New statistical prediction scheme for monthly precipitation variability in the rainy season over northeastern China

This study focused on seasonal prediction for monthly precipitation over Northeast China (NEC) in the rainy season. A statistical method combining empirical orthogonal function (EOF) decomposition and multi-linear regression was developed and tested. For each EOF mode of each month in summer, the relationship between the EOF and SSTs in the previous winter was investigated and indices were constructed to be used as predictors. The predictors were required to be physically connected to the predictand and to perform well in cross validation. Monthly rainfall was reconstructed from the predicted time series and the observed spatial load of the first three EOF modes. The results of the leave-one-out cross-validation for 1982-2010, and of the independent validation for 2011-2018, indicate that this new method provides good predictions of monthly precipitation over NEC. There was good agreement between the observed and predicted monthly precipitation, with correlation coefficients of 0.45, 0.40 and 0.55, and hit rates of 73, 62 and 73% for June, July and August precipitation, respectively, for the period 1982 to 2018.

Automated summary

This study developed a new statistical method combining EOF decomposition and regression to predict monthly precipitation over NEC in the rainy season. The results showed good agreement between predicted and observed values, with high correlation coefficients and hit rates, demonstrating the model's high predictive accuracy.