampen
outdoor winds, resulting in a higher air velocity within the
chamber and a rapid increase in sampling rates.115 This is
supported by results from DCs, which have shown much higher
sampling rates at windy, coastal, and mountain sites.36,99,101 For
XAD-PAS, sampling rates are also significantly higher under
14083
Critical Review
windy conditions.106,120 In contrast to the PUF−PAS, the
XAD-PAS sampling rates double between winds of 0 and 1 m/
s, but the effect on sampling rates is less with higher wind
speed.6,120
Temperature affects PUF and XAD-PAS in similar ways.
Temperature controls the gas-particle partitioning of SVOCs,
thus affecting the sampling rate (discussed below). However,
temperature also affects the sampling medium: for example,
higher temperature leads to higher diffusivity and thus higher
sampling rate95 and higher temperature leads to lower sorptive
capacity and thus shorter linear uptake phase. The effect of
these two factors on the overall sampling rate is complex and
varies by compound,106 but can lead to underestimation of
concentrations in warm/tropical conditions if deployment
times are the same as in colder/temperate regions. Further
complicating matters, variations in temperature also affect
ambient air concentrations (e.g., higher temperature leads to
higher volatilization and thereby higher air concentrations)
which impacts one of the major assumptions of the PAS theory,
that the air concentration is constant.
Particle-Associated Compounds. A big question when using
PAS for SVOCs is their performance for particle-associated
compounds. The XAD-PAS is considered a purely gas-phase
sampler while the PUF−PAS also accumulates particle-
associated compounds. Current findings are not consistent;
some suggest similar PUF−PAS sampling rates for gas- and
particle-phase compounds,33,96,121 others show poorer perform-
ance and up to 100× lower sampling rates for the particle-
associated compounds.93,105 The reason for inconsistent results
is not known but may be related to site characteristics, PUF
types, or analytical differences. High TSP values and high wind
speeds may favor the uptake of particle-associated compounds
to a level similar to gas-phase compounds. Another possible
reason may be differences in PUF density, as mentioned
above.93 The influence of particle-associated compounds on
overall sampling rates can be exacerbated at colder temper-
atures when a higher fraction of SVOCs are associated with
particles. The variable results imply large errors and low
comparability for particle-associated compounds between sites,
and it is difficult to draw a general conclusion that is applicable
to all or most of the sites in monitoring networks.
The ability to use PAS for particle-associated compounds is a
key need of the SVOC community, as regulated and emerging
SVOCs (PAHs, PBDEs, NFRs, PFCs, CUPs) tend to have high
particle fractions. The applicability of PAS for these compounds
is uncertain, and the subject of ongoing research.
■
CHALLENGES AND LIMITATIONS
Implications for Air Sampling. Given the aforementioned
sources of bias and potential error in AAS and PAS, it is clear
that each individual measurement is associated with some
degree of uncertainty which may be a bigger contribution than
the analytical uncertainty. However, more importantly, the
uncertainty, accuracy, and precision of the measured values vary
widely by sampling technique and sampler configuration, and
thus there may be poor comparability between different data
sets.
Key issues for comparability are (i) sampling times, sample
volumes and frequencies, (ii) PUF−PAS assumptions about
particles, (iii) PAS sampling rates, (iv) spatial coverage, and the
associated larger range of environmental variables as sampling is
expanded to more geographic regions.2
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
 Table 2. Percent of Ambient Air Concentration Captured by Given Sampler Configurationsa
% of air concentration captured by sampler
high-volume sampler PUF−PAS XAD-PAS
(24 h, 600 m3) (90 day deployment) (1 year deployment)
example vapor pressure (Pa) at log octanol-air partitioning coefficient atm. half-life (hrs) estimated particle fraction
compounds 0 °C, 25 °C (KOA) at 0 °C, 25 °C at 25 °C at 0 °C, 25 °C 0 °C 25 °C 0 °C 25 °C 0 °C 25 °C
PFOS 0.0973, 0.981 5.23, 4.84 917 0.00, 0.00 70% (1) 6% (1) 1% (3) 1% (3) 50% (3,4) 0% (3,4)
fluorene 0.630, 0.819 8.03, 6.74 15 0.01, 0.00 4% (1, 2) 3% (1, 2) 20% (2,3) 4% (2, 3) 100% 80% (3)
α-HCH 0.105, 0.376 8.37, 7.37 224 0.01, 0.00 70% (1) 20% (1) 60% (3) 20% (3) 100% 90% (3)
PCB-28 0.0314, 0.123 9.28, 8.06 108 0.03, 0.00 100% 100% (1) 90% (3) 50% (3) 100% 100%
chlorpyrifos 0.00399, 0.00536 9.56, 8.41 1.4 0.04, 0.01 0% (2) 0% (2) 80% (2,3) 60% (2, 3) 70% (2) 70% (2)
endosulfan 5.06 × 10−4, 10.58, 9.28 16 0.16, 0.03 90% (2) 90% (2) 80% (2, 3) 80% (2, 3) 80% (2) 80% (2)
2.31 × 10−5
Environmental Science & Technology

PBDE-47 4.91 × 10−7, 12.09, 10.54 128 0.64, 0.15 100% 100% 50% (4) 90% (4) 40% (4) 90% (4)
2.19 × 10−5
benzo[a]pyrene 0.00550, 6.26 × 10−6 12.29, 11.15 2.5 0.97, 0.76 60% 60% (2) 8% (2, 4) 20% (2, 4) 3% (4) 20% (4)
PBDE-209 1.01 × 10−8, 16.51, 15.26 3810 1.00, 1.00 100% 100% 10% (4) 10% (4) 0% (4) 0% (4)
2.23 × 10−4
a
The percentages represent the combined effects of breakthrough/degradation/exceedance of linear uptake phase at 0 and 25°C and 50 ppb ozone. Causes of the losses corresponding to each compound
are indicated by the numbers following the percentage. (1) indicates breakthrough, (2) is degradation, (3) is exceedance of the linear uptake phase, (4) is a high particle fraction. Full details on the
estimates are given in the SI. References: PCB, HCH, and PAH physical-chemical properties and temperature adjustments are from Beyer et al.131 and Paasivirta et al.;132 PBDE physical-chemical
properties and temperature adjustments are from Harner and Shoeib,133 Tittlemeier et al.,134 and Wang et al.;135 PFOS, chlorpyrifos and endosulfan physical-chemical properties and temperature
adjustments are from Odabasi and Cetin136 and EpiSuite.137 Particle-phase fractions were calculated using the methods of Harner and Bidleman126 using the temperature-adjusted KOA values.

14084
Breakthrough was calculated using the equations of Pankow.67 Length of the linear uptake phase was calculated using the equations of Shoeib and Harner.95 Degradation was estimated based on field study
data from Armstrong et al.,138 Schauer et al.,85 Tsapakis and Stephanou,84 Peters et al.,139 Menichini,82 and Goriaux et al.140 Full details on all assumptions/estimates are provided in the SI.
Critical Review

dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091

Environmental Science & Technology
For example, an intermittent AAS can easily be biased for
compounds with short-term temporal concentration variability
(e.g., CUPs, combustion-related SVOCs)31 and a long-term
time weighted average concentration from a PAS cannot
provide information on episodic events. If intermittent sample
timing results in sampling on the same weekday or period of a
month, this can be biased by daily, weekly or monthly activity
patterns (e.g., weekday traffic). This can lead to an over- or
underestimation of concentrations if the sample either
coincides with or completely misses a high concentration
event. Furthermore, as we have moved to measurement of
emerging compounds and lower concentration ranges (pg/m3
and fg/m3 levels), sample collection volumes have been pushed
higher in order to achieve detection, particularly in remote
environments, without thoroughly assessing the associated
breakthrough potentials.
As described earlier, there is much uncertainty in how
particles behave in PUF−PAS. Comparability issues are
introduced because users of PUF−PAS make different
assumptions about the fraction of the particle phase sampled
by the PUF. The assumptions range from treating PUF−PAS as
gas-phase only to assuming 1−10% of particles are
sampled,93,105 to assuming all particles are sampled.96 If
PUF−PAS are treated as bulk samplers when they are only
sampling a fraction of particles, measured values may be biased
low, particularly for higher molecular weight SVOCs.
The systematic variations in PAS sampling rates by physical-
chemical properties identified in some studies,32,33,93,100 reflect
situations where the conventional PAS framework does not
apply, such as when volatile compounds enter the curvilinear
uptake phase, or when only a fraction of atmospheric particles
are sampled. Thus, the use of a general sampling rate may
introduce bias in either the low molecular weight or high
molecular weight compounds and different compound/
congener distributions.33 It is also not well quantified how
much sampling rates differ between different climate zones/
seasons, thus considering the global range of average
temperatures (SI Figure S4), large biases may be introduced
when comparing PAS measurements from very different
regions or seasons.
Spatial distributions of sampling networks further bias our
global knowledge of air concentrations.1,2 While coverage is
somewhat improved when one also considers case studies,
there continue to be areas of the world with limited
measurement coverage (Figure 1). For example, if one
considers available PBDE data for the period 2003−2013,
only 2% of the global land mass is within 100 km of a reported
PBDE air concentration, and 0.1% is within 20 km. By
population, only 30% of people live within 100 km of a
reported PBDE measurement, and only 6% within 20 km.
Southeast Asia has some coverage for PBDEs measured by case
study, but these largely consist of one-time measurements,
thereby limiting temporal information. Additionally, land-based
southern hemisphere sampling is largely PAS (SI Figure S1),
furthering the spatial bias.
Examples of Bias. To demonstrate the potential combined
effects of the aforementioned sampling artifacts and biases, we
estimated the potential errors in measured concentrations for
three typical sampler configurations: (1) a high-volume active
air sampler with GFF or QFF and two PUF plugs, collecting
∼600 m3 of air over 24 h, (2) a PUF−PAS, deployed outdoors
for 90 days, and (3) an XAD-PAS, deployed outdoors for one
year, and two temperature scenarios: 0 and 25 °C. Estimates are
14085
Critical Review
summarized in Table 2 and details are given in SI Table S4.
These estimates have significant uncertainty, due to the very
different data sources, but they emphasize the large possibility
for sampling artifacts to influence measured concentrations.
The estimates suggest combined effects of breakthrough
and/or exceedance of the PAS linear uptake phase lead to large
underestimates in the reported values of the volatile SVOCs,
notably PCB-28, α-HCH, fluorene, PFOS, and by extension,
other SVOCs with similar physical-chemical properties. This
effect also influences XAD-PAS for more volatile SVOCs (e.g.,
α-HCH, fluorene, PFOS) at higher temperatures.59,106 Within-
sampler degradation, although not well-characterized, is also
estimated to contribute to large losses, particularly for fluorene,
benzo[a]pyrene, and chlorpyrifos. Similar degradation is
expected for gas-phase compounds with short reactive
atmospheric half-lives, which includes all PAHs, many CUPs
and some NFRs (SI Table S2). The particle-associated SVOCs
are under-sampled by PUF−PAS and not sampled by XAD-
PAS. Estimates in Table 2 assume PUF−PAS collect 10% of the
total particle fraction, but this may range from 1 to 100%. Thus,
reported concentrations may vary between 15 and 100% of the
actual concentration, depending on the fraction of particles that
is truly collected by the PUF−PAS.
Attention is frequently given to issues of analytical
comparability, but sampling comparability affects many
compounds which we assume have minimal analytical
uncertainties. For example, BDE-209, which has frequently
been identified to have difficulties in laboratory analysis,122 is
“correctly” sampled, for example, 100% of the air concentration
should be captured by a typical high volume air sampler,
whereas PCB-28, a compound that has been consistently
included in global monitoring networks and is one of the seven
typical indicator PCBs, may experience losses of up to 15%
based on breakthrough alone, and thus reported values may be
biased low in many situations.
Problematic Chemicals. Many “emerging” SVOCs have
different physical-chemical properties compared to the legacy
SVOCs and this may introduce challenges in sampling. Current
sampling networks and techniques were developed for legacy
SVOCs and may not be appropriate for providing representa-
tive data on these newer SVOCs. For example, the use of PAS
in the future may be more challenging because target
compounds are either too volatile (volatile methylsiloxanes,
phthalates) or too involatile (CUPs, NFRs) and thus have high
uncertainty in conventional PAS. As a consequence of
regulatory actions, newer compounds are often designed to
be less persistent in the environment, but as a result are also
less persistent in samplers (leading to more within-sampler
degradation) and thus more difficult to correctly quantify.
Degradation is also of concern when measuring nitro- and
other substituted PAHs,123 and existing data on these
compounds is likely an underestimate. The lack of data on
degradation and the potential importance for many emerging
compounds highlights the need for more studies.
Furthermore, we must be aware that chemicals currently
measured by the SVOC community are a small fraction of the
chemicals in use that may be of environmental and/or human
health concern.124 Adding more chemicals to the lists of
analytes without adapting current sampler configurations will
introduce more uncertainty in measured concentrations.
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
■ RECOMMENDATIONS AND FUTURE
DEVELOPMENTS
Recommendations to Enhance Comparability. Given
the wide range of potential biases and uncertainties in air
sampling identified above, it is clear that careful consideration
must be given to sampling approaches. This consideration
should first consider the study goals and outcomes. For
example, what sampling designs are best suited to address
global distributions vs data for human exposure assessments vs
source identification, long-term vs short-term trends, and the
target SVOCs? The sampler type, frequency, and sampling
materials should correspond to the specific question and
compounds to be addressed. But ultimately, as the GMP and
other initiatives aggregate international data, study design
should also aim for comparability with global data sets.
Overall, PAS are best for broad spatial coverage and long-
term temporal trends, while AAS are best for short-term
temporal trends and coverage of a broader set of compounds.
When the intended result of a study is long-term integrated
average concentrations (e.g., monthly to yearly scale), and
particularly for largely gas-phase compounds, PAS have a clear
advantage. The simplicity of PAS (no electricity, cheaper) make
them best suited to resolve concentration gradients where levels
vary by orders of magnitude. AAS are recommended when
absolute concentrations are needed, smaller concentration
variations are expected or high time resolution is needed.
AAS are the only choice when gas-particle partitioning is
important or particle-phase concentrations are expected to be
dominant and when short-term trends (hourly to weekly) are
of interest. Because congener/compound ratios can be biased
in PAS, AAS are best for source identification/fingerprinting
techniques. Regulations for human exposure assessment are
often based on particle-associated SVOCs and the best sampler
choice is a low volume AAS providing continuous gas- and
particle-phase concentrations, while PAS are not as well suited.
However, current research continues to demonstrate the broad
applicability of PAS,96,97 and, although they have higher
uncertainty, particularly for particle-phase compounds, the
uncertainty in a temporally integrated PAS could be lower than
uncertainty in intermittent AAS for compounds with daily
concentration variations.
Sorbent choice should be considered with respect to the
length of sampling period and target compounds. In PAS, PUF
is recommended for capturing seasonal trends for stable, legacy
compounds, while XAD-PAS are recommended for longer term
(e.g., yearly, decadal trends)31 and also to cover emerging gas-
phase SVOCs (more polar compounds). In AAS, for best
applicability to a broad range of SVOCs and conditions, we
recommend a combination of PUF and XAD, as this reduces
breakthrough concerns and allows quantification of both polar
and nonpolar SVOCs. While PUF and XAD are the current
focus of the majority of the SVOC community, new sorbents
should be investigated for their applicability to a wider range of
compounds.
Although there is a need for standardization in SVOC air
measurements, with large regional differences and equipment
constraints, implementation of standardization can be challeng-
ing. This could be addressed in the long term through the
development of international standards for SVOC sampling
(e.g., an ISO standard), but in the short term resetting of all
sampling campaigns and networks to the same configurations is
not practical, nor necessarily useful. Many networks have long-
14086
Critical Review
term established techniques and continuation of the same
techniques allows for temporal comparisons. Furthermore,
universal standardization of techniques would not be
appropriate for all regions. For example, sampling in high
temperatures tropical regions may require a different sampling
approach to low temperature polar regions. However, when
different techniques are used it is important for the SVOC
community to understand the biases from techniques/
configurations/approaches. Instead of implementing identical
sampling techniques, data from different sampling configu-
rations may be adjusted using standardization factors. Stand-
ardization factors can account for effects of breakthrough,
temperature differences, wind, etc., and can be tailored to
specific sites/sampler types/seasons/sampling approaches. For
example, if degradation was better characterized, a stand-
ardization factor could be used to account for the under-
estimation caused at sites with high levels of reactive trace
gases. For cases where large sample volumes are needed,
standardization factors can be used to account for losses of the
more volatile SVOCs. In fact, this is already done to a certain
extent when compound-specific sampling rates are used for
PAS; these rates account for nonlinear uptake of volatile
SVOCs and lower sampling of particle-associated SVOCs.
One way standardization factors can be determined,
particularly for long-term monitoring networks, is at “super-
sites” where samplers and sampling approaches from different
networks are used simultaneously.125 Quality assurance
techniques such as spiking samplers with labeled compounds
should also be used periodically across different networks to
further evaluate comparability.
However, standardization factors, once developed, should
not be treated as a perfect solution. Better consideration of
study design should be given in advance of sampler
deployments. There is sufficient theoretical and experimental
knowledge to allow researchers to estimate PAS linear uptake
phases,41,95 gas-particle partitioning of target compounds,126,127
and AAS breakthrough67 in advance of sampler deployment. A
necessary part of sampling campaign should be estimating these
effects and structuring the sampling campaign to reduce these
artifacts. New monitoring networks and case studies should aim
to harmonize, where possible, with existing networks to
maximize data comparability.
Recommendations for Future Work. Extensive exper-
imental, modeling and calibration work has and continues to be
done on PAS techniques with the goal of improving the
accuracy of measured concentrations and better understanding
uptake processes. This should be continued. However, limited
efforts have been made to improve AAS, despite knowledge of
uncertainties from degradation, breakthrough, etc. and this
should be an additional focus.
For PAS, there are some obvious improvements needed,
particularly in the determination of sampling rates. Currently,
sampling rates are determined either with calibration studies or
DCs, but sampling rates could be improved by combining these
two methods. Sampling rates (determined from calibration
studies) could be adjusted according to differential losses of
DCs between sampling sites to account for site-specific
influences. Further studies should also focus on understanding
the performance of PUF−PAS for particle-associated SVOCs
under different conditions and for different PUF densities, to
better understand why current studies obtain inconsistent
results.
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
For AAS there are also clear improvements needed,
particularly in the study of breakthrough and degradation.
Current sampling configurations are based on information from
breakthrough studies from 1980s and early 1990s24,67,128 when
different SVOCs were the primary concerns and concentrations
were often higher. Today, larger volumes are used to detect
compounds at the pg/m3 and fg/m3 level, particularly in low
concentration regions (e.g., polar regions, Figure 2) without a
good understanding of the associated impacts (breakthrough,
degradation). Since the use of denuders in conjunction with
AAS has yielded important information regarding the
degradation of PAHs, this avenue should be further investigated
to better understand within sampler degradation for PAHs and
other reactive SVOCs. As the relationship between ozone levels
and degradation appears relatively consistent,85 it may be
possible to estimate SVOC losses based on ozone levels, which
could be especially important in locations/times of year when
losses are expected to be large. Additionally, we should
reconsider the paradigm that AAS provide more accurate
results. We have summarized numerous well-known sampling
artifacts associated with PAS and AAS, but due to limited
experimental evidence, the question of the overall effect of
these artifacts on the accuracy of reported concentrations
remains a rough estimate (e.g., Table 2). The question of the
overall significance of these artifacts on PAS vs AAS could be
addressed with a well-designed PAS-AAS comparison.
In addition, as sampling expands to broader global coverage,
a more thorough understanding of the effect of environmental
conditions/meteorology on both AAS and PAS is needed.
Currently, sampler evaluation and testing is largely performed
in temperate regions; this may limit applicability in more
extreme environments.
There is also room for innovation in sampling techniques.
For example, source regions and the influence of specific air
masses can be assessed using new directional AAS129 and
PAS.130
Finally, the SVOC community must be progressive and
adaptable. How can sampling campaigns be designed to
produce more comparable data without compromising the
data needs of individual networks and case studies? What
compounds are most important to continue monitoring, and
what compounds should be added to global monitoring? We
have established that there are big challenges in comparability
between well-known SVOCs such as PAHs, PCBs, and PBDEs.
These must be addressed, but the SVOC community should
also move forward with air sampling methods that can meet the
future challenges of new chemicals. We must also be prepared
to address the future global developments, as the key questions
in SVOCs shift, e.g., responding to climate change-induced
meteorological shifts and new source balances due to melting of
glaciers and polar ice caps1 and the relationship between
SVOCs and human health impacts, including consideration of
possible synergistic effects.3
■
ASSOCIATED CONTENT
*
S Supporting Information
Maps of global distributions of sampling networks and case
studies, description of atmospheric degradation processes,
atmospheric reactive half-lives for SVOCs, and details of
calculations for Table 2. This material is available free of charge
via the Internet at http://pubs.acs.org.
14087
Critical Review
■ AUTHOR INFORMATION
Corresponding Authors
*(L.M.) Phone: +420 54949 5149; fax: +420 54949 2840; e-
mail: melymuk@recetox.muni.cz.
*(P.B.) Phone: +420 54949 5149; fax: +420 54949 2840; e-
mail: bohlin@recetox.muni.cz.
Notes
The authors declare no competing financial interest.
■ ACKNOWLEDGMENTS
This research was supported by the projects “Employment of
Best Young Scientists for International Cooperation Empower-
ment” (CZ.1.07/2.3.00/30.0037) and “Employment of Newly
Graduated Doctors of Science for Scientific Excellence”
(CZ.1.07/2.3.00/30.009) cofinanced from European Social
Fund and the state budget of the Czech Republic. The
RECETOX research infrastructure was supported by the
projects of the Ministry of Education of the Czech Republic
(LM2011028) and (LO1214). We thank Jana Borůvková and
Roman Prokeš for assistance.
■
REFERENCES
(1) Klánová, J.; Harner, T. The challenge of producing reliable results
under highly variable conditions and the role of passive air samplers in
the Global Monitoring Plan. Trends Anal. Chem. 2013, 46, 139−149.
(2) Hung, H.; MacLeod, M.; Guardans, R.; Scheringer, M.; Barra, R.;
Harner, T.; Zhang, G. Toward the next generation of air quality
monitoring: Persistent organic pollutants. Atmos. Environ. 2013, 80,
591−598.
(3) Klánová, J.; Diamond, M. L.; Jones, K. C.; Lammel, G.; Lohmann,
R.; Pirrone, N.; Scheringer, M.; Balducci, C.; Bidleman, T. F.; Bláha,
K.; et al. Identifying the research and infrastructure needs for the
global assessment of hazardous chemicals ten years after establishing
the Stockholm Convention. Environ. Sci. Technol. 2011, 45, 7617−
7619.
(4) Bergman, A.; Heindel, J. J.; Jobling, S.; Kidd, K. A.; Zoeller, R. T.
State of the Science of Endocrine Disrupting Chemicals 2012; Bergman,
A., Heindel, J. J., Jobling, S., Kidd, K. A., Zoeller, R. T., Eds.; World
Health Organization: Geneva, 2012.
(5) Harner, T.; Shoeib, M.; Diamond, M. L.; Ikonomou, M. G.;
Stern, G. A. Passive sampler derived air concentrations of PBDEs
along an urban-rural transect: Spatial and temporal trends. Chemo-
sphere 2006, 64, 262−267.
(6) Wania, F.; Shen, L.; Lei, Y. D.; Teixiera, C. F.; Muir, D. C. G.
Development and calibration of a Resin-Based Passive Sampling
System for Monitoring Persistent Organic Pollutants in the
Atmosphere. Environ. Sci. Technol. 2003, 37, 1352−1359.
(7) Zhang, T.; Huang, Y.; Chen, S.-J.; Liu, A.-M.; Xu, P.-J.; Li, N.; Qi,
L.; Ren, Y.; Zhou, Z.-G.; Mai, B.-X. PCDD/Fs, PBDD/Fs, and PBDEs
in the air of an e-waste recycling area (Taizhou) in China: Current
levels, composition profiles, and potential cancer risks. J. Environ.
Monit. 2012, 14, 3156−3163.
(8) Sharma, B. M.; Bharat, G. K.; Tayal, S.; Nizzetto, L.; Cupr, P.;
Larssen, T. Environment and human exposure to persistent organic
pollutants (POPs) in India: A systematic review of recent and
historical data. Environ. Int. 2014, 66, 48−64.
(9) Leyk, S.; Binder, C. R.; Nuckols, J. R. Spatial modeling of
personalized exposure dynamics: The case of pesticide use in small-
scale agricultural production landscapes of the developing world. Int. J.
Health Geogr. 2009, 8, 17.
(10) Pozo, K.; Harner, T.; Lee, S. C.; Sinha, R. K.; Sengupta, B.;
Loewen, M.; Geethalakshmi, V.; Kannan, K.; Volpi, V. Assessing
seasonal and spatial trends of persistent organic pollutants (POPs) in
Indian agricultural regions using PUF disk passive air samplers.
Environ. Pollut. 2011, 159, 646−653.
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
(11) Melymuk, L.; Robson, M.; Helm, P. A.; Diamond, M. L. PCBs,
PBDEs, and PAHs in Toronto air: Spatial and seasonal trends and
implications for contaminant transport. Sci. Total Environ. 2012, 429,
272−280.
(12) Hearn, L. K.; Kennedy, K. E.; Hawker, D. W.; Toms, L.-M. L.;
Alberts, V.; Müller, J. F. Spatial mapping of city-wide PBDE levels
using an exponential decay model. J. Environ. Monit. 2012, 14, 643−
650.
(13) Persoon, C.; Peters, T. M.; Kumar, N.; Hornbuckle, K. C.
Spatial distribution of airborne polychlorinated biphenyls in Cleveland,
Ohio and Chicago, Illinois. Environ. Sci. Technol. 2010, 44, 2797−2802.
(14) Simcik, M. F. Air monitoring of persistent organic pollutants in
the Great Lakes: IADN vs. AEOLOS. Environ. Monit. Assess. 2005, 100,
201−216.
(15) Blanchard, P.; Kallweit, D.; Brice, K. A.; Froude, F. A.; Chan, C.
H.; Neilson, M.; Holz, J.; Millat, H. A comparison of European and
North American atmospheric deposition networks: Polycyclic aromatic
hydrocarbons and lindane. J. Environ. Monit. 2006, 8, 465.
(16) Wu, R.; Backus, S. M.; Basu, I.; Blanchard, P.; Brice, K. A.;
Dryfhout-Clark, H.; Fowlie, P.; Hulting, M.; Hites, R. A. Findings from
quality assurance activities in the Integrated Atmospheric Deposition
Network. J. Environ. Monit. 2009, 11, 277−296.
(17) Schlabach, M.; Farag-clement, R.; Hung, H.; Kallenborn, R.
AMAP/EMEP/NCP Inter-Laboratory study for POP Analysis, EMEP/
CCC-Report 7/2011; Norwegian Institute for Air Research: Kjeller,
Norway, 2010.
(18) Gouin, T.; Harner, T.; Blanchard, P.; Mackay, D. Passive and
active air samplers as complementary methods for investigating
persistent organic pollutants in the Great Lakes Basin. Environ. Sci.
Technol. 2005, 39, 9115−9122.
(19) Klánová, J.; Kohoutek, J.; Hamplová, L.; Urbanová, P.;
Holoubek, I. Passive air sampler as a tool for long-term air pollution
monitoring: Part 1. Performance assessment for seasonal and spatial
variations. Environ. Pollut. 2006, 144, 393−405.
(20) Halse, A. K.; Schlabach, M.; Sweetman, A. J.; Jones, K. C.;
Breivik, K. Using passive air samplers to assess local sources versus
long range atmospheric transport of POPs. J. Environ. Monit. 2012, 14,
2580−2590.
(21) Dreyer, A.; Shoeib, M.; Fiedler, S.; Barber, J.; Harner, T.;
Schramm, K.-W.; Jones, K. C.; Ebinghaus, R. Field intercomparison on
the determination of volatile and semivolatile polyfluorinated
compounds in air. Environ. Chem. 2010, 7, 350.
(22) Brandsma, S. H.; de Boer, J.; Leonards, P. E. G.; Cofino, W. P.;
Covaci, A. Organophosphorus flame-retardant and plasticizer analysis,
including recommendations from the first worldwide interlaboratory
study. TrAC Trends Anal. Chem. 2013, 43, 217−228.
(23) Su, Y.; Hung, H. Inter-laboratory comparison study on
measuring semi-volatile organic chemicals in standards and air
samples. Environ. Pollut. 2010, 158, 3365−3371.
(24) Bidleman, T. F.; Simon, C. G.; Burdick, N. F.; You, F.
Theoretical plate measurements and collection efficiencies for high-
volume air samplers using polyurethane foam. J. Chromatogr. 1984,
301, 448−453.
(25) Bidleman, T. F.; Olney, C. E. High-volume collection of
atmospheric polychlorinated biphenyls. Bull. Environ. Contam. Toxicol.
1974, 11, 442−450.
(26) Orgill, M. M.; Sehmel, G. A.; Petersen, M. R. Some initial
measurements of airborne DDT over Pacific Northwest forests. Atmos.
Environ. 1976, 10, 827−834.
(27) Simon, C. G.; Bidleman, T. F. Sampling airborne PCB with
polyurethane foamChromatographic approach to determination of
collection effeciencies. Abstr. Pap. Am. Chem. Soc. 1978, 176, 187.
(28) Aas, W.; Breivik, K. Heavy Metals and POP Measurements, 2010,
EMEP Co-operative Programme for Monitoring and Evaluation of the
Long-Range Transmission of Air Pollutants in Europe, EMEP/CCC-
Report 3/2012; Norwegian Institute for Air Research: Kjeller, Norway,
2011.
(29) Hung, H.; Kallenborn, R.; Breivik, K.; Su, Y.; Brorström-
Lundén, E.; Olafsdottir, K.; Thorlacius, J. M.; Leppänen, S.; Bossi, R.;
14088
Critical Review
Skov, H.; et al. Atmospheric monitoring of organic pollutants in the
Arctic under the Arctic Monitoring and Assessment Programme
(AMAP): 1993−2006. Sci. Total Environ. 2010, 408, 2854−2873.
(30) Su, Y.; Hung, H.; Sverko, E.; Fellin, P.; Li, H. Multi-year
measurements of polybrominated diphenyl ethers (PBDEs) in the
Arctic atmosphere. Atmos. Environ. 2007, 41, 8725−8735.
(31) Hayward, S. J.; Gouin, T.; Wania, F. Comparison of four active
and passive sampling techniques for pesticides in air. Environ. Sci.
Technol. 2010, 44, 3410−3416.
(32) Hazrati, S.; Harrad, S. J. Calibration of polyurethane foam
(PUF) disk passive air samplers for quantitative measurement of
polychlorinated biphenyls (PCBs) and polybrominated diphenyl
ethers (PBDEs): Factors influencing sampling rates. Chemosphere
2007, 67, 448−455.
(33) Melymuk, L.; Robson, M.; Helm, P. A.; Diamond, M. L.
Evaluation of passive air sampler calibrations: Selection of sampling
rates and implications for the measurement of persistent organic
pollutants in air. Atmos. Environ. 2011, 45, 1867−1875.
(34) Batterman, S. A.; Chen, T.-C.; Chernyak, S. M.; Godwin, C.
Design and performance evaluation of a medium flow sampler for
airborne brominated flame retardants (BFRs). J. Environ. Monit. 2009,
11, 858−866.
(35) Aas, W.; Breivik, K. Heavy metals and POP measurements 2011;
EMEP Co-operative Programme for Monitoring and Evaluation of the
Long-Range Transmission of Air Pollutants in Europe. Norwegian
Institute for Air Research. Kjeller, Norway, 2011; EMEP/CCC-Report
4/2013.
(36) Pozo, K.; Harner, T.; Wania, F.; Muir, D. C. G.; Jones, K. C.;
Barrie, L. A. Toward a global network for persistent organic pollutants
in air: Results from the GAPS study. Environ. Sci. Technol. 2006, 40,
4867−4873.
(37) Koblizková, M.; Genualdi, S.; Lee, S. C.; Harner, T. Application
of sorbent impregnated polyurethane foam (SIP) disk passive air
samplers for investigating organochlorine pesticides and polybromi-
nated diphenyl ethers at the global scale. Environ. Sci. Technol. 2012,
46, 391−396.
(38) Farrar, N. J.; Prevedouros, K.; Harner, T.; Sweetman, A. J.;
Jones, K. C. Continental scale passive air sampling of persistent
organic pollutants using rapidly equilibrating thin films (POGs).
Environ. Pollut. 2006, 144, 423−433.
(39) Schuster, J. K.; Gioia, R.; Breivik, K.; Steinnes, E.; Scheringer,
M.; Jones, K. C. Trends in European background air reflect reductions
in primary emissions of PCBs and PBDEs. Environ. Sci. Technol. 2010,
44, 6760−6766.
(40) Mayer, P.; Wania, F.; Wong, C. S. Advancing passive sampling
of contaminants in environmental science. Environ. Sci. Process. Impacts
2014, 16, 366−368.
(41) Bartkow, M. E.; Booij, K.; Kennedy, K. E.; Müller, J. F.; Hawker,
D. W. Passive air sampling theory for semivolatile organic compounds.
Chemosphere 2005, 60, 170−176.
(42) Tuduri, L.; Millet, M.; Briand, O.; Montury, M. Passive air
sampling of semi-volatile organic compounds. Trends Anal. Chem.
2012, 31, 38−49.
(43) Bohlin, P.; Jones, K. C.; Strandberg, B. Occupational and indoor
air exposure to persistent organic pollutants: A review of passive
sampling techniques and needs. J. Environ. Monit. 2007, 9, 501−509.
(44) Esteve-Turrillas, F. A.; Yusà, V.; Pastor, A.; de la Guardia, M.
New perspectives in the use of semipermeable membrane devices as
passive samplers. Talanta 2008, 74, 443−457.
(45) Berg, T.; Dye, C.; Hanssen, J. E.; Krognes, T.; Munthe, J.;
Reissell, A.; Schaug, J.; Schmidbauer, N.; Semb, A.; Torseth, K.; et al.
EMEP Manual for Sampling and Analysis; Norwegian Institute for Air
Research: Kjeller, Norway, 2001.
(46) Finlayson-Pitts, B. J.; Pitts, J. N. J. Chapter 9: Particles in the
troposphere. In Chemistry of the Upper and Lower Atmosphere; Elsevier
Inc.: San Diego, 1999; pp 349−435.
(47) Englert, N. Fine particles and human health–a review of
epidemiological studies. Toxicol. Lett. 2004, 149, 235−242.
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
(48) Lammel, G.; Klánová, J.; Ilić, P.; Kohoutek, J.; Gasić, B.;
Kovacić, I.; Škrdlíková, L. Polycyclic aromatic hydrocarbons in air on
small spatial and temporal scales − II. Mass size distributions and gas-
particle partitioning. Atmos. Environ. 2010, 44, 5022−5027.
(49) Zhang, B.-Z.; Zhang, K.; Li, S.-M.; Wong, C. S.; Zeng, E. Y. Size-
dependent dry deposition of airborne polybrominated diphenyl ethers
in urban Guangzhou, China. Environ. Sci. Technol. 2012, 46, 7207−
7214.
(50) Chrysikou, L. P.; Samara, C. A. Seasonal variation of the size
distribution of urban particulate matter and associated organic
pollutants in the ambient air. Atmos. Environ. 2009, 43, 4557−4569.
(51) Degrendele, C.; Okonski, K.; Melymuk, L.; Landlová, L.;
Kukučka, P.; Cupr, P.; Klánová, J. Size specific distribution of the
atmospheric particulate PCDD/Fs, dl-PCBs and PAHs on a seasonal
scale: Implications for cancer risks from inhalation. Atmos. Environ.
2014, in press.
(52) Okonski, K.; Degrendele, C.; Melymuk, L.; Landlová, L.;
Kukučka, P.; Vojta, Š.; Kohoutek, J.; Klánová, J. The value of size-
segregated data: Investigation of the size-fraction distribution of flame
retardants on atmospheric aerosols. Submitted.
(53) Lee, J. J.; Huang, K.-L.; Yu, Y. Y.; Chen, M. S. Laboratory
retention of vapor-phase PAHs using XAD adsorbents. Atmos. Environ.
2004, 38, 6185−6193.
(54) Barthel, P.; Thuens, S.; Shunthirasingham, C.; Westgate, J. N.;
Wania, F.; Radke, M. Application of XAD-resin based passive air
samplers to assess local (roadside) and regional patterns of persistent
organic pollutants. Environ. Pollut. 2012, 166, 218−225.
(55) Sun, P.; Basu, I.; Hites, R. A. Temporal trends of
polychlorinated biphenyls in precipitation and air at chicago. Environ.
Sci. Technol. 2006, 40, 1178−1183.
(56) Genualdi, S.; Lee, S. C.; Shoeib, M.; Gawor, A.; Ahrens, L.;
Harner, T. Global pilot study of legacy and emergin persistent organic
pollutants using sorbent-impregnated polyurethane foam disk passive
air samplers. Environ. Sci. Technol. 2010, 44, 5534−5539.
(57) Ahrens, L.; Harner, T.; Shoeib, M.; Lane, D. A.; Murphy, J. G.
Improved characterization of gas-particle partitioning for per- and
polyfluoroalkyl substances in the atmosphere using annular diffusion
denuder samplers. Environ. Sci. Technol. 2012, 46, 7199−7206.
(58) Kamprad, I.; Goss, K.-U. Systematic investigation of the
sorption properties of polyurethane foams for organic vapors. Anal.
Chem. 2007, 79, 4222−4227.
(59) Hayward, S. J.; Lei, Y. D.; Wania, F. Sorption of a diverse set of
organic chemical vapors onto XAD-2 resin: Measurement, prediction
and implications for air sampling. Atmos. Environ. 2011, 45, 296−302.
(60) Shoeib, M.; Harner, T.; Lee, S. C.; Lane, D. A.; Zhu, J. Sorbent-
impregnated polyurethane foam disk for passive air sampling of
volatile fluorinated chemicals. Anal. Chem. 2008, 80, 675−682.
(61) Clark, C. R.; Truex, T. J.; Lee, F. S. C.; Salmeen, I. T. Influence
of sampling filter type on the mutagenicity of diesel exhaust particulate
extracts. Atmos. Environ. 1981, 15, 397−402.
(62) Brown, A. S.; Yardley, R. E.; Quincey, P. G.; Butterfield, D. M.
Studies of the effect of humidity and other factors on some different
filter materials used for gravimetric measurements of ambient
particulate matter. Atmos. Environ. 2006, 40, 4670−4678.
(63) Lee, D. S.; Nicholson, K. W. The measurement of atmospheric
concentrations and deposition of semi-volatile organic compounds.
Environ. Monit. Assess. 1994, 32, 59−91.
(64) Ruuskanen, J.; Tuch, T.; Ten Brink, H.; Peters, A.; Khlystov, A.;
Mirme, A.; Kos, G. P. A.; Brunekreef, B.; Wichmann, H. E.; Buzorius,
G.; et al. Concentrations of ultrafine, fine and PM2.5 particles in three
European cities. Atmos. Environ. 2001, 35, 3729−3738.
(65) Cabada, J. C.; Rees, S.; Takahama, S.; Khlystov, A.; Pandis, S.
N.; Davidson, C. I.; Robinson, A. L. Mass size distributions and size
resolved chemical composition of fine particulate matter at the
Pittsburgh supersite. Atmos. Environ. 2004, 38, 3127−3141.
(66) Hites, R. A.; Basu, I.; Venier, M.; Salamova, A.; Nettesheim, T.;
Blanchard, P.; Backus, S. M.; Brice, K. A.; Froude, F. A.; Galarneau, E.
Technical Summary of the Progress of the Integrated Atmospheric
14089
Critical Review
Deposition Network (IADN)2002−2008; Environment Canada,
2008; p 117.
(67) Pankow, J. F. Overview of the gas phase retention volume
behavior of organic compounds on polyurethane foam. Atmos. Environ.
1989, 23, 1107−1111.
(68) You, F.; Bidleman, T. F. Influence of volatility on the collection
of polycyclic aromatic hydrocarbon PAH vapors with polyurethane
foam. Environ. Sci. Technol. 1984, 18, 330−333.
(69) Harrad, S. J.; Hazrati, S.; Ibarra, C. Concentrations of
polychlorinated biphenyls in indoor air and polybrominated diphenyl
ethers in indoor air and dust in Birmingham, United Kingdom:
Implications for human exposure. Environ. Sci. Technol. 2006, 40,
4633−4638.
(70) Ter Schure, A. F. H.; Larsson, P.; Agrell, C.; Boon, J. P.
Atmospheric transport of polybrominated diphenyl ethers and
polychlorinated biphenyls to the Baltic Sea. Environ. Sci. Technol.
2004, 38, 1282−1287.
(71) Cotham, W. E.; Bidleman, T. F. Laboratory investigations of the
partitioning of organochlorine compounds between the gas phase and
atmospheric aerosols on glass fiber filters. Environ. Sci. Technol. 1992,
26, 469−478.
(72) Pankow, J. F.; Bidleman, T. F. Effects of temperature, TSP and
per cent nonexchangeable material in determining the gas particle
partitioning of organic compounds. Atmos. Environ. 1991, 25, 2241−
2249.
(73) Bunce, N. J.; Dryfhout, H. G. Diurnal and seasonal modelling of
the tropospheric half-lives of polycyclic aromatic hydrocarbons. Can. J.
Chem. 1992, 70, 1966−1970.
(74) Keyte, I. J.; Harrison, R. M.; Lammel, G. Chemical reactivity and
long-range transport potential of polycyclic aromatic hydrocarbons
A review. Chem. Soc. Rev. 2013, 42, 9333−9391.
(75) Mandalakis, M.; Berresheim, H.; Stephanou, E. G. Direct
evidence for destruction of polychlorobiphenyls by OH radicals in the
subtropical troposphere. Environ. Sci. Technol. 2003, 37, 542−547.
(76) Li, R.; Jin, J. Model predictions of toxaphene degradation in the
atmosphere over North America. Environ. Toxicol. Chem. 2013, 32,
2663−2671.
(77) Al Rashidi, M.; Chakir, A.; Roth, E. Heterogeneous oxidation of
folpet and dimethomorph by OH radicals: A kinetic and mechanistic
study. Atmos. Environ. 2014, 82, 164−171.
(78) Bai, J.; Yang, W.; Zhang, C.; Zhao, Y.; Gong, C.; Sun, X.; Zhang,
Q.; Wang, W. Theoretical study on the OH-initiated atmospheric
reaction of 1,1,1-trichloro-2,2-bis(4-chlorophenyl) ethane (DDT).
Atmos. Environ. 2013, 67, 177−183.
(79) Schenker, U.; Soltermann, F.; Scheringer, M.; Hungerbühler, K.
Modeling the environmental fate of polybrominated diphenyl ethers
(PBDEs): The importance of photolysis for the formation of lighter
PBDEs. Environ. Sci. Technol. 2008, 42, 9244−9249.
(80) Stapleton, H. M.; Dodder, N. G. Photodegradation of
decabromodiphenyl ether in house dust by natural sunlight. Environ.
Toxicol. Chem. 2008, 27, 306−312.
(81) Wang, J.; Chen, S.-J.; Nie, X.; Tian, M.; Luo, X.; An, T.-C.; Mai,
B. Photolytic degradation of decabromodiphenyl ethane (DBDPE).
Chemosphere 2012, 89, 844−849.
(82) Menichini, E. On-filter degradation of particle-bound benzo-
[a]pyrene by ozone during air sampling: A review of the experimental
evidence of an artefact. Chemosphere 2009, 77, 1275−1284.
(83) Liu, K.; Duan, F.; He, K.; Ma, Y.; Cheng, Y. Investigation on
sampling artifacts of particle associated PAHs using ozone denuder
systems. Front. Environ. Sci. Eng. 2013, 8, 284−292.
(84) Tsapakis, M.; Stephanou, E. G. Collection of gas and particle
semi-volatile organic compounds: Use of an oxidant denuder to
minimize polycyclic aromatic hydrocarbons degradation during high-
volume air sampling. Atmos. Environ. 2003, 37, 4935−4944.
(85) Schauer, C.; Niessner, R.; Pöschl, U. Polycyclic aromatic
hydrocarbons in urban air particulate matter: Decadal and seasonal
trends, chemical degradation, and sampling artifacts. Environ. Sci.
Technol. 2003, 37, 2861−2868.
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology
(86) Finlayson-Pitts, B. J.; Pitts, J. N. J. Chapter 10: Airborne
polycyclic aromatic hydrocarbons and their derivatives. In Chemistry of
the Upper and Lower Atmosphere; Elsevier Inc.: San Diego, 1999; pp
436−546.
(87) Fowler, D.; Amann, M.; Anderson, R.; Ashmore, M. R.;
Depledge, M.; Derwent, D.; Grennfelt, P.; Hewitt, C. N.; Hov, O.;
Jenkin, M.; et al. Ground-level ozone in the 21st century: Future
trends, impacts and policy implications Science Policy. Sci. Policy Rep. -
R. Soc. 2008, 15/08.
(88) Bahm, K.; Khalil, M. A. K. A new model of tropospheric
hydroxyl radical concentrations. Chemosphere 2004, 54, 143−166.
(89) Finlayson-Pitts, B. J.; Pitts, J. N. J. Chemistry of the Upper and
Lower Atmosphere; Elsevier Inc.: San Diego, 1999.
(90) Bloss, W. J.; Evans, M. J.; Lee, J. D.; Sommariva, R.; Heard, D.
E.; Pilling, M. J. The oxidative capacity of the troposphere: Coupling
of field measurements of OH and a global chemistry transport model.
Faraday Discuss. 2005, 130, 425.
(91) Perraudin, E.; Budzinski, H.; Villenave, E. Kinetic study of the
reactions of ozone with polycyclic aromatic hydrocarbons adsorbed on
atmospheric model particles. J. Atmos. Chem. 2006, 56, 57−82.
(92) Atkinson, R. Atmospheric lifetimes of dibenzo-para-dioxins and
dibenzofurans. Sci. Total Environ. 1991, 104, 17−33.
(93) Bohlin, P.; Audy, O.; Škrdlíková, L.; Kukučka, P.; Vojta, Š.;
Přibylová, P.; Prokeš, R.; Č upr, P.; Klánová, J. Evaluation and
guidelines for using polyurethane foam (PUF) passive air samplers in
double-dome chambers to assess semi volatile organic compounds in
non-industrial indoor environments. Environ. Sci. Process. Impacts 2014,
Accepted, DOI: 10.1039/C4EM00305E.
(94) Bohlin, P.; Audy, O.; Škrdlíková, L.; Kukučka, P.; Přibylová, P.;
Prokeš, R.; Vojta, Š.; Klánová, J. Outdoor passive air monitoring of
semi volatile organic compounds (SVOCs): A critical evaluation of
performance and limitations of polyurethane foam (PUF) disks.
Environ. Sci. Process. Impacts 2014, 16, 433−444.
(95) Shoeib, M.; Harner, T. Characterization and comparison of
three passive air samplers for persistent organic pollutants. Environ. Sci.
Technol. 2002, 36, 4142−4151.
(96) Harner, T.; Su, K.; Genualdi, S.; Karpowicz, J.; Ahrens, L.;
Mihele, C.; Schuster, J. K.; Charland, J.-P.; Narayan, J. Calibration and
application of PUF disk passive air samplers for tracking polycyclic
aromatic compounds (PACs). Atmos. Environ. 2013, 75, 123−128.
(97) Heo, J.; Lee, G. Field-measured uptake rates of PCDDs/Fs and
dl-PCBs using PUF-disk passive air samplers in Gyeonggi-do, South
Korea. Sci. Total Environ. 2014, DOI: 10.1016/j.scitotenv.2014.03.073.
(98) Mari, M.; Schuhmacher, M.; Feliubadaló, J.; Domingo, J. L. Air
concentrations of PCDD/Fs, PCBs and PCNs using active and passive
air samplers. Chemosphere 2008, 70, 1637−1643.
(99) Pozo, K.; Harner, T.; Shoeib, M.; Urrutia, R.; Barra, R.; Parra,
O.; Focardi, S. Passive-sampler derived air concentrations of persistent
organic pollutants on a north-south transect in Chile. Environ. Sci.
Technol. 2004, 38, 6529−6537.
(100) Zhang, X.; Tsurukawa, M.; Nakano, T.; Lei, Y. D.; Wania, F.
Sampling medium side resistance to uptake of semivolatile organic
compounds in passive air samplers. Environ. Sci. Technol. 2011, 45,
10509−10515.
(101) Pozo, K.; Harner, T.; Lee, S. C.; Wania, F.; Muir, D. C. G.;
Jones, K. C. Seasonally resolved concentrations of persistent organic
pollutants in the global atmosphere from the first year of the GAPS
study. Environ. Sci. Technol. 2009, 43, 796−803.
(102) Persoon, C.; Hornbuckle, K. C. Calculation of passive sampling
rates from both native PCBs and depuration compounds in indoor and
outdoor environments. Chemosphere 2009, 74, 917−923.
(103) Santiago, E. C.; Cayetano, M. G. Polycyclic aromatic
hydrocarbons in ambient air in the Philippines derived from passive
sampler with polyurethane foam disk. Atmos. Environ. 2007, 41, 4138−
4147.
(104) Wilford, B. H.; Harner, T.; Zhu, J.; Shoeib, M.; Jones, K. C.
Passive sampling survey of polybrominated diphenyl ether flame
retardants in indoor and outdoor air in Ottawa, Canada: Implications
for sources and exposure. Environ. Sci. Technol. 2004, 38, 5312−5318.
14090
Critical Review
(105) Klánová, J.; Cupr, P.; Kohoutek, J.; Harner, T. Assessing the
influence of meteorological parameters on the performance of
polyurethane foam-based passive air samplers. Environ. Sci. Technol.
2008, 42, 550−555.
(106) Armitage, J. M.; Hayward, S. J.; Wania, F. Modeling the uptake
of neutral organic chemicals on XAD passive air samplers under
variable temperatures, external wind speeds and ambient air
concentrations (PAS-SIM). Environ. Sci. Technol. 2013, 47, 13546−
13554.
(107) Petrich, N. T.; Spak, S. N.; Carmichael, G. R.; Hu, D.;
Martinez, A.; Hornbuckle, K. C. Simulating and explaining passive air
sampling rates for semivolatile compounds on polyurethane foam
passive samplers. Environ. Sci. Technol. 2013, 47, 8591−8598.
(108) Zhang, X.; Wong, C.; Lei, Y. D.; Wania, F. Influence of sampler
configuration on the uptake kinetics of a passive air sampler. Environ.
Sci. Technol. 2012, 46, 397−403.
(109) Dušek, L.; Borůvková, J.; Hůlek, R.; Jarkovský, J.; Kubásek, M.;
Gregor, J.; Šebková, K.; Hřebíček, J.; Holoubek, I. GENASISGlobal
environmental assessment and information system. In Environmental
Software Systems; Springer, 2011; online.
(110) Ahrens, L.; Shoeib, M.; Harner, T.; Koblizková, M.; Reiner, E.
J. Characterization of two passive air samplers for per- and
polyfluoroalkyl substances. Organohalogen Compd. 2012, 74, 339−342.
(111) Genualdi, S.; Harner, T.; Cheng, Y.; Macleod, M.; Hansen, K.
M.; van Egmond, R.; Shoeib, M.; Lee, S. C. Global distribution of
linear and cyclic volatile methyl siloxanes in air. Environ. Sci. Technol.
2011, 45, 3349−3354.
(112) Saini, A.; Chaudhuri, S. R.; Goosey, E.; Diamond, M. L.;
Shoeib, M. Calibration of passive air sampler for monitoring phthalates
in indoor air; Society for Environmental Toxicology and Chemistry
European Annual Meeting; SETAC: Glasgow, 2013.
(113) Robson, M. Personal communication; Toronto, Ontario,
Canada, 2014; [email communication with L. Melymuk, 21−4−2014].
(114) Chaemfa, C.; Barber, J. L.; Gocht, T.; Harner, T.; Holoubek, I.;
Klánová, J.; Jones, K. C. Field calibration of polyurethane foam (PUF)
disk passive air samplers for PCBs and OC pesticides. Environ. Pollut.
2008, 156, 1290−1297.
(115) Tuduri, L.; Harner, T.; Hung, H. Polyurethane foam (PUF)
disks passive air samplers: Wind effect on sampling rates. Environ.
Pollut. 2006, 144, 377−383.
(116) Tao, S.; Liu, Y.; Xu, W.; Lang, C.; Liu, S.; Dou, H.; Liu, W.
Calibration of a passive sampler for both gaseous and particulate phase
polycyclic aromatic hydrocarbons. Environ. Sci. Technol. 2007, 41,
568−573.
(117) Abdallah, M. A.-E.; Harrad, S. J. Modification and calibration of
a passive air sampler for monitoring vapor and particulate phase
brominated flame retardants in indoor air: Application to car interiors.
Environ. Sci. Technol. 2010, 44, 3059−3065.
(118) May, A. A.; Ashman, P.; Huang, J.; Dhaniyala, S.; Holsen, T.
M. Evaluation of the polyurethane foam (PUF) disk passive air
sampler: Computational modeling and experimental measurements.
Atmos. Environ. 2011, 45, 4354−4359.
(119) Kennedy, K. E.; Hawker, D. W.; Bartkow, M. E.; Carter, S.;
Ishikawa, Y.; Müller, J. F. The potential effect of differential ambient
and deployment chamber temperatures on PRC derived sampling
rates with polyurethane foam (PUF) passive air samplers. Environ.
Pollut. 2010, 158, 142−147.
(120) Zhang, X.; Brown, T. N.; Ansari, A.; Yeun, B.; Kitaoka, K.;
Kondo, A.; Lei, Y. D.; Wania, F. Effect of wind on the chemical uptake
kinetics of a passive air sampler. Environ. Sci. Technol. 2013, 47, 7868−
7875.
(121) Bohlin, P.; Jones, K. C.; Levin, J.-O.; Lindahl, R.; Strandberg,
B. Field evaluation of a passive personal air sampler for screening of
PAH exposure in workplaces. J. Environ. Monit. 2010, 12, 1437−1444.
(122) Covaci, A.; Voorspoels, S.; de Boer, J. Determination of
brominated flame retardants, with emphasis on polybrominated
diphenyl ethers (PBDEs) in environmental and human samplesA
review. Environ. Int. 2003, 29, 735−756.
dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091
Environmental Science & Technology Critical Review

(123) Albinet, A.; Leoz-Garziandia, E.; Budzinski, H.; ViIlenave, E.

Sampling precautions for the measurement of nitrated polycyclic
aromatic hydrocarbons in ambient air. Atmos. Environ. 2007, 41,
4988−4994.
(124) Muir, D. C. G.; Howard, P. H. Are there other persistent
organic pollutants? A challenge for environmental chemists. Environ.
Sci. Technol. 2006, 40, 7157−7166.
(125) Cape, J. N.; Bealey, W.; Coyle, M.; DiMarco, C.; Kentisbeer, J.;
Leaver, D.; Malcolm, H.; Nemitz, E.; Simmons, I.; Twigg, M.; et al.
Operation of EMEP “Supersites” in the United KingdomAnnual Report
for 2008; Centre for Ecology and Hydrology: Edinburgh, 2010; 112 p.
(126) Harner, T.; Bidleman, T. F. Octanol-air partition coefficient for
describing particle/gas partitioning of aromatic compounds in urban
air. Environ. Sci. Technol. 1998, 32, 1494−1502.
(127) Pankow, J. An absorption model of gas/particle partitioning of
organic compounds in the atmosphere. Atmos. Environ. 1994, 28, 185−
188.
(128) Hart, K. M.; Isabelle, L. M.; Pankow, J. F. High-volume air
sampler for particle and gas sampling. 1. Design and gas sampling
performance. Environ. Sci. Technol. 1992, 26, 1048−1052.
(129) Research Centre for Toxic Compounds in the Environment.
MonAirNet Project webpage. http://www.monairnet.eu/index-en.php.
(130) Ferranti, E. J. S.; Fryer, M.; Sweetman, A. J.; Solera Garcia, M.
A.; Timmis, R. J. Field-testing a new directional passive air sampler for
fugitive dust in a complex industrial source environment. Environ. Sci.
Process. Impacts 2014, 16, 159−168.
(131) Beyer, A.; Wania, F.; Gouin, T.; Mackay, D.; Matthies, M.
Selecting internally consistent physicochemical properties of organic
compounds. Environ. Toxicol. Chem. 2002, 21, 941−953.
(132) Paasivirta, J.; Sinkkonen, S. I.; Mikkelson, P.; Rantio, T.;
Wania, F. Estimation of vapor pressures, solubilities and Henry’s law
constants of selected persistent organic pollutants as functions of
temperature. Chemosphere 1999, 39, 811−832.
(133) Harner, T.; Shoeib, M. Measurements of octanol-air partition
coefficients (KOA) for polybrominated diphenyl ethers (PBDEs):
Predicting partitioning in the environment. J. Chem. Eng. Data 2002,
47, 228−232.
(134) Tittlemier, S. A.; Halldorson, T.; Stern, G. A.; Tomy, G. T.
Vapor pressures, aqueous solubilities, and Henry’s law constants of
some brominated flame retardants. Environ. Toxicol. Chem. 2002, 21,
1804−1810.
(135) Wang, Z.-Y.; Zeng, X.-L.; Zhai, Z.-C. Prediction of supercooled
liquid vapor pressures and n-octanol/air partition coefficients for
polybrominated diphenyl ethers by means of molecular descriptors
from DFT method. Sci. Total Environ. 2008, 389, 296−305.
(136) Odabasi, M.; Cetin, B. Determination of octanol-air partition
coefficients of organochlorine pesticides (OCPs) as a function of
temperature: Application to air-soil exchange. J. Environ. Manage.
2012, 113, 432−439.
(137) U.S. Environmental Protection Agency. Estimation Programs
Interface Suite (EPI Suite) for Microsoft Windows, v 4.11, 2012.
(138) Armstrong, J. L.; Fenske, R. A.; Yost, M. G.; Tchong-French,
M.; Yu, J. Comparison of polyurethane foam and XAD-2 sampling
matrices to measure airborne organophosphorus pesticides and their
oxygen analogs in an agricultural community. Chemosphere 2013, 92,
451−457.
(139) Peters, A. J.; Lane, D. A.; Gundel, L. A.; Northcott, G. L.;
Jones, K. C. A comparison of high volume and diffusion denuder
samplers for measuring semivolatile organic compounds in the
atmosphere. Environ. Sci. Technol. 2000, 34, 5001−5006.
(140) Goriaux, M.; Jourdain, B.; Temime, B.; Besombes, J.-L.;
Marchand, N.; Albinet, A.; Leoz-Garziandia, E.; Wortham, H. Field
comparison of particulate PAH measurements using a low-flow
denuder device and conventional sampling systems. Environ. Sci.
Technol. 2006, 40, 6398−6404.

14091 dx.doi.org/10.1021/es502164r | Environ. Sci. Technol. 2014, 48, 14077−14091

ANNEX 7

List of free online GIS data sources.

Annex 7

List of free online GIS data sources

Webpage Type of data Coverage Provider
http://www.diva-gis.org/climate climate World Database of of climate Data
http://www.gadm.org/ administrative boundaries World Database of Global Administrative Areas
http://www.diva-gis.org/gdata data by country World Database of Global Adminstrative Areas
http://srtm.csi.cgiar.org/ elevation World The CGIAR Consortium for Spatial Information
http://gcmd.gsfc.nasa.gov/ climate change World Global Change Master Directory, National Aeronautics and Space Administration
https://water.usgs.gov/maps.html hydrology USA US Geological Survey
http://hydrosheds.cr.usgs.gov/dataavail.php hydrology USA US Geological Survey
http://www.bafg.de/GRDC/EN hydrology World Global Runoff Data Centre
http://download.geofabrik.de/ human geography World Geofabrik - OpenStreetMap Data Extracts
http://www.eea.europa.eu/data-and-maps/ environment EU European Envirinmental Agency
http://geoportal.icimod.org/ physical geography Himalayas The International Centre for Integrated Mountain Development
http://www.cgiar-csi.org/data climate World The CGIAR Consortium for Spatial Information
http://www.webgis.com/ land Use, elevation World Lakes Enviornmental Software
http://gdem.ersdac.jspacesystems.or.jp/ elevation World Japan Space Systems
http://asterweb.jpl.nasa.gov/gdem.asp elevation World Jet Propulsion Laboratory, National Aeronautics and Space Administration
http://www.viewfinderpanoramas.org/dem3.html elevation World Database of Digital Elevation Data
http://esdac.jrc.ec.europa.eu/ soils EU Joint Research Centre, European Comission
http://www.edenextdata.com/?q=data human and physical geography World EDENext Data Portal
http://www.naturalearthdata.com/ human and physical geography World Natural Earth
http://www.iscgm.org/gm/ human and physical geography World International Steering Committee for Global Mapping
http://sedac.ciesin.columbia.edu/data/sets/browse human and physical geography World Socioeconomic Data and Applications Center, National Aeronautics and Space Administration
http://www.fao.org/geonetwork/srv/en/ human and physical geography World GeoNetwork, Food and Agriculture Organization
http://www.arcdata.cz/produkty/geograficka-data/arccr-500 human and physical geography Czech Republic ArcData Praha
http://www.dibavod.cz/ hydrology Czech Republic DIBAVOD, T. G. Masaryk Water Research Institute
http://geoportal.gov.cz/ human and physical geography Czech Republic Czech Environmental Information Agency

Other webpages offering links on another GIS data sources

Webpage Provider
http://freegisdata.rtwilson.com/ Free GIS Data webpage
http://www.edenextdata.com/?q=content/global-gis-datasets-links-0 EDENext Data Portal
https://lib.stanford.edu/gis-branner-library/gis-data-websites-test-panel Stanford University Libraries
http://www.gisinecology.com/gis_data_sources.htm GSI in Ecology webpage
http://www.terrainmap.com/rm39.html Terrain Mapping Journal for the Eartha and Beoynd
http://gisgeography.com/ GISGeography webpage