Wheat Leaf Area Index (LAI) Inversion By Using “Satellite-UAV-Ground” Multi-Source Remote Sensing Data

Leaf Area Index（LAI）determines crop growth,photosynthetically active radiation（PAR）absorption ratio,biomass and yield.Therefore,accurate and rapid estimation of LAI is helpful to better monitor crops.In previous studies,LAI inversion based on remote sensing tends to be applied to a single remote sensing data,and rarely involves multi-source remote sensing data co-inversion LAI.In view of more and more available remote sensing data and less research on crop LAI in arid areas,it is particularly important to establish a cooperative multi-source remote sensing data inversion model for crop LAI and select the best scale for crop LAI inversion in arid areas.In this study,winter wheat at different growth stages was selected as the research object,and comprehensive remote sensing monitoring experiments of"satellite-UAV-ground"were carried out to analyze the spectral characteristics of winter wheat canopy at different growth stages.Univariate and multivariate regression models for inversion of LAI of winter wheat at different growth stages were established by vegetation index method,and validated.According to the inversion results,the LAI spatial scale transformation of winter wheat was carried out,and the inversion scale suitable for this study was selected.The main conclusions of this study are as follows:1)Compared with the commonly used Hyperspectral Vegetation index,the correlation between the new optimized spectral index and LAI has been significantly improved,and the accuracy of the model has also been improved.The quadratic polynomial model based on the first derivative of the spectrum RSI_FD（627,774）（ratio spectral index）is superior to other models,and the coefficient of determination（R²）reaches 0.809,indicating that the model based on the first derivative of the spectrum is better than other models.The new spectral index calculated by order derivative has a better indication for LAI of winter wheat at jointing stage;Multivariate Partial Least Square Regression（PLSR）model based on UAV digital image parameters VARI（visible atmospheric resistance index）,RGBVI（red green vegetation index）,B（DN value of blue channel）and GLA（green leaf algorithm）has good stability and prediction ability,R²,root mean square error（RMSE）and relative prediction deviation（RPD）are0.776,0.468 and 1.838,respectively,and the RPD is between 1.4 and 2.The inversion accuracy of the model was lower than that of the inversion model based on the surface Hyperspectral Vegetation index;In the inversion model based on GF-1/2 broad-band vegetation index,the PLSR model based on GF-2 multivariable has higher precision than that based on GF-1,R² reaches 0.809,which shows that the stability of the multivariable model is better than that of the single-variable model.2)Compared with the estimation model based on single remote sensing data,the"satellite-UAV-ground"collaborative inversion model based on RSI_FD（627,774）,VARI_UAV and GNDVI_GF-1（green normalized difference vegetation index）has the best prediction effect.The validation group R² and RPD have reached 0.840 and 2.430,respectively.It shows that the winter wheat LAI collaborative inversion based on"sky-space-ground"multi-source remote sensing data has the best prediction effect.The stability and predictive ability of the model have reached a good level,which can better carry out the inversion of winter wheat LAI.3)Scale-up conversion of winter wheat LAI showed that the scale-up conversion effect of point-3cm was better,which was suitable for the scale-up conversion of this study.After verification,R² reached 0.66;the scale-up conversion effect of surface-surface was not good,and the phenomenon of underestimation of winter wheat LAI at jointing stage appeared after scale-up conversion.Considering the results of scale conversion,the 3cm spatial scale of UAV is determined to be the best one for this study.This paper mainly studies winter wheat LAI from spatial scale.In the future,we can add spectral scale to study wheat LAI by using UAV multispectral and hyperspectral data,lidar data and satellite hyperspectral image data.