Precipitation Data from CMIP5-ESMs-RCPs Experiment: in Weyib River Basin, Southeastern Ethiopia
Chapter 4: Evaluation of the ArcSWAT Model in Simulating Catchment Hydrology: in Weyib River Basin, Southeastern Ethiopia
B. Model Validation
streamflow with R2, NSE, RSR and Pbias values of 0.86, 0.83, 0.25 and 1.72 respectively for calibration period (Fig.4.3 and 4.4).
Figure 4.3 Hydrograph of the observed and simulated streamflow using calibrated parameters for the calibration period (1984-1994)
Figure 4.4 Regression line fit between observed and simulated streamflow during calibration period An intensive hydrologic calibration resulted in good SWAT predictive efficiency at the daily time step of the basin when compared to measured flow data. The hydrograph and regression line between observed and simulated flow indicated that the ArcSWAT model is capable of simulating the hydrology of Weyib River basin as shown in Fig.4.3 and 4.4.
model has the high predictive capability with R2, NSE, RSR and Pbias values of 0.84, 81, 0.31 and 2.69 respectively (Fig 4.5 and 4.6). Though rigorous calibration has undertaken for daily streamflow, simulated flow over predicts peak flow but under predicts all other time for both calibration and validation periods. The shape of the hydrograph of simulated flow was the same as the shape of hydrograph of measured daily streamflow (Fig.4.3 and 4.5).
Figure 4.5 Hydrograph of the observed and simulated streamflow using calibrated parameters for the validation period (1995-2004)
Figure 4. 6 Regression line fit between observed and simulated streamflow during validation period Streamflow Simulations are considered satisfactory if R2≥0.6, NSE˃0.5, RSR≤0.7 and Pbias are within ±25% (Moriasi et al. 2007). According to these criteria, Tables 4.1 and 4.6 indicate, for all evaluation criteria given, very satisfactory results for both calibration and
0 20 40 60 80 100 120 140 160 180 200 0
10 20 30 40 50 60
Jan-95 May-95 Sep-95 Jan-96 May-96 Sep-96 Jan-97 May-97 Sep-97 Jan-98 May-98 Sep-98 Jan-99 May-99 Sep-99 Jan-00 May-00 Sep-00 Jan-01 May-01 Sep-01 Jan-02 May-02 Sep-02 Jan-03 May-03 Sep-03 Jan-04 May-04 Sep-04 Total monthly rainfall (mm)
Average monthly streamflow (cms)
Time (month) Validation
Rainfall Observed Simulated
y = 1.0425x R² = 0.84 NSE=0.81 RSR=0.31 Pbias=2.69
0 10 20 30 40 50 60
0 10 20 30 40 50 60
Simulated streamflow (cms)
Observed streamflow (cms) Validation (N=120)
verification periods have been obtained. Therefore since the model performed as well in the validation period, as for the calibration period hence, the set of optimized parameters listed in Table 4.5 during the calibration process of Weyib River basin can be taken as the representative set of parameters for the similar basin. Altogether, when comparing the model’s performance against the model evaluation criteria based on the guidelines (Moriasi et al., 2007) presented in Table 4.1 for the monthly time step, SWAT simulated streamflow very well in all the evaluation indices (Table 4.6).
Table 4.6 Summary of model performance ratings for simulations of streamflow Model Performance Statistics Calibration for the Period
of 1984-1994
Validation for the Period of 1995-2004
Coefficient of Determination (R2) 0.86 (very good) 0.84 (very good) Nash-Sutcliffe Coefficient (NSE) 0.83 (very good) 0.81 (very good)
RSR 0.25 (very good) 0.31 (very good)
Percent Bias (Pbias) 1.72 (very good) 2.69 (very good) The Evaluation of the SWAT model in simulating catchment hydrology in case study of the Modder River basin has been investigated (Kusangaya et al., 2014) and they reported that the results of calibration and validation of the model at a monthly time step gave NSE of 0.65, Pbias of 15 and RSR of 0.4, while NSE of 0.5, Pbias of 31 and RSR of 0.5 have recorded for validation. Jha (2011) reported R2 of 0.86 and NSE of 0.85 for calibrated monthly flows, and for validation the following monthly flows statistics have reported as R2 of 0.69 and NSE of 0.61. Srinivasan et al. (2010) reported R2 of 0.75 and NSE of 0.74 for calibrated monthly flows, and for validation the following monthly flows statistics have reported as R2 of 0.58 and NSE of 0.69. Bouraoui et al. ('2005') reported R2 of between 0.62 and 0.84 and NSE of between 0.41 and 0.84 for calibrated monthly flows. Setegn (2010) said R2 of 0.80 and NSE of 0.73 for calibrated while R2 of 0.80 and NSE of 0.71 for validation period. In an interrelated study, Shawul et al (2016) also reported that the SWAT model had a good performance in simulating the monthly, seasonal, and annual mean discharges with the R2, NSE and D values of 0.81, 0.75 and 23 respectively in the calibration period, while R2, NSE and D values of 0.65, 0.59 and 20 respectively in the validation period. In this study, the results of calibration and validation of the model at daily time step gave R2 of 0.86, NSE of 0.83, and RSR of 0.25 and Pbias of 1.72 during the calibration period, while R2 of 0.84, NSE of 0.81, RSR of 0.31 and Pbias of 2.69 during the validation period. There has been observed
a slight variation in statistical values among researchers, in this particular study too; this difference might be due to mainly spatial data used (predominantly LULC data), the disparity insensitive catchment parameters that affect calibration processes, uncertainty during data handling.
4.4. Conclusion
The assessment of SWAT hydrological model and investigation of its ability to simulate reliably the different components of water balance in general and streamflow, in particular, using different efficiency criteria gave an insight into how one can successfully generate useful information in catchments where there is little data available. To do so various efficiency criteria were implemented, namely, R2, NSE, RSR and Pbias with hydrograph technique. The results suggest that the SWAT hydrological model can be a useful tool which, once calibrated effectively, can produce meaningful predictions of catchment hydrology to aid management decisions. The results obtained indicate that basin output simulated by ArcSWAT after calibration is comparatively consistent with recorded values.
This study provided a better understanding of SWAT model set-up, sensitive parameters that influence the model output, and hydrologic processes of the catchment. The most sensitive parameter found in this hydrological simulation exercise was Curve Number (CN) which is dependent on 'land management practice and soil parameters'. These parameters are found to influence hydrologic processes more than others. It is paramount that calibrated model results provide a reasonable reflection of actual hydrologic processes. Statistical evaluation criteria suggested (Moriasi et al., 2007) only provides the guidelines to which to evaluate model’s performance. Moreover, it is also essential to look at other statistical indices that can be used to assess the model's performance. It has inferred that further study is conducted to evaluate uncertainties in the model that affect model performance and the sensitivity of the 'distributed hydrologic' simulations to different calibration schemes under different catchment conditions.