摘要
时间趋势产量预报是作物产量气象预报中很重要的一部分。本文采用一种递推型模型——指数平滑法进行时间趋势产量的预报。它对近期样本给予较大权重,从而可以较好地拟合产量序列的历史演变,并具有一定的外推功能。大量产量资料的应用试验表明,我国多数地区使用该方法时可将初始时刻的产量实测值取为初始平滑值;若产量序列变化较平稳,则加权系数取值较小,样本可直接进行指数平滑,反之,加权系数变大,样本亦需进行适当的预处理。通过与其它常用方法的比较,认为该方法可作为一种作物趋势产量模拟及短期预测的方法加以应用。
The forecasting of the time-dependent trend yield is an important part of the meteorologicalforecasting of crop yield.In this paper,a kind of recursion model of exponential smoothing hasbeen used.The recent samples in a time series have bigger weighted values so as to fit historicalvariation of time series better.It has a certain extropolation function.The results of the appliedtests show that the initial actual yield can be used as the initial smoothing values in most areas ofthe country;if the variation of yield is less fluctuating,the weighted values are smaller and theexponential smoothing method can be used immediately.Otherwise,if the weighted values arebigger,the samples have to be preprocessed.By comparison with other methods commonly used,itis considered that this method can be used as an approach for fitting the time-dependent trendyield and short term forecasting of crop yield.
