Estimation Of Farmland-scale Evapotranspiration Based On Multi-source Data

Evapotranspiration is critical information in agricultural operations and research.Accurate evapotranspiration information can provide decision support for rational allocation and usage of agricultural water resources.Aiming at the lack of effective quantification of evapotranspiration in the farmland,this study builds a platform to obtain basic data of estimating evapotranspiration.A new method based on UAV for estimating evapotranspiration in farmland is created according to the existed satellite evapotranspiration models.This method estimates evapotranspiration based on data collected by a multi-spectral sensor and thermal infrared sensor mounted on a UAV.The performance of this method on the experimental farmland which locates in Yangling Area is reliable.(1)Based on the data required by the evapotranspiration model,a data acquisition platform is established to solve the problem of obtaining basic data of the evapotranspiration model.This data acquisition platform based on M100 multi-rotor UAV and equipped with thermal cameras and multispectral imagers.The platform integrates multispectral sensors with thermal infrared sensors,which can acquire image data in multiple bands and provide effective data to support for estimating evapotranspiration.(2)Researching the performance of single-source model,multi-source model and machine learning model in the test area is based on the satellite data and Google Earth Engine.Select the mapping evapotranspiration at high resolution with internalized calibration model,and the two-source energy balance model,and back propagation neural network to estimate evapotranspiration based on Landsat 8 satellite data.Comparing the test results with the measurement results of the Open Path Eddy Covariance system,the experimental results show that the results of the three modeling methods and the measured values have a certain correlation.The errors of the results of the multi-source model are relatively small.(3)The spatial and temporal accuracy of the satellite model results cannot meet the needs of precision agricultural operations,which requires more accurate data.The low-altitude remote sensing data provided by the UAV can meet this demand.In order to obtain the evapotranspiration distribution with high spatial resolution,it is necessary to use low-altitude remote sensing data as the basic data to study farmland evapotranspiration mapping methods.Registration and splicing of data from different sensors are used to obtain basic data for evapotranspiration estimation with a resolution of decimeter level.It is proposed to correct the acquired thermal imager data based on the linear relationship between the UAV thermal imager data and the actual temperature.The correction result is closer to the true value than the original measured value.(4)Since evapotranspiration models are mostly based on satellite data,the adaptability of drone data to evapotranspiration models needs to be studied.The low-altitude remote sensing data was matched to the satellite remote sensing evapotranspiration model,and the evapotranspiration model was matched to the drone data.Comparing the model calculation results with the OPEC measurement results,the RMSE estimated by the METRIC model is0.0684 mm/h,the MAE is 0.0519 mm/h.The RMSE estimated by the TSEB model is 0.0874mm/h,and the MAE is 0.0720 mm/h.There is a strong correlation between estimations and measurements,but there are still some errors.(5)This research develops and validates the URSEB(UAV Remote Sensing Energy Balance)model,which is an estimation model of farmland evapotranspiration based on UAV The URSEB model based on UAV data obtains more accurate flux data,and compares the estimated value with the measured value.The sensible heat flux RMSE is 20.013 W/m~2,MAE is 15.835W/m~2,and the latent heat flux RMSE is 40.202 W/m~2,MAE is 26.017 W/m~2.The RMSE of the estimated evapotranspiration was 0.0383 mm/h,the MAE was 0.0322 mm/h,and R~2 was 0.944.Finally,the URSEB model was used to obtain the spatial distribution map of evapotranspiration in the test area.This research has developed a method for estimating evapotranspiration based on low-altitude remote sensing by comparing different data acquisition platforms and different evapotranspiration estimation methods.After experimenting a lot in Yangling's test farmland,the test results prove that this method can effectively provide support for precision agricultural technologies,such as farmland irrigation decisions,and obtain field evapotranspiration in high-resolution spatial models.