However, uncertainties were relatively high during low flow seasons, which can be seen as a model deficiency in simulating groundwater flow ( Rostamian et al., 2008). The model performance metric values in Table 3, and P-factor, and R-factor indicate the model is reliable in simulating Brahmaputra basin streamflow. Graphical comparisons of observed and simulated streamflow at a monthly scale for calibration (1988–1997), validation (1998–2004), and baseline (1988–2004) periods are shown in Fig. 3. In general, the model accurately tracked the observed streamflow for the time periods, although some selleck inhibitor peak flow months were underpredicted during calibration, but the under-prediction was less during validation,

possibly due to less temporal variability in the precipitation. Monthly flow statistics in Table 3 suggest a strong correlation between simulated Selleckchem PLX3397 and observed streamflow in all three periods. The NS coefficients for simulated streamflows were 0.85, 0.88, and 0.73 for the calibration, validation, and baseline periods, respectively. These coefficients suggest that model performance for monthly streamflow was relatively better than daily. The model underpredicted streamflows for the calibration and validation periods by 3.2% and 4.4%, respectively. The regression lines and sum difference plots reveal that the underprediction occurred primarily during higher flows (Fig. 3b, c, e, and f). Literature suggests that SWAT

is not designed to simulate extreme events and the model usually underpredicts the largest flow events (Chu Aurora Kinase and Shirmohammadi, 2004 and Tolson and Shoemaker, 2004). However, a positive bias for simulated streamflow of 2.9% was noticeable for the baseline. The notable 1999 overprediction of peak flow may have contributed to this positive bias in simulated streamflow. Overall, the SWAT model was able to simulate well the actual hydrological conditions in the Brahmaputra basin. Ten sensitive parameters were used to calibrate the model (Table 1). These parameters primarily represented surface runoff, groundwater, snow, ET, and the routing process for the basin’s hydrology.

The values for the following parameters were found to be commonly used in other studies to calibrate the SWAT model: CN2 (SCS runoff curve number for moisture condition II), ESCO (soil evaporation compensation factor), ALPHA_BF (baseflow alpha factor), SLSUBBSN (average slope length), GWQMN (threshold depth of water in the shallow aquifer required for return flow to occur), and GW_REVAP (ground revap coefficient) (Cibin et al., 2010, Ghaffari et al., 2010, Heuvelmans et al., 1999, Mutenyo et al., 2013 and Wu et al., 2012a). While the final fitted values were optimized by the automatic calibration algorithm SUFI2, the values were checked for correspondence to the basin characteristics and their underlying hydrological processes. The average CN2 value was 61. The baseflow alpha factor value of 0.