Study On Monitoring Model Of Soil Moisture Content In Summer Maize Root Domain Based On UAV Remote Sensing

Soil moisture in the root zone is an important part of soil moisture in the field and plays a key role in the growth and development of crops.The accurate estimation of soil moisture in the root zone is of great significance for predicting crop yield,field water stress and crop growth status.In this study,field and summer maize with different water treatments in 2020and 2021 were studied.Multiple moments（9:00,11:00,13:00,15:00）of three main growth stages（panicle initiation to booting stage,heading to anthesis stage and milk to soft dough stage）of summer maize were acquired by UAV equipped with multi-spectral and thermal infrared lens.00),combined with canopy phenotype data and meteorological parameters,two methods were proposed to improve the performance of crop water stress index（CWSI）.The evaporation ratio on daily time scale is calculated based on TSEB model,and theα_PTalgorithm is improved.A variety of vegetation indices were calculated and soil water content monitoring at different depths was carried out.The main contents and conclusions of this study are as follows:（1）Two methods to improve CWSI performance were proposed by improving canopy temperature.One was to use different statistical quantiles based on normal distribution to segment canopy temperature and calculate the average canopy temperature on different statistical quantiles to calculate CWSI(denoted as CWSI_%Tc).Second,based on variance of canopy temperature,corn canopy data is divided into four parts:IntervalⅠ,IntervalⅡ,IntervalⅢ,IntervalⅣ.and the most sensitive statistical quantile on the respective interval was selected to calculate CWSI.The results showed that:1)Although CWSI_%Tc can provide a higher accuracy of crop water stress,the most sensitive canopy temperature range varies greatly between years（2020,61.17%;2021,49.38%;Two-year data,83.51%）.This may affect the application of CWSI_%Tc in practice 2)CWSI_n presented a higher correlation for crop physiological indexes.In practice,CWSI_n is less affected by the most sensitive canopy temperature interval and has higher stability（n RMSE;16.60%,27.37%and 28.49%by 2020;21.60%,18.95%,22.64%in 2021）.Therefore,compared with the original CWSI,CWSI_n can more accurately monitor crop water stress.（2）Soil moisture monitoring models at different depths were constructed by using the evaporation ratio calculated by TSEB-2T model on a daily time scale,and theα_PT algorithm was improved based on soil moisture content.The results show that the standard root-mean-square error of the Mod2 model based on microwave soil moisture downscaling is smaller and the prediction ability is stronger.But the optimal monitoring depth is not the same at different growth stages.The best soil moisture monitoring depth was 10-20cm at Panicle initiation to booting stage（R²=0.40）,10-30cm at heading to anthesis stage（10-20cm R²=0.27,20-30cm R²=0.28）,20-40cm at milk to soft dough stage（20-30cm R²=0.31,30-40cm R²=0.33）.Theα_PT algorithm was improved by using soil moisture,and the improvedα_Pt-N values based on measured soil moisture values and Mod2 model simulation values were better than the originalα_PT model in canopy temperature inversion.（3）A method of using infrared information and spectral information to coordinate monitoring soil moisture content at different depths was proposed.The method adopted the coupling method of partial least squares regression model and variable projection importance analysis.The research conclusions indicated that,CWSI is limited by the model itself,and the accuracy of monitoring surface soil moisture（0～10cm）is limited.The monitoring accuracy of multiple linear regression model constructed by multiple vegetation indices is better than that of single vegetation index,but the method based on a large number of vegetation indices is not practical.However,the coupling method can improve the monitoring accuracy of soil moisture content at different depths,and the optimal monitoring depth is different at different growth stages,among which the optimal monitoring depth at Panicle initiation to booting stage ranges from 0 to 10cm（modeling set R²=0.48,validation set R²=0.49）.The best monitoring depth at heading to anthesis stage was 20～30cm（modeling set R²=0.65,verification set R²=0.65）,and the best monitoring depth at milk to soft dough stage was 30～40cm（modeling set R²=0.30,verification set R²=0.32）.