Extended-range forecast of spring rainfall in southern China based on the Madden–Julian Oscillation

Abstract

Spring (March–May) rainfall after a dry period in winter has a substantial impact on agriculture and water management in populous southern China. The occurrence of low-frequency spring rainfall anomalies has been linked with the tropical Madden–Julian Oscillation (MJO) through its modulation of large-scale circulation and moisture supply over southern China. Using the spatial and temporal information of the MJO as a predictor, an empirical model for extended-range forecasting of spring rainfall in southern China was constructed. We first obtained the coupled patterns between the preceding MJO evolutions (real-time multivariate MJO index) and the succeeding rainfall variability in southern China based on singular value decomposition analysis. Then, a prediction was carried out by projecting the predictor onto the spatiotemporal coupled patterns. Useful skill, in terms of the temporal correlation coefficient (TCC) between the predicted and observed rainfall over southern China, persisted up to a forecast lead-time of six pentads. The forecast amplitude bias in terms of root-mean-square error was around 1.0 standard deviation. Also, the forecast skill was highly dependent on the strength of the MJO signal. During active MJO periods, the TCC skill was around twofold larger than that during weak MJO periods. The current statistical model shows encouraging ability, but additional work is required to improve its forecasting skill.