Predicting extreme rainfall over eastern Asia by using complex networks

A climate network of extreme rainfall over eastern Asia is constructed for the period of 1971–2000, employing the tools of complex networks and a measure of nonlinear correlation called event synchronization(ES). Using this network,we predict the extreme rainfall for several cases without delay and with n-day delay(1 ≤ n ≤ 10). The prediction accuracy can reach 58% without delay, 21% with 1-day delay, and 12% with n-day delay(2 ≤ n ≤ 10). The results reveal that the prediction accuracy is low in years of a weak east Asia summer monsoon(EASM) or 1 year later and high in years of a strong EASM or 1 year later. Furthermore, the prediction accuracy is higher due to the many more links that represent correlations between different grid points and a higher extreme rainfall rate during strong EASM years.