Monitoring Maize Evapotranspiration And Soil Water Content In Farmland Based On UAV Multi-Source Remote Sensing Data

Agricultural evapotranspiration and soil moisture content are key parameters in the energy balance and hydrological cycle between the land surface and the atmosphere,and are also important factors governing the growth and development of crops in arid and semi-arid regions.Timely and accurate acquisition of crop evapotranspiration and soil moisture content is the basis and key to monitoring crop growth,guiding irrigation decisions and realizing precision agriculture.With the development of precision irrigation technology in precision agriculture,the use of UAV remote sensing methods can achieve the tasks of crop information and soil moisture monitoring in agricultural fields quickly,efficiently and accurately without touching or destroying the original soil structure.In this study,we selected maize under deficit irrigation treatment in Dalat Banner,Inner Mongolia,and estimated the evapotranspiration of maize in farmland based on visible,multispectral and thermal infrared image data obtained by UAV multi-source remote sensing and ground-measured leaf area index and plant height growth index as well as meteorological data,combined with FAO56 dual crop coefficient method.In addition,based on the UAV multi-source remote sensing data and machine learning algorithm to estimate the shallow soil water content of maize throughout the reproductive period,combined with the water balance model to simulate the dynamic changes of soil water content in the root zone,aiming to provide technical support for crop irrigation management and efficient use of agricultural water resources.The main research contents and conclusions of the paper are as follows:(1)To address the problem that it is difficult to estimate the crop coefficients under water stress due to the multispectral vegetation index because the stress coefficients are not considered,this study proposes a model for estimating maize evapotranspiration in agricultural fields based on UAV multispectral remote sensing and FAO56 dual crop coefficient method.The results showed that the model based on the ratio of transformed chlorophyll absorption in reflectance index(TCARI)and renormalized difference vegetation index(RDVI)was the most accurate.The model with the highest accuracy in estimating crop coefficients compared to the model with the highest accuracy in estimating crop coefficients and the distribution of crop coefficients generated by the model can effectively monitor the variability of different irrigation treatments.In addition,the model with FAO56 dual crop coefficient method has the highest accuracy for estimating maize evapotranspiration under adequate irrigation and mild water deficit irrigation treatments.(2)To address the problem that the current evapotranspiration models based on multispectral remote sensing and crop coefficient method have obvious overestimation under water stress conditions,the maize evapotranspiration estimation models obtained from UAV multi-source remote sensing data combined with FAO56 dual crop coefficient method were developed,and its estimation accuracy under different levels of deficit irrigation was analyzed.The results showed that the accuracy of the maize evapotranspiration models based on the canopy temperature Crop water stress index(CWSI)and Normalized difference vegetation index(NDVI)was the highest,with R2of 0.84 and RMSE of 0.50 Although the accuracy of each model varied with irrigation level,the daily mean deviation error of the evapotranspiration prediction model was small,indicating that the evapotranspiration estimation method proposed in this study was robust in different years.The sensitivity analysis of each parameter in the evapotranspiration model revealed that the model was highly sensitive to the slope of linear regression between the base crop coefficient and vegetation index,followed by the soil evaporation constant.(3)To investigate the feasibility of collaborative inversion of soil water content of maize throughout the reproductive period from spectral,temperature,canopy structure and texture information obtained by UAV multi-source remote sensing,a machine learning algorithm was used to construct a shallow soil water content prediction model for maize in agricultural fields,and the effects of different deficit irrigation treatments and different growth stages on the accuracy of the model were analyzed.The results showed that the combined model with all types of sensor characteristic indices using the random forest regression algorithm had the highest accuracy of soil water content estimation(R~2=0.65,RMSE=2.1%).Also,the optimal soil water content estimation model based on the random forest regression algorithm with all types of sensor characteristic indices produced the most accurate and robust estimation accuracy during the nutritional growth period of maize,followed by the maturity period of maize.In addition,the optimal model showed similar accuracy and strong applicability in predicting soil water content for two soil depths(10 cm and 20 cm),with the highest accuracy in estimating soil water content under fully irrigated and light to moderate deficit irrigation treatments.(4)To investigate the feasibility of combining UAV multispectral remote sensing and water balance model for estimating soil water content in the root zone of maize in agricultural fields,water deficit factors calculated from rainfall irrigation and atmospheric demand information and stress coefficients estimated from vegetation index were used to constrain the actual evapotranspiration of maize under drought conditions.Finally,the effect of different water application and rainfall intensity on the accuracy of soil water content models in the root zone of maize in agricultural fields was analyzed.The results showed that the model was able to accurately estimate the soil moisture trends throughout the maize growth period and had the highest simulation accuracy under mild and moderate deficit irrigation treatments with R~2 of 0.60-0.92 and RMSE of 0.4%-2.30%.Overall,this study provides technical support for water-saving irrigation,water management and planning of farmland in arid and semi-arid areas.The combination of UAV multi-source remote sensing and related models achieved high spatial and temporal resolution farmland maize evapotranspiration and soil water content estimation,breaking through the previous limitation of using only a single sensor to estimate crop evapotranspiration and soil water content at a certain growth stage,and providing technical support for crop irrigation management and efficient use of agricultural water resources.