Summary: | 碩士 === 國立中興大學 === 水土保持學系所 === 96 === AnnAGNPS model was used to simulate runoff and sediment yields in the Pei-shih river watershed, and the suitability of the model was examined in this study. The observed runoff and sediment yield for the year 1997 and 1998 at the watershed outlet are collected to calibrate values of parameters of the model. Records suitability of the year 2002, 2004 and 2005 were taken to verify the model by coefficient efficiency(CE)and estimation difference(ED).
Performance of the model for the verify years is 0.36, -1.82 and -0.85 in CE, and is 52.54%, 23.42% and 39.65% in ED under the simulation of runoff estimation. CE values showed that monthly runoff estimation had an insignificant relationship between measured and simulated. ED values indicated that the absolute error for the annual runoff estimation is 52.54%. Due to the model ignores the status of subsurface runoff; the daily measured discharge can be neglected as the daily precipitation is zero. Both of CE and ED are well improvement, and the value will be 0.81, -1.79, -0.30, and 21.83%, 22.91%, 17.74% respectively. It means that subsurface runoff can obviously affect the efficiency of simulation. Measured sedimentation derived from daily discharge (zero daily precipitation deducted), the simulation of annual sediment yield in CE is 0.90, 0.39, 0.51, and is 37.09%, 12.26%, 35.84% in ED. There is a good relationship between measured and simulated data, and the error percentage of annual estimation for the sediment yield is less than 40%.
The errors encountered in the model can be categorized as precipitation pattern, hysteresis effect, and base flow due to less consideration of subsurface runoff in the watershed. Calibrating the model with short term data can amplify base flow induced errors, how to analyze the effects of hydrologic hysteresis and base flow can effectively help promote the accuracy of simulation.
|