香港降水的短期气候预测

利用正规化周期回归分析方法和门限自回归理论对香港1853-1995年月降水资料进行了建模和拟合，并对1996年进行了预报试验，正规化周期回归拟合效果较好，对香港7个时距的试报准确率为71%，建立了7个时距的10年滑动平均序列的门限自回归模型，模型对序列拟合效果较理想，预报误差较小，试报准确率主国100%。

SHORT-RANGE CLIMATE PREDICTION OF THE PRECIPITATION IN HONGKONG

Based on the monthly precipitation data of Hong Kong for 1853-1995, statistical models are developed by using the normalized periodic regression and threshold auto regression analyses. The normalized periodic regression model can reproduce the precipitation with a variance which makes up above 80 percent of the total variance except in former rainy season, while the periodic complex correlation coefficients reach up to 0 9. The accuracy of the prediction tests for seven seasons in 1996 is 71 percent with errors less than one and half times of the residual standard deviation. The threshold auto regression models for predicting 10 year moving average precipitation of the seven seasons are developed. The predicting residuals of the models are small, and the accuracy of the prediction tests for seven seasons in 1996 is 100 percent with errors less than one and half times of the residual standard deviation.