See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/249920452

SAGA: a decision support system for air pollution management around a coal-
fired power plant

Article  in  International Journal of Environment and Pollution · July 2009

DOI: 10.1504/IJEP.2009.027275

CITATIONS READS

5 134

4 authors, including:

Jose A Souto-Gonzalez J.J. Casares Long

University of Santiago de Compostela University of Santiago de Compostela
132 PUBLICATIONS   573 CITATIONS    131 PUBLICATIONS   968 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Anel Hernandez PhD Thesis View project

All content following this page was uploaded by Jose A Souto-Gonzalez on 11 November 2014.

The user has requested enhancement of the downloaded file.

444 Int. J. Environment and Pollution, Vol. 38, No. 4, 2009

SAGA: a decision support system for air pollution

management around a coal-fired power plant

José A. Souto*, Marcos Hermida

and Juan J. Casares
Department of Chemical Engineering,
School of Engineering,
University of Santiago de Compostela,
15782 Santiago de Compostela, Spain
Fax: +34 981 528 050
E-mail: ja.souto@usc.es
E-mail: Marcos.Hermida@meteogalicia.es
E-mail: eqjjcl@usc.es
*Corresponding author

José L. Bermúdez
Section of Environment,
As Pontes Power Plant,
Endesa Generación, S.A.,
As Pontes de García Rodríguez, Spain
E-mail: jbermudez@endesa.es

Abstract: Air quality models are currently feasible approaches to prevent air
pollution episodes. From one of the first source-oriented modelling approaches
for air pollution forecasting (Souto et al., 1994, 1996, 1998), a new decision
support system for air quality management, SAGA, was developed to provide
support to As Pontes Power Plant (APPP) staff. SAGA can provide air
pollution forecasts and manage meteorological and air quality measurements.
Power plant decisions are supported by the results of a non-hydrostatic
meteorological model (ARPS, Xue et al., 2001) to produce Meteorological
Forecasts (MFs), and to be coupled to different Lagrangian dispersion models.

Keywords: decision support software; air quality; MF; meteorological forecast;

power plant; industrial plant operation.

Reference to this paper should be made as follows: Souto, J.A., Hermida, M.,
Casares, J.J. and Bermúdez, J.L. (2009) ‘SAGA: a decision support system for
air pollution management around a coal-fired power plant’, Int. J. Environment
and Pollution, Vol. 38, No. 4, pp.444–461.

Biographical notes: José A. Souto is a Lecturer of Chemical Engineering at

the University of Santiago de Compostela. He got his PhD in Chemistry
(Industry) in 1993 at the University of Santiago de Compostela. After that, he
worked for seven years at the Supercomputing Centre of Galicia, as Software
and Projects Manager, including environmental research. In 2000, he joined the
University of Santiago de Compostela as full-time Professor. In the last

Copyright © 2009 Inderscience Enterprises Ltd.

 SAGA: a decision support system for air pollution management 445

17 years, his main research work has covered meteorological and atmospheric
dispersion modelling, photochemical air pollution modelling and solar radiation
studies.

Marcos Hermida got an MSc in Mathematics from the University of Santiago

de Compostela and a BSc in Computer Sciences from the University of
La Coruña, Spain. Currently, he is the web and graphical software manager of
the Galician Weather Service, MeteoGalicia. His research expertise includes
photochemical modelling and software tools for environmental information.

Juan J. Casares has been Professor of Chemical Engineering at the University

of Santiago de Compostela since 1993. He got his PhD in Chemistry in 1976,
University of Barcelona, Spain, and his PhD in Chemical Engineering in 1978,
Salford University, UK. His main positions have been: Assistant Professor at
the University of Petroleum and Minerals (KSA), Head of Results Analysis
(As Pontes Power Plant), Director of the Supercomputing Centre of Galicia,
Director of the Environmental Information and Technology Centre and,
currently, Chairman of the Department of Chemical Engineering, University of
Santiago de Compostela. His current research covers emissions inventories
(PRTR) for environmental risk assessment and air quality modelling.

José L. Bermúdez is Head of the Environment at the As Pontes Power Plant.

He got an MSc in Chemistry at the University of Santiago de Compostela.
His expertise includes emissions and air quality monitoring, environmental
management, environmental quality control systems and applied research in
clean combustion technologies.

1 Introduction

As Pontes Power Plant (APPP) is a 1400 MW coal-fired power plant located in Galicia in
Northwestern Spain. Before December 2007, the plant burns a mixture of local lignite
(with up to 2% sulphur) and imported sub-bituminous coal (with less than 0.1% sulphur).
The combustion gases are emitted through a 350 m stack, to prevent local air pollution
episodes. However, a complex terrain environment (see Figure 1) with sea-influence
from the Atlantic Ocean can produce meteorological conditions favourable to poor local
air quality.
The development of operational air quality forecasting systems based on single
source-oriented models started in the 1990s, at APPP; some years later, other more
complex operational systems (Kessler et al., 2001; San José et al., 2002; Blond and
Vautard, 2004; Grell et al., 2005; Monteiro et al., 2005) were either launched or proposed
by different research groups and institutions, to cover large areas with several sources,
primary and secondary pollutants. However, because of the large amount of primary
pollutants (specifically SO2) emitted by APPP, with respect to the rest of the sources in
its region, the single-source model-based approach remains valid for the primary
pollutants close to this source, as SO2 30 km around APPP.
446 J.A. Souto et al.

Figure 1 As Pontes Power Plant (APPP) environment, 30 km around it. Although heights are
below 1000 asl-m, the complexity of the topography can be observed. In addition, sea
inlets (namely, ‘rías’) are shown in the West and North of the area, which produce a
more complicated sea breeze effect in the area (see online version for colours)

Therefore, the main goal of the first APPP model-based air quality system was to
forecast, 24 h in advance at least, unfavourable meteorological and SO2 air quality
conditions owing to the dispersion of SO2 emitted by the power plant. This goal tried to
prevent SO2 air pollution alerts and to plan changes in the power plant operation,
specifically:
a Reduce the load. As APPP must report to the Spanish Electricity Network its
availability to produce energy along the next day, the forecasted power production
must be guaranteed. This reduction was followed until 1994, as the ratio of foreign
coal (without sulphur) was limited to 10% by the design of the boilers.
b Mixing of coals with less sulphur. After the modification in 1994 of APPP boilers,
a mixture of imported coal (30%) and local lignite (70%) is usually burnt. However,
specific meteorological conditions can produce SO2 air pollution episodes, so the
same boilers can burn up to 100% of foreign coal during limited periods (several
hours).
Option (b), which is currently applied, requires the planning of the coal mixing for the
following hours, because of the delayed effect of the fuel change in the SO2 emissions
and ground level concentrations. In addition, fuel changes should be done carefully,
to prevent strong stress in APPP equipment.
As focused in option (b), the development of this air pollution forecasting system
started in 1990, with the study of the state-of-the-art in meteorological and plume
dispersion modelling techniques (Souto et al., 1994), which could be operationally
applied to prevent the local air quality around an industrial plant, with a limitation in the
computational time available. From this preliminary study, the following conclusions
were achieved:
 SAGA: a decision support system for air pollution management 447

• available operational weather forecasts could not provide meteorological fields

accurate enough in the planetary boundary layer to be coupled to appropriate plume
dispersion models, so a local micro-mesoscale meteorological model should be
developed
• some advanced Lagrangian puff models and newer LPMs were the most adequate for
estimating the local plume dispersion
• the complexity of the models (especially, the meteorological model) was limited by
the available computational time in a workstation environment (<1 Mflop).
With these considerations in mind, a plume dispersion forecasting system was developed
(Souto et al., 1996, 1998) and started to run in 1994. This system included an emissions
module (based on mass balances for the boilers of the coal-fired power plant),
a hydrostatic meteorological model (coupled to the Spanish National Weather Service
forecast), and an operational version of the APM (Ludwig et al., 1989) for plume
dispersion estimation. Model results could be analysed by using an X-window Graphical
User Interface (GUI), as time-series and maps of meteorological and air quality data.
The use of this system and the analysis of its results required a good knowledge
and experience in both GUI (over UNIX systems) and air quality modelling. Therefore,
an expert in these issues located in the power plant provided the recommendations to the
power plant operation, to prevent future SO2 air pollution episodes.
With the increment of personal computers performance, the possibility of porting this
system to an MS-Windows environment was considered. At the same time, the
accumulated experience in data analysis for air quality decision support allowed to select
the most significant MFs and air quality data, to obtain automatic reports adapted to the
requirements of the power plant staff. With this base, SAGA Decision Support System
for Air Quality Management was developed and launched (for testing) in 2003, with its
first interface, SAGAwin, based on MS-Windows. After that, a new interface SAGAweb
was operational in 2004, to provide automatic reports through a private webpage to the
power plant staff.
In this work, a description of SAGA software and its modules is introduced, including
the capabilities of both interfaces (SAGAwin and SAGAweb), and the possibility to
adapt SAGA to different environments and numerical models. In what follows, the use of
SAGA at APPP is considered. Next, some results from the most useful parameters
provided by the operational SAGA software at APPP are shown. Finally, main
conclusions and future work are presented.

2 SAGA

As a model-based decision support system oriented to air quality management around a

point source, SAGA is configured by the combination of three types of models:
• An emissions model, which estimates the emissions and flow of the flue gas from the
point source (typically, an industrial plant), depending on the planned operation, at
least one day before.
In APPP, this model is based on a coal combustion balance, which estimates the flow
and composition of flue gas depending on the selected coals composition and the
448 J.A. Souto et al.

electric power that is planned to produce for the next day. In addition, from historical
records, a constant emission temperature of 190ºC is adopted.
• A meso-β meteorological model, which provides a high-resolution numerical
weather forecast, with specific interest in the meteorological parameters having
significant effect in the plume dispersion.
At APPP, two meteorological models, ARPS (Balseiro et al., 2001) and MM5
(Grell et al., 1995), have been tested. Currently, ARPS is the operational model,
providing a high-resolution numerical weather forecast (24 h in advance), for the
surrounding area 30 km around the power plant. As different 2 km and 10 km
horizontal resolutions were compared without any significant improvement in their
results (Saavedra et al., 2005), the operational version runs with the lowest resolution
of 10 km.
• A Lagrangian atmospheric diffusion model that estimates the local primary
pollutant’s concentrations around the point source. Although secondary pollutants
(like O3) are a significant air quality issue nowadays, the relationships between
secondary pollutant levels and local point source emissions are usually more specific
to each environment, depending on multiple source contributions. Therefore, SAGA
is currently limited to primary gaseous pollutants.
SO2 emissions control at APPP is focused on local dry deposition episodes, when SO2
conversion is low (<10%). Therefore, SO2 emissions dispersion can be estimated by
using two different Lagrangian diffusion models: APM2 (Souto et al., 2000) and LPM
(Souto et al., 2001). Comparison of both models (Penabad et al., 2002) around the
APPP environment shows a good qualitative agreement to observed SO2 episodes.
A quantitative and episodic testing was done but, because of its poor results owing to the
high sensitivity of the APPP plume dispersion to the meteorological conditions, it was
not recommended to be provided operationally to the power plant staff. Particularly, the
complexity of the APPP environment, with complex terrain and sea breezes affecting the
plume rise and plume growth, has been proved in a modelling intercomparison exercise
(Davakis et al., 1998). Therefore, more effort should be put to achieve an operational
application of these Lagrangian models in the APPP environment.
Anyway, because of the experience of APPP staff in the relationship between specific
local meteorological measurements and SO2 episodes, an MF report can be provided by
SAGA, including the most significant parameters that affect the air quality in the local
environment. Using the ARPS weather forecast, selected meteorological parameters,
from one- to three-day forecasts (in 1 h and 6 h intervals), at the point source and
surrounding locations, are obtained.

2.1 SAGA interfaces

SAGA is a modular software package, which includes two different interfaces:
• an MS-Windows interface (SAGAwin), which provides general reports about
meteorological and air quality measurements, and forecasts at the selected point
source environment
• a web-based interface (SAGAweb), which provides specific reports adapted to the
requirements of the air quality control around the selected point source.
 SAGA: a decision support system for air pollution management 449

Because of its simplicity of use, SAGAweb is currently the main interface used by the
power plant staff. SAGAwin can provide more information in specific cases (including
comparisons of measurements and past forecasts) to understand the SAGA results in
some complex episodes.
Figure 2 shows the main window of SAGAwin, with the typical MS Windows menu
bar, a Control Panel to get text or graphical reports, an Active Window (with the APPP
environment) for showing the selected reports, and a Status bar for showing any
incidence.

Figure 2 SAGAwin: Main window, including the MS-Windows menu bar, the Control Panel,
the Active Window and the Status bar. In the map, blue square shows the APPP
location (close to the A Mourela meteorological station) and red squares show
the glc monitoring stations’ locations, up to 30 km around the power plant.
Five of them include 10 m meteorological towers too (see online version for colours)

As an example of SAGAwin text report, Figure 3 shows a ground level concentration

(glc) text report. It can include hourly glc values either measured or forecasted along a
selected date for the 17 locations where the power plant has glc monitoring stations
installed.
Graphical reports in SAGAwin combine either meteorological or air quality data,
both measured and forecasted. As an example, Figure 4 shows SO2 glc (5 min. averages)
measurements along one day at a specific monitoring station. Figure 5 includes
temperature measurements at four levels in the A Mourela meteorological tower
(close to APPP), for the same day. These measurements can be compared in real time
with the forecasted values, to evaluate the accuracy of the forecast and, as a consequence,
to apply corrections to the planned operation of the plant.
450 J.A. Souto et al.

Figure 3 SAGAwin: An example of ground level concentration text report, showing some of the
forecasted glc hourly values for 7 September, 2003 (see online version for colours)

Figure 4 SAGAwin: Measured SO2 ground level concentration (5 min. averages),

on 9 October, 2003 (see online version for colours)
 SAGA: a decision support system for air pollution management 451

Figure 5 SAGAwin: Measured temperatures (5 min. averages) at 4 agl-levels (2, 10, 30 and
80 m), on 9 October, 2003 (see online version for colours)

2.2 SAGA reports

Because of the huge amount of information generated by SAGA models
(especially, meteorological and dispersion models), specific graphical and text reports
were designed to summarise the appropriate information to the power plant staff.
From this specific information, and applying their own experience in the management of
the power plant operation, changes in the fuel (coal mixing) for the next day can
be previewed, to guarantee that the SO2 glc legal limits are accomplished in the
surrounding area.
Two main reports were designed to answer the power plant staff requirements
(Figure 6):
a An MF report, with selected meteorological parameters that are significant to the
local air quality, 30 km around the power plant.
b Meteorological and Air Quality Forecast (MAQF) report, with the
same meteorological parameters plus some information about location,
duration and intensity of glc episodes forecasted by SAGA, around the power
plant.
452 J.A. Souto et al.

Figure 6 (a) Meteorological Forecast (MF) and (b) Meteorological and Air Quality Forecast
(MAQF) reports of 23 September 2003, provided by SAGA to the APPP staff
on 22 September, 2003 at 7:14 a.m. (see online version for colours)

(a)

(b)

Although initially the MAQF report would seem to be more useful, as it includes
direct information of the impact of the power plant emissions in the local air quality,
the uncertainties in the estimation of the local plume dispersion show that
meteorological parameters are a better guide for the power plant staff. In fact, before the
 SAGA: a decision support system for air pollution management 453

availability of SAGA, the APPP staff used to analyse the real-time meteorological
measurements, to consider their influence in the air quality for the same day. Therefore,
with the MF report, they can apply their experience directly to a qualitative estimation of
the influence of the power plant emissions in the local air quality, for the next day.
Specifically, the APPP staff can consider the most appropriate coal mixing on every 6 h
interval included in the MF report, to prevent any air pollution episode.
Both reports, MF and MAQF, are available by using the SAGAwin interface, only
accessible by authorised staff, to guarantee the appropriate use of the SAGA reports.
These reports are thoroughly explained in Sections 2.4 and 3.
After the use of SAGAwin for more than one year (between 2003 and 2004), a new
interface approach was proposed to make easier the access to the MF report and to extend
it as the most useful of the specific reports provided by SAGA. The new interface,
namely SAGAweb (Figure 7) provides MF reports of six different locations around the
power plant (at six different meteorological stations’ locations), with a three-day forecast
(today, tomorrow and the day after tomorrow). In addition, it allows access to old MF
reports (historical forecasts).

Figure 7 SAGAweb main page, showing the MF report at A Mourela location for 21 June 2005
(obtained on 20 June, 2005). MF reports for 20 and 22 June are available too,
at six different locations (see online version for colours)

This new interface, launched in 2004, allows to access to the MF reports from any PC
terminal of a private intranet located in APPP, by using a web navigator (Internet
Explorer, Netscape, Firefox, …). Therefore, any personnel in charge of this issue at the
power plant can access the SAGA final results from their own office, send the report to
other people in PDF format and print it. These facilities are a significant improvement in
the exploitation of SAGA in the operation of an industrial plant, where the time available
for decisions is limited and the operation staff can change regularly.
454 J.A. Souto et al.

2.3 SAGA operational weather forecast

The operational weather forecast in SAGA is currently obtained by using the ARPS
(Xue et al., 2001), a non-hydrostatic meteorological model adjusted to Galicia.
ARPS was developed in the 1990s by the Centre for Analysis and Prediction of Storms
(CAPS) of the Oklahoma University. Its current version includes solutions at boundary
layer and mesoscale; in spite of the complexity of the Galician environment, this model
was tested in different applications and spatial resolutions, as follows:
• an operational weather forecast for Galicia
• study of the wind patterns in the port and estuary of a local town
• estimation of wind in the forecasting of the Prestige oil spill
• forecast of wind energy production
• high-resolution coastal marine forecast, coupled to an oceanographic model.
Depending on the application, ARPS has been tested in Galicia with horizontal
resolutions from 50 km to 300 m. In SAGA, the operational resolution is 10 km; although
higher resolutions were tested (the highest one, 2 km, with a huge extra computational
cost), the results obtained (Saavedra et al., 2005) are not improved, so its application
is not justified. Therefore, 10 km resolution seems to be a reasonable operational
approach in this environment, although it could be changed in other locations.
Taking D (as the day of tomorrow, with respect to the day when the MF report is
published), the ARPS operation in SAGA starts at 20:00 LST of D – 2, when the model
reads the NCEP numerical weather forecast, covering the Southwest of Europe (including
the whole Iberian Peninsula), North of Africa and the Atlantic coast (Figure 8(a)) with
a horizontal resolution of 50 km. In addition, during the start-up of the model, a static
assimilation of the rawinsonde measurements available in the region is done, using the
INTRP tool included in ARPS.
Results of this coarse simulation are nested to a finer grid (Figure 8(b)) with the final
resolution of 10 km, covering the Northwest of the Iberian Peninsula, including
Galicia as the region of interest. One-way nesting is applied, because the forecast period
(three days) is not so long, and the regions considered are not so big to consider feedback
effects between grids. In addition, standard surface meteorological measurements
available in the region (currently, 23 stations) are assimilated in the start-up of this finer
grid simulation. Finally, before 9:00 LST of D – 1, a numerical weather simulation
for the region covering from D – 2 to D + 1 is obtained. From these results, values for the
meteorological parameters included in the MF report are derived for days D – 1, D, D + 1
and D + 2.
In summary, around 9:00 LST, a weather forecast adapted to the requirements of the
power plant is available, for today (D – 1), tomorrow (D), and the day after tomorrow
(D + 1). Usually D forecast is considered for decision support.
 SAGA: a decision support system for air pollution management 455

Figure 8 Simulation grids in the operational ARPS numerical weather forecast applied in SAGA:
(a) coarse grid and (b) fine grid. Scale shows the mean asl-height included in the ARPS
model for each simulation grid (see online version for colours)

(a) (b)

2.4 SAGA air quality forecast

As it was mentioned before, SAGA can only provide a qualitative air quality forecast
with either APM2 or LPM models. Assuming the use of a coal mixing ratio at APPP to
guarantee that SO2 emission concentration remains below the current legal emissions
threshold along the next day, the emissions model estimates the SO2 emissions
concentration and, additionally, the gas flow and temperature. With these data, usually
APM2 is coupled to the ARPS forecast over the finest grid considering 30 min. averages
meteorological results, to simulate the SO2 emissions dispersion along the next day.
Finally, the contribution of the simulated emissions to hourly SO2 glc along the next day
is estimated, over a 60 × 60 km2 grid centred in APPP, with a 1 km2 resolution.
After this APM2 simulation, which usually requires less than 10 min. of CPU time,
hourly SO2 glc forecasts are processed, considering four 6 h intervals (starting at 0:00,
6:00, 12:00 and 18:00 LST), and eight geographical sectors. These sectors (Figure 9) are
defined considering the risk of air pollution episodes in the past, that is, taking into
account the measurements available from 17 SO2 glc monitoring stations, and including
in the same sector a group of stations that usually registered in the past an air pollution
episode at the same time, that is, in the same 6 h interval.
For every 6 h interval and sector, an automatic analysis of the hourly SO2 glc forecast
over the grid cells is done, considering two different conditions:
• Intensity. Low (100–350 µg/m3), medium (350–500 µg/m3), and high (>500 µg/m3)
SO2 glc are considered. All of them start below the legal threshold of 700 µg/m3
established by the Spanish Government for hourly SO2 glc, but the glc intervals
associated with the risk levels can be adapted to the legal thresholds, if it is
necessary.
• Duration. The number of hours when the intensity level appears is considered;
therefore, episodes are classified into short (<1 h), medium (from 1 h to 2 h) and
durable (more than 2 h).
456 J.A. Souto et al.

Figure 9 Geographical sectors considered around APPP, to estimate the potential risk of SO2
episodes in the MAQF report. Every sector includes at least one SO2 glc monitoring
station, as the actual risk is evaluated by APPP staff using their measurements
(see online version for colours)

Results were daily published for testing in an MAQF report, as it was shown in
Figure 6(b). In this case, on 23 September, 2003, a medium level intensity air quality
episode was forecasted, affecting from 8:00 to 24:00 LST, and extended from SSW
sector to WSW sector. However, the comparison of the air quality forecast to observed
SO2 glc shows that only a short and low level episode was detected between 18:00 and
24:00 LST, at the SSW sector. This behaviour was usual during the operational testing of
the MAQF report (both using APM2 and LPM models), proving the difficulty of an
appropriate calibration of the modelling system to forecast air quality episodes over
specific locations, as it was shown in the non-operational testing of both models
presented in Penabad et al. (2002).

3 Results

SAGA system has been applied in APPP as a decision support tool for daily fuel changes,
to prevent local air pollution episodes. Currently, this coal-fired power plant burns a
mixing local lignite (with up to 2% of sulphur) and foreign coal (with less than 0.1% of
sulphur), so one of the most relevant air pollutants emitted is SO2. In some specific local
meteorological conditions, SO2 ground level concentration episodes can be produced;
therefore, any forecast of the meteorological parameters that affect the SO2 plume
dispersion is an interesting help to the power plant staff.
Everyday, the power plant needs to evaluate to convenience of changing the
coal mixing for the next day, taking into account the risk of local poor air quality
(due to SO2 glc). This risk is evaluated by the power plant staff considering the
meteorological parameters included in the MF report, as follows:
 SAGA: a decision support system for air pollution management 457

• Wind speed. Low winds are favourable for mixing a high level emission, such as
APPP emissions. Therefore, winds above 9 m/s are usually required for causing poor
local air quality.
• Wind direction. Southwest and Northeast sectors around the power plant are usually
more affected by the SO2 plume, because of their complex topography and the
typical sunny conditions during them.
• Temperature. High surface temperatures (with respect to aloft temperatures) are
usually favourable to create convective plumes, with higher risk of plume down.
• Cloudiness and solar radiation. High solar radiation and clear sky are more
favourable to create convective plumes, with the same result.
• Surface lapse rate. Lapse rate from 80 m to 30 m is a direct magnitude of the
atmospheric stability near the surface; as this parameter is historically measured and
studied by the power plant staff, they are able to use it to identify the risk of poor air
quality. Both lapse rate and corresponding Pasquill class are included in the SAGA
reports.
The selection of these parameters is not casual, because the APPP staff has a long
experience (more than 15 years) in analysing these specific parameters from
real-time measurements; for instance, staff prefers to classify the atmospheric
stability by using the Pasquill classes, instead of other more accurate parameters,
as wind fluctuations or Richardson gradient. In addition, taking advantage
of the capabilities of an operational weather forecast, a new aloft parameter was
included.
• Aloft lapse rate. Lapse rate from 800 m to 500 m is a direct magnitude of the
atmospheric stability around the plume transport layer. Jointly to the surface lapse
rate, it can help to consider the influence of thermal inversions in the plume
dispersion and, consequently, in the local air quality. Both lapse rate and
corresponding Pasquill class are included in the SAGA reports.
From each new MF report available everyday before 9:00 LST, the power plant staff
analyses it and, before the end of the morning, they consider some possible changes in
the coal mixing ratio to be applied to the next day. Usually, they keep a 70 : 30
(lignite : foreign coal) mixing ratio, but the power plant is prepared to change to 100% of
foreign coal for a limited time interval, if it is necessary. However, this severe change
cannot be applied quickly to prevent failures in the combustion systems; therefore,
mixing ratio should be changed several hours before the meteorological conditions for
poor local air quality are expected.
As the most useful result of SAGA, a systematic evaluation of the operational MF
report was performed, by comparison of the meteorological results with the available
measurements from May to August 2005, as follows.
All the meteorological parameters included in the MF report, except the aloft lapse
rate, can be compared with measurements available in the A Mourela meteorological
station, close to the APPP. This site includes an 80 m tower with measurements of wind
(10 m and 80 m aloft), temperature (2, 10, 30 and 80 m aloft), surface moisture, pressure,
458 J.A. Souto et al.

precipitation and global solar radiation. Some of these parameters are applied in a short
period testing of the MF report, in order to evaluate:
• the agreement between measurements and modelling values, considering the 6 h
intervals applied in the MF report
• the accuracy required in the forecasted values included in the MF report, to evaluate
its usefulness as part of the decision support tool SAGA.
Therefore, comparisons were made for the following parameters:
• wind speed and wind direction at 80 m
• surface temperature
• surface lapse rate
• Pasquill class, from the surface lapse rate.
A statistical analysis by events was considered more appropriate for wind and Pasquill
class testing, because both parameters are included as discrete values (wind intervals,
Class) in the MF report. Only for surface temperature, a classical statistical testing was
done, obtaining values of RMSE, Mean Absolute Error (MAE) and Bias.
The statistical analysis by events is based on the comparison of intervals (for wind) or
Classes (for stability), considering three different results:
• agreement: forecasted and measured values are in the same interval or Class
• error level 1: a difference of one consecutive either interval or Class between
forecasted and measured values
• error level 2: a difference of two consecutive either interval or Class between
forecasted and measured values.
Table 1 shows the wind speed and wind direction intervals considered, and the
well-known Pasquill classes, applied in this statistical analysis.

Table 1 Wind speed and wind direction intervals, and Pasquill classes, applied in the statistical
analysis by events for the comparison of measured and forecasted values

Wind speed intervals (m/s) [0, 1], (1, 3], (3, 5], (5, 7], (7, 9], (9, 13], (13, …)
Wind direction intervals (degrees) N, NE, E, SE, S, SW, W, NW
Pasquill classes A, B, C, D, E, F

Table 2 shows results for the surface temperature comparison: a significant deviation of
minimum temperatures is observed, with the forecasted values higher than observed;
however, maximum temperatures are in agreement. The systematic error in minima
seems to be extended to the measuring levels in the tower (10, 30 and 80 m), as the
comparison of surface lapse rate in Table 3 shows a poor quantitative agreement between
observed and forecasted values.
 SAGA: a decision support system for air pollution management 459

Table 2 Root Mean Square Error (RMSE), Mean Absolute Errors (MAEs) and Bias (BIAS)
of the maximum (MaDST) and minimum (MiDST) daily surface temperature
estimated by ARPS in SAGA, compared with the measurements of the
A Mourela meteorological station

RMSE MAE BIAS

MiDST (ºC) 4.84 4.25 3.17
MaDST (ºC) 4.25 3.44 –0.48

Table 3 Mean Absolute Errors (MAEs) of the surface lapse rate (80–30 m) estimated
by ARPS in SAGA, compared with the measurements of the A Mourela
meteorological station

RMSE MAE BIAS

Minimum 80–30 lapse rate 0.74 0.41 0.20
Maximum 80–30 lapse rate 1.11 0.76 –0.24

Better agreement is obtained in Pasquill classes, as it is shown in Table 4. At the same

time, wind speed and direction (Table 5) shows an agreement of around 50%, which is
poor for a quantitative use in plume dispersion forecasting. However, jointly to the
Pasquill class, it can be enough for a qualitative evaluation of the risk of poor local air
quality due to SO2 fumigation episode.

Table 4 Statistical events analysis of Pasquill stability classes at surface level estimated
by ARPS in SAGA and obtained from the surface lapse rate measurements
of the A Mourela meteorological station

Agreement Error level 1 Error level 2

Most unstable class 0.64 0.34 0.02
Most stable class 0.35 0.59 0.06

Table 5 Statistical events analysis of wind speed and wind direction intervals at surface
level (10 m) estimated by ARPS in SAGA and obtained from the surface lapse
rate measurements of the A Mourela meteorological station

Agreement Error level 1 Error level 2

Wind speed interval 0.22 0.34 0.26
Wind direction interval 0.25 0.41 0.23

4 Conclusions

The development and application of a decision support system for air quality
management around a point source is presented. The system, namely SAGA, can provide
MAQF, one-day before, to a 1400 MW coal-fired power plant staff, to help in the
planning of the plant operation. A MF report specifically adapted to these point source
requirements is the most useful result provided by SAGA. The meteorological parameters
included in this report (related to the local plume transport and dispersion) are daily
460 J.A. Souto et al.

analysed by the power plant staff to decide the changing of the coal mixing for the next
day. This coal mixing can be programmed to reduce the SO2 emissions during short
periods, to prevent local air pollution episodes.
Regarding the statistical evaluation of the MF report, the surface temperatures show
significant differences in the minimum values, which are repeated in the surface lapse
rate. However, the Pasquill class shows a good agreement and, jointly to the wind speed
and direction intervals included in the report, provides a set of good values for decision
support in air quality management at APPP. Although some improvements in the
numerical weather forecast were considered and achieved in some specific events
(Saavedra et al., 2006), they could lead to a poor operational response time of the model.
Therefore, a balance between accuracy of results and operational application of SAGA is
necessary.
SAGA is nowadays focused in primary pollutants (like SO2), but it is planned to
consider other pollutants (as NOx and O3), by including different comprehensive
air quality models. STEM-II (Carmichael et al., 1986) for acid deposition, CAMx
(ENVIRON International Corporation, 2005) for photochemical oxidants and, more
recently, WRF-CHEM (Grell et al., 2005) and CHIMERE (Vautard et al., 2006) are
possible candidates to be included in SAGA. However, this extension of SAGA will
require a new evaluation of the computational resources needed, because of the high
requirements of these models.

Acknowledgements

This work was financially supported by Xunta de Galicia and Endesa under project
PGDIT1C05E.

References
Balseiro, C.F., Souto, M.J., Pérez-Muñuzuri, V., Xue, M. and Brewster, K. (2001) ‘Operational
numerical weather forecast in Galician region (Spain) by using a non-hydrostatic numerical
model’, 26th EGS General Assembly, March, Nice, France.
Blond, N. and Vautard, R. (2004) ‘Three-dimensional ozone analyses and their use for short-term
ozone forecasts’, Journal of Geophysical Research, Vol. 109, p.D17303.
Carmichael, G.R., Peters, L.K. and Kitada, T. (1986) ‘A second generation model for regional-scale
transport/chemistry/deposition’, Atmospheric Environment, Vol. 20, pp.173–188.
Davakis, E., Deligiannis, P. and Souto, J.A. (1998) ‘Dispersion modelling intercomparison
exercise’, 5th International Conference on Harmonisation within Atmospheric Modeling for
Regulatory Purposes, Rhodes, Greece, 18–21 May.
ENVIRON International Corporation (2005) User’s Guide – CAMx – Comprehensive Air Quality
Model with Extensions, Novato, California.
Grell, G.A., Dudhia, J. and Stauffer, D.R. (1995) A Description of the Fifth-Generation Penn
State/NCAR Mesoscale Model (MM5), Tech. Rep. NCAR/TN-398+STR, National Center for
Atmospheric Research (NCAR), Boulder (Co), USA.
Grell, G.A., Peckham, S.E., Schmitz, R., McKeen, S.A., Frost, G., Skamarock, W.C. and Eder, B.
(2005) ‘Fully coupled ‘online’ chemistry within the WRF model’, Atmospheric Environment,
Vol. 39, pp.6957–6975.
 SAGA: a decision support system for air pollution management 461

Kessler, Ch., Brücher, W., Memmesheimer, M., Kerschgens, M. and Ebel, A. (2001) ‘Simulation
of air pollution with nested models in North Rhine-Westphalia’, Atmospheric Environment,
Vol. 35, No. 1, pp.S3–S12.
Ludwig, F.L., Salvador, R. and Bornstein, R. (1989) ‘An adaptive volume plume model’,
Atmospheric Environment, Vol. 23, pp.127–138.
Monteiro, A., Vautard, R., Lopes, M., Miranda, A.I. and Borrego, C. (2005) ‘Air pollution forecast
in Portugal: a demand from the new air quality framework directive’, International Journal of
Environment and Pollution, Vol. 25, No. 2, pp.4–15.
Penabad, E., Pérez-Muñuzuri, V., Souto, J.A., Casares, J.J., Bermúdez, J.L. and Ludwig, F.L.
(2002) ‘A comparison of Lagrangian dispersion models coupled to a meteorological model for
high stack air pollution forecast’, Air Pollution, Segovia, Spain, 1–3 July.
Saavedra, S., Penabad, E. and Souto, J.A. (2005) ‘Testing of local turbulence closures in mesoscale
model to be applied in the SAGA decision support system’, Atmospheric Sciences and Air
Quality Conference (ASAAQ), San Francisco, USA, 27–29 April.
Saavedra, S., Souto, J.A. and Vila, J. (2006) ‘Impact of meteorological factors on turbulent
dispersion over complex terrain’, 28th NATO/CCMS International Technical Meeting on Air
Pollution Modelling and its Application, 15–19 May, Leipzig.
San José, R., Pérez, J.L., Blanco, J.F., Barquin, R. and González, R.M. (2002) ‘An operational
version of MM5-CMAQ modelling system over Madrid city’, Fourth Symposium on the
Urban Environment, 19–24 May, American Meteorological Society.
Souto, J.A., de Castro, M., Casares, J.J., Souto, M.J., Pérez-Muñuzuri, V. and Bermúdez, J.L.
(2000) ‘Testing of an adaptive puff model for regulatory purposes around as pontes power
plant’, International Journal of Environment and Pollution, Vol. 14, pp.198–207.
Souto, J.A., Pérez-Muñuzuri, V., de Castro, M., Souto, M.J. and Casares, J.J. (1996) ‘Forecasting
and real-time simulation of plume transport around a power plant’, 5th International
Conference on Atmospheric Sciences and its Application to Air Quality, Seattle, USA, June.
Souto, J.A., Pérez-Muñuzuri, V., de Castro, M., Souto, M.J., Casares, J.J. and Lucas, T. (1998)
‘Forecasting and diagnostic analysis of plume transport around a power plant’, Journal of
Applied Meteorology, Vol. 37, pp.1068–1083.
Souto, J.A., Pérez-Muñuzuri, V., Ludwig, F.L. and Casares, J.J. (1994) ‘Meteorological and
atmospheric diffusion modelling for air pollution forecasting’, Envirosoft, San Francisco,
USA, Vol. 94, November.
Souto, M.J., Souto, J.A., Pérez-Muñuzuri, V., Casares, J.J. and Bermúdez, J.L. (2001)
‘A comparison of operational Lagrangian particle and adaptive puff models for plume
dispersion forecasting’, Atmospheric Environment, Vol. 35, pp.2349–2360.
Vautard, R., Bessagnet, B., Menut, L. and Monge, J.L. (2006) CHIMERE. Documentation of the
Chemistry-Transport Model, http://euler.lmd.polytechnique.fr/chimere/downl-chimere.html
Xue, M., Droegemeier, K.K., Wong, V., Shapiro, A., Brewster, K., Carr, F., Weber, D., Liu, Y. and
Wang, D. (2001) ‘The Advanced Regional Prediction System (ARPS) – a multi-scale
nonhydrostatic atmospheric simulation and prediction model, part I: model physics and
applications’, Meteorology and Atmospheric Physics, Vol. 76, pp.143–165.

View publication stats