干旱区生态环境知识资源中心

Hydrologic fluxes in mountainous watersheds are particularly important as these areas often provide a significant source of freshwater for more arid surrounding lowlands. The state of Colorado in the United States comprises a principal snow catchment area, with all major headwater river basins in Colorado providing substantial water flows to surrounding western and midwestern states. The ability to represent and quantify hydrologic processes controlling the generation and movement of water in headwater basins of Colorado therefore has significant implications for effective management of water resources in the western United States under varying climatic and land-use conditions. In the research reported in this paper, hydrologic modeling was applied to four snow-dominated, mountainous basins of Colorado [i.e.,the river basins of (1)Cache la Poudre, (2)Gunnison, (3)San Juan, and (4)Yampa] to evaluate the relevance of specific hydrologic components (i.e.,evapotranspiration, snow processes, groundwater processes, surface runoff, and so on) in the complex, high-elevation watersheds. The soil and water assessment tool (SWAT) model was calibrated and tested for multiple river locations within each basin using monthly naturalized flows over the 1990-2005 period. The model was able to adequately simulate streamflows at all locations within the four basins. Monthly patterns of precipitation, snowfall, evapotranspiration (ET), and total water yield were similar for all the basins, while subsurface lateral flow was the dominant hydrologic pathway, contributing between 64 and 82% to gross basin water yields on an average annual basis. Overall, results indicated the strong influence of snowmelt and groundwater processes on amounts and timing of streamflows in the study basins. Hence, enhanced representation of these processes may be essential to improve hydrological estimation using computer software in snowmelt-driven mountainous basins. In particular, examination of monthly streamflow residuals indicated that the normality and independence of model residuals, which are often assumed in parameter estimation and uncertainty analysis, were not always satisfied.