Decadal prediction skill for spring and summer surface air-temperature over India and its association with SST patterns in CFSv2 and CNRM coupled models

In this study, we have assessed the skill of decadal prediction of boreal spring (March–May) and summer (June–September) Surface Air Temperature (SAT) over India and its relation with Sea Surface Temperature (SST) in the National Centers for Environmental Prediction (NCEP) Climate Forecast System Version 2 (CFSv2) coupled model. The skill of CFSv2 is compared with CNRM (Centre National de Recherches Meteorologiques) coupled model, which is the best among the selected CMIP5 (Coupled Model Intercomparison Project 5) models for long-lead forecasts (6–9 years) of global SSTs (with high skill). It is found that both models show significant skill in predicting 4-year mean SAT over central and southern peninsular India at 1–5 year leads in spring. During summer, significant skills for SAT over the northwest and southeast India are seen in CFSv2, whereas CNRM displayed significant skills over the north and central India at 1–5 years lead. The first Empirical Orthogonal Function (EOF) mode of SAT variability over India indicates a country-wide warming/cooling pattern in both observations and models for spring and summer. The analysis reveals that the decadal variability of SAT (EOF-1) over India is highly related to SST variations over the Indo-Pacific and North Atlantic regions. The strong convergence of low-level winds over the equatorial Indian Ocean and maritime continent accompanied by warm SST anomalies drive the northerly dry winds over India and favour warm SAT in both spring and summer. Further, changes in shortwave radiation also contributed to SAT variability over India. In general, the SAT relationship with SST in different parts of tropical and sub-tropical regions is underestimated in CFSv2 compared to the observations and CNRM in both boreal spring and summer. The models are able to represent the changes in the atmospheric circulation and related Indo-Western Pacific SST patterns reasonably well at the 1–5 years lead with some discrepancy. However, both models showed relatively low skills in capturing the relationship between SAT over India and equatorial Pacific SSTs. This might limit the skills of models in predicting decadal variations of SAT over India.