| Soil moisture,as a key indicator of the global water cycle,is the focus of research in modern agriculture,meteorology,and hydrology.Accurate and real-time acquisition of soil moisture is an important reference value for agricultural irrigation,meteorological forecasting,and water circulation.GNSS-IR as a new microwave remote sensing technology,emits L-band signals that are very sensitive to soil moisture,and it mainly uses GNSS direct It mainly uses the interference effect between the GNSS direct signal and the signal reflected by the ground surface at the receiver antenna to invert the physical parameters of the ground surface through the interference signal characteristics,which is widely used in various fields of remote sensing monitoring and is gradually becoming an important branch of GNSS remote sensing technology.Due to the low accuracy and reliability of the existing GNSS-IR soil moisture inversion models and the difficulty to avoid the error caused by seasonal vegetation growth,many problems of soil moisture inversion by GNSS-IR technology need to be solved and improved,so the implementation of multi-star data fusion GNSS-IR soil moisture inversion taking into account the influence of vegetation is a problem to be solved nowadays.The research results will be of reference significance to the development of GNSS-IR land surface remote sensing.This thesis starts from the basic theory of the mathematical and physical properties of the reflection signal,and conducts research on the single-antenna GNSS-IR soil moisture inversion model and the mechanism of the influence of the vegetation error term on the reflection signal,and finally obtains the GNSS-IR soil moisture inversion model applicable to the correction of the vegetation moisture content perturbation.On the one hand,this paper proposes a dual-frequency data fusion vegetation error correction method based on the entropy value method,which overcomes the limitations of accuracy and reliability of the NDVI inversion using only single-frequency data,and solves the problem of phase correction of the reflection signal in the absence of measured vegetation water content data.The method introduces the entropy method to measure the uncertainty of multi-path data calculated by GPS L1,L2 dual-frequency pseudorange and carrier phase observations,evaluates the information entropy of each index,reverses the weighting to achieve the optimal weighted fusion and calculates the NMRI to improve the accuracy and reliability of NDVI.In order to ensure the consistency of the mathematical scales of the phases of NDVI and reflection signals,the calculated values of the median zeroing process of NDVI in the first 15% of the annual time series are approximated instead of the phase shift caused by vegetation,and finally the above algorithm is verified by using the observation data from four stations of PBO in the United States.On the other hand,this thesis carries out in-depth research on soil moisture inversion model based on GNSS-IR multisatellite data fusion,with a view to solving the problem of low degree of automated combined optimal inversion soil moisture satellites of existing multisatellite soil moisture inversion algorithms,and proposes a MARS soil moisture inversion model is proposed.The method is divided into three steps: forward stepwise,backward pruning,and model selection to adaptively extract the satellite combination that can reflect the optimal soil moisture.In addition,to verify the accuracy and reliability of the MARS model,a comparison analysis with BP neural network,support vector regression machine,and multiple linear regression soil moisture inversion model was conducted.For the vegetation error correction problem of dual-frequency data fusion,this paper carries out experimental analysis by four sites of PBO observation data,and the experimental results show that(1)the correlation of NDVI inversion model after dual-frequency data fusion is 0.828 0.826,0.800,and 0.816 at four sites,and the root mean square error is 0.052,0.037,0.050,and 0.064,and their correlations increased by 9.4%,28.5%,17.6%,and 7.4%,respectively,and the root mean square errors decreased by 17.5%,15.9%,16.6%,and 15.8%,respectively,compared with the conventional L1 carrier model.(2)The correlation between the soil moisture inversion results and the actual values after correcting the vegetation error term was improved at all four sites,and the correlation between the four model inversion results and the actual values before and after the correction increased by 7.4% on average.For the problem of soil moisture inversion models with multi-star data fusion,a comparative analysis of the four models was carried out using the PBO experimental data,and the experimental results showed that(1)the correlations of the MARS soil moisture inversion models reached 0.930,0.912,0.903,and 0.921 at the four stations P036,P037,P041,and P049,compared with the uncorrected vegetation moisture content The correlations of 0.836,0.823,0.842,and 0.844 increased by 11.2%,10.8%,7.2%,and 9.1%,respectively,and the RMSE decreased by 47.7%,45.5%,26.9%,and 37.9%,respectively.(2)The correlations of the three soil moisture inversion results of BPNN,SVRM,and MLR ranged from 0.803 to 0.874 after correcting the reflected signals at four stations,and none of them was better than the MARS soil moisture inversion model.Compared with the traditional methods,this study can improve the accuracy,reliability and generalization ability of the surface soil moisture inversion model in vegetation covered areas,and can provide typical reference data,basic theories and supporting technologies for research work related to climate change,precision agriculture,droughts and floods. |
PDF Full Text Request