Soil Salinity Monitoring Model Based On Upscaling Of UAV-satellite Remote Sensing Data

Soil salinization seriously restricts the sustainable development of agriculture.The use of remote sensing technology can obtain soil salinity information,but scale effects occur when remote sensing data of different scales monitor soil salinity in the same area.Therefore,it is important to solve the scale effect between different remote sensing data and construct an accurate monitoring model for soil salinity control and irrigated agriculture.In this study,field experiments were conducted in the irrigation area of Sha Trench Canal in Inner Mongolia to obtain soil salinity data,satellite remote sensing images and UAV remote sensing images,construct various spectral variables,and combine index screening and machine learning algorithms to build a soil salinity estimation model.Firstly,the soil salinity monitoring model was constructed by using ground-UAV-satellite data synergistically;secondly,the scale effects of UAV and satellite monitoring salinity models in the irrigated area of sand trench canals under different vegetation cover were analyzed;thirdly,a scale conversion was carried out by combining the traditional and improved Chen NDVI methods,and a fractal calculation model was constructed by using the NDVI fractal scale conversion method to construct a soil salinity monitoring model under the best scale.The main conclusions of this study are as follows.The main conclusions obtained from this study are as follows:(1)A soil salinity monitoring model was constructed based on ground-UAV-satellite dataThe soil salinity monitoring model based on ground-UAV-satellite data has improved the accuracy of satellite data monitoring of salinity and expanded the scope of UAV data monitoring.The SMR-UAV model constructed based on ground-UAV data can accurately monitor soil salinity at the UAV scale(R~2of both modeling set and validation set are greater than 0.811,and RMSE are less than 0.112);the salt results SMR-UAV-upscaling by resampling SMR-UAV to 16-meter scale and SMR-UAV in terms of spatial distribution of salt and salt ratio are basically consistent,which can represent the real situation of salinity at the 16-m scale;the salt monitoring results of the soil salinity monitoring model SMR-UAV-Satellite constructed based on ground-UAV-satellite data are basically consistent with those of SMR-UAV-upscaling in terms of spatial distribution of salt and salt ratio,which effectively improves the salt monitoring by satellite data accuracy;the area of salt monitoring using satellite data is 86 times larger than that of UAV data monitoring,which expands the scope of experimental monitoring.(2)Constructing a soil salinity monitoring model based on vegetation cover by upscalingThe soil salinity monitoring model based on vegetation cover with upscaling correction corrects the scale effect between UAV and satellite remote sensing data and improves the accuracy of the soil salinity monitoring model.The vegetation cover Fr obtained from UAV remote sensing data and the vegetation cover FVC obtained from satellite have good correlation,and the R~2of fitting accuracy of Fr and FVC can reach 0.72;the accuracy of salt monitoring by UAV is higher than that of satellite,and the R~2of modeling accuracy of UAV as a whole can reach more than 0.440,and the R~2of modeling accuracy of satellite data limit learning machine can reach 0.412 above;when the vegetation cover CFVC is in the range of 0.1-0.4,the salt monitoring results of UAV are higher than the monitoring results of satellite data,when the CFVC is in the range of 0.4-0.8,some of the salt data obtained from UAV and satellite inversion overlap with each other,and when the CFVC is greater than 0.8,some of the results of SSC no are higher than the estimated results of SSC guard;using vegetation After correcting the scale effect between the UAV and satellite remote sensing data using vegetation cover,both salt distributions were on both sides of the 1:1 line,and the scale correction effectively reduced the overestimation of salt in the range of0.1-0.4.(3)Scale conversion and soil salinity monitoring model based on fractal theoryThe accuracy of the soil salinity monitoring model based on fractal theory was significantly improved compared with that of the model without fractal operation;NDVI had significant fractal characteristics as the scale increased,but the fractal dimension D of each model was relatively small,indicating that the curve structure of the fractal model was not complicated;the difference between the simulated NDVI value and the actual value obtained after scale upward extrapolation was acceptable and met the requirement of model veracity test.For the vegetation,bare land and mixed feature types in the study area,the average spatial heterogeneity index of the images increased and then decreased during the up-scaling process;the accuracy of the UAV model obtained based on the fractal theory improved significantly compared with that without the fractal operation,and compared with the original UAV images,after fractal correction,the R~2of the modeling set increased by0.123,the RMSE decreased by 0.04,and the validation set R~2improved by 0.057 and RMSE decreased by 0.04;fractal processing of high-resolution UAV data helps to achieve the improvement of soil salinity monitoring accuracy.This study constructs a soil salinity monitoring model using different remote sensing data and provides different methods to solve the scale effect problem arising from salinity monitoring,which can provide a reference for the integrated monitoring of soil salinity in agricultural fields by ground data,UAV and satellite multispectral remote sensing.