基于无偏灰色马尔可夫链的吉林省降水量预测

Predication of Precipitation Based on Unbiased Gray Markov Chain in Jilin Province

为了更准确地对吉林省降水量进行预测,分析其时空变化特征,应用无偏灰色马尔可夫链模型对8个具有代表性的雨量站进行降水量预测,并根据预报结果讨论历史数据波动性与预报精度的关系。其中:83%以上预测结果合格,白城、乾安、长春、蛟河、四平、通化6个地区降水量多年呈递减趋势,减幅分别为0.23%、0.09%、0.24%、1.01%、0.51%、0.54%;延吉、靖宇2个地区降水量多年呈递增趋势,增幅分别为2.60%、0.54%。结果表明:无偏灰色马尔可夫链模型预测精度较高,说明该方法适用于吉林省的降水量预测;吉林省中西部地区降水量呈递减趋势,东部地区呈递增趋势,但变幅不大;在波动性与预报精度的关系方面,时间序列的波动性越大预测所产生的误差越大。

For more accurately predication and analysis of the spatial and temporal variation characteristics of precipitation of eight representative stations of Jilin Province,we used the unbiased grey Markov chain model to discuss the relationship between historical data volatility and the forecast accuracy.More than 83% predicted results are qualified.Baicheng,Qian’an,Changchun,Jiaohe, Siping,Tonghua six regional yearly precipitation shows a trend of decline,the damping ranges are 0.23%,0.09%,0.24%,1.01%,0.5 1%,0.54%;while Yanji and Jingyu two regional precipitation is increasing,with growth rates of 2.6% and 0.54% respectively.The results show that the modified unbiased gray Markov chain model is suitable for Jilin Province’s precipitation forecast with higher accuracy.It shows that the precipitation has a trend of decreasing in Midwest of Jilin Province and the precipitation has a trend of increasing in eastern of Jilin Province.In the relation between volatility and forecast precision,the study finds that the greater the volatility of time series is,the larger theprediction error is.