Leaf Area Index Inversion of Winter Wheat Based on Sentinel-2 Data and Machine Learning
Gao Rui
Feng Wenjie
Zhang Junyong
Zhang Zhuoran
Luo Xiubin
Liang Guoxin
Guo Hongyan
Yan Shuai
Wang Fei
Ma Shuang
Abstract：Leaf area index(LAI)is an important indicator for remote sensing monitoring of crop growth and yield.In this study,the Sentinel-2 satellite images of Feicheng City,Shandong Province were used as the data source to explore remote sensing inversion methods for monitoring winter wheat LAI on a large scale.Win-ter wheat LAI inversion models were constructed based on Random Forest(RF),Back Propagation Artificial Neural Networks(BP-ANN),and Support Vector Regression(SVR)methods.We selected highly correlated band reflectance information and vegetation indexes based on Sentinel-2 as input variables for three models.By evaluating the accuracy of each model,the optimal inversion model was selected to map winter wheat LAI on county level.The results showed that the correlation coefficients of difference vegetation index(DVI),reverse difference vegetation index(IDVI),ratio vegetation index(RVI),and chlorophyll vegetation index(CI-green)with LAI were over 0.8,while the correlation coefficients of red light band B4,red edge band B7,and near-infrared bands B8 with B8A with LAI were all over 0.7.The LAI-RF inversion model had the highest ac-curacy among the three inversion models with R2 values above 0.85 for both the training and testing sets,and compared with LAI-BP-ANN and LAI-SVR,the R2 of the test set increased by 8.9%and 26.5%respectively,while the RMSE decreased by 23.7%and 41.5%respectively.These results showed that the LAI-RF model based on Sentinel-2 data could better reflect the actual LAI situation of winter wheat and was suitable for fore-casting and mapping LAI of winter wheat at county level.Further obtaining the winter wheat LAI distribution in the county-level study area through the LAI-RF model,it was found that the spatial distribution of winter wheat LAI in Feicheng showed a trend of higher in north and south region and lower in middle region.These results could provide scientific references for growth monitoring and yield evaluation of winter wheat on county scale,further provide serves for modern agricultural production and ensure food security.
Keywords：Leaf area index(LAI)Winter wheatRandom ForestBack Propagation Artificial Neural NetworksSupport Vector RegressionRemote sensing monitoring
Publication Date：2025-12-30
Online Publishing Date：2026-05-22(First online date of this platform, not the publication date of the document)
Pages：8( 145-152 )