摘要
为进一步促进机器学习技术在玉米单产估测中的应用,以河北中部平原为研究区域,选取与玉米长势和产量密切相关的条件植被温度指数(Vegetation temperature condition index,VTCI)和叶面积指数(Leaf area index,LAI)为特征变量,通过极限梯度提升(Extreme gradient boosting,XGBoost)算法和随机森林(Random forest,RF)算法分别对玉米单产进行估测。基于组合预测思想与Shapley值理论,分别确定组合预测模型中XGBoost与RF模型权重,进而得到组合预测模型,结果表明,基于Shapley值确定的组合估产模型精度较高(R^(2)=0.32),达极显著水平(P<0.001)。同时将组合预测模型应用于河北中部平原2012年各县(区)玉米的单产估测,结果表明,模型精度较高(R^(2)=0.52),玉米估测单产与实际单产的平均相对误差和均方根误差分别为9.86%、831.14 kg/km^(2),达到极显著水平(P<0.001),且组合预测模型的精度均优于单一估测模型。研究发现,河北中部平原玉米估测单产随年份发生波动变化,呈先降低后升高的趋势。玉米估测单产以西部地区最高,其次是北部和南部地区,东部地区最低。
Aiming to promote the application of machine learning in agriculture field and improve accuracy of the maize yield estimation,the central plain of Hebei Province was selected as the study area,which includes fifty-three counties(districts).Vegetation temperature condition index(VTCI)and leaf area index(LAI)at the main growth stages of maize were selected as key crop growth indicators for estimating the maize yield by using two machine learning methods,extreme gradient boosting(XGBoost)and random forest(RF),and as well as their combination.Firstly,the XGBoost and RF were used to estimate yield of maize from^(2)010 to 2017,then the XGBoost and RF’s weights were determined by combination forecasting model by using the Shapley value method,and finally maize yield of each county in 2012 was estimated based on the combination forecasting model.The results showed that the mean relative error(MRE)and root mean square error(RMSE)between the estimated yield of maize and the actual yield were 9.86%and 831.14 kg/km^(2),respectively.The accuracy of the combination forecasting model(R^(2)=0.52,P<0.001)was better than that of the XGBoost model and RF model,which can be applied to estimate the yield of maize in the study area.The combination model was used to estimate the maize yield of the central plain of Hebei Province pixel by pixel from^(2)010 to 2018.The estimated yield of maize showed a trend of decrease first and then increase over time.The spatial distribution of maize yield was the highest in the western region,followed by the northern and southern regions,and the eastern region was the lowest.The results showed that the temporal and spatial changes of maize in the central plain of Hebei Province were in line with reality,and the research result can provide guidance for the growth monitoring and yield estimation of maize in the study area.
作者
王鹏新
乔琛
李俐
周西嘉
许连香
胡亚京
WANG Pengxin;QIAO Chen;LI Li;ZHOU Xijia;XU Lianxiang;HU Yajing(College of Information and Electrical Engineering,China Agricultural University,Beijing 100083,China;College of Land Science and Technology,China Agricultural University,Beijing 100083,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2021年第9期221-229,共9页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家重点研发计划项目(2016YFD0300603-3)。
关键词
玉米
估产
条件植被温度指数
叶面积指数
SHAPLEY值
组合预测
maize
estimation of yield
vegetation temperature condition index
leaf area index
Shapley value
combination forecasting
