区域气候数值模式预测误差的非线性预估及订正应用试验

THE BIAS CORRECTION METHODS BASED ON THE PREDICTIONS OF NON-LINEAR TIME SERIES AND THEIR APPLICATIONS

本文基于区域气候模式(RegCM4.3)对我国东南部地区进行降水距平预测,并对预测结果进行误差订正.通过构建以经验正交函数分解、集合经验模态分解和最小二乘支持向量机为基础的非线性误差预测订正模型,对我国东南部地区的模式降水预测误差进行预估,并将误差预估结果订正到模式预测输出结果中.试验结果表明,基于误差场的非线性预估模型可以较好地订正我国东南部地区20个侯和3个月以内的模式预测结果.对比未经订正的结果,订正后的距平相关系数有了较为明显的改善,区域标准化均方根误差有了一定幅度的降低,预估订正模型的效果较为理想.

Based on the regional climate modal RegCM4.3, the prediction of the precipitation in South of China was obtained, and the modification was applied on the results. Based on the Empirical Orthogonal Decomposition （EOF）, Ensemble Empirical Mode Decomposition （EEMD） method and Least Square Support Vector Machine （LS-SVM）, a new predictionmethod was established for the non-linear bias filed. Then based on that method, the bias filed in south of china was predicted and the model＇ s results was corrected. Using the Temporal Anomaly Correlation Coefficients （ACC）, the Standard Root Mean Error （SRME） to evaluate the results obtained by the new method, which shows that it is much well in 20 pentads and 3 months, the ACC and SRME is much well improved. It indicates that it much well improve the effect of predictions of RegCM4.3 based on the prediction of non-linear time series.