A period of 1970–2010 is used to model the hydrology, from which the first 2 years are left for the model warm-up. The SWAT model is calibrated (1972–1997) and validated (1998–2010) for all study basins. Qualitative and quantitative statistical methods are utilized for model performance evaluation. Calibration and validation results reveal a good fit between the observed and simulated daily discharges (Table 2). Statistical criteria NSE values range from 0.71 to 0.77. The best fit between the observed and simulated flows are found in the Keila basin (Figure 8), where NSE values are 0.77 and 0.72 for calibration and validation, respectively.
Table 2
SWAT model calibration and validation results
| Keila | Valgejõe | Kunda | |
|---|---|---|---|
| R2 | 0.78 | 0.75 | 0.77 |
| NSE | 0.77 | 0.74 | 0.73 |
| PBIAS | 0.5% | −1.4% | −3.2% |
| R2 | 0.73 | 0.72 | 0.75 |
| NSE | 0.72 | 0.71 | 0.72 |
| PBIAS | 1.9% | −3.3% | −8.5% |
| Keila | Valgejõe | Kunda | |
|---|---|---|---|
| R2 | 0.78 | 0.75 | 0.77 |
| NSE | 0.77 | 0.74 | 0.73 |
| PBIAS | 0.5% | −1.4% | −3.2% |
| R2 | 0.73 | 0.72 | 0.75 |
| NSE | 0.72 | 0.71 | 0.72 |
| PBIAS | 1.9% | −3.3% | −8.5% |
Figure 8
Observed and simulated mean monthly discharge in Keila River for the calibration (1972–1997) and validation (1998–2010) period.