Study On Soil Dielectric Properties And Multisource Remote Sensing Moisture Inversion In Salinization Irrigated Area

Soil moisture is the basic condition for plant growth in land surface ecosystem.It is also the important indicator for the works of soil moisture monitoring and crop yield estimation.The traditional method is water monitoring through the ground observation network.Although the obtained soil moisture information is very accurate,the number of sampling points is limited and the timeliness of data collection is poor,which makes it difficult to achieve large-scale continuous monitoring.However,microwave remote sensing with all-day,all-weather and strong penetrability characteristics can fast obtain surface temporal and spatial information,which provides the possibility for comprehensive observation.Dielectric constant is a macroscopic parameter to describe the interaction between electromagnetic field and matter.Different soil moisture content has obvious different dielectric properties,and then the scattering coefficients are different.So it is very important to study soil dielectric properties.In this paper,soil dielectric properties are simulated through RADARSAT-2 radar data,and then the soil moisture properties are studied and inverted.At the same time,in order to verify the wide applicability of microwave remote sensing for moisture inversion,analyze the relationship between spectral reflectivity and moisture of Landsat8 remote sensing image data,a moisture inversion model is established.The seriously salinization affected Jiefangzha Irrigation Field in Inner Mongolia Hetao Irrigation Area is selected as experimental area:(1)Section plate method is used to measure surface roughness.Combined roughness neural network inversion model of root mean square height and relevant length is built.(2)Regression analysis of the relax relationship among real part of dielectric constant and SAR four polarization backscattering coefficient and surface roughness is conducted.It is also compared with oh empirical model,and the relativities are respectively R2=0.8597,R2=0.8209 and significantly related.Dielectric constant inversion model is built.(3)Dielectric constant model is verified.The relativity of the real part of dielectric constant verified by Dobson semi empirical model and Hallikainen simplified real part empirical model and the measured value are respectively R2=0.9359 and R2=0.8690,which shows that the two models can both simulate the close relationship between the surface soil moisture and the real part of the dielectric constant;at last,the soil moisture content model is built by inversion of Dobson model and Dobson simplified real part empirical model,and the correlation between simulated values and measured values are respectively R2=0.8038 and R2=0.7374,and the root mean square errors are respectively RMSE=5.2%and RMSE=5.7%.(4)According to the AIEM forward modeling,the selection work of moisture inversion parameters is conducted.The moisture inversion regression model is constructed through HH?HV?VH?VV backscattering coefficients and their combination,dielectric constant and combined surface roughness.The relativity of simulation value and the measured value of soil is R2=0.8421,and the root mean square error is RMSE=5%.Based on the sampling point data of 2015,the correlation between the simulated values and soil measured values was R2 = 0.7971,and the mean square root error was RMSE = 5.4%.(5)According to the Landsat8 remote sensing image data,the exponential relationship between spectral reflectivity and soil moisture is established,the correlation is R2=0.6793;finally,the Landsat8 remote sensing image data moisture inversion model is established,the correlation is R2=0.6317,and the root-mean-square error is RMSE=6.3%.By comparing and analyzing the four moisture inversion models,it is found that the inversion results of Dobson model and statistical regression model are in good agreement with the field soil moisture distribution,which all have good accuracy and applicability,however,the Dobson model needs more parameters,which is less convenient than the statistical regression model,therefore,the statistical regression model has better applicability,thus laying the foundation for monitoring soil moisture by microwave remote sensing.