Environmental Modelling & Software 73 (2015) 76e89

Contents lists available at ScienceDirect

Environmental Modelling & Software

journal homepage: www.elsevier.com/locate/envsoft

A fast mobile early warning system for water quality emergency risk
in ungauged river basins
Yonggui Wang a, Wanshun Zhang a, *, Bernard A. Engel b, Hong Peng c, Larry Theller b,
Yingyuan Shi a, Shan Hu d
a
School of Resource and Environmental Sciences, Wuhan University, Wuhan 430079, PR China
b
Agricultural and Biological Engineering, Purdue University, West Lafayette, IN 47907-2093, USA
c
School of Water Resources and Hydropower, State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University,
Wuhan 430072, PR China
d
Hydroco Science and Technology Ltd, Wuhan 430000, PR China

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

Article history: Environmental decision support system (EDSS) has become an important research topic in the fields of
Received 20 April 2015 water environment protection and emergency early warning. Most existing systems require plenty of
Received in revised form measured data, which are typically unavailable in ungauged basins. In order to develop an early warning
31 May 2015
system that can work with few measured data, a mobile environment decision support system (MEW-
Accepted 4 August 2015
SUB) for early warning and emergency assessment in ungauged river basins was developed. A one-
Available online 22 August 2015
dimensional water quality model was built for the system and the terrain module and pollutant mod-
ule were developed for terrain digitization and pollutant release process data generation. MEWSUB can
Keywords:
Mobile environment decision support
quickly create water quality model input files, simulate the temporal release of pollutants, and display
system the results in a dynamically rendered map and trend line diagram via HTML5. The system has been used
Early warning successfully to support early warning efforts during accidents on several ungauged rivers in China.
Emergency risk © 2015 Elsevier Ltd. All rights reserved.
Ungauged river basin
Water quality accident

Software and data availability

Software Name MEWSUB 1. Introduction

Developers This paper's authors, Hydroco Science and
Technology Ltd Sudden water pollution accidents have taken place more
Hardware Requirements Smartphone or tablet with GPS frequently around the world in recent years. The water envi-
Software Requirements the APP on Android 4.0 and above or ronment emergency risks cannot cause damage to humans' daily
IOS6 and above, browser supporting lives and the natural environment. These water environment
HTML5. pollution incidents suddenly or unexpectedly occur, that may
Programming Language Java, Fortran lead to catastrophic environmental problems and grave social
Download instructions Requires authorization. consequences. After water quality accidents, it is important to
Availability and cost: Android APP may be requested free of warn downstream areas. For downstream locations, in order to
charge by e-mailing the author evaluate the pollutant impact and carry out emergency actions
(wangyonggui@whu.edu.cn); or in a timely, accurate and reasonable manner, it is necessary to
directly browser (http://115.28.11.12: gain answers to four important questions about when the
8085/ewup/) with account name and pollutant will arrive at sensitive sites downstream, what the
password. peak value of the pollutant will be, how large the pollution areas
will be, and how long the pollutant concentration value will
remain at or above the security value (herein called 2W2H
questions) (Hou et al., 2014).
* Corresponding author. With the development of GIS and distributed environmental
E-mail address: wszhang@whu.edu.cn (W. Zhang). numerical models (e.g., hydrodynamic and water quality models),

http://dx.doi.org/10.1016/j.envsoft.2015.08.003
1364-8152/© 2015 Elsevier Ltd. All rights reserved.
 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89
these 2W2H questions can be answered easily. Early warning sys-
tems (EWS) for emergency response of sudden water pollution
accidents coupled with GIS and water quality models have been
studied and put into practical application in many countries (e.g.,
the British raw water quality monitoring system established by
Severn Trent Water company on the Trent River (Adams et al.,
2006), the American water quality early warning system
managed by the Ohio River Valley Water Sanitation Commission on
the Ohio River (Grayman et al., 2001), and the Chinese drinking
water quality early warning and control system (Hou et al., 2013)).
These early warning systems can be applied to real-time alarm
monitoring and emergency response (Song et al., 2013); water
quality EWS can provide important and essential pollution warning
information for the adjustment of treatment operations to clean
pollutant or decrease the contamination concentration and protect
human lives (Quansah et al., 2010). Most of these existing systems
are based on client-server architecture software approaches, and
few have a Web services structure. With abundant input data, they
can be used for precisely simulating contaminant transport. These
EWS need a large number of measured data for numerical
computation, but many of these data may be unavailable in unga-
uged river basins. Ungauged river basins abound in developing
countries and often have high occurrences of water pollution. A
new framework for water quality early warning to answer the
2W2H questions in ungauged river basins is urgently needed,
especially in China.
Due to the rapid economic development and the growth of
agriculture, industry and social construction activities, China has
many occurrences of river water pollution contamination events,
such as the nitrobenzene spill accident in the Songhua River in
2005 (Hu et al., 2009; Zhang et al., 2011), the odorous tap water
crisis in Wuxi in 2007 (Zhang et al., 2010b), and the phenol leakage
incidents in the Qiantang River in 2011 (Hou et al., 2014). This series
of water pollution accidents attracted great attention from the
government and academia, which prompted the development of a
rapid and effective emergency response framework and building
up the capability of early accident detection (Zhang et al., 2011).
Early warning systems for water emergency risk have become an
important research topic in water environment protection and risk
assessment fields. Although many systems like DEWS (Hou et al.,
2013), WPMS_ER (He et al., 2011), EWS-FDWQ (Li et al., 2009)
and FDEWS (Li et al., 2006) are designed to address specific prob-
lems in a specific water body, they may still require large, spatially
distributed inputs for their coupled complicated numerical models.
Environmental monitoring and science research in developing
countries like China lags that in developed countries. Due to time
and financial resource constraints to support monitoring stations in
all waters, government and research institutions typically focus on
demonstrations in big rivers and lakes. Numerous data may be
collected in such areas, while little data are collected in other
smaller rivers or channels. Insufficient terrain, hydro-
meteorological, pollution source and water quality parameter
data needed for distributed models in China (Reckhow, 1994) is
effectively putting a brake on the development and practical
application of EWS (Burchard-Levine et al., 2014).
With the rapid development of industrial transfer (high
polluting industries have been moved from the east China cities to
west China cities or from the big cities to the small towns) and the
construction of traffic routes in China, a number of ungauged river
basins have become areas accepting traditional industry and have
accepted many new factories. Most of these new factories are
operated in a traditional industrial production style with high en-
ergy consumption and high pollution. Together with the impact of
traffic lines and factories along rivers, these ungauged river basins
face severe water quality threats from factory discharges and traffic
77
accidents. In these areas, as the river water may be the main or even
the only water source for life and production, once it is polluted,
people's daily lives will be seriously impacted. Therefore, an
effective tool that can answer the 2W2H questions and be applied
in ungauged river basins is urgently needed. Using this tool for
assistance, the environment management branch or the environ-
ment protection operator can simulate fate and transport of
chemical pollution, create an emergency response plan for pollu-
tion cleanup, and inform the relevant people when to store
adequate amounts of fresh water.
Furthermore, early warning and emergency planning involves
significant work in the field, where computer equipment may not
be readily available. Equipped with professional mobile systems,
smartphones are strong candidates used for the work of field in-
vestigations, data collection and decision making (Dong and Li,
2014). Water pollution accident evaluation and early warning of
the fate and flow distribution of contaminates and the rate and
trend of water pollution in ungauged river basins is an example of
such field work. In order to meet these requirements of convenient
and rapid early warning of water environment emergency risks in
ungauged rivers, this paper explores early warning technology of
sudden water pollution accidents in ungauged rivers, proposes a
framework for mobile water quality early warning systems, and
presents its recent application in an ungauged river in China.
The paper is structured as follows: Section 2 analyzes related
work about the risk in ungauged basins to identify the key points of
early warning in ungauged basins. The One-dimensional water
quality model for pollutant simulation, and the features of EWS and
Web services technology for mobile early warning systems are also
discussed in this section. Section 3 presents the architecture of
MEWSUB as well as the main content and functions of the input
files for the model and the system. The two technologies for digital
terrain data generation and accident release data creation are
described in Section 3.3. Section 3.4 describes the method to cor-
rect the model results, and Section 3.5 describes the output data
from the system. A case study using the system for early warning in
the Xin'an River of China is presented in the Section 4. Finally, the
paper provides conclusions in Section 5.
2. Related work
2.1. Risk and early warning in ungauged river basins
Risk is defined as “the combined measure of the degree of
detriment to society (in this case concerning people's health)
caused by a defined event (or combination of events), and the
probability of that event occurring” (McIntyre et al., 2003). These
risks arise for numerous reasons such as unlawful or accidental
pollution discharges (point discharges), transportation spills and
chemical or radioactive spills. As pressure on water resources
globally is rising, small catchments and tributaries of big rivers are
becoming increasingly more important for water supply and eco-
nomic development (Smakhtin, 2001). Numerous factories and
transport lines have been built in these areas. Unfortunately, due to
historical and current causes, the systematically recorded flow and
environment data in these basins are often sparse or non-existent,
and these phenomena are especially prominent in many devel-
oping countries (Niadas, 2005). These basins are ungauged or
poorly gauged basins. An ungauged basin was defined in the
“predictions in ungauged basins (PUB) project” by the International
Association for Hydrological Sciences (IAHS) as “an ungauged basin
is one with inadequate records (in terms of both data quantity and
quality) of hydrological observations to enable computation of
hydrological variables of interest (both water quantity or quality) at
the appropriate spatial and temporal scales, and to the accuracy
78 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89
acceptable for practical applications” (Sivapalan et al., 2003).
Numerous research efforts have been conducted in the field of
ungauged basins, including extrapolation of response information,
hydrologic response prediction, rainfall-runoff modeling, estima-
tion of flow characteristics, and reduction of predictive uncertainty
(Hrachowitz et al., 2013). Most previous studies mainly focused on
the field of hydrology but few on water quality. While factories are
typically set up close to water and roads are also built along coasts
or dikes, the water quality risk in these ungauged basins is great,
especially in rivers. With poor data, a precise numerical model for
early warning and response will be difficult to create. The
numerous emergency events show that a simple and running
quickly tool is urgently needed which can use limited data but help
emergency workers answer the 2W2H questions and make emer-
gency decisions (Storey et al., 2011). For such a tool, selecting a
special and appropriate water quality model is required. With the
results of the model, emergency workers can estimate the scale of
pollution damage and its trend once the accidents happens, which
will be helpful for managers to make scientific and effective
emergency decisions about where and when to conduct emergency
monitoring and pollutant cleaning up, and how to best arrange
personnel and emergency supplies.
2.2. Description of the water quality model in MEWSUB
Numerical models in water quality management have diverse
structures and dimensions. Normally, there are four categories
related to dimension for water quality models: Zero-dimensional
(0D), One-dimensional (1D), Two-dimensional (2D) and Three-
dimensional (3D). Different dimension models have different
applicability in water quality simulation. Zero-dimensional models
may be useful for preliminary estimation of water quality condi-
tions in well-mixed water bodies, but they can't simulate water
quality changes along rivers. One-dimensional models are capable
of simulating spatial change in a single dimension, typically ori-
ented longitudinally down the length of a river or a narrow es-
tuary. Compared with Two-dimensional or Three-dimensional
models, the input requirements of One-dimensional models are
fewer and simpler, which appropriately meets the 2W2H de-
mands of an ungauged river basin's emergency risk simulation.
Two-dimensional models have the capacity for simulating the
spatial variation in lateral and longitudinal directions or in vertical
and longitudinal directions. Three-dimensional models can
describe changes in all three spatial dimensions and provide the
most detailed assessment of pollutant distributions. Two-
dimensional and Three-dimensional water quality models
require large amounts of detailed hydrology and water quality
data, which is unlikely to be available in ungauged basins. As a
consequence, a One-dimensional model was selected as the pri-
mary numerical model for water quality simulation in ungauged
river basins.
For a river, its depth and width are small compared with its
length, and therefore it can be assumed that pollutants will be
mixed in the cross section after flowing a limited distance from
their entry point. For many rivers (especially river segments), the
problem of calculating water quality can therefore be simplified to a
1D chemical transport problem, which assumes that the pollution
concentration in the cross section is homogeneous and changes
only in the flow direction (Hou et al., 2014).
A common water quality model containing the continuity
equation, momentum equation and pollutant convection diffusion
equation (Zhang et al., 2010a) was built as the server model in
MEWSUB. The SainteVenant equations are employed to describe
the dynamic processes in the river.
Continuity equation:
vA vQ
þ ¼q (1)
vt vx
Momentum equation:
vAu vQu vz n2 jQ jjuj
þ þ gA þ g 1d 4=3 ¼ 0 (2)
vt vx vx R
Pollutant convection diffusion equation:


vc vuc v vc
þ ¼ Ex  kc þ Sd (3)
vt vx vx vx
where A (m2) is the discharge section area; q (m2/s) is the inflow of
the side-stream; z (m) is water level; h (m) is water depth; Q (m3/s)
is water quantity;g (g/m2) is gravitational acceleration;u (m/s)is
flow velocity; n is channel roughness; R is hydraulic radius; t (s) is
time; x (m) is the distance the pollutant moves; Ex (m2/s) is river
longitudinal dispersion coefficient; vz=vx is gradient; c (mg/l) is the
concentration of water quality pollutant; Sd is the sink/source, and k
is the pollutant decay rate.
2.3. Early warning system and web services
Early warning (EW) was defined as “the provision of timely and
effective information, through identified institutions, that allows
individuals exposed to hazard to take actions to avoid or reduce
their risk and prepare for effective response” by the United Nations
Office for Disaster Risk Reduction (UNISDR, 2004). The early
warning system (EWS) is such a systems that can provide useful
services for disaster and risk reduction strategies in fields that
range from natural physical phenomena and public health, to man-
made systems (ISCRAM and French, 2011). The UN Inter-Agency
Secretariat of the International Strategy for Disaster Reduction
(UN/ISDR) third International Early Warning Conference (EWCIII) in
2006 defined four key elements-risk knowledge, monitoring and
warning service, dissemination and communication and response
capability for a complete and effective early warning system (UN/
ISDR, 2006). From the Global survey of early warning systems of
UN/ISDR, it showed that the existing EWSs (up to 2006) were rarely
set on the four key elements but often set on one or two of these
elements only (Meissen and Voisard, 2008).
To improve the capacity of EWS, the research of technology for
EWS development has become a hot area in recent years. In 2008,
the 5th International Conference on Information Systems for Crisis
Response and Management (ISCRAM) held a special session on
early warning systems to discuss the benefits of using the geo-
information technologies, information management and
modeling and simulation methods for EWS (Fiedrich and Van de
Walle, 2009). The outcomes report of the 5th International
Disaster and Risk Conference (IDRC Davos 2014) declared that
research on early warning systems should be continued and further
strengthened, and it also suggested that EWS may provide real-
time access to reliable data and use information and communica-
tion technology (ICT) innovations to enhance collection, analysis
and dissemination of data (Pasquini-Descombs et al., 2014). The
start of the art collection report about the recent development in
early warning in the International Disaster Risk Conference in
Davos in August 2014 had discussed the EWS by analyzing the four
elements-risk knowledge, monitoring and warning service,
communication and response capacity, which concluded that the
existing EWSs are not operational in many areas and the smaller
and local EWS should be established (Usla €nder, 2014). Recently,
considerable progress has been made in technologies such as
remote sensing, GIS, ICT innovations, Internet of Things, geospatial
 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89
sensor web and Web Services. The remote sensing and GIS are
widely used in EWS for several decades (Gupta and Mishra, 2015).
The development of the Internet of Things and ICT innovations can
make the devices like smart phones cameras, all kinds of sensors,
satellites and computers become seamless interconnection for data
monitoring, collecting, delivering and analyzing to improve
recognizing and assessing risks (Usla €nder, 2015). The Sensor web
has been proved a useful tool for integrating heterogeneous
observation resources, and could serve an important function for
dynamic monitoring and early warning of water disaster events
(Chen et al., 2014; Kussul et al., 2012). With these technologies, the
early warning system has become a very important tool for floods,
droughts, earthquake, storms, water quality pollution, bushfires
and other hazard assessment and risk alert. Many EWSs have been
built to help people for early warning in conjunction with disaster
management, like the flood early warning system (Katiyar and
Hossain, 2007; Krzhizhanovskaya et al., 2011), German Indone-
sian Tsunami early warning system (Rudloff et al., 2009), Istanbul
earthquake early warning system (Erdik et al., 2003), the USGS (U.S.
Geological Survey) globe earthquake notification service system
(USGS ENS) (Wald et al., 2008), and landfill operations early
warning and emergency response system (Fulton and Fulton, 2013),
global disaster alert and coordination system (GDAS) (De Groeve
et al., 2006) and so on. Some of these systems, such as USGS ENS
and GDAS, are global scale multi-hazard systems and can provide
information to users through Web services, Email, SMS and reports.
While, it still requires significant work to improve the existing
technology (Grasso and Singh, 2011). There is still a technological
and national capacity divide between systems in developed and
developing countries, and gaps in communication and geographical
coverage among different systems are still big (Grasso and Singh,
2011). To create software systems with unified standards for use
in other regions, universal frameworks, basic structure and inter-
national standards, such as Open Geospatial Consortium (OGC) and
Sensor Web Enablement standards, which can be used for real-time
observation achieving and event alerting (Broring et al., 2011; Chen
et al., 2013), were developed for the integration of heterogeneous
sensor systems (Lendholt and Hammitzsch, 2011). For EWS of water
quality, the US EPA (Environmental Protection Agency) has put
forward a framework which includes 12 desirable features (shown
in Table 1) (Hasan et al., 2005).
Based on the US EPA framework, many early warning systems
have been designed, but most of them do not exhibit the 12
desirable features and have several limitations including focusing
on individual components, pursuing data-mining and long-term
seasonal forecasting weather, climate, land degradation trends
(Hou et al., 2013), but most of them can't reliably predict short-term
impacts of water pollution and unable to respond rapidly (Quansah
Table 1
The 12 desirable features of the water quality EWS.
Contents Desirable characteristics
Speed (1) Responding rapidly
Function (2) Identify pollution source
(3) Accurate prediction of contaminants
(4) Sensitive analysis
Capable (5) Wide range of potential contaminants
(6) Continuous, predictable and year-round functioning
(7) Third party inspection
(8) Exhibit robustness and ruggedness
Cost (9) Affordable for acquisition, maintenance and upgrades
Operability (10) Requiring low skill and training
(11) Allowing remote operation and adjustment
(12) Automation
Adapted from Hasan et al., 2005.
79
et al., 2010). Because of limited data, it is difficult to obtain “accu-
rate prediction of contaminants” in ungauged river basins. Ac-
cording to the aim of answering the 2W2H questions and the
specific requirements of the application field, a fast simulation with
a relatively accurate trend of the spread of pollutants is needed. As a
result, the EWS should have: (1) an appropriate algorithm or
models for fast simulation, (2) a backend database with model
parameters and the characteristics data of pollutants and other
empirical data for system initialization, (3) reasonable methods of
generating digital terrain data for the models, (4) a rapid pollution
transport analytic method to quickly identify pollution source and
analyze the accidental release process, (5) intermediate results
verification method to improve the accuracy of the simulation, and
(6) graphical ways to clearly present the simulation results. (7)
Furthermore, the work of responding to the risk in ungauged river
basins needs to be able to respond to events rapidly and be avail-
able to decision makers during the event, which demands the EWS
be easy to use and convenient to access.
The EWS based on mobile devices makes sense; herein we name
it Mobile Early Warning System in Ungauged River Basins (MEW-
SUB). With a browser-based interface and Web services model, it
allows risk response personnel to use the system on any mobile
client. To make the model module in the system allow remote
operation and adjustment, the model should be designed as
service-oriented. The innovative middleware technology, especially
the service-oriented architectures (SOA) and Web services make it
possible. Web services is a standardized method, which is a
promising choice to implement a service oriented architecture and
its strategic objectives for the communication between client
application and web resource (Castronova et al., 2013; Lupp, 2008;
Usl€ander et al., 2010). For the design of environmental software
applications, the service oriented architecture style and the Web
services technology for backstage service and to provide online
services is increasingly used (Swain et al., 2015). The Hydro Desktop
built by CUAHSI is a Web services based software for hydrologic
data discovery, download, visualization, and analysis (Ames et al.,
2012). Also, another discovery platform is the Web services based
hydrologic data distribution system built for delivering remote
sensing data to the hydrologic community (Kanwar et al., 2010).
Recently, environmental modelers are using service-based ar-
chitectures to perform quantitative water quality modeling as well
(Kim et al., 2012; Park, 2014). With the technology of Web services,
internal business processes can be exposed as services and be made
accessible via the Internet, which allows users to operate MEWSUB
on client devices but does not require installation of the models and
their supporting database. To make the MEWSUB meet interna-
tional standards, the data delivery over the internet will be
normalized as a web mapping service (OGC, 2006) and web feature
service (OGC, 2014). The simulation data should be encoded for
communications between the client and server according to web
processing service (OGC, 2007). Once a client device is accessible to
the Internet, users can carry out the work of early warning with
MEWSUB.
3. Architecture and key capabilities of MEWSUB
3.1. Architecture of MEWSUB
The general architecture of MEWSUB and its system modules
including client and server is shown in Fig. 1.
The server adopts J2EE architecture and contains six modules.
(1) The database module stores the project data that contains the
user information, the historical input and output data of every
registered user and the empirical data like water quality model
parameters, pollutant chemical and physical property parameters
80 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89

Time (5)Correction Results

Module correcting acceptable
-Cross sections Output
-Concentration Data unacceptable
(1)Database (6)Management
(4)Model
S Numerical modeling odu e
Module Module
Module
E Projects Managing
R
Model
V (3)Pollutant Polluant release data Analyzing
results
E Module generation
R Pollutant
database Showing
(2)Terrain Terrain digitization River
Module data generating samples Sharing
database

Pre-processing
Result Showing

C Google Earth
L
I Web Search
E
N
T Professional User Ordinary Authenticated User

Fig. 1. The Architecture of MEWSUB.

and all kinds of river sample data. (2) The terrain module with the
automatic terrain generation method is responsible for generating
digital terrain data for the model. (3) The pollutant module with the
pollution source analytic method is responsible for generating the
time-series data of pollutant emissions. (4) The model module
integrating the One-dimensional water quality models is used to
perform numerical simulation and print out time-series of
pollutant concentration in every cross section. (5) The correction
module is responsible for model result error analysis and correc-
tion. (6) The management module is responsible for project and
model result management including creation, maintenance the
project data and model output data files, analyzing and showing
the result to answer the 2W2H questions.
HTML5 is the fifth revision of the HTML standard, and it contains
new features. The HTML5 Canvas is a new feature that can draw
graphics via scripting in JavaScript (Fulton and Fulton, 2013), which
is helpful for model results display. The client interface is based on
browsers supporting HTML5 to operate MEWSUB and call back-end
services. Generally, there are two kinds of users: professional users,
such as authorized administration staff, early warning experts, and
emergency managers who have super user permission to establish
individual user space, simulate the accident and store the simula-
tion results and execute all operations in the five modules; ordinary
authenticated users only have access to browse the shared results.
To be authenticated, the users should contact the system admin-
istrator to request an account.
The MEWSUB flow chart is shown in Fig. 2. As soon as a pro-
fessional user logs into the system, a project data directory will be
created under its personal database on the server. At the same time,
a user on the client interface will be asked to set the characteristics
of the river basin and the pollutant according to their field judg-
ment from a series of selections. The client submits the user's data
to the server, the server checks and analyzes the data, and then
returns the river sample data, pollution data and other
recommended initial data through a wireless network to the user.
At the same time, the client has interfaces to Google Earth and web
search engines, from which users can obtain data like the location
of the river, river width, elevation and other parameters of the
pollutant. According to the actual situation and the data users
obtain, professional users can change the digital data and pollutant
data (detailed data type will be described in Section 3.2) and send
them to the server. The server uses these data to create digital data,
to generate pollutant time-series data and parameter data, and to
start the model. When the intermediate results are computed, the
server sends them to the client for graphical representation or re-
view if the user thinks it is necessary (according to the error
assessment). After the project finishes, the user will be asked
whether to share the results or not. The shared results can be
browsed by other users. With the results, users can answer the
2W2H questions and make risk assessment and early warning
decisions.
To meet the requirements of convenient and instant access for
early warning in ungauged rivers, MEWSUB differs from other
existing systems, as listed in Table 2.
Unlike the other systems, MEWSUB aims to provide early
warning in ungauged river basins. And compared to other sensor
web systems, MEWSUB can work with few monitor data and focus
on smaller rivers’ water quality early warning. It is integrated with a
simple One-dimensional water quality model and the methods for
semi-automatically creating input data. MEWSUB can quickly
answer the 2W2H questions and meets the requirement of early
warning in the ungauged river basins.
3.2. Content of data files and databases in MEWSUB
The goal of MEWSUB is to answer the 2W2H questions for early
warning, with a simulation model serving as the main computing
center to provide results. A complete processes of operating
 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89 81

Professional User

Success Sign In

Success

Characteristics Setting

Selection
Creating
Accident
Place
& Recommendation Pollutant
River
Basin
Modification

Digital Terrain Data Creation Pollutant Data Generation

Project Directory Storing

Input Data Directory

Web services
Model Simulation

Web services

Output Data Directory

Disply Sharing Checking

Legend
Risk Assessment
On Client

Early Warning On Server

Fig. 2. The flow chart of early warning with MEWSUB.

MEWSUB is that the client makes a request by supplying input data
to the Web services and then the service interprets this input data
and executes its service-oriented computing model to produce the
desired output data (Castronova et al., 2013). These processes can
be divided into pre-processing for input data and post-processing
of output data with the model. These kinds of input data that can
describe the characteristics of the river, the accident, the pollutant
and control the model are needed. The output data being sent back
to the client application for further processing or visualization are
also requisite. As shown in Table 3, there are six basic data files for
MEWSUB.
As the data in ungauged river basins are limited, one of the most
important capabilities of MEWSUB is being able to quickly generate
the input files through different approaches. Historically, users
acquire the data in ungauged river basins through field observation,
online search and personal experience. Through field observation
or individual's experience, some parameter data and digital terrain
data that the model needs are difficult to obtain. The database
together with the automatic generation methods in different
modules are helpful for input data file generation.
There are two main databases in the system. ①The river sample
database contains data including cross section shape, longitudinal
slope, roughness, river width and other hydrology parameters
about rivers from existing investigation data of different basins
throughout China. Users can find a similar river where pollution
accidents once occurred to select basic attribute characteristics
including the position (South, North and etc.), regional terrain
(plain or mountain) and riverbed characteristics. ②The pollutant
database contains various common pollutants with their chemical
name, molecular weight, quality, density, solubility and volatility
properties, degradability, poisonousness and treatment methods in
different environments and different regions; all these character-
istics are parameterized in advance. Unlike other early warning
systems, in which to make the early warning the topographic data
82 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89

Table 2
Differences between MEWSUB and other existing early warning systems.

Category Early warning systems

MEWSUB DEWS (Hou et al., 2013) EWSs recommended by USEPA (Hasan et al.,
2005; Quansah et al., 2010)

Goal
Input data requirement
Scope of main application
Computational methods/
models integrated
Results
Early earning in ungauged river basins
Semi-Automatically created in field,
data from field observation and existing
database
Ungauged river basins
One-dimensional water quality model
Arriving time, peak value of the
pollutant; duration and area of the over
standard pollutant
Early warning for event-driven,
urban, drinking water quality
Must prepared in advance, collect
data using monitoring networks
Rivers and lakes
One-dimensional models for river
and two dimensional models for
lake; two-dimensional risk matrix
Probability of a water quality
contamination event;
concentration and risk value of the
pollutant
Reliably monitoring surface water systems for
traces of chemicals, radioactivity, pathogens,
and biotoxins.
Must prepared in advance, received data from
monitoring sites.
Lakes, reservoirs, rivers, and groundwater
systems
Flow and spatial models, hydraulic model and
water quality models, mathematical/statistical
software.
Accurate location and concentration value at
the downstream of the detection point and etc.

and pollutant data should be prepared in advance, MEWSUB pro-
vides automated methods to create input data.
3.3. Methods for creating input data
3.3.1. Digital terrain data generation
The geo.dat file describes the longitudinal and vertical
morphology of a river. The data in the geo.dat file are organized in
two groups: group A describes the number of sections; group B
describes characteristics of every section which contains the dis-
tance from the current section to the first section and the number of
points in the section. Group C describes every point in the section.
Group A : NumS
Group B : IDS DistS NumP
Group C : IDP DisP ElevP
Where NumS is the number of sections, IDS is the serial number
of sections; DistS is the distance from section IDS to the first section,
NumP is the number of points in the IDS section, IDP is the serial
number of points from the river's left to the right, DisP is the dis-
tance from point IDP to the first point, ElevP is the elevation of the
point IDP. Group C loops constantly until IDP equals NumP in the
same section, Group B loops until all sections are described.
For the geo.dat file, the most important part is to obtain the data
of DisP and ElevP which are unknown in ungauged river basins.
Sometimes we cannot accurately measure the river bed elevation
but we can accurately measure the river bank elevation using GPS
equipment or even a mobile phone from where we stand. Once the
river bank elevation is obtained, the river bed elevation will be
calculated from the river bank elevation minus the river depth. By
Table 3
Data files for MEWSUB.
Name Contents and meanings
geo.dat Location, width and elevation of every points in cross sections
kp.dat Pollutant degradation coefficient and diffusion coefficient parameter
cn.dat Channel roughness parameter of every cross section
qz.dat Water quantity process of the upstream and water level process of
downstream
acc.dat Pollution discharge process in accidents
res.dat Simulation results with pollutant concentrations at every
cross section
con.dat Control parameters (time step, output time interval and etc.)
setting the place where the user is standing as the first section, then
the elevation of other sections can be estimated with the following
function.
elk ¼ el0  dlk  i (4)
Where elk is the elevation of section k, el0 is the elevation of the
first section, dlk is the distance from section k to the first section,
and i is the longitudinal slope. The recommended river depth and
longitudinal slope will be given when the river topology type is
selected or obtained through field observation. Meanwhile, the
section form style of the ungauged rivers can be conceptualized
into typical Shape U or Shape W or Shape V (Broekert and
Sandiford, 2005) as shown in Fig. 3. To quickly get DisP and ElevP
for every point in the section, these three section shapes are
described with three generalized methods.
As shown in Fig. 3, assuming point A is the starting point of
sections at the river's left bank, dh is the current channel depth, dB
is the channel width, Len is the distance from the current section to
the first section, el0 is the river bank elevation of first section, i is
the longitudinal slope, a is the angle between the river bank and
river bed. The coordinates of each point in sections are expressed as
P (DisP, ElevP), where DisP is the distance from the current point to
the starting point A, ElevP is the elevation of the current point.
As shown in Fig. 3, a V shape is divided into three points: A, B
and C; the default DisP value of B is dB/2 and default a is ATAN
(2*dh/dB); U shape is divided into four points: A, B, C and D; the
default DisP value of B is dB/4, the default DisP value of C is 3*dB/4
and default a is TAN (4*dh/dB); W shape is generalized as five
points: A, B, C, D and E, the default DisP value of B is dB/4, the
default DisP value of C is dB/2, the default DisP value of D is 3*dB/4
and default ElevP of C is dh/4. Every point in U, V, and W will be
calculated with the following functions.
8
< Að0; el0  i*LenÞ;
V Bðdh=TANðvÞ; el0  i*Len  dhÞ (5)
:
CðdB; el0  i*LenÞ
8
>
> Að0; el0  i*LenÞ;
<
Bðdh=TANðvÞ; el0  i*Len  dhÞ
U (6)
>
> CðdB  dh=TANðvÞ; el0  i*Len  dhÞ
:
DðdB; el0  i*LenÞ
 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89 83

dB dB dB
Z C A Z Z
A D A E

dh
dh dh

B B C B D
V Shape U Shape W Shape

Fig. 3. The three profile shapes of rivers.

8 Table 4
>
> Að0; el0  i*LenÞ;
>
>
>
> Bðdh=TANðvÞ; ; el0  i*Len  dhÞ
>
< Emission patterns
Cðdisc; ðel0  i*Len  dh  ðdisc*TANðbÞÞÞ
W (7)
>
> disc ¼ ðdB  2dh=TANðvÞÞ=2 þ dh=TANðaÞ
>
> ②
>
> DðdB  dh=TANðvÞ; el0  i*Len  dhÞ
>
: EðdB; el0  i*LenÞ
The three types above allow users to add, move and modify
control points and their related values. Users can select the type of
the cross-section for the entire river at once. The system also pro-
vides a function for users to define the type of the cross-sections
one by one or reach by reach (one river reach contains a set of
sequentially numbered cross sections).
3.3.2. Accident temporal release process data generation
When an accident pollution plume is detected, to conduct water
quality simulation, the quantity of pollutant leaked is required
(Hazart et al., 2014). The pollutant release process plays the most
important role in the process of pollutant transport in the water.
There are three basic emission models of emergency pollutant
emissions defined in the system, as shown in Fig. 4. (1) The trap-
ezoidal discharge pattern, in which the discharge amount of pol-
lutants increases from zero, then maintains a fixed value in the later
stage, and finally reduces to zero (for example, the enterprises
abidingly discharge pollutants without permission). (2) The
parabola discharge pattern, which shows that the pollutants are
dumped into the water from zero increase to the peak and then
reduces to zero like the parabolic curve (such as the pollutants leak
into the water in a sudden transport pollution accident). (3) The
triangular discharge pattern, it displays the amount of pollutants
released linearly increases to the peak without duration and reduce
to zero (for instance, the enterprises instantaneously discharge
pollutants without permission in a short time).
In a field emergency disposal activity, the probable value of the
total mass of pollutant (mp), pollutant density (r), total emission
time (tem), and pollutant emission pattern can be estimated by the
Recommend selection of basic pollutant emission patterns.
Pollution sources/situations
Trapezoidal ① Sewage outlet continuing emission;
hazardous chemical leak from tanker
Parabola ③ Pollution caused by rainstorm or traffic accident;
④ non-point source pollution
Triangular ⑤ Sewage outlet instantaneous emission;
⑥ fuel tanker explosion
emergency personnel. It is very difficult to judge detailed release-
rate or mass of pollutants in the temporal release process.
Through generalizing the above three emission patterns shown in
Fig. 4 and using control points to calculate the emission process, it is
possible to obtain detailed emission rates at different times during
the emission process. The trapezoidal pattern is controlled by four
pollution points, the parabolic pattern is generalized as eight
pollution control points, and the triangle pattern includes three
pollution control points. The default time of all points in the time
axis is shown in Fig. 4, and the area under the line pattern is equal
to mp. As a result, the emission process with every point (time,
emission rate) will be created. There are several pollution sources/
situation defined by the system including the pollution caused by
sewage outlet, tanker, rainstorm and so on as shown in Table 4. In
an emergency situation, the pollution sources/situation should be
first selected by the users, then the system recommends the deci-
sion of pollutant discharge patterns for different pollution sources
as shown in Table 4.
To create the detailed emission data, the total mass of pollutant
(mp), pollutant density (r), total emission time (tem) are necessary
and should be inputted by the users. Then, the system will create
the series pollutant emission data according to the selected
discharge pattern and show the data in a curve diagram. The users
can change the data by adding, deleting or modifying t data points
according to the practical situations. If the modified discharge data
create acceptable results, which will be described in Section 3.4, the

mass mass B
B C E
D F

C
kg/min

kg/min
kg/min

B H

A tem/4 3*tem/4 D A I A C
0 min tem 0 0 tem/2 min
0 0 tem/4 min tem 0
(a) (b) (c)

Fig. 4. Three basic patterns of pollutant emission process: (a) is the trapezoidal discharge pattern, (b) is the parabola discharge pattern and (c) is the triangular discharge pattern.
84 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89
system will store the time series pollutant emission data together
with the mp, r, tem and the pollution situation (as a key field) to the
pollutant database as a pollutant emission sub-database. When the
other users select the pollution situation and input the mp, r, tem, in
the next incident, the system will first search the sub-database. If
there is one pollution situation stored in the sub-database match-
ing the new pollution situation, the time series pollutant emission
data of that situation will be recommended to the users. If there is
no matched pollution situation, one of the basic emission patterns
will be recommended.
In addition, if there are monitoring sites downstream of an
incident position, the changing pollutant concentration at different
times can be measured. The system provides a custom discharge
module in which users can input monitoring concentration values
at different times, and the system can convert the concentration
value to pollutant loads according to the current flow. The model
can be started by regarding the current monitoring position as a
starting calculation section and the monitoring data as the emis-
sion process data.
3.4. Intermediate results correction
The aim of Intermediate Results Correction is to check the ac-
curacy of simulation results of models to enhance performance
indices of models. As soon as an emergency incident happens, the
preliminary simulation can quickly calculate the polluted area
range and the pollutant concentration downstream, which is
helpful for risk assessment and emergency monitoring site selec-
tion. The result error of the models using the uncertain condition
data is bigger than which using the certain condition data
(Reckhow, 1994). The data from the monitoring sites between the
accident place and important regions will be helpful for providing
data and reducing modeling errors. It will take some time before
the pollutant arrives at the important regions downstream, which
makes it possible to decide where and when to set up monitoring
sites according to the results in the previous simulation. As shown
in Fig. 5, when the monitoring data (time ranges from t1 to tin) at the
monitor point of downstream is obtained, pollutant concentration
and the monitor time can be input to the system and be compared
with simulation results (time range from 0 to T). The system will
automatically compute the relative error and absolute error and
give a comprehensive error evaluation conclusion. If users are
satisfied with simulation results, the system will go on running the
model following the previous conditions. If users think the errors
between simulation result and monitoring values are unacceptable,
they can reset the pollutant release process data and the pollutant
degradation coefficient or the diffusion coefficient parameter and
restart the simulation. At this time (defined as tind), pollutants have
arrived at the monitoring point, so it doesn't make sense to monitor
up stream.
For early warning, the most important job is to simulate
pollutant concentration for downstream locations, where the
pollutant has not yet arrived. The monitoring section can be treated
as a new pollutant source and the time series monitor data can be
treated as new pollutant emission process data. Compared with the
data gained through emission models, the monitoring data is more
accurate. Restarting the simulation with the new and more accurate
pollutant release process data will result in error reduction and
simulation accuracy improvement. At the same time, the results
will reflect pollutant transfer processes more practically.
3.5. Output data demonstration
There are two basic methods to display the simulation results on
the client. One is dynamic maps which can display the changing
simulation results
monitoring value
error
acceptable go on
0 tem t1 tind running
emission process T
important regions
river bed start section
monitoring section simulation results
monitoring value
0 t1 tind T
error unacceptable
monitoring value
t1 tind T
monitoring value as new pollutant emission process data
important regions
river bed
monitoring section as new calculating start section
Fig. 5. Schematic diagram of intermediate results correction progress. Using the
monitoring value for model results error analysis: ① if the error is acceptable, the
simulation will go on running with the previous condition data, ② if the error is un-
acceptable, the monitoring value data will be used as the new pollutant emission
process data and the monitoring section will be set as a new start section for starting a
new simulation.
process of pollution in the river dynamically, the other is the trend
line. The realization of dynamic maps is based on HTML5 tech-
nology. Unlike the previous version, HTML5 technology can not
only be used to represent texts and pictures, but also offers seam-
less integration of multimedia and graphical content to the web, on
which videos, audios, images and animations are standardized
(Grensing-Pophal, 2011). The canvas tag provided by HTML 5 makes
it possible to display maps dynamically in the browser, which
makes it unnecessary for users to install flash plug-ins or GIS plug-
ins.
First, the server will scan the user's project data directory, read
res.dat files and send them to the client through the Web services
interface. The Web services interface was developed in Java lan-
guage, including the createFiles function used for receiving model
parameter data including terrain file data, roughness data, hydro-
dynamic data, pollutant discharge process data and simulated
operation time from the client and the getResult function which is a
non-parameter constructor and used for reading the results and
sending them back to the client.
Second, according to the predetermined relationship between
water quality grade and color bar, the colors corresponding to each
pollutant concentration in the cross section will be calculated.
Third, on the basis of HTML5 Canvas technology and colors for
each section, results will be visualized for the river section layer, so
as to present animation effects.
In this process, the browser will show concentration trend lines
of all sections at different times below the dynamic map. Users can
clearly see the pollutants peak on the current moment and specific
density of each section. At the same time, the system will deliver
the area of contamination and pollutant disposal method to the
customers. According to this information, users can quickly assess
risk and develop an early warning plan.
 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89
4. Case studies
China is a developing country in which a large number of wa-
tersheds are ungauged. In the most populous country of the world,
many Chinese people live in ungauged river basins. Once an acci-
dent occurs in these areas, many people's daily lives will be
affected. Most of the rivers in China have their own river man-
agement institutions, and as soon as an accident happens, the
institution will quickly organize an emergency response team to
analyze the accident and put forward the emergency response
scheme. On 23:00 06/04 2011, 20 metric tons of toxic chemical
phenol leaked into the Xin'an River in a tanker truck accident (as
shown in Fig. 5), which brought great panic among residents and
managers for there are several drinking water intakes downstream
(Hou et al., 2013). The accident was soon reported to the related
government institution, and an expert group was quickly orga-
nized. MEWSUB was applied in this emergency for early warning.
4.1. Process of using MEWSUB
The Xin'an River is an important source to the Fuchun River; it
meets with the other branch of Qiangtang River, called Lanjiang, at
Qianmen. The Fuchun River flows to Hangzhou city and is also
named Qiantang River (Zhou et al., 2006). There are hydrological
monitoring stations which can provide flow velocity and quantity
data in the river. After the accident happened, the government
quickly deployed eight water sampling sites (shown in Fig. 6), and
the water quality parameter data were monitored hourly. These
data were used in MEWSUB and during the emergency situation,
additional information was obtained through the internet, Google
Earth and field observation. The pre-processing operation process
in MEWSUB is shown in Fig. 7.
(1) Digital terrain data generation: Input the coordinates of the
accident place or select the place through the coupled Google
Earth. If the river is clearly shown in Google Earth, it is
possible to measure its width, length and even elevation
which will be used in the next step. Xin'an River is located in
a mountainous area, while the Fuchun River and the Qian-
tang River are located in the plain area. The database stores
Fig. 6. Application river map, accident location and the sampling sites in the river.
85
the recommended longitudinal slope data that the mountain
is 30%e40%, and the plain area is from 2.5% to 4%. A value of
33% was selected as the longitudinal slope of Xin'an River
and 3% as the longitudinal slope of the Fuchun and Qiantang
River. From Google Maps, a distance of 148 km was measured
from the accident place to Hangzhou City. Setting the sepa-
ration distance of each cross section at 500 m, 296 cross
sections were created. The width of the section was
measured from Google Maps ranging from 600 m to 1200 m.
Input for the river bank elevation of the first cross section
was 40.2 m and its water depth was 18 m. The U shape was
selected as the section shape style. After modifying the
characteristic data of some cross sections, the geo.dat file was
created.
(2) Hydrology parameter determination: After selecting a river
with similar vegetated cover and geologic structure of the
river bed and river bank similar to the river where the ac-
cident occurred, the channel roughness parameter from
0.010 to 0.015 was recommended. The upstream water
quantity process of the Xin'an River (range from 250 to
400 m3/s), and downstream water level process of Qiantang
River (range from 3.5 to 4.2 m) were provided by the hy-
drometric station. With these data inputted, the qz.dat file
was generated.
(3) Accident pollution data creation: After selecting the chemical
phenol as the accident pollutant, MEWSUB recommended a
degradation coefficient ranging from 0.012 to 0.018, a vola-
tilization coefficient of 1.52, a degradation coefficient of
0.0198 and an absorption coefficient of 0.002. When the user
accepted these data, then the kp.dat (the parameter data file
as shown in Table 3.) would be created. As the accident was a
traffic accident, parabolic mode was selected as its discharge
mode, the total mass of 20 mt was set and the discharge time
of 6 h was selected, then acc.dat (the pollution discharge
process data file as shown in Table 3.) was created.
(4) Model simulation time: A simulation time of 2 days was
selected, the start time of 06/04 2011 23:00 selected, and the
model was run.
4.2. Results and early warning
(1) Results Display and Correction
According to the Chinese National Surface Water Quality Stan-
dards (GB3838-2002), concentrations of the chemical phenol can
be categorized into five levels (I and II  0.002 mg/L, III  0.005 mg/
L, IV  0.01 mg/L and V  0.1 mg/L), which will be displayed in five
colors. As shown in Fig. 8, the upper map shows the phenol con-
centration of every cross section with colors. The time value 11:02
means the current time is 11 h and 2 min after the start time. The
peak value presents the maximum value of current pollutant con-
centration among all the cross sections, and the distance value
shows the distance from polluted section whose water quality
concentration is changed after the accident to the starting section.
The trend line below shows the value of current pollutant con-
centrations at every cross section from the location of the accident
to the last section. The two results can change dynamically through
time. Furthermore, users can click “search” to query the detailed
values including pollutant concentration at every output time,
when the pollutant will arrive, and how long the pollutant con-
centration value above the warning value will last by clicking the
mouse pointer.
As shown in Fig. 6, A is the accident location. 16 km away from A
to Mamu Bridge, and Yanlingwu is located 40 km away from A. The
86 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89

Fig. 7. The pre-process operation progress in MEWSUB: interface (a) is the initial page; interface (b) is the page for selecting the regional terrain of rivers; interface (c) is the page for
selecting the profile shapes of rivers, interface (d) is the page for pollutant emission process setting.

Fig. 8. Simulation result and measured values at Mamu Bridge cross section and Yanlingyu cross section.

dynamic map shows that the concentration of the pollutant at
section 32 will peak beyond 0.1 mg/l within almost 10 h, and the
water quality in level V will last 30 h. The concentration peak
beyond 0.005 mg/l will reach section 80 in 18 h. The two sections'
simulation results and measured values are shown in Fig. 8. It
demonstrates that the simulation results of the measured value
shows the same trend at Mamu Bridge and Yanlingwu, but the error
at Yanlingwu is unacceptable. This phenomenon in Yanlingwu may
be caused by a tributary named Lanjiang River flowing into Xin'an
River, which is not considered in the model.
(2) Water quality in the water intake area
The public water system water intakes are the highest priority
locations during the accident. As soon as the accident happens, the
people want to know whether the pollution will reach the water
intake. There are many water intakes in the river, and three sig-
nificant locations were chosen as sample sites. Jiangbei
waterworks, which is 58 km away from the new starting section, is
the main waterworks of Fuyang City. Xiaoshan waterworks, the
most important waterworks in Xiaoshan District is 87 km away.
Jiuxi waterworks is 98 km away which is the biggest waterworks in
Hangzhou and takes the responsibility for providing water to half of
the Hangzhou people. At the water intake area, the safe concen-
tration is 0.005 mg/l, beyond which there will be a warning and the
waterworks should stop pumping water from the river. To correct
the results, Yanlingwu was set as the new calculating start section,
the measured value was used as the discharge value of the accident
pollution, and the starting time was set as 06/05 2011 10:00 am and
the simulation time as 2 days. The results are shown in Fig. 9.
The phenol concentrations in the three waterworks are all
below 0.005 mg/l in the next two days, which means that the ac-
cident would not cause an adverse impact on the water quality of
the three drinking water sources. These data display the same trend
as the measured data. The phenol is only slightly soluble in water
and some of the phenol may sink into the river sediment, and
 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89 87

Fig. 9. The correction simulation results at three different times. The horizontal axis “Distance (km)” is the distance from the downstream cross sections to the accident location, the
vertical axis represents the pollutant concentration (mg/l).

therefore construction projects or river dredging work which will 5. Conclusions

stir the river water and sediment in the accident river from A to
Mamu Bridge should be controlled to avoid secondary pollution. Rapid early warning for water quality emergency risk in unga-
uged river basins is important for accident evaluation and emer-
4.3. CPU time needed for early warning using MEWSUB gency response. Due to data limitations, it is very difficult to
provide early warning in ungauged river basins with existing sys-
To evaluate the speed of the MEWSUB system, we designed tems quickly. According to the analysis of characteristics of water
three different situations, different river lengths and simulation quality emergency risks and the requirements of early warning in
times, with 1 km distance between cross sections and other pa- ungauged river basins, a series of technologies for quickly obtaining
rameters were the same as shown in Section 4.1, to test the system. and creating model data, water quality model correction, and
The MEWSUB was tested using a laptop-ThinkPad E440, with model results visual dynamic demonstration in mobile devices was
operating system Windows 7 Professional, dual processor (2.5 GHz proposed. Based on these technologies, the architecture of a mobile
Core i5-4200M), RAM(4 GB DDR3) and graphics coprocessor (Intel device early warning system (named MEWSUB) and successful
HD Graphics 4600). The system was running in the Google Chrome warning early application in ungauged river basins were described.
(version 43.0.2357.78). Table 5 shows the CPU time needed for the MEWSUB is aimed at water quality early warning in ungauged river
different simulation situations. basins. MEWSUB provides a browser-based interface and Web
From Table 5, simulating a 500 km river with 500 cross sections services model, and it allows risk response personnel to use the
for 2 days requires 6.898 s, and for 6 days simulation, 19.4 s is system anyplace, even if data are limited and on any mobile client.
needed. This indicates that for early warning through MEWSUB, the MEWSUB has five modules and the following capabilities.
time consumed is acceptable.
(1) Interface to Google Earth, it can quickly determine the inci-
dent location on the map according to GPS information in
Table 5 mobile devices and measure the width, distance and eleva-
CPU time needed for different simulation situation. tion of rivers.
River length Simulation CPU time needed (2) River sample database to provide recommended data to in-
(kilometer) time (day) (milliseconds) crease data in ungauged river basins.
100 2 2366 (3) Based on the automated terrain generalized method and
4 4474 pollution source analytic method, it can quickly generate
6 6520 digital terrain data, pollutant discharge process data, and
200 2 3740
coefficient parameter data for the model. With the inter-
4 6618
6 9523 mediate results verification method, it can correct simulation
500 2 6898 results and improve accuracy of the simulation.
4 13,332 (4) Quickly simulate the change tendency of the pollutant, pro-
6 19,423
vide results of when the pollutant will arrive at downstream
88 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89
sensitive sites, what the peak value of the pollutant will be,
how large area the pollution zone will be, and how long the
pollutant concentration value will keep beyond the security
value. These results can be presented through a dynamic
render map and a trend line diagram via HTML5.
(5) Evaluate the water environment risk, provide early warning
information, and put forward emergency response schemes
to the environment management.
MEWSUB has been successfully used in the early warning of
several accidents in rivers including Xin'an River, Fujiang River, and
Jialing River in China. It was inspired by the Yangtze River Water
Resources Protection Bureau and installed on mobile devices of
managers in the water resources protection bureaus. Recently,
MEWSUB has been packaged into an Android app and is free for
authorized users to download and use.
There are still modules which need to be improved in the future.
For instance, priorities for the MEWSUB development and main-
tenance team would include extending the river sample database
and pollutant database, improving the automated terrain general-
ized method for Two-dimensional grid generation, strengthening
the model to include a Two-dimensional option, and consideration
of tributary influence in the model. Many EWSs integrate with each
other using open OGC geospatial standards to address more com-
plex problems. Combining MEWSUB with other early warning
systems will be the target of ongoing efforts.
Acknowledgments
This work is funded by the major science and technology pro-
gram for water pollution control and treatment of China (No.
2013ZX07503-001) “Water Environment Risk Assessment and
Early warning Technology Research and Demonstration Project in
The three Gorges Reservoir Area and Upstream Watershed”. The
authors would like to thank the Hydroco Science and Technology
Ltd, for their excellent work and support to MEWSUB, and all users
of MEWSUB, for their feedback regarding application and software
improvement proposals. We gratefully acknowledge the three
anonymous reviewers for their comments to the research and
manuscript. Specially, we also wish to thank the Agricultural &
Biological Engineering, Purdue University for providing an excel-
lent research environment and facilities.
References
Adams, J.A., McCarty, D., Crousore, K., 2006. A real-time early warning system for
pathogens in water. Chem. Biol. Sens. VII 6218, 2180e2180.
Ames, D.P., Horsburgh, J.S., Cao, Y., Kadlec, J., Whiteaker, T., Valentine, D., 2012.
HydroDesktop: web services-based software for hydrologic data discovery,
download, visualization, and analysis. Environ. Model. Softw. 37, 146e156.
Broekert, P., Sandiford, M., 2005. Buried Inset-Valleys in the Eastern Yilgarn Craton,
Western Australia: geomorphology, age, and allogenic control. J. Geol. 113 (4),
471e493.
Broring, A., Echterhoff, J., Jirka, S., Simonis, I., Everding, T., Stasch, C., Liang, S.,
Lemmens, R., 2011. New generation sensor web enablement. Sensors (Basel) 11
(3), 2652e2699.
Burchard-Levine, A., Liu, S., Vince, F., Li, M., Ostfeld, A., 2014. A hybrid evolutionary
data driven model for river water quality early warning. J. Environ. Manag. 143,
8e16.
Castronova, A.M., Goodall, J.L., Elag, M.M., 2013. Models as web services using the
Open Geospatial Consortium (OGC) Web Processing Service (WPS) standard.
Environ. Model. Softw. 41, 72e83.
Chen, N., Chen, X., Wang, K., Niu, X., 2013. Progress and challenges in the archi-
tecture and service pattern of Earth Observation Sensor Web for Digital Earth.
Int. J. Digital Earth 7 (12), 935e951.
Chen, N., Wang, K., Xiao, C., Gong, J., 2014. A heterogeneous sensor web node meta-
model for the management of a flood monitoring system. Environ. Model.
Softw. 54, 222e237.
De Groeve, T., Vernaccini, L., Annunziato, A., 2006. Global disaster alert and coor-
dination system. In: Van de Walle, B., Turoff, M. (Eds.), Proceedings of the 3rd
International ISCRAM Conference, pp. 1e10. Newark.
Dong, Y., Li, G., 2014. Mobile application for hydrogeologic field investigations.
Environ. Model. Softw. 53, 62e64.
Erdik, M., Fahjan, Y., Ozel, O., Alcik, H., Mert, A., Gul, M., 2003. Istanbul earthquake
rapid response and the early warning system. Bull. Earthq. Eng. 1 (1), 157e163.
Fiedrich, F., Van de Walle, B., 2009. The fifth international conference on informa-
tion systems for crisis response and management, Washington DC, May 4e7
2008. Int. J. Inf. Syst. Crisis Response Manag. (IJISCRAM) 1 (1), 70e74.
Fulton, S., Fulton, J., 2013. HTML5 Canvas. O'Reilly Media, Inc.
Grasso, V.F., Singh, A., 2011. Early Warning Systems: State-of-art Analysis and Future
Directions. Draft report. UNEP.
Grayman, W., Deininger, R., Males, R., 2001. Design of Early Warning and Predictive
Source-water Monitoring Systems. AWWA Research Foundation Report,
(Denver).
Grensing-Pophal, L., 2011. HTML5 options and opportunities. Econtent 34 (2),
10e12.
Gupta, C., Mishra, S.K., 2015. Disaster management: role of remote sensing and GIS
technology. RIET-IJSET Int. J. Sci. Eng. Technol. 2 (1), 1e10.
Hasan, J., Goldbloom-Helzner, D., Ichida, A., Rouse, T., Gibson, M., 2005. Technolo-
gies and Techniques for Early Warning Systems to Monitor and Evaluate
Drinking Water Quality: a State-of-the-art Review (DTIC Document).
Hazart, A., Giovannelli, J.-F., Dubost, S., Chatellier, L., 2014. Inverse transport prob-
lem of estimating point-like source using a Bayesian parametric method with
MCMC. Signal Process. 96, 346e361.
He, Q., Peng, S., Zhai, J., Xiao, H., 2011. Development and application of a water
pollution emergency response system for the Three Gorges Reservoir in the
Yangtze River, China. J. Environ. Sci. (China) 23 (4), 595e600.
Hou, D.B., Ge, X.F., Huang, P.J., Zhang, G.X., Loaiciga, H., 2014. A real-time, dynamic
early-warning model based on uncertainty analysis and risk assessment for
sudden water pollution accidents. Environ. Sci. Pollut. Res. 21 (14), 8878e8892.
Hou, D.B., Song, X.X., Zhang, G.X., Zhang, H.J., Loaiciga, H., 2013. An early warning
and control system for urban, drinking water quality protection: China's
experience. Environ. Sci. Pollut. Res. 20 (7), 4496e4508.
Hrachowitz, M., Savenije, H., Blo €schl, G., McDonnell, J., Sivapalan, M., Pomeroy, J.,
Arheimer, B., Blume, T., Clark, M., Ehret, U., 2013. A decade of Predictions in
Ungauged Basins (PUB)da review. Hydrol. Sci. J. 58 (6), 1198e1255.
Hu, B., Hu, B., Wana, J.Z., Nie, H.L., Zhai, C.Z., 2009. Safe river water: a ubiquitous and
collaborative water quality monitoring solution. Pervasive Mob. Comput. 5 (5),
419e431.
ISCRAM, E., French, S., 2011. Welcome to ISCRAM2011. Newsletter 1 (2).
Kanwar, R., Narayan, U., Lakshmi, V., 2010. Web service based hydrologic data
distribution system. Comput. Geosci. 36 (7), 819e826.
Katiyar, N., Hossain, F., 2007. An open-book watershed model for prototyping space-
borne flood monitoring systems in International River Basins. Environ. Model.
Softw. 22 (12), 1720e1731.
Kim, J., Engel, B.A., Park, Y.S., Theller, L., Chaubey, I., Kong, D.S., Lim, K.J., 2012.
Development of web-based load duration curve system for analysis of total
maximum daily load and water quality characteristics in a waterbody.
J. Environ. Manag. 97, 46e55.
Krzhizhanovskaya, V.V., Shirshov, G., Melnikova, N., Belleman, R.G., Rusadi, F.,
Broekhuijsen, B., Gouldby, B., Lhomme, J., Balis, B., Bubak, M., 2011. Flood early
warning system: design, implementation and computational modules. Procedia
Comput. Sci. 4, 106e115.
Kussul, N., Mandl, D., Moe, K., Mund, J.-P., Post, J., Shelestov, A., Skakun, Sergii,
Szarzynski, J., Langenhove, G.V., Handy, M., 2012. Interoperable infrastructure
for flood monitoring: SensorWeb, grid and cloud. IEEE J. Sel. Top. Appl. Earth
Observ. Remote Sens. 5 (6), 1740e1745.
Lendholt, M., Hammitzsch, M., 2011. Generic information logistics for early warning
systems. In: Proceedings of the 8th International ISCRAM Conference, Lisbon.
Li, N., Fu, Z., Wang, B.Q., Zhang, X.S., 2006. FDEWS 2.0: a Web-GIS-based early
warning system for fish disease via water quality management. In: WCICA
2006: Sixth World Congress on Intelligent Control and Automation. Conference
Proceedings, vols. 1e12, pp. 2570e2574.
Li, N., Wang, R.M., Zhang, J., Fu, Z.T., Zhang, X.S., 2009. Developing a knowledge-
based early warning system for fish disease/health via water quality manage-
ment. Expert Syst. Appl. 36 (3), 6500e6511.
Lupp, M., 2008. OGC Web Services. In: Encyclopedia of GIS. Springer, pp. 799e800.
Meissen, U., Voisard, A., 2008. Increasing the effectiveness of early warning via
context-aware alerting. In: Proceedings of the 5th International Conference, on
Information Systems for Crisis Response and Management (ISCRAM),
pp. 431e440.
McIntyre, Neil R., Wagener, T., Wheater, Howard S., Yu, Zeng Si, 2003. Uncertainty
and risk in water quality modelling and management. J. Hydroinf. 05 (4),
259e274.
Niadas, I.A., 2005. Regional flow duration curve estimation in small ungauged
catchments using instantaneous flow measurements and a censored data
approach. J. Hydrol. 314 (1e4), 48e66.
OGC, 2006. OpenGIS Web Map Server Implementation Specification.
OGC, 2007. OpenGIS Web Processing Service.
OGC, 2014. OGC Web Feature Service 2.0 Interface Standard.
Park, Y.S., 2014. Development and Enhancement of Web-based Tools to Develop
Total Maximum Daily Load. Purdue University.
Pasquini-Descombs, H., Brender, N., Maradan, D., 2014. Economic considerations in
risk management. In: IDRC Davos 2014: Extended Abstracts, Oral Presentations,
Special Panels, Sessions and Workshops, Davos, 24e28 August 2014,
pp. 112e115.
 Y. Wang et al. / Environmental Modelling & Software 73 (2015) 76e89
Quansah, J.E., Engel, B., Rochon, G.L., 2010. Early warning systems: a review. J. Terr.
Observ. 2 (2), 5.
Reckhow, K.H., 1994. Water quality simulation modeling and uncertainty analysis
for risk assessment and decision making. Ecol. Model. 72 (1), 1e20.
Rudloff, A., Lauterjung, J., Münch, U., Tinti, S., 2009. Preface “the GITEWS Project
(German-Indonesian Tsunami Early Warning System)”. Nat. Hazards Earth Syst.
Sci. 9 (4), 1381e1382.
Sivapalan, M., Takeuchi, K., Franks, S., Gupta, V., Karambiri, H., Lakshmi, V., Liang, X.,
McDonnell, J., Mendiondo, E., O'connell, P., 2003. IAHS Decade on Predictions in
Ungauged Basins (PUB), 2003e2012: shaping an exciting future for the hy-
drological sciences. Hydrol. Sci. J. 48 (6), 857e880.
Smakhtin, V., 2001. Low flow hydrology: a review. J. Hydrol. 240 (3), 147e186.
Song, X.X., Hou, D.B., Huang, P.J., Zhang, G.X., 2013. Research on dynamic early
warning technology of water pollution emergency based on pollutant disper-
sion simulation. Energy Eng. Environ. Eng. 316e317, 682e685. Pts 1 and 2.
Storey, M.V., van der Gaag, B., Burns, B.P., 2011. Advances in on-line drinking water
quality monitoring and early warning systems. Water Res. 45 (2), 741e747.
Swain, N.R., Latu, K., Christensen, S.D., Jones, N.L., Nelson, E.J., Ames, D.P.,
Williams, G.P., 2015. A review of open source software solutions for developing
water resources web applications. Environ. Model. Softw. 67, 108e117.
UN/ISDR, 2006. Developing early warning systems: a checklist. In: Third Interna-
tional Conference on Early Warning EWC III: Bonn, Germany.
UNISDR, 2004. Terminology: Basic Terms of Disaster Risk Reduction. United Na-
tions, Geneva.
89
Usl€
ander, O., 2014. Early warning e some recent developments. Planet@Risk 3 (1),
24e32.
Usl€
ander, T., 2015. The trend towards the internet of things: what does it help in
disaster and risk management? Planet@Risk 3 (1), 140e145.
Usl€
ander, T., Jacques, P., Simonis, I., Watson, K., 2010. Designing environmental
software applications based upon an open sensor service architecture. Environ.
Model. Softw. 25 (9), 977e987.
Wald, L.A., Wald, D.J., Schwarz, S., Presgrave, B., Earle, P.S., Martinez, E.,
Oppenheimer, D., 2008. The USGS earthquake notification service (ENS): cus-
tomizable notifications of earthquakes around the globe. Seismol. Res. Lett. 79
(1), 103e110.
Zhang, W., Wang, Y., Peng, H., Li, Y., Tang, J., Wu, K.B., 2010a. A coupled water
quantityequality model for water allocation analysis. Water Resour. Manag. 24
(3), 485e511.
Zhang, X.J., Chen, C., Ding, J.Q., Hou, A.X., Li, Y., Niu, Z.B., Su, X.Y., Xu, Y.J., Laws, E.A.,
2010b. The 2007 water crisis in Wuxi, China: analysis of the origin. J. Hazard.
Mater. 182 (1e3), 130e135.
Zhang, X.J., Chen, C., Lin, P.F., Hou, A.X., Niu, Z.B., Wang, J., 2011. Emergency drinking
water treatment during source water pollution accidents in China: origin
analysis, framework and technologies. Environ. Sci. Technol. 45 (1), 161e167.
Zhou, R., Zhu, L., Yang, K., Chen, Y., 2006. Distribution of organochlorine pesticides
in surface water and sediments from Qiantang River, East China. J. Hazard.
Mater. 137 (1), 68e75.