一个降水预报的时间序列线性平稳预测模型

本文利用兰州１９４４～１９９７年的月平均降水资料，建立了线性平稳序列的降水预测模型，该模型使用了功谱密度函数中的最大熵法（或叫全极模型），并将特征多项式模大于１的根反射到单位圆内，再返回修正后的线性预测的系数。并对１９８６～１９９７年１１年的月降水做了预测试验，试验结果表明，该模型具有一定的预报能力，其中取１５阶预报效果较好。此方法在短期气候预测业务中，可作为台站月、季、年降水预测走向客观化、定量化方法的一种初步尝试

A Linear stable forecasting model of time series for monthly, seasonal and yearly precipitation forecasting of station

By using maximum entropy method the precipitation forecasting model of linear stationary time series wasconstructed.In order to make the results stably,the roots of thecharacteristic polynomial,whose modes are more than l,areprojected to the unit ciit circle to modify the coefficients.On tye base of Lanzhou monthly average precipitation data in the range of 1944 to 1997,the monthly preciptation forecasting tests were made from 1986 to 1997.The orders of autoregresion model are chosen as 15,is the best. This model is objective and quantitative and is suitable for the monthly,seasonal or yearly precipitaton forecasting of any station.