Research On Monitoring Method Of Soil Salinization Based On Scale Conversion Of Space-air Remote Sensing Data

Soil salinization has become an important factor restricting the development of irrigated agriculture,which affects the growth and development of crops.Therefore,accurate and rapid access to soil salinization information is of great significance to the sustainable development of modern agriculture.Because of its non-destructive and high-efficiency characteristics,remote sensing technology can monitor soil conditions in real time,and has become an important means of modern agriculture precision.Among them,satellite remote sensing has been widely used at home and abroad because of its advantages of large-scale monitoring.As a new remote sensing platform,UAV remote sensing has the advantages of high time efficiency,high spatial resolution,low-altitude flight under clouds,and high maneuverability.Multi-temporal and multi-scale high-spatial-resolution remote sensing images can be obtained by setting the flight altitude,but the UAV has defects in monitoring soil salinization on a large scale.How to combine the high spatial resolution of UAV remote sensing with satellite remote sensing The combination of large-scale monitoring has attracted the attention of many scholars.In addition,with the improvement of technology and the diversification of spatial sensors,a large amount of data has been brought to agricultural remote sensing,which provides rich data support for the monitoring of soil salinization,but it also brings a lot of information redundancy while providing rich information.The monitoring accuracy may not be improved.Therefore,under multi-source data,how to select the appropriate scale for target monitoring is of great theoretical significance for the study of the conversion from a drone to a reasonable scale of satellites,and monitoring of soil salinization.Research has rarely been reported.Therefore,in this study,the surface soil with different degrees of salinization was used as the research object in May?bare soil period?and July?vegetation cover?in the Shahaoqu Irrigation Area of Hetao Irrigation District,using drone and satellite observe the spectral characteristics of soil and vegetation,analyze the correlation between the spectral reflectance and the measured soil salinity,use mathematical modeling methods to build a soil salinization monitoring model under the UAV data and GF-1 satellite remote sensing data source,and verify the accuracy On the basis of the comparison,the estimation accuracy of different models for soil salinity is compared.And through the scale conversion method to achieve the inversion accuracy of satellite remote sensing monitoring of soil salinization.The research results are expected to provide some theoretical support for improving the accuracy of remote sensing monitoring salinization,and hope to promote the remote sensing scale.The results obtained are as follows:?1?Soil inversion models under different spectral bands and spectral indexes were constructed.The spectral band reflectance and soil salinity corresponding to the UAV and GF-1 data showed a trend of increasing first and then decreasing.Both of the optimal inversion models constructed by the selected spectral bands are multiple linear regression models.When the spectral band and the spectral index together form the spectral factor for retrieving soil salinity,the optimal model under GF-1 satellite data is a multiple linear regression model with a coefficient of determination R² of 0.38;the optimal model under UAV data For the stepwise regression model,its decision coefficient R² can reach 0.45.It can be seen that the modeling effect after adding the spectral index is better than that of the individual spectral bands.?2?The point spread function method is superior to the traditional resampling method.Through the analysis of scale effects such as average value and standard deviation,it is found that the image after upscaling by the point spread function method is closest to the original image,and the image after the nearest neighbor upscaling is greatly different from the original image.Through the comparison and analysis of different scale conversion methods on the inversion of soil salinization accuracy,it is found that different scale conversion methods have different optimal monitoring models,and finally determine the optimal conversion of monitoring soil salinization under bare soil period conditions Method and optimal model.In general,the quantile regression model under the point diffusion function conversion method has the best effect.The R² of the modeling set and the verification set are above 0.53,and the RMSE is less than 0.15.?3?The upscaling method based on Ts HARP can improve the model accuracy of GF-1satellite remote sensing inversion of soil salinity.Among the spectral bands and salinity indexes of the two remote sensing data,the blue band B1,the near-infrared band B5,the salinity index SI,the salinity index S5,and the improved spectral index NDVI-S1 have a good correlation with the surface soil salinity,and the correlation coefficients are both Above0.35;based on these five variable factors,the soil salinity inversion models of UAV multi-data and GF-1 data were established respectively.By comparison and analysis,the optimal models of UAV inversion of surface soil salinity were stepwise regression.In the model,the modeling set and validation set R² are 0.452 and 0.473,respectively.The GF-1satellite data retrieval surface soil salt is a multivariate linear model,and the modeling set and validation set R² are 0.383 and 0.347,respectively.Through the Ts HARP upscaling method,the trend surface established by the drone data and the measured salt data is applied to the GF-1 data.After the conversion residual correction,the modeling set and verification set R² of the model after upscaling can reach 0.72 and 0.71.It can be seen that the upscaling method based on Ts HARP can improve the model accuracy of GF-1 satellite remote sensing inversion of soil salinity.?4?Through fractal theory,the optimal scale?spatial resolution?between the UAV image resolution and the GF-1 satellite image resolution was selected,and the soil salt inversion model at the optimal scale was established.Using the common vegetation index NDVI,a spatial scale conversion model with an applicable range of 0.1m to 16m was determined:log₂NDVI=0.012238Scale-3.1485,and it showed that the overall change trend of NDVI in this range increased first and then decreased,and determined The optimal spatial resolution is 2m.Compared with the model directly established by the 0.1m resolution UAV data source,the inversion model established at the optimal resolution can increase the modeling set and verification set R² by at least 0.1 and the RMSE by about 0.07.Compared with the GF-1 satellite data with a resolution of 16m,the R² of the modeling set and the verification set is increased by at least 0.16,and the RMSE and Bias are also significantly reduced.