The Statistical Downscaling Method of Summer Rainfall Prediction over the Huang-Huai Valley

The statistical downscaling method and predictability of summer rainfall anomaly over the Huang-Huai Valley (SRAHV) is studied based on station precipitation data, NCEP/NCAR reanalysis data and BCC_CSM1.1m hindcasts from 1991 to 2011.Firstly, correlation coefficients between SRAHV and seasonal circulations in troposphere are calculated. In the high troposphere, significant circulation patterns are the South Asia high, the westerly over Eurasia, 200 hPa zonal wind over the southern of South China Sea and the Philippines. In the middle level, significant predictors are blocking high over Ural and west Pacific subtropical high. In the low level, southern anomaly wind over South China is the key factor. These predictors show clearly positive relationship with SRAHV and may lead to more rainfall.Secondly, the performance of BCC_CSM1.1m is diagnosed on the basis of summer hindcast circulations. Skills of 200 hPa and 500 hPa potential heights, 200 hPa zonal wind, 850 hPa meridional wind by BCC_CSM1.1m are relatively high in some key regions which may affect the SRAHV in reasonable physical mechanism. Six key factors are selected based on the consistent anomaly ratio of factors between BCC_CSM1.1m and reanalysis data, as well as the ratio between SRAHV and predictors from reanalysis data. The optimal sub-tree regression (OSR) is used as transfer function in the statistical downscaling model. Six predictors are tested by one-year-out cross validation sample tests. The consistent ratio between observation of SRAHV and prediction is 61%. By deleting dependent factors, three independent predictors (200 hPa potential height over the Ural, 200 hPa potential height over the South Asia high region to South China, 200 hPa zonal wind over South China Sea to South Philippines) are used to make the statistical downscaling model again, and the accuracy is improved to 72%.Further studies show that the predictability of statistical downscaling model comes from the skill of three key predictors by BCC_CSM1.1m, representing the strength of blocking activity over the Ural, the strength and position of the South Asia high, and the strength of west anomaly wind over the west tropical Pacific. When model output show high skill on three factors, skills of downscaling model are also high and predictions of SRAHV are close to observations in the years of 1994, 1995, 1998, 2004 and 2010. In the years of 1991, 1996 and 1997, BCC_CSM1.1m performs poorly especially on west anomaly wind over the west tropical Pacific. The correlation coefficient of west anomaly wind over the west tropical Pacific and SRAHV is 0.55 which passing the test of 0.01 level, indicating BCC_CSM1.1m's important role in the statistical downscaling model, which determines the prediction skill of SRAHV.