资源环境科技发展态势分析平台

The present study assess the potential predictability of boreal summer (June through September, JJAS) tropical sea surface temperature (SST) and Indian summer monsoon rainfall (ISMR) using high resolution climate forecast system (CFSv2-T382) hindcasts. Potential predictability is computed using relative entropy (RE), which is the combined effect of signal strength and model spread, while the correlation between ensemble mean and observations represents the actual skill. Both actual and potential skills increase as lead time decreases for Nio3 index and equatorial East Indian Ocean (EEIO) SST anomaly and both the skills are close to each other for May IC hindcasts at zero lead. At the same time the actual skill of ISMR and El Nio Modoki index (EMI) are close to potential skill for Feb IC hindcasts (3 month lead). It is interesting to note that, both actual and potential skills are nearly equal, when RE has maximum contribution to individual year's prediction skill and its relationship with absolute error is insignificant or out of phase. The major contribution to potential predictability is from ensemble mean and the role of ensemble spread is limited for Pacific SST and ISMR hindcasts. RE values are able to capture the predictability contribution from both initial SST and simultaneous boundary forcing better than ensemble mean, resulting in higher potential skill compared to actual skill for all ICs. For Feb IC hindcasts at 3 month lead time, initial month SST (Feb SST) has important predictive component for El Nio Modoki and ISMR leading to higher value of actual skill which is close to potential skill. This study points out that even though the simultaneous relationship between ensemble mean ISMR and global SST is similar for all ICs, the predictive component from initial SST anomalies are captured well by Feb IC (3 month lead) hindcasts only. This resulted in better skill of ISMR for Feb IC (3 month lead) hindcasts compared to May IC (0 month lead) hindcasts. Lack of proper contribution from initial SST and teleconnections induces large absolute error for ISMR in May IC hindcasts resulting in very low actual skill. Thus the use of potential predictability skill and actual skill collectively help to understand the fidelity of the model for further improvement by differentiating the role of initial SST and simultaneous boundary forcing to some extent.