ANALYSIS OF COASTAL GEORGIA ECOSYSTEM STRESSORS USING GIS INTEGRATED REMOTELY SENSED IMAGERY AND MODELING:
A PILOT STUDY FOR LOWER ALTAMAHA BASIN
INTERIM REPORT
Dr. M. M. Aral and Orhan Gunduz
Multimedia Environmental Simulations Laboratory
School of Civil and Environmental Engineering
Georgia Institute of Technology
Atlanta, Georgia 30332
March 2001
Table of Contents
This interim report presents the progress made in the University of Georgia Sea Grant College Program Project entitled "Analysis of Coastal Georgia Ecosystem Stressors using GIS Integrated Remotely Sensed Imagery and Modeling: A Pilot Study for Lower Altamaha Basin". This report highlights the steps taken in the first phase of the project and discusses the fundamental issues addressed during project formulation. The main activities performed in this first phase of the project included: (1) a thorough review of the literature pertinent to the various project components, (2) data collection and compilation activities, (3) model formulation and development tasks and (4) planning of the future stages of the project. The following sections of this report introduce the details of these project activities.
The main project activity during the first phase of the study focused on a comprehensive literature review. This review process mainly aimed studying the current developments in the field of (i) spatially-distributed hydrologic modeling, (ii) remote sensing applications in hydrological science, (iii) computer models of watershed hydrology and hydraulics as well as (iv) geographic information system (GIS) implementation of these models. According to the information base gathered from this review process, data collection and compilation efforts are initiated and are still in effect. Most of the efforts were directed towards locating the source of a particular data, analyzing and downloading this data and compiling the data according to the needs of the project. One other direct outcome of the literature review process was to identify the current status of hydrologic and hydraulic modeling, which constitutes the main core of this project. Currently, several models are under development as modules of an integrated modeling system. The following sections present the details associated with the literature review process and introduce the associated model development and data collection/compilation efforts undertaken during the past phase of the project.
2.1. Review of Literature
The literature review process for the project concentrated on four fundamental aspects. The most crucial one aimed at understanding of the current status of spatially distributed hydrologic modeling that forms the basis for the planned project objectives. A direct consequence of using spatially distributed modeling is the demand for using remotely sensed pixel-based data such as digital terrain models or land cover/use data. Therefore, literature review process also covered the details that help us understand the various uses of remotely sensed data in hydrologic modeling. The other parts of the review focused on the application of computer models used in hydrologic modeling and the use of GIS systems to orchestrate the different components of integrated modeling environment. The details of this review are discussed in the following sections.
2.1.1. Spatially Distributed Hydrologic Modeling
Spatially distributed hydrologic modeling is one of the most recent developments in hydrological modeling, which gained tremendous importance starting from the late eighties. Not surprisingly, the timing of early advances in distributed hydrologic modeling coincides with the increase in computer resources such as processing speed, memory and data storage. This link between computing and spatially distributed hydrological modeling is strongly related to the fact that spatial distribution of hydrological processes enormously increased the numerical computations performed even for a relatively small-scale watershed. Comparing spatially-distributed models with lumped models, which do not reflect the spatially non-uniformity and heterogeneity, clearly reveals that grid-based modeling of hydrological processes require much more input data and process parameters. This increased data requirement necessitates (at least for some data types) a more convenient way of data collection compared with the traditional techniques like on-site field measurement, cartographic analysis, etc. It was this motivation that originated the attempts to use remote-sensing data in spatially distributed hydrological modeling applications.
Spatially distributed hydrologic models originate from the idea to use more physically based processes to define the hydrologic phenomena, rather than approximating them with empirical formulas. Thus, considering spatial heterogeneity directed the research towards using mathematically well defined conservation laws such as conservation of mass, momentum and energy to simulate the natural system (Abbott and Refsgaard, 1996). Accordingly, using advanced physically based distributed models became the current trend in hydrological modeling. The traditional hydrologic models of lumped conceptual type were based on averaging the processed in the so-called 'hydrologic response units', giving average responses to hydrologic events. In order to extract more detailed information, the modeler should switch to grid-based distributed models that can solve this hydrologic response in much smaller spatial scales.
One of the pioneering works on distributed modeling was completed by Abbott et al. (1986a,b), resulting the development of the SHE (Systéme Hydrologique Européen) model which is renown to be one of the earliest of distributed models. Similarly, Beven (1984) has developed TOPMODEL to simulate runoff production from a watershed. As models become sophisticated in years, the data requirements increased dramatically and remote sensing and digital elevation models found widespread use in spatially distributed models. Numerous authors such as Tao and Kouwen (1989), Sharma et al. (1996), Johnson and Miller (1997) and Yao and Terakawa (1999) developed more recent examples of spatially distributed models.
2.1.2 Remote Sensing and Applications in Hydrologic Modeling
Physically, remote sensing can be accomplished via several platforms such as satellites, radars or aircrafts. Each of these platforms uses different sensors operating on different ranges of electromagnetic spectrum and hence provides useful information for various earth sciences covering a wide range of disciplines from meteorology to hydrology and geology to biology. The use of remote sensing data in hydrological modeling involves a broad scope information such as digital topography data, land cover, vegetation and soil data as well as meteorological data. A suitable combination of these data resources is applied to a hydrological model to simulate the rainfall-runoff process.
Using remote sensing as a new approach to collect hydrological data dates back to early seventies when the National Aeronautics and Space Administration (NASA) launched the first of a series of Earth-monitoring satellites that provide repetitive global coverage of land masses. Initially named as "Earth Resources Technology Satellite" (ERTS), this attempt initiated the NASA's on-going remote sensing program for earth sciences, which is then renamed as the well-known LANDSAT Program. This program provides temporal and spatial coverage of earth from the early seventies up to now. Following this successful endeavor by NASA, several other agencies in or outside the United States initiated several other remote-sensing programs such as GOES and AVHRR programs of National Atmospheric and Oceanographic Agency (NOAA) as well as national programs of France (SPOT), Canada (RADARSAT) and India (IRS) (Jensen, 1996). All of these programs provide different aspects of the data bank, which is required for hydrological modeling.
One of the most promising aspects of remote sensing in hydrological modeling is that remote sensing techniques offer areal measurements rather than point measurements obtained by the traditional techniques. The areal data coverage is one of the keystones of spatially distributed modeling. Remote sensing techniques also provide high resolution in space and/or time serving the temporal and spatial needs of distributed models. Furthermore, remotely sensed data is generally in digital form allowing easy processing and handling under GIS platform. Last but not the least, remote sensing has the most powerful aspect of supplying information from "remote" areas of the earth, where no measurements would have been taken otherwise (Schultz, 1988).
Today, one common application of remote sensing in hydrology is to measure precipitation. Recognizing the practical limitations of rain gages for measuring spatially averaged rainfall over large and/or inaccessible areas, hydrologists turned to remote sensing to quantify precipitation input. In this regard, radar is being extensive used in precipitation measurement. Snow, however, is treated differently from rain due to the lag time between when it falls and when it produces runoff. There is extensive research conducted on remote sensing of snow coverage and snow-water equivalent mainly by satellites and high altitude air photography. Recently, remote sensing of soil moisture became another and important application in hydrological modeling. Being important parts of evaporation and runoff modeling, new developments related to remote sensing of soil moisture allows the use of more physically sophisticated models to simulate rainfall-runoff process. One last but extremely important application of remote sensing is to determine the land use/cover on the watershed. Distributed models particularly need specific data on land use/cover to predict the land classifications, resistance factors and other runoff related parameters (Eagleman, 1995).
2.1.3 Computer Models, Data Types and GIS
Even though computer-based hydrologic modeling has been carried out for more than three decades, extensive use of distributed models entered hydrology literature in the last 10-15 years. Increasing availability of data and software for processing spatial information changed the way hydrologic systems are being modeled. With these technological advances, it is now possible to describe the watershed characteristics more accurately when determining its runoff response to rainfall input. Computer models are now capable of simulating the watershed in a finer time and spatial scale.
In general, different models address different aspects of the hydrological events occurring in the watershed. Models of precipitation excess, overland flow and river routing all tackle a distinct portion of the hydrological cycle. Our current understanding of the physical processes underlying these events help us develop fairly complex models of watershed hydrology. Singh (1995) compiled numerous computer models of watershed hydrology, some of which are the TOPMODEL (Beven, et al., 1995), the SLURP model (Kite, 1995), MIKE SHE (Refsgaard and Storm, 1995), SWMM (Huber, 1995) and HEC-1 (Feldman, 1995). All of these models incorporate the hydrological processes mentioned above to a different level of physical description as well as mathematical and computational difficulty. But common to all of them is the integrated modeling approach for the principal watershed processes.
One major data type that majority of spatially distributed models use is the topographic information. This data had previously been supplied by analyzing topographic maps of various scales. While this approach worked fine for long years in small watersheds and lumped models, it became impractical with the extended watershed sizes that are being analyzed and the increased spatial resolution of distributed models applied nowadays. Consequently, hydrologic modelers started to use digital topographic data, called Digital Elevation Models (DEM) or Digital Terrain Models (DTM). Digital elevation data is digital representation of cartographic information. It can follow one of the three commonly available forms; (i) raster or grid-based data, (ii) triangular irregular networks (TIN), and (iii) vector or contour line networks (DeVantier and Feldman, 1993). Each data structure has several unique features, which make them applicable to different set of conditions and models. According to the results of literature survey, however,
The results of the literature survey reflect the fact that grid-based DEMs are the most commonly used forms of digital terrain data. It is a fairly new data type, which entered hydrologic literature in the mid-eighties. Several researchers such as O'Callaghan and Mark (1984), Jenson and Dominique (1988), Tarboton et al. (1991), Jenson (1991) and Marts and Garbrecht (1992) proposed techniques to extract useful information from DEMs for hydrologic models. These studies formed the basis of digital elevation data analysis and set the stage to extract hydrology-related information such as watershed delineation and drainage network extraction. As more and more hydrologic information became available everyday, modelers started to develop models that can use such new information. Wang and Hjelmfelt (1998) used DEM data in an overland flow routing model. Johnson and Miller (1997) and Yao and Terakawa (1999) incorporated DEMs, raster data structures and distributed hydrologic models.
Even though the data requirements of distributed models are strongly correlated to the complexity of the model developed, most of the models mentioned above also require data on soil characteristics, land use/cover distribution and vegetation indices as well as temperature and precipitation data. In a typical example of such a model, Dutta et al. (2000) simulates the flood inundation in a river basin using surface and subsurface components of the hydrologic cycle (i.e., precipitation, evaporation, saturated and unsaturated zone flow, overland flow and channel flow) in separate modules. Similarly, Fortin et al. (2001) developed a new watershed model called HYDROTEL that uses all of the above mentioned data types within a GIS environment.
This extensive use of data in distributed modeling necessitated a platform that links spatial data to other information concerning the processes and properties related to geographic location (DeVantier and Feldman, 1992). Apparently, one such platform is the geographic information system (GIS) operating on personal or mainframe computers, which offers an improved and efficient way of data storage, manipulation and retrieval. A comprehensive study by Singh and Fiorentino (1996) covered a wide range of hydrological modeling applications using GIS platform. Today, GIS-based platforms such as ArcView and ArcInfo of ESRI and GRASS of USACE became like an unofficial industry standard for hydrologic modeling.
2.2. Data Collection and Compilation
Parallel to the findings of the literature review process, some of the project data associated particularly with the lower Altamaha basin are being collected, compiled and stored in the so-called Altamaha River database located in the MESL laboratory. This database includes different data types (i.e., DEMs, land cover/use maps, soil maps, satellite imagery and other coverages) in both raster and vector format. Raster and vector data are the two primary ways of storing spatial information in a computer. It is possible to use either or both when developing hydrologic models using GIS. Raster data are made up of individual grid cells (pixels) of a specified dimension. Vector data, on the other hand, is more "maplike" and comprised of points, lines and polygons. Vector data is more capable of representing small objects and maintains the three fundamental aspects of topology (i.e., connectivity, adjacency and area definition). Raster data, however, tend to distort the actual size and shape of the object and cause a stair-step formation along linear edges. In terms of processing efficiency, raster data is quite advantageous over vector data due to its matrix structure that favors mathematical computations. Nevertheless, both data types have different places where using one favors over the other and are both used extensively in GIS-integrated modeling of hydrologic systems. A mixture of raster and vector data will be used in the proposed modeling system. The details of these data sets are provided in the following sections.
2.2.1. Digital Elevation Models
Topography plays an important role in the distribution and flux of water and energy within the natural landscape. The quantitative assessment of hydrological processes such as surface runoff, evaporation and infiltration depend on the topographic configuration of the landscape, which is one of the several controlling boundary conditions (ASCE, 1999). Digitizing the natural landscape into an array of elevation values is called a Digital Elevation Model (DEM) and is a viable alternative to traditional field surveys and manual evaluation of topographic maps.
DEMs are generally stored in one of the three data structures mentioned previously. The raster or grid structure consists of a square matrix and the elevation of each grid square (pixel) is stored in a matrix node. In TIN structures, a continuous surface is generated from interconnected triangles with known elevation values at the vertices of the triangles. Contour-based structures, on the other hand, consist of digitized contour lines of specified elevation, which is also referred to as a Digital Line Graph (DLG).
Each of the DEM structures has its advantages and disadvantages. Square grid DEMs are most widely used because of their simplicity, processing efficiency and compatibility with other image storage formats. Disadvantages include grid size dependency of certain computed landscape parameter and the inability to adjust the grid size to changes in the size of the landscape. TINs overcome these disadvantages to some extent but they are limited by the computational processing difficulty. Finally, contour-based structures provide better outlines of landscape features but they are limited with their one-dimensional nature to represent two dimensional landscape continuum without the use of extra data (ASCE, 1999). Due to the reasons mentioned above, most of the research on hydrologic modeling is performed by using grid structures. The limitation of spatial resolution is overcome by the decreasing grid size of the latest products. Even though no firm guidelines are available for the selection of DEM and selection is mostly based on the particular application, data availability and experience, 30 by 30 meter DEM data by USGS (level 1 and 2) is applied successfully in many applications.
The Georgia GIS Data Clearinghouse, which is a part of the Georgia Spatial Data Infrastructure (http://gis.state.ga.us), is used as the main source of DEM data for the project. Clearinghouse provides access to GIS resources of Georgia for use by different parties such as government, academia, and the private sector. The database contains large amounts of data with different theme, scale, areal extent and data originator selection options. For the scope of this project, 7.5-minute DEM files at county and quadrangle level in 1:24000 scale are downloaded from the Clearinghouse web page. Another resource for the DEM data is the USGS ftp servers, which allow transfer of files in the Spatial Data Transfer Standard (SDTS) format.
2.2.2 Land Use/Cover Data
Land use/cover information is used in hydrologic modeling to estimate the surface roughness or friction as it affects the velocity of the overland flow of the water. Land use information is also used to estimate the amount of rainfall infiltration on a surface. Moreover, land use information is further coupled with the soil characteristics of a land surface to compute the percolation and water holding capacity values (ASCE, 1999). Leaf area index (LAI) values can sometimes be computed from land cover information depending on the quality and the resolution of the data.
Large-extent land use/cover information is almost always obtained by remote sensing techniques such as aircrafts and satellites. Interpretation of the remotely sensed digital images is generally performed by supervised and unsupervised classification methods. These classification techniques yield clusters of different spectral classes, which are then interpreted into different land use types.
The required land use/cover information is obtained from the data sets supplied by the Multi-Resolution Land Characteristics (MRLC) project, which is a collaborative effort of four U.S. Government agencies: U.S. Environmental Protection Agency, U.S. Geological Survey, National Oceanic and Atmospheric Administration and the U.S. Department of Interior. These four agencies formed the MRLC Consortium to purchase Landsat-TM images covering the conterminous U.S. and sponsored different programs to develop a national 30-meter land characteristics database using Landsat thematic mapper (TM) data. Of the several programs sponsored by MRLC Consortium, the National Land Cover Data (NLCD) program and the North American Landscape Characterization (NALC) program provided useful land cover/use data that can be used in hydrological modeling.
Similar to the DEM data, county-level land use/cover data is also available at the Georgia GIS Data Clearinghouse. This data is an outcome of MRLC-NLCD program and uses the national land cover data key, which has 9 major classes and 21 subclasses. For the scope of this project, county-level data is downloaded in different formats (i.e., ERDAS and TIFF formats of USGS and Georgia Department of Natural Resources). In addition, NALC data sets for the lower Altamaha Basin are acquired from the USGS Earth Resources Observation Systems (EROS) Data Center (EDC). The data is provided in 8-mm cartridges, which are read, processed and downloaded to the Altamaha River database located in MESL laboratory. The image analysis software "Environment for Visualizing Images" (ENVI v3.2) is purchased and installed on the MESL database in order to retrieve, process and visualize the NALC data set. This software is specifically developed to analyze satellite and aircraft remote sensing data (ENVI, 1999).
3. FUTURE PROJECT ACTIVITY2.2.3. Soil Data
Information on soil characteristics is vital in hydrologic modeling. Most components of the hydrologic cycle require specific soil data like soil type, porosity and hydraulic conductivity. Despite this key role of soil data in modeling, county-level soil data is generally not available throughout the U.S. in digital form. National Resources Conservation Service (NRCS) is in the process of digitizing this data from the already available hardcopy soil maps. Nevertheless, soil data is available in state level from the State Soil Geographic (STATSGO) database and Soil Survey Geographic (SSURGO) database. The STATSGO database provides soil maps at the 1:250000 scale, where as SSURGO database has maps at varying scales from 1:12000 to 1:63360 (ASCE, 1999).
Currently, only the large-scale soil maps of STATSGO and SSURGO databases are used in model development studies due to the limitations of small-scale county-level data at the Georgia GIS Data Clearinghouse database and any other resource in digital form.
2.2.4 Precipitation Data
Precipitation data is one of the most important data inputs of hydrologic modeling. Confidence in the hydrologic modeling effort depends largely on the availability of high quality rainfall and runoff data for model calibration and verification. Despite the recent developments in radar measuring of precipitation, rain gage data is still accepted to be the most readily available and reliable form of data for hydrologic modeling applications. The most significant precipitation data source for the lower Altamaha basin is the data archive operated by NOAA National Climatic Data Center (NCDC). This database is an extremely useful source of information as it, not only provides precipitation data but also supplies other important climatic data such as temperature and radiation, which are used in evaporation studies. In this regard, data from the meteorological stations operating in the project area are being compiled and downloaded to the Altamaha River database located in MESL laboratory.
2.2.5. Stream Data
Stream data for hydrologic modeling purposes are either used as readily available stream data in vector form or systematically derived from the DEM. There are two major sources of stream data in vector from. One of these sources is the digitized stream and channel network data obtained from the USGS in the form of DLGs, which are created from maps and other related sources. These are available as in several scales and categories. Commonly used DLGs are digitized from 7.5-minute USGS 1:24000 quadrangles. Another source of stream data in vector form is the data developed by the USEPA. The USEPA stream and channel network databases are called the River Reach Files. River Reach Files are a series of national hydrologic databases that uniquely identify and interconnect the stream segments or reaches that comprise the nation's surface water drainage system. Three versions of the Reach File currently exist, known as the RF1, RF2 and RF3, created from increasingly detailed sets of hydrography data created by USGS (ASCE, 1999). The USGS DLGs at 1:24000 scale are available at the Georgia GIS Data Clearinghouse database and are downloaded to the Altamaha River database at MESL Laboratory.
Surface drainage and channel network can also be automatically extracted from DEMs. In addition to extracting stream data, DEM processing also allows delineation of watersheds and other hydrologic data using GIS platform. Numerous researchers have contributed to this interesting feature of DEMs (Mark, 1984; Jenson and Dominigue, 1988; Garbrecht and Martz, 1993). Automated extraction techniques produce fast and accurate stream networks as long as the input DEM has good resolution and quality (ASCE, 1999).
Derivation of hydrography data from USGS DEMs and raster processing methods have been the core of many hydrologic model such as Wang and Hjelmfelt (1998), Johnson and Miller (1997) and Polarski (1997). Most of these methods incorporate the so-called D-8 approach in which flow from a cell is directed towards one of its 8 neighbors with the steepest slope. The method assigns a flow direction to each cell in the raster grid and accumulates area downslope along the flow paths connecting adjacent cells. The drainage network is identified by selecting a threshold catchment area at the bottom of which a source channel originates. All the cells with a greater catchment area are classified as a part of a drainage network (ASCE, 1999). This approach is fairly straightforward and generates connected networks. However, the use of D-8 method has certain drawbacks as it permits flow only in one direction away from a DEM cell. It is reported that this representation fails when there is divergent flow on convex slopes (Quinn et al., 1991).
The accuracy of the D-8 method is directly linked to the quality of the DEM and the method faces difficulties identifying the drainage pattern in the presence of depressions, flat areas and flow blockages. These features are often the result of noisy data, interpolation errors and systematic production errors in DEM elevation values. Therefore, any extraction algorithm requires that these sinks and problematic cells be removed from the data set prior to drainage extraction.
Drainage network extraction techniques that operates on the DEM data of the Lower Altamaha River Basin are currently under development in GIS system. The systems grid processing capabilities of GIS platform allows the development of these algorithms as preliminary steps of a hydrological model.
2.2.6. Other Data
Apart from the basic data requirements discussed above, hydrologic modeling efforts also require several other auxiliary data types. Most of these data (i.e., county and city boundaries, locations of monitoring stations, etc.) are in vector format and available in the Georgia GIS Data Clearinghouse archives. Such data is readily downloadable if the need be.
2.3. Model Development
Model development activities are initiated in accordance with the findings of the literature review. One of the most important steps taken during the model development stage was the selection of the hydrologic cycle components that will be included in the simulation. Preliminary assessments reveal that the following components are to be included in the model:
Precipitation
Infiltration
Evaporation
Overland flow
Unsaturated zone flow
Channel flow
Currently, temporal, spatial and dimensional analysis on these components are underway, which, in the end, form the core of the proposed modeling system. Physical and mathematical details of these model components will be provided in future reports.
4. REFERENCESFor the second phase of the project, literature review and data collection efforts will continue depending on the needs of the research. However, the main focus of the future project activity will be directed towards understanding the physical and mathematical details of the model components as well as formulating the links between each of these components. These activities will form the core of the entire modeling system.
Future project activities also include platform development under GIS using and coding using Visual Basic, ANSI C and Fortran programming languages. In addition, several specific software packages (i.e., image analysis, data conversion, etc.) will be used to process and modify some of the data.
Abbott, M.B., J.C. Bathurst, J.A. Cunge, P.E. O'Connel, and J. Rasmussen, 1986a. An introduction to European Hydrological System - Systéme Hydrologique Européen, "SHE", 1: History and philosophy of a physically based distributed modeling system. Journal of Hydrology, vol. 87, pp. 45-59
Abbott, M.B., J.C. Bathurst, J.A. Cunge, P.E. O'Connel, and J. Rasmussen, 1986b. An introduction to European Hydrological System - Systéme Hydrologique Européen, "SHE", 2: Structure of a physically based distributed modeling system. Journal of Hydrology, vol. 87, pp. 61-77
Abbott, M.B. and J.C. Refsgaard, 1996. Distributed Hydrological Modeling, Kluwer Academic Publishers.
ASCE, 1999. GIS Modules and Distributed Models of the Watershed. A report by the ASCE Task Committee on GIS Modules and Distributed Models of Watershed, P.A. DeBarray and R.G. Quimpo (Eds.), Reston, VI.
Beven, K.J., M.J. Kirkby, N. Schoffield, and A. Tagg, 1984. Testing a physically-based flood forecasting model (TOPMODEL) for three UK catchments. Journal of Hydrology, vol. 69, pp. 119-143
Beven, K.J., R. Lamb, P. Quinn, R. Romanowicz and J. Freer, 1995. TOPMODEL. In V.P. Singh (Ed.), Computer Models of Watershed Hydrology: 627-668, Water Resources Publications, Highlands Ranch, CO.
DeVantier, B.A. and A.D. Feldman, 1993. Review of GIS Applications in Hydrologic Modeling. Journal of Water resources Planning and Management, vol. 119, no. 2, pp. 246-261
Dutta, D., S. Herath and K. Musiake, 2000. Flood inundation simulation in a river basin using a physically based distributed hydrologic model. Hydrological Processes, vol. 14, pp. 497-519
Eagleman, E.T., 1995. Recent advances in remote sensing in hydrology. Rev.Geophys, vol. 33, suppl.
ENVI, 1999. The Environment for Visualizing Images - User's Guide. Research Systems Inc.
Feldman, A.D., 1995. HEC-1 Flood Hydrograph Package. In V.P. Singh (Ed.), Computer Models of Watershed Hydrology: 119-150, Water Resources Publications, Highlands Ranch, CO.
Fortin, J., R. Turcotte, S. Massicotte, R. Moussa, J. Fitzback and J. Villeneuve, 2001. Distributed Watershed Model Compatible with Remote Sensing and GIS Data, I: Description of Model. Journal of Hydrologic Engineering, vol. 6, no. 2, pp. 91-99
Garbrecht, J. and L.W. Martz, 1993. Network and sub-watershed parameters extracted from digital elevation models: The Bills Creek experience. Water Resources Research, vol. 29, no. 6, pp. 909-916
Huber, W.C., 1995. EPA Storm Water Management Model - SWMM. In V.P. Singh (Ed.), Computer Models of Watershed Hydrology: 783-808, Water Resources Publications, Highlands Ranch, CO.
Jensen, J.R., 1996. Introductory Digital Image Processing: A Remote Sensing Perspective, Prentice-Hall Inc.
Jenson, S.K. and J.O. Dominigue, 1988. Extracting topographic structure from digital elevation data for geographic information system analysis. Photogrammetric Engineering and Remote Sensing, vol. 54, no. 11, pp. 1593-1600
Jenson, S.K., 1991. Applications of hydrologic information automatically extracted from digital elevation models. Hydrological Processes, vol. 5, no. 1, pp. 31-44
Johnson, D.L. and A.C. Miller, 1997. A spatially distributed hydrologic model utilizing raster data structures. Computers and Geosciences, vol. 23, no. 3, pp. 267-272
Kite, G.W., 1995. The SLURP Model. In V.P. Singh (Ed.), Computer Models of Watershed Hydrology: 521-562, Water Resources Publications, Highlands Ranch, CO.
Mark, D.M., 1984. Automatic detection of drainage networks from digital elevation models. Cartographica, vol. 21, no. 2/3, pp. 168-178
O'Callaghan, J.F. and D.M. Mark, 1984. The extraction of drainage networks from digital elevation data. Computer Vision, Graphics and Image Processing, vol. 28, pp. 323-344
Polarski, M., 1997. Distributed rainfall-runoff model incorporating channel extension and gridded digital maps. Hydrological Processes, vol. 11, pp. 1-11
Quinn, P., K. Beven, P. Chevallier and O. Planchon, 1991. The prediction of hillslope flow paths for distributed hydrological modeling using digital terrain models. Hydrological Processes, vol. 5, no. 1, pp. 59-79
Refsgaard, J.C. and B. Storm, 1995. MIKE SHE. In V.P. Singh (Ed.), Computer Models of Watershed Hydrology: 809-846, Water Resources Publications, Highlands Ranch, CO.
Sharma, K.D., M. Menenti, J. Huygen, and P.C. Fernandez, 1996. Distributed rainfall-runoff modeling in an arid region using thematic mapper data and a geographical information system. Hydrological Processes, vol. 10, pp. 1229-1242
Shultz, G.A., 1988. Remote Sensing in Hydrology. Journal of Hydrology, vol. 100, pp. 239-265
Singh, V.P. (Ed.), 1995. Computer Models of Watershed Hydrology. Water Resources Publications, Highlands Ranch, CO.
Singh, V.P. and M. Fiorentino (Eds.), 1996. Geographical Information Systems in Hydrology. Kluwer Academic Publishers, Dordrecht, The Netherlands.
Tao, T. and N. Kouwen, 1989. Remote sensing and fully distributed modeling for flood forecasting. Journal of Water Resources Planning and Management, vol. 115, no. 6, pp. 809-823
Tarboton, D.G., R.L. Bras, and I. Rodriguez-Iturbe, 1991. On the extraction of channel networks from digital elevation data. Hydrological Processes, vol. 5, pp. 81-100
Wang, M. and A.T. Hjelmfelt, 1998. DEM based overland flow routing model. Journal of Hydrologic Engineering, vol. 3, no. 1, pp. 1-8
Yao, H. and A. Terakawa, 1999. Distributed hydrological modeling for Fuji River Basin. Journal of Hydrologic Engineering, vol. 4, no. 2, pp. 108-116