Prediction of monthly-seasonal precipitation using coupled SVD patterns between soil moisture and subsequent precipitation

It was suggested in a recent statistical correlation analysis that predictability of monthly-seasonal precipitation could be improved by using coupled singular value decomposition (SVD) patterns between soil moisture and precipitation instead of their values at individual locations. This study provides predictive evidence for this suggestion by comparing skills of two statistical prediction models based on the coupled SVD patterns and local relationships. The data used for model development and validation are obtained from a simulation over East Asia with a regional climate model. The results show a much improved skill with the prediction model using the coupled SVD patterns. The seasonal prediction skill is higher than the monthly one. The most remarkable contribution of soil moisture to the prediction skill is found in warm seasons, opposite to that of sea surface temperature.