Table 5 presents the total SI and its cumulative sum for all 23 parameters. If contributing 90% of the total output variance is used as a threshold to identify the important parameters, ten parameters have to be kept for model optimization. Comparing these ten parameters (the last row of Table 5) from the Sobol’ method with the ten important parameters from the MOGA (the last row of Table 4), it can be found that seven parameters are identified to be commonly important by both methods, even though the importance order is not exactly the same. Especially, eight parameters are identified to be commonly important for the same watershed and metric (row 2 of Table 4 versus the last row of Table 5).