Research On Inversion Of Surface Soil Moisture Based On Wavelet Analysis By Multi-satellite Fusion GNSS Remote Sensing

Soil moisture plays an important role in research in fields such as hydrology,meteorology and agro-environment,and can be used as a measure of the water cycle in environmental sciences,among others.In recent years,soil moisture inversion using GNSS reflection signal has become a new monitoring method with the advantages of high precision,good continuity and wide application,therefore,ground soil moisture inversion based on GNSS-IR has a broader practical significance and application prospect.To address the problems of anomalous jump of individual annual product days and large fluctuation of single-star inversion accuracy in soil moisture based on GNSS-IR inversion,Use wavelet analysis and LS-SVM algorithm to optimize the inversion model,discuss the feasibility and effectiveness of wavelet analysis and LS-SVM algorithm in GNSS-IR soil moisture inversion.The main research contents and conclusions of this paper are as follows:(1)This paper proposes a method of separating satellite reflection signals using wavelet analysis.The observation files of P041 station in 2011 and 2012 are processed separately,and the soil moisture data retrieved by each satellite are obtained respectively.The soil moisture experiment is compared and analyzed,and it is found that most of the satellite inversions have good accuracy,but the inversion is prone to system errors and abnormal jitter problems.In response to this problem,the use of wavelet multi-scale decomposition of satellite reflection signals to improve the problem of poor local fit.For the accuracy of local fit,the wavelet analysis method is relatively low-order polynomial method,and its RMSE can be increased by at least 7.98%;MAE At least an increase of 25.2%.Weaken the abnormal jump values of some satellites in individual years,and the average correlation coefficient of the inversion results obtained is 0.803,reaching a higher level of accuracy in single-satellite inversion.(2)This paper studies the selection of wavelet basis functions in the separation of satellite reflection signals.Using traditional low-order polynomial methods and 4 wavelet basis functions,the reflection signals of the observation data of 10 satellites at P041 station are separated,and the final result is obtained.The comparative analysis between the inversion result and the reference value of soil moisture found that the use of wavelet analysis of satellite reflection signals can effectively reduce system errors and jumps,and the accuracy of the inversion results obtained from 10 satellites has been improved,and further reduced Compared with the traditional method,the error of RMSE and MAE is obtained,and a better correlation coefficient is obtained.Through the comparative analysis of the five schemes,using the linear regression model to invert the soil moisture can better show the changing trend and law.The average correlation coefficient of coif5 wavelet single-satellite inversion is0.836,which is an increase of at least 20.1% compared with the traditional method,and10.4%,5.2% and 11.5% respectively compared with bior6.8,dmey and sym10 wavelets.It shows that the application of coif5 wavelet to GNSS-IR soil moisture inversion is feasible and effective.(3)This paper uses least squares support vector machine(least Squares-Support Veotor Maohine,LS-SVM)for multi-satellite combination rolling to estimate soil moisture changes.Due to the difference in the azimuth and angle of each satellite relative to the station,at the same time,each satellite’s There are also certain differences in parameters and performance.Therefore,only using a certain satellite pair to retrieve changes in soil moisture cannot fully and systematically grasp the changes in soil moisture in a certain area.Soil moisture changes are non-linear,and through different numbers of The combination of satellites can obtain different accuracy inversion results.Experiments and analysis have found that the use of LS-SVM for multi-satellite fusion can significantly improve the inversion accuracy and effectively integrate the advantages of each satellite.With the number of satellite combinations As the number of satellites increases,the accuracy of the inversion is continuously improved;when the number of satellites in combination reaches seven,the correlation coefficient of the inverted soil moisture results reaches 0.916,which is at least16.1% higher than that of a single satellite;the RMSE is 0.039.Therefore,the combination The wavelet analysis method,the establishment of a satellite reflection signal separation model and multi-satellite fusion can significantly improve the accuracy of soil moisture inversion,and it is a new method for monitoring soil moisture with high actual spatial and temporal resolution.