基于马尔可夫模型的南京市年降水量预测试验

鉴于降水过程存在大量不确定性的特点，中长期降水预报一直是气象科学领域的一个难题。主要采用平滑处理消除异常值，对年降水量进行分级，建立滑动平均马尔可夫链预测模型。通过建立的预测模型，对南京市2009、2010年年降水量进行了试预测。结果表明，2009年预测值与实测值相对误差为2．17％，2010年预测值与实测值相对误差为3．09％，预测精度较高。同时，为弥补滑动平均马尔可夫链模型对极端年份无法预测准确的缺陷，利用灰色灾变系统，对南京市极端降水年份进行预测。两者结合能更好地完成南京市年降水量的预测工作。

Forecast Experiment of Nanjing Annual Precipitation Based on Markov Model

Given the uncertainties in the precipitation process,the forecast of long-term precipitation has always been a problem in meteorological field.Here through the use of smoothing to eliminate outliers and the classification of annual precipitation,a moving average Markov forecast model was established to carry out a trial forecast of Nanjing precipitation in 2009 and 2010.As shown by the results,the relative error between the forecasted value and measured value was 2.17% in 2009 and 3.09% in 2010,the forecast precision was high.Moreover,in order to make up for a deficiency of moving average Markov model that it could not predict the year with extreme precipitation,the grey system was used to forecast the extreme rainfall years.Combination of both could better accomplish the forecasting of annual precipitation in Nanjing.