Surface Soil Moisture Estimated From Dual-frequency GNSS Interferometric Reflectometry

Soil moisture has a great significance for the global water cycle,meteorology and agricultural production.Soil moisture,as an important observation factor reflecting soil moisture content,has received extensive attention.Methods of existing for soil moisture detection have their advantages and disadvantages.GNSS-R(Global Navigation Satellite System Reflectome-try)is a new technology of microwave remote sensing,which has developed rapidly for its advantages of rich signal source,low cost and no limitation of climate and so on.In this thesis,the method and model of soil moisture retrieval using SNR data observed by GNSS receiver are studied from the view of reducing cost,improving data quality and inverting precision.In this thesis,we mainly improve the inversion accuracy from two aspects.Firstly,we reduce the influence of error data on the inversion results.In order to improve the accuracy of the data inversion,we propose a robust estimation method to reduce the influence of error data.Secondly,the inversion accuracy is improved by data fusion.The data have different quantity.The reasonable data fusion method is used to make more efficient information remain to improve the inversion precision.The research contents and conclusions are as follows:(1)Robust estimation method for detecting outliers is proposedThe observed data are easily affected by vegetation water content and surface roughness in the process of soil moisture detection by GNSS-IR(Global Navigation Satellite System Interference and Reflectometry)technology,which makes the observed results appear abnormal.Usually,the error data are judged according the figure with changes of degree from the signal-to-noise ratio time series or the correlation degree between the characteristic parameters and the observed objects.Then the error data are eliminated manually.In this thesis,the robust estimation method is proposed to detect the error data.Robust estimation can effectively to reduce or eliminate the influence of the observed data on the observed results.In this thesis,we can effectively reduce the results of soil moisture retrieval by the characteristic parameters,and improve the accuracy of the results of the inversion.(2)A method of entropy value data fusion is proposedWith the addition of new wavebands,new systems and multiple observation platforms,data fusion has become an important research direction in the field of GNSS-R.The traditional method is to process the single frequency data to retrieve the soil moisture.In this thesis,soil moisture retrieval by combining different frequency data is presented.The contribution weights of each frequency are determined by entropy method.Comparing with data fusion result the correlation increased 20%and the RMSE decreased 19%.(3)A robust multiple regression method Establishing robust multivariate regression modelMost things in reality are influenced by many factors.The single factors often can't fully reflect the characteristics of things.The observation data with SNR for soil moisture detection are affected by frequency,amplitude and phase.Therefore,the robust multivariate regression model is established based on the robust regression theory to improve the accuracy of inversion.The experimental results show that the robust multivariate regression model can effectively improve the accuracy of model inversion and root mean square error RMSE(Root Mean Squared Error)decrease 54%;The average decision factor R~2 was increase 35%.Through the research of the above three aspects,the problem of the accuracy of GNSS-IR inversion affected by error data is solved.The inversion accuracy is improved by data fusion method,which provides a reliable basis for the application of GNSS-IR in water cycle,meteorology and agricultural production.