School Work

2006 ArcGIS SWAT a Geodata Model and GIS Interface for SWAT

Description
gis geodata
Categories
Published
of 15
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  ABSTRACT: This paper presents ArcGIS-SWAT, a geodatamodel and geographic information system (GIS) interface forthe Soil and Water Assessment Tool (SWAT). The ArcGIS-SWATdata model is a system of geodatabases that store SWAT geo-graphic, numeric, and text input data and results in an orga-nized fashion. Thus, it is proposed that a single andcomprehensive geodatabase be used as the repository of aSWAT simulation. The ArcGIS-SWAT interface uses program-ming objects that conform to the Component Object Model(COM) design standard, which facilitate the use of functionalityof other Windows-based applications within ArcGIS-SWAT. Inparticular, the use of MS Excel and MATLAB functionality fordata analysis and visualization of results is demonstrated. Like-wise, it is proposed to conduct hydrologic model integrationthrough the sharing of information with a not-model-specifichub data model where information common to different modelscan be stored and from which it can be retrieved. As an exam-ple, it is demonstrated how the Hydrologic Modeling System(HMS) – a computer application for flood analysis – can useinformation srcinally developed by ArcGIS-SWAT for SWAT.The application of ArcGIS-SWAT to the Seco Creek watershedin Texas is presented.(KEY TERMS: watershed management; geographic informationsystems (GIS); water quality; hydrologic modeling; computa-tional methods; computer interface.)Olivera, Francisco, Milver Valenzuela, R. Srinivasan, JanghwoanChoi, Hiudae Cho, Srikanth Koka, and Ashish Agrawal, 2006.ArcGIS-SWAT: A Geodata Model and GIS Interface for SWAT.Journal of the American Water Resources Association (JAWRA)42(2):295-309. INTRODUCTIONThis paper presents ArcGIS-SWAT, a geodatamodel and GIS interface for the Soil and Water Assessment Tool (SWAT) (Neitsch  et al., 2002a,b).SWAT is a physically based, continuous time, riverbasin scale model that quantifies the impact of landmanagement practices on flows, sediment loads, andchemical yields. It models the entire hydrologic cycle,including the evapotranspiration, shallow infiltration,percolation to deep aquifers, and lateral flow process-es (Arnold  et al., 1998).Since a significant amount of SWAT's input dataare of a spatial character (such as those derived fromstream network, drainage divide, land use, and soiltype maps), GIS tools for extracting information forSWAT from readily available digital spatial data havebeen developed. Srinivasan and Arnold (1994), forexample, developed an interface in the GRASS plat-form, which was a front-end preprocessor that wroteSWAT input files. Bian  et al. (1996), in turn, devel-oped an interface that worked in the ARC/INFO plat-form (ESRI, Redlands, California) and ran in UNIX systems. Di Luzio  et al. (1998) likewise developed acomprehensive ArcView 3.x (ESRI, Redlands, Califor-nia) interface for SWAT that took advantage of itsgraphical user interface (GUI) and ran in Windowssystems. Di Luzio  et al. (2000, 2002) further devel-oped Di Luzio  et al. ’s (1998) interface by adding to itcapabilities for terrain analysis based on digital eleva-tion models (DEM). 1 Paper No. 04087 of the  Journal of the American Water Resources Association (JAWRA) (Copyright © 2006) . Discussions are open untilOctober 1, 2006. 2 Respectively, (Olivera) Assistant Professor of Civil Engineering, (Valenzuela, Choi, Cho, Koka, Agrawal) Graduate Research Assistants,Texas A&M University, 3136 TAMU, College Station, Texas 77843; (Srinivasan) Associate Professor of Biological and Agricultural Engineer-ing, Texas A&M University, 1500 Research Parkway, Suite 221E, College Station, Texas 77845 (E-Mail/Olivera: folivera@civil.tamu. edu). J OURNAL OF THE  A MERICAN  W ATER  R ESOURCES  A SSOCIATION  295 JAWRA JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION APRILAMERICAN WATER RESOURCES ASSOCIATION2006  A  RC GIS-SWAT: A GEODATA MODEL AND GIS INTERFACE FOR SWAT 1  Francisco Olivera, Milver Valenzuela, R. Srinivasan, Janghwoan Choi, Hiudae Cho, Srikanth Koka, and Ashish Agrawal 2   ArcGIS-SWAT has been developed for the ArcGISplatform. The ArcGIS-SWAT data model stores SWATgeographic, numeric, and text input data and results.The geodatabase data structure, which is that of arelational database with the capability of storing geo-graphic information in addition to numbers and text,was used for the data model. Therefore, a geodatabaseis proposed as the repository of all the spatial andtemporal information of a SWAT simulation, asopposed to a series of text files. Likewise, the ArcGIS-SWAT interface uses ArcObjects (Zeiler, 2001), whichconform to the Component Object Model (COM) proto-col and therefore facilitate the use, within ArcGIS-SWAT, of already available functionalities in otherWindows-based applications. In particular, the use of Microsoft Excel and MATLAB for result visualizationand statistical analysis is demonstrated. Another fea-ture of ArcGIS-SWAT is its capability to georeferencethe hydrologic response units (HRUs), which allows amore accurate calculation of the model parametersthan what would be obtained if they were averagedover the subbasins.Integration with other hydrologic models is accom-plished through the sharing of geographic and hydro-logic data. For this purpose, hydrologic objects (e.g.,watershed polygons) and their corresponding parame-ters (e.g., watershed times of concentration) areexported from the ArcGIS-SWAT geodatabase to anot-model-specific hub geodatabase, from which theycan be imported by another interface for use in a dif-ferent model. By implementing the hub geodatabaseconcept, model integration consists of interfacing eachmodel to the hub rather than to each of the othermodels. As an example, it is presented how HMS –  acomputer application for flood analysis developed bythe Hydrologic Engineering Center (HEC) (USACE,2005) and not related to SWAT –  uses informationsrcinally developed by the ArcGIS-SWAT interfaceand stored in the ArcGIS-SWAT data model.THE GEODATABASE DATA STRUCTURE AND ARCHYDROThe ArcGIS-SWAT data model is based on the geo-database data structure. Geodatabases are relationaldatabases that can also store geographic features(MacDonald, 1999). That is, a geodatabase is a collec-tion of tables whose fields can store a geographicshape (i.e., a point, a line, or a polygon), a string, or anumber and that are related to each other throughkey fields. Regardless of the number of tables andrelationships in a geodatabase, it is stored in a singlefile, and its contents can be explored using databasemanagement systems (DBMS). Non-GIS DBMS (e.g.,Microsoft Access), however, cannot access the geo-graphic information in the geodatabases. Geodatabasetables are object classes in which each row representsan object and each column stores an attribute of theobject. Feature classes are particular cases of objectclasses in which each object is additionally attributedwith a feature (i.e., a geographic shape). Featureclasses collect features of a single type, so that theyare either point feature classes, line feature classes,or polygon feature classes. Furthermore, featureclasses that share the same spatial extent can be col-lected in feature datasets. Within feature datasets,geometric networks can be built based on line featureclasses, which establish topologic relationships among their elements. Finally, relationship classes can becreated to associate objects of different classesthrough key fields and can be stored as part of thegeodatabase.Maidment (2002) pioneered the use of geodatabasedata structures in water resources engineering bydeveloping the ArcHydro data model. ArcHydro is ageospatial and temporal data model that supportshydrologic simulation models but is not a simulationmodel itself. In ArcHydro, the structure of a hydrolog-ic system is defined around the stream geometric net-work, in which links and junctions are defined.Drainage areas are related to network junctions, andtime series are related to monitoring points in themap, which in turn are related to network junctions.Thus, the stream network and the relationshipsamong the different hydrologic elements and the net-work junctions constitute the backbone of the ArcHy-dro representation of a system (Olivera  et al., 2002).The ArcHydro data model was developed with theenvisioned purpose of storing in an organized fashionspatial and time series data to support most (if notall) hydrologic models. However, the SWAT datastructure does not include all of the elements of  ArcHydro and, more importantly, includes a numberof elements not considered in it. Among the elementsof ArcHydro not included in the SWAT data structureare the channel cross sections and profile lines, thestream geometric network (since the network topologyin SWAT is based on attributes), and those that areredundantly stored as hydrographic elements and asdrainage elements. The SWAT data structure, on theother hand, includes many land and stream parame-ters to model the soil water balance, plant growth,irrigation and fertilization practices, and pollutanttransport that are not included in ArcHydro. Hence,using ArcHydro as the SWAT data model would haverequired such a level of customization that it wouldhave left little of its srcinal design, and it was decid-ed to develop a new geodata model specifically forSWAT. JAWRA 296  J OURNAL OF THE  A MERICAN  W ATER  R ESOURCES  A SSOCIATION O LIVERA , V ALENZUELA , S RINIVASAN , C HOI , C HO , K OKA , AND A GRAWAL  The SWAT data model presented in this paper is asystem of geodatabases with object classes, featureclasses, and feature datasets from which all inputdata needed by SWAT can be retrieved and in whichall output generated by SWAT can be stored. Withrespect to the SWAT input and output text files, geo-databases have the advantage that they can storegeographic data to describe subbasins, reaches, out-lets, reservoirs, and inlets and benefit from existing database technology. (Geodatabases are a particularcase of databases.) Database technology would beneeded, for example, to query the information todescribe the changes of a hydrologic variable overtime at a given location or to describe the status of the system at different locations at a given time. Among other capabilities, database technologyincludes cross-referencing of records in differenttables by means of common attributes, speeding upthe query of records that match a given criteria andextending that query to cross-referenced records inother tables, updating records in bulk, and perform-ing complex aggregate calculations. Most importantly,database technology has been developed to work withlarge amounts of information otherwise impossible tohandle with text files. In the specific case of SWAT,the amount of information is significant because thegeodatabase stores geographic information, hydrologicparameters, and time series of each hydrologic featureof the system.METHODOLOGY The ArcGIS-SWAT data model consists of a dynam-ic geodatabase that stores information of the studyarea and a static geodatabase that stores non-project-specific information such as lookup tables anddatabases of default parameter values. The ArcGIS-SWAT interface includes modules for watershed delin-eation, HRU definition, synthetic weather generation,exporting data from the geodatabases to prepareSWAT input files, importing SWAT results from theoutput files to the dynamic geodatabase, analysis of propagation of uncertainty, data visualization andstatistical analysis, and model integration. The firstthree modules include spatial analysis using topo-graphic, land use, soil type, and weather data. Theother modules connect the SWAT data model toSWAT and support hydrologic analysis and modelintegration. For clarity purposes, in the following,object classes and feature classes are referred to as<( Object class name )> and [(  Feature class name )]respectively, in which italic font in parenthesis indi-cates the text has to be replaced with the correspond-ing name. Watershed Delineation The watershed delineation module identifiesstreams and drainage divides from DEMs using theeight-direction pour point algorithm (Jensen andDomingue, 1988). It follows the procedure presentedby Olivera (2001) for DEM-based stream and water-shed delineation but adapted to the SWAT data struc-ture. Reaches are defined wherever drainage areasare greater than a user defined threshold value; sub-basin outlets are automatically defined on each of thereaches right upstream of the confluences and at userdefined points; and subbasins are defined as theincremental drainage area of each outlet. Thus, asrequired by SWAT, one-to-one relationships are estab-lished between reaches, outlets, and subbasins (i.e.,no subbasin has more than one reach, and no reachlies in more than one subbasin). Additionally, otherelements can be interactively defined on the map suchas inlet points to the system (which allow one toexclude upstream drainage areas and isolate the por-tion of the watershed to be modeled), reservoirs(which are also subbasin outlets), and point sourcedischarges. Likewise, the subbasin longest path isused by SWAT as a surrogate of the residence time inthe subbasin. Figure 1 shows the hydrologic elementsof the Seco Creek watershed in Texas. The watersheddelineation module creates six feature classes in afeature dataset: [Watershed], [SubBasin], [Reach],[LongestPath], [Outlet], and [MonitoringPoint] (seeFigure 2). [PolyHRU] is created by the HRU Defini-tion module that will be discussed below. [Watershed]stores the polygon that represents the entire studyarea; [SubBasin] stores the subbasin polygons;[Reach] stores the segments of the channel network;[LongestPath] stores the longest flow path withineach subbasin; [Outlet] stores the subbasin outletpoints; and [MonitoringPoint] stores inlet points tothe watershed, reservoir points, point source dis-charges, and a copy of the subbasin outlet points,among others points that are appended by otherinterface modules. In Figure 1, note that white circlesare used to represent those monitoring points thatcoincide with outlets, while black circles to representthose that do not coincide with outlets, such as raingauges. [MonitoringPoint] is also related to objectclass <TimeSeries>, which stores all observed and cal-culated hydrologic time series for all the features of the system. By storing a copy of the outlet points in[MonitoringPoint], time series of any feature of thesystem can be related to its corresponding location inthe map. Additionally, all features have a unique identifica-tion number stored in the field HYDROID, which isused to establish relationships between the different J OURNAL OF THE  A MERICAN  W ATER  R ESOURCES  A SSOCIATION  297 JAWRA A RC GIS-SWAT: A G EODATA M ODELAND GIS I NTERFACEFOR SWAT  classes. These relationships are represented byarrows in Figure 2. For example, the relationshipbetween [Outlet] and [Subbasin] is established bystoring the HYDROID value of the subbasin outletpoint in the OUTLET field of the subbasin polygon.Note that in <TimeSeries>, the field FEATUREIDstores the HYDROID value of the corresponding fea-ture. The feature classes also have fields related totheir geometry (i.e., AREA for polygons and LENGTHfor lines), and [Reach], in particular, has FNODE andTNODE that store the upstream and downstreampoints of each reach segment, which are used to estab-lish the stream network topology. Other attributessuch as elevation, slope, and location (i.e., longitudeand latitude) are also included.  HRU Definition The HRU definition module identifies unique com-binations of soil and land use within each subbasin. Soil and land use data can be provided by the user;however, tools have been included for easy use of soildata from the State Soil Geographic (STATSGO)database (USDA-NRCS, 1995) and lookup tables forconverting different land use classifications to theSWAT classification.The STATSGO database defines map units, each of which consists of one or more polygonal areas of thesame soil type. This database, whose format is differ-ent from the one defined in USDA-NRCS (1995),includes one feature class of map-unit polygons perstate, called [(State Name)], and one object class permap unit, called <(Map unit)>. Texas, for example,has 633 map units represented by 4,031 polygons, andconsequently if the study area were in Texas, a fea-ture class [Texas] with 4,031 map unit polygons and633 map unit object classes <TX001>, <TX002>, … <TX633> would be included in the static geodatabase.In these object classes, the records refer to the differ-ent soil components of the map unit, and the fieldsstore soil properties of the component and of up to 10layers of the component, as well as the percentage of the component in the map unit. To speed up thedatabase operations, only the feature and object class-es of the states needed are added to the static geo-database from a source set of 48 geodatabases of STATSGO data (i.e., one geodatabase per state of theconterminous United States). The reader is referredto USDA-NRCS (1995) for detailed STATSGO docu-mentation on soil map units, components, and layers.Similarly, land use data consist of polygonal areaswithin which a single land use is found. Land usescan be classified with any land use classification sys-tem but should be converted to the SWAT classifica-tion system for implementation with SWAT. Thus, alookup table, that the user can modify if needed, hasbeen included to define the equivalence of the Ander-son  et al. (1976), National Land Cover Dataset(NLCD) (USGS, 2004a), and SWAT land use classifi-cation systems.Regardless of the sources of the input data, one soiltype grid and one land use grid are created. Afteridentifying the grid cells that have the same soil typeand land use within each subbasin, they are groupedtogether and converted into polygons that representHRUs. Thus, in the resulting polygon feature class,called [PolyHRU], the features have a unique combi-nation of input data. As can be seen in Figure 2,[PolyHRU] polygons have a unique identificationnumber stored in field HYDROID, which is used to establish relationships with [SubBasin] and <TimeSeries>. Additionally, ArcGIS-SWAT calculates runoff curvenumbers (USDA-SCS, 1972) and terrain slopes perHRU. Runoff curve numbers are calculated based on JAWRA 298  J OURNAL OF THE  A MERICAN  W ATER  R ESOURCES  A SSOCIATION O LIVERA , V ALENZUELA , S RINIVASAN , C HOI , C HO , K OKA , AND A GRAWAL Figure 1. Feature Classes Generatedby the Watershed Delineator.
Search
Similar documents
View more...
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks