欢迎访问《麦类作物学报》编辑部网站

近年来，机器学习算法逐渐被运用到作物估产中，但现有研究仅对比不同方法的估产精度，很少分析估产的最佳时间。本研究基于随机森林算法，对2001-2013年河南省八个时间段的冬小麦遥感、土壤、气候数据进行训练并预测2014、2015年产量，对比实际产量确定最适合河南省小麦产量的训练时间段，探讨影响因子对产量预测的影响程度。结果表明：（1）随机森林算法适用于河南省冬小麦产量预测，能取得较好效果；（2）12-3月为河南省随机森林算法预测产量的最佳时间段，两年的r均达到0.8，且该算法在河南省更适用于短时间序列预测；（3）在影响因子中，月降水对模型精度的影响最大，月最高温度影响最小。

英文摘要:

Estimating crop yield accurately is crucial for making farming decisions and is related to the national economy and people’s livelihood. In recent years, machine learning algorithms have been gradually applied to crop estimation. However, the existing researches only compare the effects of different methods on the accuracy, but there is very little discussion on the estimation accuracy for different time periods. Based on the random forest algorithm,the remote sensing,soil and climate datas winter wheat at eight time periods from 2001 to 2013 in Henan province were trained and the wheat yields in 2014 and 2015 were predicted. By comparing with the actual yields in 2014 and 2015,the suitable training time period for wheat yield prediction in Henan province was determined, and the influence of impact factors on yield prediction was discussed. The results showed that,the random forest algorithm is suitable for winter wheat yield prediction in Henan province, and can achieve good results.The period from December to March is the best time for predicting the yield with r reaching to 0.8 in both years. Moreover this algorithm is more suitable for short-time sequence prediction in Henan province. Among the influence factors of prediction, precipitation has the greatest influence on the accuracy of the model, and the highest temperature has the minimum effect.

查看全文查看/发表评论下载PDF阅读器

关闭