4.3. Sensitivity analyses
The above results indicate the Method 2 predicts runoff with greater accuracy than the other SCS-CN methods. A sensitivity analyses can identify the primary importance parameters that affecting the performance of the model. In this study, sensitivity analyses were conducted by observing the effect of variation of the calibrated parameter on the runoff prediction in term of NSE with the datasets of the plots in XDG watershed.
Figure 5 illustrates the sensitivity of the predictions to the various parameters of Method 2. The sensitive variables are parameters for whichNSE changes dramatically when the parameter is increased or decreased from the calibrated value. Parameter b1is apparently rather sensitive to variation, with the NSEdecreasing from 80.58% to 47.83% when the b1value varies in a narrow range from 130% to 80% of the calibrated value, whereas the initial ratio λ appears to be the least sensitive. In general, the parameters of soil moisture (b1 andb2 ) and storm duration equations (c ) are more sensitive than those parameters of slope equation (a1 and a2 ) and the initial abstraction ratio λ .