efficient gis-based model-driven method for flood …...open access efficient gis-based...

Nat Hazards Earth Syst Sci 14 331ndash346 2014wwwnat-hazards-earth-syst-scinet143312014doi105194nhess-14-331-2014copy Author(s) 2014 CC Attribution 30 License

Natural Hazards and Earth System

SciencesO

pen Access

Efficient GIS-based model-driven method for flood riskmanagement and its application in central China

Y Liu 123 J Zhou13 L Song4 Q Zou13 J Guo13 and Y Wang5

1School of Hydropower and Information Engineering Huazhong University of Science and TechnologyWuhan 430074 China2School of Electrical and Electronic Engineering Hubei University ofTechnology Wuhan 430068 China3Hubei Key Laboratory of Digital Valley Science and Technology Huazhong University of Science and Technology Wuhan430074 China4Laboratory of Numerical Modeling Technique for Water Resources Department of Water Resources and Environment PearlRiver Water Resources Research Institute Guangzhou 510623 China5School of Engineering and Technology Hubei University of Technology Wuhan 430068 China

Correspondence toJ Zhou (jzzhoumailhusteducn)

Received 13 March 2013 ndash Published in Nat Hazards Earth Syst Sci Discuss 24 April 2013Revised 2 January 2014 ndash Accepted 17 January 2014 ndash Published 21 February 2014

Abstract In recent years an important development in floodmanagement has been the focal shift from flood protectiontowards flood risk management This change greatly pro-moted the progress of flood control research in a multidis-ciplinary way Moreover given the growing complexity anduncertainty in many decision situations of flood risk man-agement traditional methods eg tight-coupling integrationof one or more quantitative models are not enough to pro-vide decision support for managers Within this context thispaper presents a beneficial methodological framework to en-hance the effectiveness of decision support systems throughthe dynamic adaptation of support regarding the needs of thedecision-maker In addition we illustrate a loose-couplingtechnical prototype for integrating heterogeneous elementssuch as multi-source data multidisciplinary models GIStools and existing systems The main innovation is the appli-cation of model-driven concepts which put the system in astate of continuous iterative optimization We define the newsystem as a model-driven decision support system (MDSS )Two characteristics that differentiate the MDSS are as fol-lows (1) it is made accessible to non-technical specialistsand (2) it has a higher level of adaptability and compatibilityFurthermore the MDSS was employed to manage the floodrisk in the Jingjiang flood diversion area located in centralChina near the Yangtze River Compared with traditional so-lutions we believe that this model-driven method is efficient

adaptable and flexible and thus has bright prospects of appli-cation for comprehensive flood risk management

1 Introduction

Flood disasters are among the most frequent and devastat-ing types of disasters in the world (International Federationof Red Cross and Red Crescent Societies 1998) In Chinaup to two-thirds of the land subject to different types andlevels of risk from river and coastal flooding (Feng et al2001 Wang et al 2004) Within this context we are facinga paradigm shift in dealing with flood issues from flood pro-tection towards flood risk management (Evers et al 2012)Moreover given the growing complexity and uncertainty inmany decision situations of flood risk management personalexperience and single models are not enough to provide de-cision support for managers Therefore comprehensive floodrisk management has widely been used in flood insuranceflood plains management disaster evacuation disaster warn-ing disaster evaluation flood influence evaluation and im-proving the publicrsquos flood risk awareness and understandingof flood disasters (Zeng et al 2007 Ray 2007 Escuder-Bueno et al 2012 Chen et al 2011 Schinke et al 2012Li et al 2012 Evers et al 2012)

Published by Copernicus Publications on behalf of the European Geosciences Union

332 Y Liu et al Efficient GIS-based model-driven method for flood risk management

Traditionally tight-coupling integrations of one or morequantitative models were the dominant of the flood riskmanagement approaches However unique aspects of floodrisk management problems require a special approach fordynamic adaptation of support regarding the needs of thedecision-maker Bijan et al (1997) believed that there isan additional intelligent ldquoadaptationrdquo component that canachieve absolute adaptation We cannot prove that this as-sumption is feasible but our approach is to ensure that thealgorithms and models are in a loose-coupling frameworkeg can be easily reconstructed to accommodate the de-mand Over the past 10 yr we have focused on the need forflood control by integrating remote sensing and GIS tech-niques with qualitative or quantitative models including hy-drologic analysis flood simulation flood risk analysis anddisaster assessment (Zou et al 2012a b Guo et al 2011Song et al 2011b)

In recent years the purpose of our research has increas-ingly become to propose a universal approach for flood riskmanagement by the model-driven method which providesan accessible interface to non-technical specialists such asmanagers Furthermore the method is intended for repeateduse in the same or a similar decision situation (Power andSharda 2007) In order to achieve these objectives many re-search studies have been carried out on this topic They canbe outlined as follows (1) model-driven method for floodrisk management (2) loose-coupling framework of hardwareand software system (3) multidisciplinary optimization andmathematical programming models

For the readerrsquos convenience the remainder of this paperis organized as follows

ndash Section 2 provides an overview and analysis of pastmodel-driven method research then identifies researchchallenges related to behavioral and technical aspectsof implementing and using the method for flood riskmanagement

ndash Section 3 defines our model-driven methodologicalframework and proposes its systems life cycle for de-signing and developing a MDSS

ndash Section 4 illustrates the loose-coupling technical pro-totype through identifying and incorporating the keycomponents for our model-driven method

ndash Section 5 presents how we adopt the method in theJingjiang flood diversion area where we are commis-sioned by the government to assist the operations ofengineering and controlling the flood risk

ndash Section 6 demonstrates the benefits of our model-driven method through comparing with some tradi-tional solutions

ndash The final section provides a summary of observationsand recommendations for future directions of researchin model-driven methods for flood risk management

2 Background

21 Overview of model-driven method research

Nowadays developing applications for non-technical endusers is becoming increasingly complex as they want toaccess these applications everywhere at every moment ofthe day and night and with many different devices (Gates2008) On the other hand the model which address the de-cision support problem for flood risk management is be-coming multidisciplinary (as discussed last section) There-fore economists geographers water resource specialists op-eration researchers and management scientists had to inte-grate their respective models together to support managersrsquodecision-making and planning

Traditionally by using object-oriented software engineer-ing many computer programs which are based on a tightlycoupled integration of models and interface is presented tohelp the decision-maker make effective decisions It is com-monly believed that this way can guarantee the stability andhigh performance of the system However given the grow-ing complexity and uncertainty in many decision situationsof flood risk management we need a special approach tothe dynamic adaptation of support regarding the needs ofthe decision-maker This approach needs to provide an in-tegration platform for multidisciplinary models not subjectto geographical distribution research area development lan-guage software tool and other constraints and to guaran-tee that the system has a higher level of adaptability andreusability Taking the above factors into consideration thefocus of the study will be divided into two parts (1) model-centric decision-making process ensures DSS adaptability(2) loose-coupling implementation ensures DSS flexibility

Model-driven method research can be distributed in twoaspects The first is as DSS foundations research which hasbeen one of the most important topics since 1969 (Powerand Sharda 2007) the method uses algebraic decision an-alytic financial simulation and optimization models to pro-vide decision support The second software engineering re-search model-driven method (or model-driven engineering)and recent trend whose main proposal is to focus on mod-els rather than on computer programs (Beacutezivin 2004) Andit is being increasingly applied to enhance system develop-ment from perspectives of maintainability extensibility andreusability (Mehmood and Jawawi 2013) With the rapid de-velopment within the field of Web service technologies andservice-oriented computing the model-driven method hasbeen widely used in the area of business process manage-ment (Watson 2008)

Eom (2003) identified a wide variety of DSS applica-tions reported in the academic literature There are more than1800 related articles and many of them focused on our re-search direction with good results However by querying thislarge number of scientific papers we noticed that the model-driven research within the field of natural disasters is at the

Nat Hazards Earth Syst Sci 14 331ndash346 2014 wwwnat-hazards-earth-syst-scinet143312014

Y Liu et al Efficient GIS-based model-driven method for flood risk management 333

beginning stage especially the rare loose-coupling frame-work for resolving behavioral and technical challenges Fur-thermore we believe that this approach would address theissues raised by the managers end users and scientists fromdifferent disciplines Therefore we started a study to validatethis hypothesis

22 Behavioral and technical challenges

Behavioral and technical research on model-driven methodsneeds to address many unresolved issues which is identifiedby Power et al (2007) as follows (1) construction of specificquantitative models (2) storage and retrieval of data neededby different types of models (3) communication of param-eters among models and other DSS (decision support sys-tem) components (4) multi-participant interaction in modeluse and value elicitation (5) impact of user interface de-sign alternatives on model-driven DSS effectiveness and easeof use and (6) building deploying and using model-drivenDSS In addition with the remarkable improvement of spa-tial representation research there is a new challenge associ-ated with the integration of geographic information systems(GISs) with multidisciplinary models (Merz et al 2010)

3 Model-driven methodological framework forflood risk management

We define our model-driven method as a methodology thatsupports human decision-making judgements by a loose-coupling platform with multidisciplinary optimization andmathematical programming models In this section we willdescribe the behavioral aspects of this method ie how itworks The technical aspects of implementation will be dis-cussed in the Sect 4

All in all the general types of models used in a model-driven DSS is numerous They include algebraic and differ-ential equation models various decision-analysis tools in-cluding analytical hierarchy process decision matrix and de-cision trees multi-attribute and multi-criteria models fore-casting models network and optimization models MonteCarlo and discrete event simulation models and quantita-tive behavioral models for multi-agent simulations There-fore the focus of research work is placed on how to providea simplified representation of a situation that is understand-able to a decision-maker (Power and Sharda 2007)

The main innovation of this paper is that our methodmakes the DSS in a continuous-integrated state (see Fig 1)which means (1) the system platforms and decision-makingprocesses are iterative optimization (2) the latest client willbe published simultaneously in real time In other wordsbased on the loose-coupling platform these systems can pro-vide the dynamic adaptation of support regarding the needsof the decision-maker In order to achieve the continuous-integrated state the systems life cycle is designed to be a

novel double-loop optimization structure which consists oftwo parts the technical-loop and the behavioral-loop (seeFig 2) The technical-loop is used to build deployable soft-ware at any time which can provide a dynamic integrationplatform for data models decision-making processes and theexisting systems We can illustrate the loose-coupling tech-nical prototype in the following sections Interacting with thetechnical-loop the behavioral-loop is developed for connect-ing the user groups including managers practitioners andscientists from different disciplines Scientists can registertheir models to address a new decision-making task at anytime or improve the adaptability of old models for reusabil-ity problems Practitioners can customize the decision pro-cess by combining models and design the client by systeminterface (ie Interface Service which will be referred to inSect 45) Managers who have the right to make decisionscan use the common client or the recommended client bypractitioners Furthermore the feedback mechanisms makethe requirements clearer and systems are more reasonableand reliable

4 Loose-coupling technical prototype for themodel-driven method

41 Overall framework

In this section we will illustrate a technical solution ieloose-coupling prototype for the model-driven developmentthrough identifying and incorporating the key componentsAll in all loose-coupling is the concept typically employed todeal with the requirements of scalability flexibility and faulttolerance The aim of loose coupling is to minimize depen-dencies When there are fewer dependencies modificationsto or faults in one system will have fewer consequences onother systems On the other hand it makes the joint analysisof cross-regional models possible eg mechanical experts inBeijing address the risk assessment at the same time hydro-dynamic experts in Wuhan deal with the damage estimationof flood routing

The software is worked in a simple and clear structureClient is only responsible for the function of user interac-tion which collects userrsquos command by events of variouscontrols then translates the order into an XML data streamand pushes it to the central communication service in the ser-vice layer the load balancing server deals with this commandand assigns it to the right server Finally the process ends inthe database changes Therefore scientists can register theirmodels into the service layer for easy uses and managers canget results via a beautiful interface

The software architecture is composed of a three-layerstructure (see Fig 3) which is identified as follows (1) abase layer supporting the distributed systems including net-works softwares operating systems databases etc and (2)a services layer for service-oriented specific functionalities

wwwnat-hazards-earth-syst-scinet143312014 Nat Hazards Earth Syst Sci 14 331ndash346 2014


Geoinformation modelsof GIS software

Uniform platform for data sharing

Loose-coupling platform

Service-orientedlinkage between GIS

and models

Open library ofmultidisciplinary models

Flexible user interfaces with expanded GIS

Scientists fromdifferent disciplines

Practitioners

Managers

Register andmanage theirown models

CommonClient

Customize the adaptive decision-making process and design client

RecommendedClient

Support decision-making judgments with

accessible interface

or

User interfaces

Feedback mechanismssuch as meeting E-mail phone call BBS

Provide therevised views

Upload theassessment reportof decision effect

Continuous-Integrated state

User group Key components of ourmodel-driven method

Third partyplatforms BehaviorLegend

Fig 1Behavioral aspects of the model-driven method for flood risk management

Technical-loop

Architecture

development

Continuous

integration for Data

models and systems

Continuous testing

Continuous

deploymentContinuous feedback

Decision making

goals

Requirements

analysis

Platform design

Problem identify

Problem resolutionBehavioral-loop

Decision-making

processes and interface

customization

Decision scheme set

generation

AssessmentModel modification

FeedbackSupport

Fig 2 Systems life cycle for model-driven methodological frame-work

Among them a service provider publishes its description ofservice and interface in formation to the service registry Aservice requester will find relevant service from this registryand then establish a link to invoke that service and (3) im-plementation layer It provides web service applications toimplement the service-oriented architecture (SOA) Detailson this layer are discussed in later sections (see Sect 43) Inaddition The Services layer or its implementation layer canalso project itself onto an operational layer which consists ofworkflow components of publishing discovery compositionbinding and execution (Hu et al 2011)

We define the loose-coupling DSS as a model-driven de-cision support system (MDSS) In a MDSS first a uni-form data platform (as database service) is presented to sharemulti-source data Next an open library (as multidisciplinarymodel service) is presented to integrate cross-platform mod-

els A universal linkage (as GIS Service) between GIS andmodels is then presented by use of service-oriented approachFinally a flexible interface system (as interface service) ispresented to help practitioners customize the application fordecision-makers

42 Uniform platform for data sharing

A GIS-based MDSS needs to deal with numerous raw datafor decision-making which includes the traditional struc-tured data and other special data The special data originatefrom remote sensing images live video live audio sensorstream and so on In essence it is multi-modal data with thecharacteristics of semi-structured or unstructured

Many challenges are associated with the integration ofthe multi-source data They include efficient structured treat-ment heterogeneous data fusion and cross-platform com-munication By considering the specific problem in dis-tributed systems and following the design concepts of SOAthis paper proposes a common unified data platform whichis divided into two parts monographic data support andgeographic data support We will focus on the integrationof these two in the following section Finally for differentsources of the data the database service provides varioustypes of packaged web services interface to enable efficientinteraction between the model library and user interface (seeFig 4)

421 Similarity-based monographic data integration

The discrepancies in multi-source data occur both duringeliminating duplicates from semantic overlapping sourcesand during combining complementary data from differentsources (Schallehn et al 2003) A similarity-based data



GIS DatabaseData FileSystem Database

Communication Class

Interface Class

Various Client

Data Class

GIS Service

Main Class

Central Communication Service

Services Layer

Database Service Multidisciplinary Model Service

Database Admin GIS Class

Interface Service

Communication Class

Service Provider

Implementation Layer

UDDI

Service Consumer

Service Registry and WSDL Interface

Publishing

Discovery

Composition

Binding

Execution

Discovery (SOAP) Publish (SOAP)

Invoke (SOAP)

Base Layer

Fig 3Loose-coupling framework for the model-driven method

integration model (SDIM) is used to deal with discrepanciesin multi-source data The discrepancies in structured data canbe directly solved by existing operations of common rela-tional database such as SQL operations like grouping or joinHowever for the unstructured or semi-structured data the in-tegration process will be much more complicated The pro-cessing steps of the unstructured data are as follows

1 Extracts the features of data automatically generatesthe data source configuration file and sets the indexitem

2 Supported by XML document conversion model pre-processes the raw data in order to get the proposedkeyword and generates the XML index document InXML contexts are given through element and attributenames and relationships by organizing data into treesand subtrees and through id references

3 Groups the XML information with similarity measuresby its attribute and logical combination The similarityformula is determined by the user preference whichwill not be discussed here It can be a simple thresholdvalue or a complex comprehensive approach

4 Adds formatting information and uses the relationaldatabase management software operations to sort theprocessed data

422 Demand-oriented geographic data integration

The data for representing geology is diverse heterogeneousand multi-source so it is difficult to identify the correlation

Relational database

E-mail

Notepads

Web Pages Text files

Real-time Sensor data

Picture files

Structured data

Economic data

Population data

Resource data

Building model

Environment model

Feature model

DEM

DOMRoad

network

Property database

Digital model libraries

Basemap database

JavaFortran net ArcGISGoogle MapsBing Maps GeoServer

Uniform data sharing platform

Demand index library

RESTful Web Service interface

Demand-oriented geographic data integration model

Similarity-based monographic data integration model

Data conversion interface

XML document conversion model

Fig 4Architecture of the uniform data sharing platform

between them The traditional integration methods are ded-icated to the technical problems of data integration perfor-mance (Wu et al 2005) such as data transmission and dataquery In this paper a demand-oriented geographic data in-tegration model is used to establish a logical association be-tween these data First of all the demand for geographicaldata in GIS-based DSS is divided into two spatial analysisand visual simulation Then the proposed model builds thedata schema from multisource data in a stepwise manner andinvolves six steps (1) digitizing various geological data us-ing existing software systems like ArcGIS or Microsoft SQLServer (2) registering and filling in the details (3) makingsure they have uniform dimensions and data types (4) build-ing a hierarchical link list for each demand of models (5)creating the geographic cache and (6) modifying the indexinformation for service access

43 Open library of optimization and mathematicalprogramming models

In this section we will describe how we address the loose-coupling integration problem for multidisciplinary model

The service-oriented architecture (SOA) is employed tosolve the integration problems In a SOA each model canbe published as a service and can communicate with an-other model even though it has a different underlying op-erating system These services are virtually unrelated func-tional units but loose-coupling spread over interconnectednetworks (Hu et al 2011) It uses the Web Service Descrip-tion Language (WSDL) for describing services a mecha-nism called Universal Description Discovery and Integration(UDDI) for service registry and service discovery then thesimple object access protocol (SOAP) for exchange of mes-sages Web services use XML-based standards as a format

The implementation layer consists of various web serviceapplications and is used to establish the loose-coupling openlibrary which accepts heterogeneity and leads to decentral-ization The core of the implementation layer is a specificinfrastructure called the enterprise service bus (ESB) that



allows us to combine the multidisciplinary models in an easyand flexible manner

It is the responsibility of the ESB to enable managers tocall the modelsrsquo providers supply (see Fig 5) Dependingon the technical and organizational approaches taken to im-plement the ESB this responsibility may involve (but is notlimited to) the following tasks providing connectivity datatransformation (intelligent) routing dealing with securitydealing with reliability monitoring and logging and modelcomposition management model registry and deposit

There are two ways for scientists to manage their modelspersonal management or delegated management Both meth-ods only need scientists to register models on the server byfilling some parameter information The managers still use anofficial endpoint which delegates the real task when mes-sages arrive the load balancer sends them to the differentphysical service providers that it knows about

44 Service-oriented link between GIS and models

Regional flood risk generally considered at the river-basinscale is a decision-making problem based on physical andsocioeconomic conditions (McKinney and Cai 2002) Spa-tial representations of the region consist of spatial objects andthematic objects Spatial objects represent real-world enti-ties with both geographical and physical environmental andsocioeconomic attributes Thematic objects represent meth-ods and topics relevant to the spatial objects Therefore itis the responsibility of the GIS to represent the real spatialentities and provide spatial analysis and data processing Ar-cGIS Server 93 is used to provide technical support The keycomponents can be listed as follows (ArcGIS Resource Cen-ter 2008)

ndash Mapping service It serves cached maps and dynamicmaps from ArcGIS

ndash Geocode service It finds address locations

ndash Geodata service It provides geodatabase accessquery updates and management services

ndash Globe service It serves digital globes authored in Ar-cGIS

ndash Image service It provides access to imagery collec-tions

ndash Network analysis service It performs transportationnetwork analysis such as routing closest facility andservice area

ndash Geometry service It provides geometric calculationssuch as buffering simplifying and projecting

ndash Geoprocessing service It provides spatial analysis anddata processing services

Managers

ESB

InterceptorMonitoring and

logging server

Service registry and

WSDL interface

Model server

Load balancer

Service

management

Deposit model server

Send requests

Reliability and

security server

Scientists from

different disciplines

Register their models

Publish models by

themselves

Practitioners

Customize model

composition

Upload code packageor

Fig 5 Open library of optimization and mathematical program-ming models

In our model-driven method the first seven services are usedto interact with managers via the GIS-based client to repre-sent the real spatial entities The last one (ie geoprocessingservice) is re-developed and integrated into the open librarysimilar in behavior to a particular model A geoprocessingservice takes the simple data and turns it into something ex-traordinary the probable evacuation area for flood hazards aland cover map going five years back the risk-sensitive areasafter a dam break and so on Because the service is executedon the server computer using resources of the server com-puter the clients can be lightweight applications and we donot need to cover the cost of GIS for each individual applica-tion

45 Flexible user interfaces with expanded GIS

The goal of making model-driven DSS accessible to non-technical specialists implies that the design and capabilitiesof the user interface are important to the success within thesystem (Power and Sharda 2007) The user interface controlshow the user views results and influences how the user under-stands results and hence influences choices In our model-driven method the clients are lightweight ie they onlyknow how to send packets of simple requests to a serverand show the results through rich controls Therefore ourclient can be developed by a variety of technologies such asJava Server Pages (JSP) Active Server Page (ASP) Qt Win-form Windows Presentation Foundation (WPF) Objective-C (ObjC) and so on In the following section we have a WPFwhich is described as an example Developed by Microsoftthe WPF is graphical subsystem computer software for ren-dering user interfaces in Windows-based applications Exten-sible Application Markup Language (XAML) is employedto define and link various user interface elements BecauseXAML is a XML-based language it can be designed withany text editor In our model-driven method the interfaceservice is presented to help practitioners customize the ap-plication for decision-makers



Via interface service practitioners can define in how manysteps the decision-making process will be completed thendesign the controlrsquos layout and operation logic for each stepFinally designs will be translated into text sent to the serverand automatically be generated into corresponding XMALdocuments Decision-makers run the program by download-ing clients or landing web pages

A section on GIS and ArcGIS API (application program-ming interface) for WPF is employed to create rich desk-top applications that utilize the powerful mapping geocod-ing and geoprocessing capabilities provided by the ArcGISserver Similarly considered to be a normal control practi-tioners can generate instances of GIS by XMAL

5 MDSS in the Jingjiang flood diversion area

51 Basic information about the case study area

The methodology and the model was performed and tested inthe Jingjiang flood diversion area situated in central Chinanear the Yangtze River It covers an area of 92134 km2 (seeFig 6) It was the first large-scale hydraulic engineeringbuilt by the government of the Peoplersquos Republic of Chinain 1952 Its main composition contains the incoming floodgate (North Gate) sluice (South Gate) and 20838 km-longembankments The projectrsquos principal role is to ensure thesafety of Jianghan Plain and the Wuhan City Distinct fromdeveloped countries Chinarsquos flood diversion area is inhabitedby a large number of people its social and economic cost isvery high with 5818 thousand people living in eight townsand the annual production value is USD 700 million We arecommissioned by the government to assist the operation ofthe engineering management and control the flood risk Themodel validation is based on the successful application ofthis project in 1954

52 A feasible decision process for floodrisk management

In this section we will show an example of the feasible deci-sion process (as shown in Fig 7) Multisource remote sens-ing images DEM and hydrological data were coupled withthe uniform data-sharing platform Flood forecasting modelscan derive the flood process in for two subsequent weeksBased on hydrodynamic calculations of flood routing sim-ulation models flood risk assessment and damage estima-tion models can provide the qualitative or quantitative resultsgenerate various thematic maps to describe the distribution offlood risk provide guidance on evacuation and assist floodpath control for government policy-makers

53 Hydrology analysis and forecasting

The flood characteristics such as the frequency return pe-riod coefficient of skew coefficient of variation and so on

Fig 6 Location of the study area(A) map of China in which theannotated area is Hubei province(B) Map of Hubei in which theannotated area is Jianghan Plain Wuhan City and the Jingjiangflood diversion area(C) Map of the Jingjiang flood diversion area

are analyzed based on historic observed streamflow dataIt can provide the basis for choosing those typical floodprocesses Meanwhile to dynamically evaluate the risk anddamage of flood hazard in real time the actual-time floodforecasting model is integrated into our system At first weanalyze the characteristics of the hydrological system in thesensitive area of flood Based on this a conceptual modelthe famous Xinanjiang model (Zhao 1984) is constructed tosimulate the flood processes Currently the traditional cali-bration of hydrological models with a single objective can-not properly measure all the behaviors within the system Tocircumvent this an efficient evolution algorithm entitled themulti-objective culture shuffled complex differential evolu-tion (MOCSCDE) algorithm (see Appendix A) is proposed tooptimize the model parameters This algorithm can get betterconvergence and spread performance and can significantlyimprove the calibration accuracy (Guo et al 2012) Theseresults provide an important basis for studying the mecha-nism of flood routing

54 Flood routing simulation

The flood dynamic routing simulation and the flood riskmaps drawing are based on DEM (digital elevation model)The hydrodynamic member of our research team developeda well-balanced Godunov-type finite volume algorithm formodelling free-surface shallow flows over irregular topogra-phy with complex geometry The research results have beenpublished in many academic journals (Song et al 2011a Liuet al 2013) This two-dimensional hydrodynamic model isbased on a new formulation of the classical shallow waterequation in hyperbolic conservation form (see Appendix B)which describes the state of a flood such as water depth ve-locity duration and submerged area

The computational domain was triangulated with 77 741elements 39 389 nodes and 117 129 lines The area of thesmallest unit is 4246 m2 the area of the biggest unit is



Fig 7A feasible decision process for flood risk assessment and damage estimation

20 000 m2 and length of the shortest line is 69 m In addi-tion triangles were more densely organized along the riverto ensure the calculation accuracy The total area of the floodis 235 billion m3 and the largest flood diversion flow is6190 m3 sminus1

We obtained the bottom elevation of computational gridby bilinear interpolation with high-resolution DEM data andset the roughness values of each computing unit by land usedata Based on hydrodynamic calculations the risk mappingmodule generates various thematic maps to describe the dis-tribution of flood risk provide guidance on the evacuationand assist control flood path for government policy-makersFigure 8 shows the simulation results eg the water depthchanges within the flood routing which lasted about 131 h

55 Flood risk assessment

As we have seen flood risk assessment is a synthetic evalu-ation and consists of many factors some factors are natural

Fig 8 Filled contours plot of initial and computed water depthgiven by the present model at timet =167 1125 25 3584 58347084 8417 and 13083 h



and socioeconomic in various areas and several of them donot have unified quantitative criteria which makes the floodrisk assessment index system complex and difficult to oper-ate (Du et al 2006 Li et al 2008 Jiang et al 2008) Basedon the disaster system theory which takes into considerationthe factors that follow the systematic quantitative and oper-ability universality principle the primary flood risk assess-ment index system of a flood diversion district is establishedAccording to local investigations references to historical in-formation public participation expert judgments and so on(Du et al 2006 Li et al 2008) we obtain the grading crite-ria for each factor

Several methods such as a fuzzy comprehensive assess-ment method (Jiang et al 2008) variable fuzzy sets theory(Chen and Guo 2006) and the attribute interval recognitiontheory (Zou et al 2011) are employed for flood hazard as-sessment and flood vulnerability assessment

Recently we presented a new model for comprehensiveflood risk assessment the set pair analysis-variable fuzzy setsmodel (SPAVFS) (Zou et al 2012c) which is based on setpair analysis (SPA) and the variable fuzzy sets (VFS) theoryThis model determines the relative membership degree func-tion of VFS by using an SPA method and has the advantagesof intuitionist course simple calculation and good generalityapplication Its flow chart is shown in Fig 9

The flood risk analysis module is made up of the manage-ment of risk analysis data the risk assessment models forflood risk analysis and flood risk maps and it can show re-sults with a map and chart mode The specific technical routeof this module is as follows

1 Based on GIS statistical data and historical informa-tion we collect the precipitation data elevation dataland use data social-economic data and so on

2 Then taking the towns as the basic assessment unitswe count out four hazard assessment indexesrsquo charac-teristic values (see Table 1 and Fig 10) via the simu-lation results of hydrodynamic models including av-erage maximum flow velocity and flood depth floodsubmerging range and flood arriving time Moreoverconsidering the fact that the weather terrain and riverdistribution had a greater impact on flood hazard as-sessment we add up the other three hazard assess-ment indexes eg annual average precipitation aver-age ground elevation and land use rate Whatrsquos morefor flood vulnerability assessment there are a corre-sponding six indexes (see Table 2) consisting of pop-ulation density industrial output density agriculturalproduction densities breeding area percentage animaldensity and road network density Hence according tothe situation of the Jingjiang flood diversion area wecollect data for flood hazard and vulnerability analysis

3 After the establishment of the flood risk assessmentindex system and the grading criteria the flood hazard

and vulnerability assessment results can then be ob-tained by SPAVFS method Here the flood hazard vul-nerability and risk are divided into five grades notedas very low low medium high and very high

4 With the flood hazard and vulnerability assessmentresults and the flood risk grade-classification-matrix(Du et al 2006) we access the corresponding riskgrade for each assessment unit Finally we are ableto carry out the flood risk maps (see Fig 11) for theJingjiang flood diversion district

56 Flood damage estimation

The final part is the flood damage estimation First we collectthe data of flood ranges As Fig 12 shows affected areaswith different depths of water are marked by diverse colorsand we can see the estimation results of each town At thesame time we can also see the relation of time and damageestimation via the grid-based flood analysis model

The processing steps of grid-based flood analysis model(see Fig 13) are as follows

1 Using the result of the two-dimensional hydrodynamicmodel the water depth and the flow velocity arerecorded in the nodes then use the GIS spatial analystcomponent IDW (inverse distance weighted) interpo-lation method to obtain the flood submerge raster

2 Damages of different aspects including the loss ofhouses industry affected population and economicloss (see Table 3) can easily be quantified by a flooddisaster evaluation model (Xie 2011) Its equation isas follows

W =

sumi

sumj

sumk

αiAijηjk (1)

whereijk is the index of calculation cell land usetype and water depth level respectivelyAij is theoriginal output value of thej th land use type in theith cell ηjk is the loss rate of thej th land use type inthe kth water depth levelαi is a sign code (1) whenαi = 1 it means the ith cell needs to calculate the loss(2) whenαi = 0 it means the ith cell does not need tocalculate the loss

3 By using the Geoprocessing Service (which mentionedin the Sect 44) the flood feature rasters are reclassi-fied by water depth level and then mixed up with thesubmerge information by time sequence

6 Method benefits and validation

61 Method benefits

The method described in this paper was applied to fiveprojects over the past 4 yr It has been playing a substitute



Assess the risk grades

Calculation of the weights forthe indices based on fuzzy AHP

Establishment of flood riskassessment index system

Calculation of synthetic relativemembership degree

based on VFS

Rank the order and assess thegrades for flood hazard

vulnerability assessment basedon VFS

Integrated flood risk maps

Fig 9Flow chart of flood disaster risk assessment

Km

EmbankmentCofferdam123 6 180

Low 7850

High 0

Flood arriving time (min)

EmbankmentCofferdam

123 6 18Km

0

Water depth (m)0

0001-05

05-10

10-15

15-20

20-25

25-30

30-35

35-40

40-50

50-70

Km

Embankment

180

Low 0

High 6

Flood flow velocity (ms)

Km


Flood submerging range

Affected area

A B

C D

Cofferdam123 6

Fig 10 Four hazard assessment indexes characteristic values from the simulation results of hydrodynamic models(A) flood submergingrange(B) flood flow velocity(C) flood water depth(D) flood arriving time



Table 1Hazard characteristic by analyzing submerged condition

Town Flow Flood Arriving Submerging Precipitation Elevationvelocity depth time range(m sminus1) (m) (h) (km2) (mm) (m)

Buhe 043 185 0 17773 1064 3618Douhudi 03 23 1041 8461 1064 3327Yangjiachang 004 07 3666 6135 1064 3377Mahaokou 02 314 3791 16915 1064 3227Ouchi 028 325 4874 9301 1211 3227Huangshantou 032 388 5958 8976 1400 3368Zhakou 035 345 2833 12575 1064 321Jiazhuyuan 037 215 1583 12854 1064 3351

Table 2Vulnerability characteristic within the unit area (eg 1 km2)

Town Population Road Industry Agricultural Animals Breeding(m) (104U) (104U) area

Buhe 44086 24117 110496 10143 2519 353 Douhudi 152018 4251 382432 13013 7779 601 Yangjiachang 39385 10018 15346 11001 2458 556 Mahaokou 34238 25545 76295 10879 8794 805 Ouchi 47549 31474 253923 10619 6235 209 Huangshantou 33416 24376 29255 8138 9965 1192 Zhakou 4041 28233 79859 9234 7842 1048 Jiazhuyuan 4273 2429 77178 1285 7211 742

Flood risk mapVery low

Low

Medium

High

Very high


Km0

BA C

Fig 11Integrated flood risk maps for the Jingjiang flood diversiondistrict (A) flood vulnerability grade map(B) flood hazard grademap(C) flood risk grade map

role for existing programmes Compared with traditional so-lutions this subsection identifies and discusses the benefitsof our model-driven method

611 Supporting heterogeneity

By nature all large systems are heterogeneous These sys-tems have different purposes times of implementation andages and it is commonly found that the system landscapesare accretions of different platforms programming lan-

guages programming paradigms and even middleware Inthe past traditional solutions attempted to solve the problemsof scalability by harmonization sometimes they worked butthe maintenance cost became incredibly expensive with theexpansion of the system (Josuttis 2009)

In our model-driven method ESB (see Sect 43) is em-ployed The idea behind it is that instead of creating andmaintaining individual communication channels each modelonly has to connect to the bus to be able to connect to allother models In order to be extended for the larger sys-tem we have identified the research directions related to thetechnical and organizational rules of implementing a flexi-ble framework The heterogeneous models are virtually unre-lated functional units but rather loose-coupling spread overinterconnected networks

612 Dynamic adaptation of support regarding theneeds of the decision-maker

Dynamic adaptation which is defined in this paper meansthat the system can be flexible and reliable restructuring tomeet the different needs of the decision-maker Comparedto the traditional method the main advantage of our methodis the software life cycle which at its core consists of phasesof design implementation integration and production Mosttraditional methods are waterfall-like approaches ie they



Fig 12Interface of submerged condition analysis in quantitative flood damage estimation module

Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process

Land use data Data convert

Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit

Maps of damageGIS platform

Fig 13The process of the flood submerging analyst and flood hazard loss evaluation

Table 3 The result of the quantitative flood damage estimationmodule

Town Affected Industry Affected Economichouses (104U) population loss (104U)

Buhe 24 614 744173 22 26049 29 93168DouhuDi 37 315 903257 961781 18 90333Yangjiachang 4287 146452 668363 829380Mahaokou 27 484 592971 20 30079 26 41828Ouchi 17279 573739 970107 15 97435Huangshantou 11 768 767 633081 654533Zhakou 17574 235512 864918 11 41663Jiazhuyuan 14388 242480 14 27033 17 02173

gradually complete the process in one way It does not workbecause of requirements as well as contexts changes overtime (Josuttis 2009)

Our method makes the DSS in a continuous-integratedstate (see Sect 3 and Fig 1) Open library (see Sect 43)can provide the suitable model composition by iterative ser-vice development with managersrsquo feedback New clients arealso available at any time through a simple XAML design bypractitioners (see Sect 45) Therefore our method can pro-vide the dynamic adaptation of support regarding the needsof the decision-maker



Table 4Comparison between traditional tight-coupling systemslowast and MDSS

Traditional tight-coupling systems MDSS

Development efficiency Complex development process Two-stage development processincluding requirements design construction (1) deploy the universal loose-coupling technical prototypeintegration testing and de-bugging (2) iteration optimization withinstallation and maintenance technical-loop and behavioral-loop

Distributed integration They solve the problems It providesof scalability by harmonization flexible integration methods withthe maintenance cost becomes incredibly expensive open library based on SOA

Emergency needs They cannot be provided It publishes deployablewithin the prescribed time software at any time by loose-coupling

technical prototype

Deployment flexibility They provide only It provides flexibleone type or a few similar user interfaces with expandedtypes of clients GIS by WPF and Interface Service

Model creditability It is difficult to verify According to the reviewor maintain a single model in from practitioners and managers modelsthe tight-coupling architecture will be modified to increase the adaptability

new client will be deployed to verify the model creditabilityautomatically loop

lowast These traditional systems use some fixed interfaces tightly integrated data models GISs clients Therefore the change of any part will affect the other elements

613 Protection of intellectual property rights

In traditional object-oriented programs the models run asplugins extensions or code block Illegal software techni-cians can get the source code by some simple decompilertools This is a major hazard of the interdisciplinary cooper-ation

In our model-driven method by using SOA the web ser-vice mode restricts the behavior of user access models Whena user accesses the service we focus on three aspects of se-curity (1) authentication which has to do with verifying anidentity This means finding out who is calling the service(2) Authorization which has to do with determining whatan identity is allowed to do this involves checking whetherthe caller is allowed to call the service andor see the result(3) Auditing which involves recording all security-relevantinformation so that we can detect or analyze security holesand attacks This mechanism ensures that users cannot usethe model until they pass the legal validation

62 Method validation

In this subsection we compare the MDSS model with otherson a set of criteria We chose the following methodologicalframework (Zeng et al 2007 Chen et al 2011 Qi and Alti-nakar 2011) because it is representative of methods for riskmanagement in some facets Table 4 shows the comparison

Feature comparison between our method and other meth-ods shows that our method is an integrated methodologywhich ensures the adaptability of the effective loop optimiza-

tion mechanism while other methods focus on one specificissue or a part of the issue

Zeng et al (2007) built a GIS-based DSS for assessingthe risk of wind damage in forests The method integratesthe models into GIS software (ArcGIS) as a toolbar whichis the most representative method in related fields Howeverbecause the models are integrated in a fixed workflow thesystem looks more like a calculator than a DSS In addi-tion other problems can be outlined as follows (1) softwaredevelopment takes a lot of time (2) this mode is very hardto support the multidisciplinary multi-department collabora-tion (3) user interface can not be customized to individualneeds and differences impact the subsequent use of a DSS(4) it is difficult to verify or maintain a single model

Two other methods (Chen et al 2011 Qi and Altinakar2011) are similar to that of Zeng et al (2007) The major dif-ference is that they deploy a standalone application as an ex-tension of ArcGIS under the Windows environment There-fore these two methods have the same problem with Zeng etal (2007)

Superior from other simpler DSS the most important qual-ities of MDSS are flexibility and adaptability In this sectionwe will analyze a real example (which is described in Sect 5)with quantitative results to compare MDSS with other DSS

The decision situations of flood risk management in theJingjiang flood diversion area are complex and uncertainThere are several basic problems in the implement of sucha huge system which can be outlined as follows (1) thewhole decision-making process is composed of 49 multidis-ciplinary models (2) difficulties in joint integration becauseof the model researchers are distributed across four cities



(3) copyright and security issues regarding the use of origi-nal data (4) the ever-changing needs of project managers

The research team began work with traditional methodsfrom 2007 The whole decision-making process was splitinto several parts which were assigned to many groups withsynchronized communication once a month This methodwas maintained for two years The decision processes andcomputer systems were becoming too complex to harmonizeor maintain control For this reason we present the model-driven method that accepts heterogeneity and leads to adapt-ability and developed a loose-coupling model framework formodel integration Under the new framework integration ofa single model takes from 3 h to 2 days and implementa-tion of the whole project took 2 months The first genera-tion of MDSS has been running in the Jingjiang flood di-version area for three years During this period managerschanged the decision-making processes five times and theaverage maintenance time is 2 weeks The comparative stud-ies posed above demonstrate that MDSS is efficient adapt-able and flexible

7 Conclusions

Quantitative models embedded in a DSS can help managersmake better decisions (Power and Sharda 2007) and the ef-fectiveness of DSS is enhanced through dynamic adaptationof support regarding the needs of the decision-maker (Bijanet al 1997) This paper presents a beneficial approach to en-sure the adaptability of DSS for flood risk management andthe main innovation is the application of model-driven con-cepts which are promising for loose-coupling of GIS andmultidisciplinary models

During the course of the model-driven method several im-portant issues specific to the implementation effort becameapparent and had to be resolved The service-oriented archi-tecture (SOA) is employed to solve the integration problemsSOA accepts that the only way to maintain flexibility in largedistributed systems is to support heterogeneity decentraliza-tion and fault tolerance (Josuttis 2009) It is not a specifictechnology but a way of thinking Our model-driven methodis a strategy that includes both technical and organizationalaspects Organizationally we need appropriate processes sothat it is clear how to design new solutions and identify newservices (see Sect 3) Technically we need some infrastruc-ture to provide interoperability (see Sect 4) In addition ourMDSS is running successfully to manage the flood risk inthe Jingjiang flood diversion area located in central Chinanear the Yangtze River (see Sect 5) Section 6 demonstratesthe benefits of our model-driven method through comparisonwith some traditional solutions It demonstrates that MDSS isefficient adaptable and flexible and thus has bright prospectsof application for comprehensive flood risk assessment

Appendix A

MOCSCDE algorithm for hydrology forecasting

Suppose the optimization is done under the framework ofmulti-objectivity If the pre-defined objective functions con-flict with each other then the final optimization results can-not be one individual but a set of individuals who are the so-called non-dominated results If all the objective functionsaim to be minimized this optimization problem can be de-fined as follows

minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)

whereM is the number of objective functionsX denotes themodel parameters andfi is theith objective function In thispaper the objective functions are

MSLE =1

N

Nsumi=1

(lnQi minus lnQi)2 (A2)

M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)

whereN is the sample lengthQi denotes theith observedstreamflow value andQi is the ith simulated streamflowvalue The flowchart of the proposed algorithm MOCSCDEis shown in Fig A1

Appendix B

Shallow water equations in hyperbolic conservation form

The classical shallow water equations refer to the formula-tion of hydrostatic pressure and bottom slope effects and theirdivision between fluxes and source terms Song et al (2011a)proposed a formulation of the classical two-dimensionalshallow water equation in conservation form

partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+

0minusghSfxminusghSfy

(B2)



Start

Initialize

Devide the population into NC Complexes Ci(i=12hellipNC)

Set t=0

tltMax

Combine NC Complexes

t=t+1

Update the situational knowledge( 12 )c

iP i NC=

Update the situational knowledge P

Update the normative knowledge

Update the history knowledge

t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No

Fig A1 Flowchart of Multi-objective Culture Shuffled ComplexDifferential Evolution

whereh is the water depthu andv are the velocity compo-nents inx- andy directions respectivelyb is the bed eleva-tion S0x andS0y are bed slopes in thex- andy directionsrespectively assuming the bed is fixed ieb = b(xy) thenS0x = minuspartb

partx andS0y = minuspartb

party g is the gravity accel-

erationSf x andSfy are the friction terms in thex- andy

directions respectively The friction terms are estimated bythe Manning formulae

Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)

AcknowledgementsThis work is supported by the State KeyProgram of National Natural Science of China (No 51239004) theSpecial Research Foundation for the Public Welfare Industry ofthe Ministry of Science and Technology and the Ministry of WaterResources of China (No 201001080 201001034 200901010)National Natural Science Foundation of China (51107047) and theresearch funds of University and College PhD discipline of China(No 20100142110012)

Edited by P TarolliReviewed by E I Nikolopoulos andtwo anonymous referees

References

ArcGIS Resource Center About the ArcGIS Server 93 avail-able athttpresourcesarcgiscomzh-cncontentarcgisserver93about(last access 4 January 2013) 2008


Beacutezivin J In Search of a Basic Principle for Model Driven Engi-neering NovaticaUpgrade 2 21ndash24 2004

Bijan F Mihir A and Sameer V Adaptive decision support sys-tems Decis Support Syst 20 297ndash315 1997

Chen H Wood M D Linstead C and Maltby E Uncer-tainty analysis in a GIS-based multicriteria analysis tool for rivercatchment management Environ Modell Softw 26 395ndash405doi101016jenvsoft201009005 2011

Chen S and Guo Y Variable fuzzy sets and its application in com-prehensive risk evaluation for flood-control engineering systemFuzzy Optim Decis Ma 5 153ndash162 doi101007s10700-006-7333-y 2006

Du J He F and Shi P Integrated flood risk assessment of Xi-angjiang River Basin in China Journal of Natural Disaster 1538ndash44 2006 (in Chinese)

Eom S B Decision Support Systems Research and Reference Dis-ciplines (1970ndash2001) A Research Guide to the Literature andan Unobtrusive Bibliography with Citation Frequency EdwinMellen Press Lewiston NY 2003

Escuder-Bueno I Castillo-Rodriacuteguez J T Zechner S JoumlbstlC Perales-Momparler S and Petaccia G A quantitative floodrisk analysis methodology for urban areas with integration of so-cial research data Nat Hazards Earth Syst Sci 12 2843ndash2863doi105194nhess-12-2843-2012 2012

Evers M Jonoski A Maksimovic C Lange L Ochoa Ro-driguez S Teklesadik A Cortes Arevalo J Almoradie AEduardo Simotildees N Wang L and Makropoulos C Collabo-rative modelling for active involvement of stakeholders in urbanflood risk management Nat Hazards Earth Syst Sci 12 2821ndash2842 doi105194nhess-12-2821-2012 2012

Feng P Cui G T and Zhong Y On the evaluation and predictionof urban flood economic loss J Hydraul Eng 8 64ndash68 2001(in Chinese)

Gates B Bill Gates Keynote Microsoft Tech-Ed 2008 ndash Develop-ers available athttpwwwmicrosoftcompresspassexecbillgspeeches200806-03techedmspx(last access 4 January 2013)2008

Guo J Zhou J Qin H Zou Q and Li Q Monthlystreamflow forecasting based on improved support vec-tor machine model Expert Syst Appl 38 13073ndash13081doi101016jeswa201104114 2011

Guo J Zhou J Zhou CWang G and Zhang Y Multi-objectiveoptimization for conceptual hydrological models Advances inWater Science 23 447ndash456 2012 (in Chinese)

Hu J Khalil I Han S and Mahmood A Seamless integrationof depend ability and security concepts in SOA A feedback con-trol system based framework and taxonomy J Netw ComputAppl 34 1150ndash1159 doi101016jjnca201011013 2011

International Federation of Red Cross and Red Crescent SocietiesWorld disaster report Oxford University Press Oxford 1998



Jiang W Li J Chen Y Sheng S and Zhou G Risk assessmentsystem for regional flood disaster principle and method Journalof natural disaster 17 53ndash59 2008 (in Chinese)

Josuttis N M SOA in Practice ndash The Art of Distributed SystemDesign OrsquoReilly Media Inc 2009

Ray J J A web-based spatial decision support system optimizesroutes for oversizeoverweight vehicles in Delaware Decis Sup-port Syst 43 1171ndash1185 doi101016jdss200507007 2007

Li S Yu P and Sun S Fuzzy Risk Assessment of Flood HazardBased on Artificial Neural Network for Detention Basin Chinarural water and hydropower 6 60ndash64 2008 (in Chinese)

Li Y Gong J H Zhu J Ye L Song Y Q and Yue Y JEfficient dam break flood simulation methods for developing apreliminary evacuation plan after the Wenchuan Earthquake NatHazards Earth Syst Sci 12 97ndash106 doi105194nhess-12-97-2012 2012

Liu Y Zhou J Z Song L X Zou Q Liao L and Wang Y RNumerical modelling of free-surface shallow flows over irregulartopography with complex geometry Applied Math Model 379482ndash9498 doi101016japm201305001 2013

McKinney D C and Cai X Linking GIS and water resourcesmanagement models an objectoriented method Environ Mod-ell Softw 17 413ndash425 2002

Mehmood A and Jawawi D N A Aspect-oriented model-drivencode generation A systematic mapping study Inform SoftwareTech 55 395ndash411 doi101016jinfsof201209003 2013

Merz B Kreibich H Schwarze R and Thieken A Reviewarticle ldquoAssessment of economic flood damagerdquo Nat HazardsEarth Syst Sci 10 1697ndash1724 doi105194nhess-10-1697-2010 2010

Power D J and Sharda R Model-driven decision support sys-tems Concepts and research directions Decis Support Syst 431044ndash1061 doi101016jdss200505030 2007

Qi H and Altinakar M S A GIS-based decision support systemfor integrated flood management under uncertainty with two di-mensional numerical simulations Environ Modell Softw 26817ndash821 doi101016jenvsoft201011006 2011

Schallehn E Sattler K and Saake G Efficient similarity-basedoperations for data integration Data Knowl Eng 48 361ndash387doi101016jdatak200308004 2003

Schinke R Neubert M Hennersdorf J Stodolny U SommerT and Naumann T Damage estimation of subterranean build-ing constructions due to groundwater inundation ndash the GIS-basedmodel approach GRUWAD Nat Hazards Earth Syst Sci 122865ndash2877 doi105194nhess-12-2865-2012 2012

Song L Zhou J Guo J Zou Q and Liu Y A robust well-balanced finite volume model for shallow water flows with wet-ting and drying over irregular terrain Adv Water Resour 34915ndash932 doi101016jadvwatres201104017 2011a

Song L Zhou J Li Q Yang X and Zhang Y An unstructuredfinite volume model for dam-break floods with wetdry frontsover complex topography Int J Numer Meth Fl 67 960ndash980doi101002fld2397 2011b

Wang X Zhang X and Lai G Over-standard integrated riskanalysis of flood control system J Hydraul Eng 2 83ndash87 2004(in Chinese)

Watson A A brief history of MDA upgrade The European Jour-nal for the Informatics Professional IX 7ndash11 2008

Wu Q Xu H and Zou X An effective method for 3D geologicalmodeling with multi-source data integration Comput Geosci31 35ndash43 doi101016jcageo200409005 2005

Xie T Zhou J Song L Wang Y and Zou Q Dynamic Evalua-tion and Implementation of Flood Loss Based on GIS Grid DataComm Com In Sc 228 558ndash565 2011

Zeng H Talkkari A Peltola H and Kellomaki S A GIS-baseddecision support system for risk assessment of wind damagein forest management Environ Modell Softw 22 1240ndash1249doi101016jenvsoft200607002 2007

Zhao R Watershed Hydrological Modelling Beijing ChinaWaterPower Press 109ndash118 1984 (in Chinese)

Zou Q Zhou J Yang X He Y and Zhang Y A comprehensiveassessment model for severity degree of dam failure impact basedon attribute interval recognition theory Journal of Sichuan Uni-versity (engineering science edition) 43 45ndash50 2011 (in Chi-nese)

Zou Q Zhou J Zhou C Song L Guo J and Liu Y The prac-tical research on flood risk analysis based on IIOSM and fuzzy-cut technique Appl Math Model 36 3271ndash3282 2012a

Zou Q Zhou J Zhou C Guo J Deng W Yang M andLiao L Fuzzy risk analysis of flood disasters based on diffused-interior-outer-set model Expert Syst Appl 39 6213ndash6220doi101016jeswa201112008 2012b

Zou Q Zhou J Zhou C Song L and Guo J Comprehensiveflood risk assessment based on set pair analysis-variable fuzzysets model and fuzzy AHP Stoch Env Res Risk A 27 525ndash546 doi101007s00477-012-0598-5 2012c



Traditionally tight-coupling integrations of one or morequantitative models were the dominant of the flood riskmanagement approaches However unique aspects of floodrisk management problems require a special approach fordynamic adaptation of support regarding the needs of thedecision-maker Bijan et al (1997) believed that there isan additional intelligent ldquoadaptationrdquo component that canachieve absolute adaptation We cannot prove that this as-sumption is feasible but our approach is to ensure that thealgorithms and models are in a loose-coupling frameworkeg can be easily reconstructed to accommodate the de-mand Over the past 10 yr we have focused on the need forflood control by integrating remote sensing and GIS tech-niques with qualitative or quantitative models including hy-drologic analysis flood simulation flood risk analysis anddisaster assessment (Zou et al 2012a b Guo et al 2011Song et al 2011b)

In recent years the purpose of our research has increas-ingly become to propose a universal approach for flood riskmanagement by the model-driven method which providesan accessible interface to non-technical specialists such asmanagers Furthermore the method is intended for repeateduse in the same or a similar decision situation (Power andSharda 2007) In order to achieve these objectives many re-search studies have been carried out on this topic They canbe outlined as follows (1) model-driven method for floodrisk management (2) loose-coupling framework of hardwareand software system (3) multidisciplinary optimization andmathematical programming models

For the readerrsquos convenience the remainder of this paperis organized as follows

ndash Section 2 provides an overview and analysis of pastmodel-driven method research then identifies researchchallenges related to behavioral and technical aspectsof implementing and using the method for flood riskmanagement

ndash Section 3 defines our model-driven methodologicalframework and proposes its systems life cycle for de-signing and developing a MDSS

ndash Section 4 illustrates the loose-coupling technical pro-totype through identifying and incorporating the keycomponents for our model-driven method

ndash Section 5 presents how we adopt the method in theJingjiang flood diversion area where we are commis-sioned by the government to assist the operations ofengineering and controlling the flood risk

ndash Section 6 demonstrates the benefits of our model-driven method through comparing with some tradi-tional solutions

ndash The final section provides a summary of observationsand recommendations for future directions of researchin model-driven methods for flood risk management

2 Background

21 Overview of model-driven method research

Nowadays developing applications for non-technical endusers is becoming increasingly complex as they want toaccess these applications everywhere at every moment ofthe day and night and with many different devices (Gates2008) On the other hand the model which address the de-cision support problem for flood risk management is be-coming multidisciplinary (as discussed last section) There-fore economists geographers water resource specialists op-eration researchers and management scientists had to inte-grate their respective models together to support managersrsquodecision-making and planning

Traditionally by using object-oriented software engineer-ing many computer programs which are based on a tightlycoupled integration of models and interface is presented tohelp the decision-maker make effective decisions It is com-monly believed that this way can guarantee the stability andhigh performance of the system However given the grow-ing complexity and uncertainty in many decision situationsof flood risk management we need a special approach tothe dynamic adaptation of support regarding the needs ofthe decision-maker This approach needs to provide an in-tegration platform for multidisciplinary models not subjectto geographical distribution research area development lan-guage software tool and other constraints and to guaran-tee that the system has a higher level of adaptability andreusability Taking the above factors into consideration thefocus of the study will be divided into two parts (1) model-centric decision-making process ensures DSS adaptability(2) loose-coupling implementation ensures DSS flexibility

Model-driven method research can be distributed in twoaspects The first is as DSS foundations research which hasbeen one of the most important topics since 1969 (Powerand Sharda 2007) the method uses algebraic decision an-alytic financial simulation and optimization models to pro-vide decision support The second software engineering re-search model-driven method (or model-driven engineering)and recent trend whose main proposal is to focus on mod-els rather than on computer programs (Beacutezivin 2004) Andit is being increasingly applied to enhance system develop-ment from perspectives of maintainability extensibility andreusability (Mehmood and Jawawi 2013) With the rapid de-velopment within the field of Web service technologies andservice-oriented computing the model-driven method hasbeen widely used in the area of business process manage-ment (Watson 2008)

Eom (2003) identified a wide variety of DSS applica-tions reported in the academic literature There are more than1800 related articles and many of them focused on our re-search direction with good results However by querying thislarge number of scientific papers we noticed that the model-driven research within the field of natural disasters is at the






















and models




Practitioners

Managers


CommonClient


RecommendedClient



or

User interfaces








Technical-loop

Architecture

development

Continuous


models and systems

Continuous testing

Continuous


Decision making

goals

Requirements

analysis

Platform design

Problem identify


Decision-making


customization

Decision scheme set

generation


FeedbackSupport













Communication Class

Interface Class

Various Client

Data Class

GIS Service

Main Class


Services Layer



Interface Service

Communication Class

Service Provider


UDDI

Service Consumer


Publishing

Discovery

Composition

Binding

Execution


Invoke (SOAP)

Base Layer









Relational database

E-mail

Notepads



Picture files

Structured data

Economic data

Population data

Resource data

Building model

Environment model

Feature model

DEM

DOMRoad

network

Property database


Basemap database






























Managers

ESB


logging server


WSDL interface

Model server

Load balancer

Service

management


Send requests

Reliability and

security server

Scientists from



Publish models by

themselves

Practitioners

Customize model

composition














































W =

sumi

sumj

sumk

αiAijηjk (1)




61 Method benefits








based on VFS





Km


Low 7850

High 0


EmbankmentCofferdam

123 6 18Km

0

Water depth (m)0

0001-05

05-10

10-15

15-20

20-25

25-30

30-35

35-40

40-50

50-70

Km

Embankment

180

Low 0

High 6


Km



Affected area

A B

C D

Cofferdam123 6











Low

Medium

High

Very high


Km0

BA C












Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process


Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit















technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References

































































and models




Practitioners

Managers


CommonClient


RecommendedClient



or

User interfaces








Technical-loop

Architecture

development

Continuous


models and systems

Continuous testing

Continuous


Decision making

goals

Requirements

analysis

Platform design

Problem identify


Decision-making


customization

Decision scheme set

generation


FeedbackSupport













Communication Class

Interface Class

Various Client

Data Class

GIS Service

Main Class


Services Layer



Interface Service

Communication Class

Service Provider


UDDI

Service Consumer


Publishing

Discovery

Composition

Binding

Execution


Invoke (SOAP)

Base Layer









Relational database

E-mail

Notepads



Picture files

Structured data

Economic data

Population data

Resource data

Building model

Environment model

Feature model

DEM

DOMRoad

network

Property database


Basemap database






























Managers

ESB


logging server


WSDL interface

Model server

Load balancer

Service

management


Send requests

Reliability and

security server

Scientists from



Publish models by

themselves

Practitioners

Customize model

composition














































W =

sumi

sumj

sumk

αiAijηjk (1)




61 Method benefits








based on VFS





Km


Low 7850

High 0


EmbankmentCofferdam

123 6 18Km

0

Water depth (m)0

0001-05

05-10

10-15

15-20

20-25

25-30

30-35

35-40

40-50

50-70

Km

Embankment

180

Low 0

High 6


Km



Affected area

A B

C D

Cofferdam123 6











Low

Medium

High

Very high


Km0

BA C












Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process


Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit















technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References















































Communication Class

Interface Class

Various Client

Data Class

GIS Service

Main Class


Services Layer



Interface Service

Communication Class

Service Provider


UDDI

Service Consumer


Publishing

Discovery

Composition

Binding

Execution


Invoke (SOAP)

Base Layer









Relational database

E-mail

Notepads



Picture files

Structured data

Economic data

Population data

Resource data

Building model

Environment model

Feature model

DEM

DOMRoad

network

Property database


Basemap database






























Managers

ESB


logging server


WSDL interface

Model server

Load balancer

Service

management


Send requests

Reliability and

security server

Scientists from



Publish models by

themselves

Practitioners

Customize model

composition














































W =

sumi

sumj

sumk

αiAijηjk (1)




61 Method benefits








based on VFS





Km


Low 7850

High 0


EmbankmentCofferdam

123 6 18Km

0

Water depth (m)0

0001-05

05-10

10-15

15-20

20-25

25-30

30-35

35-40

40-50

50-70

Km

Embankment

180

Low 0

High 6


Km



Affected area

A B

C D

Cofferdam123 6











Low

Medium

High

Very high


Km0

BA C












Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process


Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit















technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References



























































Managers

ESB


logging server


WSDL interface

Model server

Load balancer

Service

management


Send requests

Reliability and

security server

Scientists from



Publish models by

themselves

Practitioners

Customize model

composition














































W =

sumi

sumj

sumk

αiAijηjk (1)




61 Method benefits








based on VFS





Km


Low 7850

High 0


EmbankmentCofferdam

123 6 18Km

0

Water depth (m)0

0001-05

05-10

10-15

15-20

20-25

25-30

30-35

35-40

40-50

50-70

Km

Embankment

180

Low 0

High 6


Km



Affected area

A B

C D

Cofferdam123 6











Low

Medium

High

Very high


Km0

BA C












Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process


Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit















technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References




















































































W =

sumi

sumj

sumk

αiAijηjk (1)




61 Method benefits








based on VFS





Km


Low 7850

High 0


EmbankmentCofferdam

123 6 18Km

0

Water depth (m)0

0001-05

05-10

10-15

15-20

20-25

25-30

30-35

35-40

40-50

50-70

Km

Embankment

180

Low 0

High 6


Km



Affected area

A B

C D

Cofferdam123 6











Low

Medium

High

Very high


Km0

BA C












Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process


Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit















technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References


















































based on VFS





Km


Low 7850

High 0


EmbankmentCofferdam

123 6 18Km

0

Water depth (m)0

0001-05

05-10

10-15

15-20

20-25

25-30

30-35

35-40

40-50

50-70

Km

Embankment

180

Low 0

High 6


Km



Affected area

A B

C D

Cofferdam123 6











Low

Medium

High

Very high


Km0

BA C












Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process


Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit















technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References





















































Low

Medium

High

Very high


Km0

BA C












Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process


Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit















technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References















































Social-economic

data

Hydrodynamic

simulation results

Depth-duration-

damage equations

Reclassify

and rasterize

Algebra

process


Raster

calculate

Grid-based disaster

estimation models

Damage per

reference unit















technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References



















































technical prototype




















7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References
















































7 Conclusions



Appendix A



minf1(X)f2(X)fM(X)

X = [x1x2xD](A1)


MSLE =1

N

Nsumi=1


M4E =1

N

Nsumi=1

(Qi minus Qi)4 (A3)


Appendix B



partU

partt+

partE

partx+

partG

party= S (B1)

in which

U =

h

hu

hv

E =

hu

hu2+ g(h2

minus b2)2huv

G =

hv

huv

hv2+ g(h2

minus b2)2

S = S0 + Sf

=

0g(h + b)S0x

g(h + b)S0y

+


(B2)



Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References














































Start

Initialize


Set t=0

tltMax


t=t+1


iP i NC=




t mod NM=0

End

Set n=0

n=n+1

nltNBS

Mutation

Crossover

Selection

No

Yes

Yes

Yes

No

Yes

No







Sf x =n2u

radicu2 + v2

h43 Sfy =

n2vradic

u2 + v2

h43 (B3)



References













































efficient gis-based model-driven method for flood …...open access efficient gis-based...

Documents