Sensitivity analysis across scales and watershed discretization schemes using ARDBSN hydrological model and GIS

Abstract

The goal of this study is to assess ARDBSN model performance across watershed sizes, and to assess the effect of using different mapset resolutions and basin configurations on runoff volume. Arid basin (ARDBSN) (Lane, 1982) is a distributed parameter, continuous hydrologic simulation model designed to simulate the effects of land-use practices on runoff volume, soil erosion and sediment yield on rangeland watersheds. Four subwatersheds located within the Walnut Gulch Experimental Watershed, were used in this study. Subwatershed areas ranged from 1.4 to 630 hectares. A large-scale GIS database (1:5000) developed for the Walnut Gulch Experimental Watershed, and one of the most common mapset scales (1:24 000) used in GIS analysis, were used to parameterize the ARDBSN model. Four thresholds were used to discretize watershed ST-223 and WS-11. These thresholds correspond to 1.5, 2.5, 5, and 10 percent of the watershed area. Sensitivity analysis showed that Curve Number (CN) parameter was the most important in defining runoff volume and peak runoff rate. Model calibration performance was measured primarily by the coefficient of efficiency (R², Nash and Sutcliff, 1970) for total annual runoff volume and for maximum annual peak runoff rate. Runoff volumes model efficiencies obtained were very good for LH-101, LH-103, and WS-11, and reasonably good for ST-223. In general, this model trends to over predict runoff volume for small rainfall events, and under predict for large rainfall events. The number of overland flow elements delineated within the 1:24,000 and 1:5,000 scales scarcely varies. The number of channels diminishes as the threshold values decreases. Map scale has a strong effect on the length of the channel network. Overall, high resolution maps show a better runoff volume model efficiency. Watershed 11 and watershed 223 had an average model efficiency seven and six percent higher than those obtained by these subwatersheds under the 1:24,000 map scale. Considering this, the author concludes that 1:24,000 map scale can be used with high confidence in hydrologic simulation modeling in areas with similar characteristics to those of the Walnut Gulch Experimental Watershed. The results showed that watershed configuration complexity significantly alters the model results on large watersheds, and can be attributed to the improved representation of spatially distributed watershed features with increased geometric complexity.