| Soil moisture refers to the water content in soil,which has an important impact on plant growth,ecological health,climate,environment,and water cycle.Traditional methods for measuring soil moisture have shortcomings such as high cost,limited applicability,and vulnerability to environmental impacts,which pose great difficulties for data collection in long time series and large areas.In recent years,more and more studies have shown that GNSS-IR(Global Navigation Satellites System Reflection Interchange Reflection Entry)technology can accurately retrieve soil moisture,and has high spatiotemporal resolution and low cost and operational difficulty.This technology can calculate the reflection coefficient and humidity of soil surface by measuring the phase delay and amplitude change of GNSS signals reflected from the ground.Because GNSS signals can penetrate clouds,rain,snow,leaves,and other obstructions,GNSS-IR technology can measure soil moisture under various weather conditions.Today,this technology has been widely used worldwide,with some achievements in monitoring and predicting drought,floods,land degradation,and crop production.However,GNSS-IR technology also has some drawbacks.When retrieving surface parameters,there are often negative impacts caused by noisy signals,rough surface,vegetation cover,and other environments,resulting in unsatisfactory retrieval accuracy.In response to this phenomenon,this paper proposes a soil moisture retrieval method that combines Complete EEMD with Adaptive Noise(CEEMDAN)and BP neural network models,aiming to minimize negative impacts and improve retrieval accuracy.The specific experimental research contents are as follows:1.This article briefly summarizes the current development status of GNSS systems and the application research of GNSS-R technology.It provides a detailed introduction to the concepts of multipath error interference and signal-to-noise ratio,as well as the relationship between the two.It analyzes the relevant characteristics of the Fresnel reflection zone and the motion trajectory of satellite reflection points.It also provides a detailed introduction to the basic mathematical principles of the CEEMDAN decomposition method,explains the basic steps and practical methods of constructing a BP neural network model,and a functional relationship of accuracy evaluation indicators was provided to evaluate the accuracy of the model inversion results.2.This paper downloaded the observation file of P043 station in 2015,and screened the effective satellite,altitude angle,azimuth angle and other information.The satellite retrieval results after screening are roughly in line with the trend of the true value of soil moisture,but the retrieval accuracy is not ideal.Therefore,this paper introduces CEEMDAN decomposition to obtain more accurate reflection signal components,and compares the results of quadratic polynomial fitting.The results show that CEEMDAN has certain advantages in separating direct reflection signals.3.Aiming at the noise caused by complex surface environment and vegetation and trees,a multi-satellite fusion soil moisture prediction method based on CEEMDAN-BP was proposed.The experiment shows that the inversion results based on CEEMDAN-BP multi-satellite model are generally consistent with the measured data,with the root mean square error of 0.016 and the determination coefficient of R~2 reaching 0.934,which proves the feasibility and reliability of this method. |
