Prediction Of East Asia Summer Rainfall:Statistical Model Establishment And Predictability Analysis

East Asian summer monsoon (EASM) is one of the most active climate systems in Asia. The rainfall predictability is still a long-standing challenge which needs to be solved greatly. Base on observational datasets and predictable mode analysis, this paper works on the dynamical characteristics, prediction and predictability of rainfall variability over East Asia (100°E-140°E,5°N-50°N) during peak summer (July-August mean, JA). The prediction skill of coupled models’multi-model ensemble mean is also compared. In order to meet the demand of operational forecast in China, the possibility of long-lead prediction of Chinese summer (June-July-August mean, JJA) rainfall is also discussed in this paper using partial least square regression. Some new results are found as follows:(1) There are two rainfall maxima over EA during JA which is divided by the mean ridge line of the western Pacific subtropical High (WPSH). One is located over tropical EA (SEA,5°N-26.5°N) and the other is in subtropical-extratropical EA (NEA, 26.5°N-50°N). The first four modes of.rainfall variability over SEA and NEA are regarded as predictable modes because they have robust physical processes and can be reconstructed by physical-empirical model (P-E model) and/or dynamical models while the higher modes tend to be noisy and uncertain from observational point of view.The physical meanings of most principle modes of rainfall variability over SEA and NEA show obvious differences except the first mode of SEA and the second mode of NEA.Thus the sources of predictability and precursors that can affect the rainfall over SEA and NEA are quite different. The prediction skill can be improved when predicting the rainfall in the two sub-domains separately rather than dealing with the whole domain.(2) According to the physical process, the physically meaningful predictors are selected for the principal components (PCs) in each mode. Then a set of P-E prediction models is established using stepwise multi-linear regression to predict each PC. The predicted field can be derived from the observed spatial patterns and predicted PCs.The forecast skill can be estimated from the correlation between the observed precipitation anomaly and the reconstructed predicted field. High prediction skill regions are found over several climatological rainfall centers while the dry regions have relatively low skills. However, the domain-averaged prediction skill using P-E models is much higher than that using dynamic models’multi-model ensemble mean. This reveals that the dynamical models have large rooms to improve. Besides, the prediction skill increases with the decreasing forecast lead time. Therefore, it is suggested that by the late June a more reliable forecast of July-August rainfall should be implemented in the operational centers.The observed predictable part is reconstructed by the linear combination of the first four modes. Supposing that these modes can be predicted perfectly, the potentially maximum attainable forecast skill can be estimated from the correlation between the observed precipitation anomaly and the reconstructed predictable part of the rainfall using the four predictable modes. The predictability, prediction skill using P-E models and dynamical models of SEA is higher than that over NEA.(3) The possibility of long-lead prediction of Chinese summer (June-July-August mean, JJA) rainfall is discussed. An independent test is applied to check the forecast ability of the latest 9 years (2005-2013) using EOF-PLS statistical model. The independent prediction test is carried out by utilizing only the predictor field (SST over ocean and 2m air temperature over continent) before the target year. It is demonstrated that most regions of floods and droughts over eastern and southern China can be generally predicted.The above studies investigate the dynamical process of EASM rainfall variability and reveal the precursors that can affect the rainfall anomaly. The prediction skill of summer rainfall over East Asia including China has been significantly improved compared with dynamic models, which has profound significance for promoting the development of EA.