短期气候预测的多时间序列相空间相似模型

为了充分利用对气候系统进行定时观测而积累的关于它过去和现在情况的大量数据，本文提出了短期气候预测的一种多时间序列相空间相似模型，并从与云南气候变化有某种物理关系的２０多个时间序列出发，建立了一个云南短期气候变化的年度预测模型。模型对云南８０个气象观测站点１９９１～１９９８年的５月雨量距平和５月气温距平预测检验的距平同号率分别达到４２４／６４０＝ ６６．３％ 和４２６／６４０＝ ６６．６％ ；对云南１９９０～１９９７年的６～８月雨量距平、６～８月气温距平、年雨量距平和年气温距平预测检验的距平同号率则可分别达到４０４／６４０＝ ６３．１％ 、４６３／６４０＝ ７２．３％ 、４５６／６４０＝ ７１．３％ 和４５５／６４０＝ ７１．１％ 。

A PHASE SPACE ANALOGY MODEL OF SHORT-RANGE CLIMATIC FORECAST BASED ON MULTITIME SERIES

In order to make full use of the accumulative huge climatic data, this paper presents a phase space analogy model of short range climatic forecast with multitime series.By using over 20 time series which are physically related with the climate changes of Yunnan Province , an yearly short range climate forecast model for Yunnan Province is established.The verification of precipitation and temperature forecasts for Yunnan Province shows that the accuracy of anomaly sign of model forecasts of the May precipitation and temperature from 1991~1998,and the June~August and yearly precipitation and temperature from 1990~1997 is 66.3% , 66.6%, and 63.1% ,72.3% ,71.3%,71.1% respectively.