欢迎访问《麦类作物学报》编辑部网站

作者	单位
李石波，朱秀芳，侯陈瑶，郭锐，刘莹	（1.地表过程与资源生态国家重点实验室，北京师范大学，北京 100875；2.土地科学技术学院，中国地质大学（北京），北京 100083；3.地理科学学部遥感科学与工程研究院，北京师范大学,北京 100875）

摘要点击次数: 483

全文下载次数: 698

中文摘要:

为了探究相关气象因子对小麦生产的影响，获取极端气候条件下满足精度需求的作物估产模型，以河南省为研究区，对研究区的历史小麦单产数据进行时间序列分解，得到趋势单产，再结合不同时间尺度干旱指数作为输入变量，以实际单产作为输出变量，建立随机森林回归单产估计模型。选择典型干旱年份（2011年）与非干旱年份（2015年）进行小麦单产模型的精度验证，并对输入变量的重要性进行了分析。结果显示，随机森林回归单产估计模型拟合精度整体较高，各市模型的决定系数平均为0.87，平均绝对值误差的均值为17.69 kg·hm^-2,平均绝对相对误差的均值为0.07。面积加权和简单平均估计得到的各市小麦估产的精度平均值在2011年分别为96.16%和95.12%，在2015年分别为92.99%和92.26%。干旱年份估产精度整体上高于非干旱年份估产精度，面积加权后的小麦单产估计精度略高于简单平均的小麦单产估计精度。对建模贡献最大的输入变量是趋势单产。就干旱指数来说，1个月时间尺度的干旱指数重要性整体高于2和3个月时间尺度的干旱指数；4月份的干旱指数重要性整体高于生长季其他月份的干旱指数。该模型能够准确及时地获取极端气候下小麦产量信息以及变量对小麦产量的影响，可以为研究气候变化对小麦产量的影响和提高极端气候条件下的估产精度提供方法参考。

英文摘要:

In order to explore the impact of relevant meteorological factors on wheat production, and to obtain crop yield estimation models that meet the accuracy requirements under extreme climate conditions,this paper took Henan province as the research area,decomposed the historical wheat yield data to obtain the trend yield,then combined with different time scales drought index to establish the random forest regression yield estimation model.The accuracy of wheat yield model was verified in typical dry year(2011) and non-dry year(2015), and the importance of input variables was analyzed. The results showed that:1)the fitting accuracy of the random forest regression yield estimation model was generally higher. The average value of the decision coefficient was 0.87; the mean value of mean absolute error was 17.69 kg·hm^-2, and the mean value of mean absolute relative error was 0.07.2)The average precision of wheat yield estimated by area weighting and simple average estimation in each city was 96.16% and 95.12% respectively in 2011,92.99% and 92.26% respectively in 2015.The precision of estimated wheat yield in the dry year was higher than that in the non-dry year,and the precision of yield estimated after weighted estimation was higher than that of simple estimation. 3)The input variable that contributed the most to modeling was trend yield. In terms of drought index, the importance of the drought index at the time scale of 1 month was higher than that at the time scale of 2 or 3 months.The overall importance of drought index in April was higher than that of the other months of the growing season. This model can accurately and timely obtain wheat yield information and the impact of variables on wheat yield under extreme climate. It can provide reference for studying the image of climate change on wheat yield and improving the precision of yield estimation under extreme climate.

查看全文查看/发表评论下载PDF阅读器

关闭