GIS and Hydrologic Models

Watershed database development usually is the first important stage in a hydrologic modeling study. Remotely sensed data might be employed to generate thematic maps and also to serve as map basis when no other reliable data are available. Landsat TM and SPOT images data are suitable for production of digital map at scales ranging from 1 :50,000 to 1 : 100,000 (Welch et al., 1985; Swann et al., 1988; Gugan and Dowman, 1988; Konecny et al., 1988). Base maps, produced from remote sensing and integrated within a GIS, hold promise in terms of greater reliability, i.e., lower meta-uncertainty (uncertainty about uncertainty) for map information because errors are known and tracked throughout the map generation process. Overlaying, merging, and performing map calculations are key GIS features often used in many hydro-logical applications. Schultz (1993) presents an example in which soil water storage information was derived by merging plant root depth data (derived from land-use classification of Landsat image) and soil porosity data (derived from digitized soil maps).

Historically, runoff modeling at the river basin scale has lumped rainfall, infiltration, and other hydraulic parameters to apply everywhere in the basin. With the advent of distributed modeling, a basin is subdivided into computational elements at a smaller scale. A distributed simulation model allows a user to simulate spatially variable parameters without lumping. However, setting up such a model with spatially distributed data and parameters is a time-consuming and laborious task. If a GIS is integrated with the model, these chores become much easier and often transparent to the user. An additional advantage of integrating distributed numerical models with a GIS includes calculation and display of runoff flow depths across watershed sub-basins.

The runoff curve number (CN) approach (USDA, 1972) to rainfall-runoff modeling is appealing for an integrated remote-sensing and GIS environment. This approach estimates volume of direct runoff (Q) in terms of volume of rainfall (P) and potential maximum storage (S), which is derived from the CN, a coefficient that is directly related to watershed land use, land management, and soil properties. Since land use can be routinely monitored using remote sensing, it is possible to analyze the effects of land-use changes (e.g., urbanization) on watershed runoff. Figure 4 shows various stages of computation of this approach implemented within a GIS. Mattikalli et al. (1996) employed Arc/Info to store various input parameters as thematic layers and generated flood hydrographs in a predominantly rural watershed. This approach has also been used to generate single-event flood hydrographs and synthetic flood frequency curves (Muzik and Chang, 1993).

Land-use Data Soils Data

Land-use Data Soils Data

Digitized Images Gis
Figure 4 Schematic diagram of a GIS approach for prediction of river discharge using the SCS curve numbers and water quality using the export coefficient model (Mattikalli et al., 1996).

In urban watersheds, the spatial analysis capabilities of a GIS can be used for hydrological analysis. Watershed attributes such as soils information (infiltration rates, hydraulic conductivity, and storage capacities), surface characteristics (pervious, impervious, slope, roughness), geometry and dimensions of flow planes, routing lengths (overland, gutter, and sewer), and geometry and characteristics of routing segments can be efficiently stored and utilized for urban runoff calculations. Most of the earlier studies have used GISs to derive parameters of lumped models. For example, Johnson (1989) used GIS for the generation of input data for a digital map-based modeling system that supports lumped parameter models such as unit hydrograph, time-area, and cascade of reservoirs. The advent of distributed modeling and powerful GIS allows modelers to simulate spatially variable parameters. To date, several hydrological models such TOPMODEL and CREAMS have been integrated to operate within GIS environments (Chairat and Delleur, 1993; Romanowicz et al., 1993). Moeller (1991) used GIS to determine input parameters for the HEC-1 model, Sircar et al. (1991) used a GIS to determine time-area curves. Djokic and Maidment (1991) used Arc/Info with the rational method to determine inlet and pipe capacity of an urban storm sewer system. Kim and Ventura (1993) used a GIS to manage and manipulate the land-use data for modeling the non-point-source pollution of an urban basin using an empirical urban water quality model. Greene and Cruise (1996) employed Arc/Info GIS to derive urban watershed feature attributes (location coordinates, parameters of runoff generating polygons, gutters and storm drains) for input into hydrologic modeling procedures to estimate runoff. Vieux (1991) developed a method for modeling direct surface runoff using a combination of the finite-element method and GIS. Schultz (1994) presents three different examples on hydrological modeling using remote sensing in the framework of ILWIS and Arc/Info GIS. These examples demonstrate merging of Landsat TM and Meteosat geostationary image products and ancillary data (viz. DEM and its derived products) stored in a GIS for rainfall/runoff modeling and water balance parameter computation at 30 m, 5 km, and at HRU spatial scales. Mattikalli et al. (1996) employed the runoff curve number (CN) approach to compute direct runoff depth and its spatial and temporal variations based on historic remotely sensed data within a GIS framework.

Water quality modeling applications using remote sensing and GIS have concentrated mainly on non-point source (NPS) pollution. To date, several water quality models (AGNPS, ANSWERS, USLE, export coefficient model, etc.) have been interfaced with GIS. The spatially distributed agricultural non-point-source (AGNPS) model integrated with GIS (Srinivasan and Engel, 1994) allows modelers to handle each point source, pesticide, and channel information in a decision support system, WATERSHEDSS (Water, Soil, and Hydro-Environmental Decision Support System) (Osmond et al., 1997). Using such a system, one can determine critical areas within a watershed and evaluate effects of alternative land treatment scenarios on water quality. Mattikalli et al. (1996) implemented an export coefficient model within a vector-based GIS to quantify spatial and temporal changes of total nitrogen loading in surface water as a response to changes in watershed land use, management, and fertilizer application rates. Although this method is based on empirical export coefficients derived from the literature, more accurate coefficients can be derived by inverse solution to a physical based model.

Management and modeling of groundwater and its quality have also been explored (e.g., Maidment, 1993; Merchant, 1994). In the majority of studies, spatial models designed to evaluate groundwater vulnerability for contamination have been implemented in GIS. However, these approaches have not employed data derived from remote sensing, probably because of the specific nature of the input parameters. The models need to be adapted to incorporate remotely sensed products and then implemented within a GIS.

Monitoring and/or prediction soil erosion computed using the universal soil loss equation (USLE) is another application of integrated GIS (e.g., Pelletier, 1985). Slope steepness (S) and slope length (L) factors are derived using DEM, and rainfall factors are assigned using the triangular irregular network (TIN) structure for the rainfall gaging stations. Erosion control practice and land-use/land-cover (or cropping management) factors are estimated using Landsat (Multi-Spectal Scanner (MSS) and TM) and SPOT sensor data via land-use/land-cover classification and associated land management information (Jurgens and Fander, 1993). In the revised USLE (Renard et al., 1991), the L factor has been modified for influence of profile convexity/concavity using segmentation of irregular slopes of a complex terrain. Mitasova et al. (1996) integrated regularized spline with tension for computation of S and L factors and used a unit stream power and directional derivative approach for modeling spatial distribution of areas with topographic potential for erosion or deposition.

Was this article helpful?

0 0

Post a comment