Gain of one-month lead time in seasonal prediction of Indian summer monsoon prediction: comparison of initialization strategies

Reasonable seasonal prediction skill for the Indian summer monsoon rainfall has been achieved using the Monsoon Mission (MM) Seasonal Forecast model, at a lead time of 3 months. The ensembles in the MM model are generated by utilizing lagged initial conditions. The possibility of enhancing the lead time is explored by using the burst ensemble approach. Comprehensive seasonal hindcast experiments carried out in this study reveal that the two methods exhibit similar skill scores for the major tropical phenomenon which govern ISMR variability. In general, the model forecasts are slightly under-dispersive but satisfactorily represent the spread-error relationship for major tropical oceanic climate modes. The ratio between the spread and RMSE is small for ISMR forecasts. Though the skill scores for the majority of indices are similar, the monsoon teleconnections seem to be quite sensitive to the initialization strategy. It is found that the burst initialization method provides a gain of 1-month lead time compared to lagged initialization strategy employed in previous studies without compromising the prediction skill. The gain of a months' lead time with the burst ensemble approach is a tempting and useful proposition, which can be crucial for the policy- and decision-makers.

Automated summary

Reasonable seasonal prediction skill for the Indian summer monsoon rainfall has been achieved using the burst ensemble approach to enhance lead time. The model forecasts, although slightly under-dispersive, satisfactorily represent the spread-error relationship for major tropical oceanic climate modes. Monsoon teleconnections are found to be sensitive to the initialization strategy, with the burst initialization method providing a gain of 1-month lead time without compromising prediction skill.