基于转移概率法的南方晚稻寒露风初日预测

Prediction on Cold Dew Wind First Day of Late Rice in Southern China Based on Method of Transfer Probability

为保障南方晚稻的增产稳产，减轻寒露风对晚稻生育的危害，基于1961-2012年南方晚稻种植区707个气象站的逐日气象资料、晚稻寒露风发生的气象行业标准和晚稻生育期资料，将寒露风初日划分为10种状态，利用转移概率法以1961—2011年的资料建立预测模型，2012年的资料用于预测检验，进行晚稻寒露风初日的预测分析．结果表明：预测模型回代准确率为77％，且70．6％的站点回代准确率在70％以上，对2012年的预测检验中，有74％的站点预测准确；预测模型在寒露风历年发生相对较少的区域预测效果较好，在寒露风历年发生相对较多的区域，预测准确率有待进一步提高；预测模型在寒露风发生频次较少，单次灾害较重的区域具有更好的应用价值．

In order to guarantee the increasing and stable production of late rice in southern China and reduce the harm to fertility of late rice by cold dew wind, prediction analysis on cold dew wind first day of late rice was carried out in this article based on the daily meteorological data of 707 weather stations located in the planting region of late rice in the south from 1961 to 2012, meteorological industry standards on cold dew wind and growth period data of late rice. The prediction model was established though dividing cold dew wind first day into 10 kinds of state and using method of transfer probability with the data from 1961 to 2011. Besides, the data of 2012 was used to prediction test. The results were as follows. Back substitution accuracy of prediction model was 77%, which was equal to or more than 70% of 70.6% stations and there were 74% stations predicting accurately in the prediction test of 2012. Meanwhile, the effect of prediction was better in the regions where cold dew wind was relatively less occurred over the years. However, the prediction accuracy needed to be further improved in the regions where cold dew wind was relatively more occurred over the years. In a word, the prediction model had better application value in the regions where cold dew wind occurred less frequently and single disaster was heavier.