Remotely sensed data, for both model input and validation, has motivated new modelling approaches that use this information, has stimulated development of methods for use of these data, and has facilitated characterization of landscape heterogeneity. For example, the large-scale hydrological experiments conducted in the southern plains region of the United States have provided extensive remotely sensed and ground truth data for this region (for more information, see Kustas, this issue). Nicks and Scheibe (1992) modelled runoff from the Little Washita River watershed in southern Oklahoma, located within range of several of the first WSR-88D installations; they used the SWRRB (Simulator for Water Resources in Rural Basins) model with NEXRAD radar information providing accurate rainfall data for model input.
Duchon et al. (1992) also implemented the data available for the Little Washita basin within the SWRRB model in their analysis of water budgets, updating model parameters with remotely sensed data. They determined daily insolation from GOES images, and estimated temperatures from the satellite data. To simplify model runs, they divided the landscape into homogenous land cover types on the basis of Landsat MSS imagery. Although comparison of ground truth data to satellite insolation and temperature data revealed some systematic errors, Duchon et al. (1992) concluded that these can be taken into account through calibration and that reasonable water budget estimates can be obtained.
In the Canadian Rockies, Kite and Kouwen (1992) used a land cover classification similar to that presented by Duchon et al. (1992) to improve simulations from the Simple Lumped Reservoir Parametric (SLURP) model. Kite and Kouwen (1992) showed that hydrograph components of the basin were reproduced more accurately (with improved calibration and verification statistics) using a modelling approach that subdivided the area on the basis of Landsat-derived land cover classes than with the lumped version of this model.
Elder et al. (1991) used information on slope, elevation, and net radiation to classify the Emerald Lake basin, located in the Sierra Nevada, into zones that affect snow distribution. From these classes they inferred the spatial distribution of SWE. Variability in net radiation was shown to be the major factor controlling SWE. The results point to the utility of data on terrain and radiation for modelling snow cover and snowmelt, but also to the need for additional information that can account for the redistribution of snow. Marks and Dozier (1992) comprehensively analyzed climate, energy exchange, and snowmelt at the Emerald Lake basin. Their study revealed that net radiation was the dominant factor providing energy for snowmelt by a factor of 5 to 10 over all other forms of heat transfer combined. Marks and Dozier (1992) point out that this result is favorable for catchment modelling in such an environment, because net radiation can be monitored with ease and the spatial distribution of radiation can be modelled.
Distributed radiation data are required as input for many watershed models. Incoming solar radiation for each cell in a grid defined over a catchment can be computed taking into account the slope, aspect, vegetation, and elevation of each cell. Dubayah (1992) linked a radiation transfer algorithm with both satellite reflectance and digital terrain data to model the spatial variability in net solar radiation for FIFE (First ISLSCP Field Experiment, where ISLSCP is an initialism for International Satellite Land Surface Climatology Project). In this prairie environment, Dubayah (1992) concluded that topographic variability was the dominant factor affecting the variability of net incoming radiation. Blöschl et al. (1991) consider the topographic effects on radiation distribution and use this information to compute spatio-temporal patterns of snow accumulation and melt in an alpine basin. The simulated snow cover patterns compared favorably with distributed data obtained from aerial photos ( Blöschl et al. 1991).
The Snowmelt Runoff Model (SRM) requires input of remotely sensed areal extent of snow cover data (Martinec et al. 1983). Although versions of this model have been used successfully for decades, improved data availability and model refinements have made this simple modelling approach more and more useful. Rango (1992b) reports on application and testing of this modelling approach for 50 basins worldwide and discusses the potential for use of SRM to address effects of climate change scenarios. Kustas et al. (1994) altered the approach used to compute melt in SRM, combining a radiation component with the traditional degree-day approach.
Over the past four years the popularity and development of procedures of coupling remotely sensed data with GIS procedures and digital terrain analysis have increased. For example, the U.S. Geological Survey's Precipitation-Runoff Modeling System (PRMS) makes use of automated methods to derive required model parameters. Hydrological response units (HRU's) are defined by partitioning the basin into hydrologically similar categories using terrain analysis routines ( Leavesley and Stannard 1990). A GIS system is then used to compute the necessary model parameters within each HRU and create a data input file for PRMS. This data-parameter interface is beneficial because it imposes consistency on the method of model parameterization and expedites calibration to a series of basins ( Battaglin et al. 1993). Battaglin et al. (1993) use this method to parameterize and calibrate sub-basins within the Gunnison River basin in Colorado. Vieux (1991) also discusses aspects of the use of GIS in watershed modelling.