| Popis: |
Prediction accuracy of physically based hydrologic models largely depends on how well the models and input data represent the real world in terms of precipitation, topography, soils, land use/land cover, and land management. Precipitation is, by far, the largest source of uncertainty; misestimating it by a small percentage leads to inadequate soil-moisture accounting and subsequent inaccuracy in runoff and streamflow estimation. The objectives of this study were (1) to determine the optimum local bias adjustment factor (BAF) threshold limits (TLs) by using SWAT-modeled streamflow values and frequency distributions with NEXRAD or raingauge precipitation inputs, and (2) to assess the impact of BAF TLs derived from 40 raingauges in and around a large watershed in west central Kansas on hydrologic response of the SWAT model. NEXRAD overestimated precipitation depths (45% to 184%) in warm months including April to September and underestimated depths (33% to 64%) in cold months including December to February; the remaining months had fairly close estimates (within ±20%). Assessment of bias-adjusted NEXRAD data based on daily raingauge comparisons found the best model efficiency (Ef) using 0.15 lower TL, whereas the lowest bias was for 0.33 lower TL, both with 2.0 upper TL. However, assessment of NEXRAD data based on predicted versus observed daily streamflow found the best Ef using 0.15 lower TL and the lowest bias using 0.25 lower TL. NEXRAD data bias-adjusted using TLs of 2.0 (upper) combined with 0.15 (lower) produced satisfactory streamflow results, similar to the default nearest-raingauge method used in SWAT. This result is encouraging considering the substantial seasonal bias evident in the Stage III NEXRAD data and it indicates the potential for future improvements in streamflow simulation accuracy as NEXRAD accuracy increases. This study demonstrated that TLs are needed in bias-adjustment of NEXRAD data for reasonable hydrologic simulation, and that better streamflow simulation accuracy is obtained if the optimum TLs are based on streamflow assessment rather than assessment of the precipitation data. |
