Research On Soil Moisture Retrieval Based On GNSS-IR Considering Correction And Repair Of Multisatellite Track Abnormal Phases

Soil moisture is a crucial parameter for predicting floods and droughts,simulating weather and climate,and optimizing agricultural management.It plays a significant role in fields such as agriculture,climate,and water cycling.Global Navigation Satellite System Interferometric Reflectometry(GNSS-IR)is a remote sensing technique that utilizes the modulation term in the Signal-to-Noise Ratio(SNR)to monitor surface soil moisture.It offers advantages such as all-weather capability,continuous monitoring,multiple signal sources,and wide coverage.In view of the shortcomings of non-linear least squares method in estimating GNSS satellite orbital phases,and the problem of optimal combination of multi-satellite orbits,this paper conducted research on correction of full-cycle abnormal phases for different satellite orbits,detection and repair of non-full-cycle abnormal phases,and studied and constructed a soil moisture retrieval method based on a combination of multi-satellite orbits.The main research findings and conclusions are as follows:(1)In order to solve the problem of abnormal jumps that are very likely to occur when solving the phases by non-linear least squares,a SNR period method is proposed based on the periodicity of the standard cosine function.The experiments show that the method can effectively correct the phase with full-cycle or half-cycle jumps for each satellite orbit.The correlation coefficient(R)between the phases and soil moisture for each satellite orbit after correction is significantly improved,and the average value of R for all satellite orbits increases from 0.502 to 0.724(44.2% improvement).It was also found that for the abnormal phases caused by low quality SNR,the jump size tends to be lower than half a cycle.These anomalous phases can be effectively detected and repaired with the interquartile range and moving average filter,thus further improving the quality of the phases of each satellite orbit.(2)To analyze the effect of different single-satellite orbits on soil moisture retrieval,a single satellite orbit linear regression retrieval model was constructed based on the corrected and repaired satellite orbit phases.The experiments show that the model retrieval results have stronger stability relative to the original satellite orbits when soil moisture is retrieved using the corrected,detected and repaired single satellite orbits.In addition,satellite orbits with higher R are usually able to provide better soil moisture.(3)The key to multi-satellite orbit combination for soil moisture retrieval is the reasonable selection of satellite orbits.Based on the above research results,a GNSS satellite orbit selection method based on Random Forest(RF)assistance is proposed,and a multi-satellite orbit linear regression retrieval model is constructed.The experiments show that the importance analysis of RF can provide an effective reference basis for multi-satellite orbit combination.For soil moisture retrieval,when the number of satellite orbit combinations reached four,the correlation coefficient,root mean square error and mean absolute error between the model retrieval results and the reference values improved by 7.9%,14.7% and 11.1%,respectively,relative to the optimal single-satellite orbit.When the number of available satellite orbits is large,using RF to select the top 15 satellite orbits for combination can obtain higher accuracy of soil moisture.