The five prescribed evaluation criteria in the section ‘Evaluation criteria for calculated outflows' are calculated according to the estimated outflows and shown in Table 3. In general, the VEP-NLMM-L produced the most accurate routing outflows in terms of all criteria and the forecasting accuracy increased with the number of inflow dividing levels. All the models except the LMM-L estimate the peak outflow on the correct time interval (DPOT = 0). The VEP-NLMM-L with L = 4 estimates the closest peak outflow to the observed value among all models.
Evaluation criteria of the estimated outflows for the data set of Wilson (1974)
| Model . | DPO . | DPOT . | η . | MARE . | MAE . | |
|---|---|---|---|---|---|---|
| LMM-L (O'Donnell 1985) | 5.80 | −1 | 93.33 | 0.137 | 4.918 | |
| NLMM2 (Niazkar & Afzali 2014) | 0.70 | 0 | 99.70 | 0.028 | 0.994 | |
| VEP-NLMM-L (this study) | L = 1 | 0.20 | 0 | 99.91 | 0.017 | 0.593 |
| L = 2 | 0.31 | 0 | 99.94 | 0.014 | 0.481 | |
| L = 3 | 0.10 | 0 | 99.95 | 0.012 | 0.390 | |
| L = 4 | 0.03 | 0 | 99.96 | 0.012 | 0.354 | |
| L = 5 | 0.26 | 0 | 99.96 | 0.010 | 0.334 | |
| Model . | DPO . | DPOT . | η . | MARE . | MAE . | |
|---|---|---|---|---|---|---|
| LMM-L (O'Donnell 1985) | 5.80 | −1 | 93.33 | 0.137 | 4.918 | |
| NLMM2 (Niazkar & Afzali 2014) | 0.70 | 0 | 99.70 | 0.028 | 0.994 | |
| VEP-NLMM-L (this study) | L = 1 | 0.20 | 0 | 99.91 | 0.017 | 0.593 |
| L = 2 | 0.31 | 0 | 99.94 | 0.014 | 0.481 | |
| L = 3 | 0.10 | 0 | 99.95 | 0.012 | 0.390 | |
| L = 4 | 0.03 | 0 | 99.96 | 0.012 | 0.354 | |
| L = 5 | 0.26 | 0 | 99.96 | 0.010 | 0.334 | |