Extended Range Forecast Experiment for Rainfall Based on the Real-time Intraseasonal Oscillation

1 State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing 1000812 Beijing Meteorological Observatory, Beijing 100089Intraseasonal oscillation (ISO) in South China Sea summer monsoon (SCSSM) plays a key role in controlling the intraseasonal variations of rainfalls over southern China, and it can be described with the leading pair of empirical orthogonal functions (EOFs) for the 110°-120°E averaged 850 hPa zonal wind (U850). An index for monitoring the SCSSM ISO is built on a pair of principal component (PC) time series of EOFs mentioned above, and then the NCEP Climate Forecast System Version 2 (NCEP/CFSv2) hindcasts and stepwise regression statistical method are employed, to explore extended range forecast (ERF) of rainfall intraseasonal variations.First, southern China is divided into three regions using rotated empirical orthogonal functions (REOFs), where the incidence rate of regionally persistent heavy rainfall (RPHR) is closely linked to the intraseasonal variation in rainfall. Based on the spatial structure of the first three REOFs, three intraseasonal rainfall indices are constructed by averaging the 30-60-day filtered precipitation over the typical regions and taken as predictands. Second, EOF1 of the 850 hPa zonal wind over the SCS and southern China mainly represent the ISO mode controlling the intraseasonal rainfall south of the Yangtze River, while EOF2 leads to the intraseasonal out-of-phase rainfall over South China and the Yangtze-Huai River Basins. Projection of the daily data onto the leading pair EOFs of 850 hPa zonal wind yields PC time series that serves as an effective filter for ISO without the need for bandpass filtering and making the PC time series two effective indices for real-time use. The pair of PC time series that form indices are called real-time indices for SCSSM ISO. Finally, 28 years of NCEP/CFSv2 reforecasts are used which include wind at 850 hPa and grid values of rainfall extending up to 30-day lead time. Characteristics of SCSSM ISO are also found similar to observations in the NCEP/CFSv2 reforecasts. Forecast models are built on the historical reforecast values of indices (predictors) and rainfall time series (predictands), and use forecast values of indices to predict the future values of rainfall time series varying mostly on the intraseasonal time scale. This method can significantly improve the ERF results (10-30 days) of intraseasonal variations of rainfalls in southern China, reduce system errors, avoid losses of forecasting data and correct negative correlations between forecasts and observations caused by bandpass filtering.