Narrow Results

This study determines the suitable general circulation models (GCMs) for the prediction of future precipitation of Upper Godavari sub-basin, India. Five performance indicators (PIs) namely correlation coefficient (CC), normalized root mean square deviation (NRMSD), absolute normalized mean biased deviation (ANMBD), skill score (SS), Nash Sutcliffe efficiency (NSE), and three different combinations (Case 1: all performance indicators, Case 2: CC, SS and ANMBD, and Case 3: CC, SS, and NRMSD) were considered to evaluate the performance of 38 GCM models for the study area. The observed precipitation data for 12 grid points covering the Upper Godavari sub-basin along with eight districts of Maharashtra were used for the selection of the suitable GCMs. The weights of the indicators were determined by the entropy method. Compromise programming (CP) and the technique for order preference to the similarity to ideal solution (TOPSIS) methods were used for ranking the GCMs. The group decision-making approach was employed to make a collective decision about the rank of 38 GCMS considering all the grid points. In view of all the three combinations of PIs, the study suggests that the effect of the performance indicator NSE on the ranking of GCM models is the most significant (weights for the grid points varying in the range 22.75%–78%) followed by ANMBD, CC, NRMSD, and SS. Including the maximum number of PIs and considering their combinations is found to be much helpful to enhance the credibility of the ranking of GCMs. From the group decision-making approach, it was observed that the ensemble of MPI-ESM-P, CNRM-CM5-2, and CNRM-CM5 is suitable for the prediction of precipitation for the study area.