| Soil water content monitoring is an important research content of agricultural water resources planning and drought warning,and can provide basic data support for the efficient use of water resources.With the maturity of remote sensing technology,multi-sensor multimethod remote sensing inversion of soil water content has become one of the main methods of soil water content monitoring.Based on this,many soil moisture content products have been produced,but the existing soil moisture content products have high time resolution and generally low spatial resolution.With the increase of high spatial resolution remote sensing sensors,the production of soil moisture content products with high spatial and temporal resolution becomes possible.At present,there are many high-resolution optical sensors that can support the production of more continuous soil water content data,but optical sensors are susceptible to weather,which is an inherent disadvantage of optical sensors.Compared with optical sensors,microwaves are not easily affected by clouds,but there are relatively few highresolution microwave sensors currently available.Therefore,how to integrate the soil water content data of different sensors into time series has become a hot topic in the research direction of soil water content.One of the problems.However,due to the low accuracy of the soil water content inversion model,the unclear differences between the soil water content data of different sensors,and the immature fusion method,the research on the construction of high-resolution soil water content data time series is relatively Therefore,this article intends to explore a highresolution,time-continuous soil water content construction method suitable for this study area.In this paper,the southern part of Hebei Province is used as the research area,GF-1 and GF-6 are used as the data source for optical soil water content inversion,and Sentinel-1 is used as the data source for microwave soil water content inversion,and the limit of soil water content is carried out.Learning model inversion.First,the input of the microwave soil moisture content model is calibrated;for the optical input,this paper selects multiple factors that affect the soil moisture content as sample input,selects the corresponding evaluation index of the inversion result,and compares the microwave and optical soil moisture content The inversion was evaluated.Secondly,choose the corresponding difference evaluation index to analyze the difference between optical and microwave soil water content data.Finally,the corresponding error elimination method is used to eliminate the difference between optical and microwave soil moisture data,and realize the fusion of high-resolution optical and optical,optical and microwave soil moisture data.The main conclusions of the article are as follows:(1)(1)Using the extreme learning machine model to construct the microwave soil water content inversion model,the VV polarization,VH polarization and the polarization difference DIF related to the surface roughness calculated by the water cloud model were combined according to different combinations Enter the model and determine through a series of contrast ratios.It is found that when the input sample is VV+VH+DIF,the microwave soil moisture content inversion result based on the extreme learning machine model is the best,and the correlation coefficient can reach 0.75.(2)Using the extreme learning machine model to construct the optical soil water content inversion model,using b1,b3,b4,MSAVI,PDI and DEM as the input of the neural network model in this paper,and the soil water content as the output,to carry out the optical data of this study area Inversion of soil water content.The experimental results show that under the multi-input of this paper,the correlation coefficients of the optical soil water content inversion model in this paper are all higher than 0.75.(3)Comparing the soil water content inversion results of GF-1,GF-6 and Sentinel-1,it is found that the correlation coefficients of the soil water content inversion results based on GF-1 are better than those of GF-6 and Sentinel-1.The inversion results,so in the follow-up fusion method research,this article will use GF-1 as the benchmark to convert GF-6 and Sentinel-1soil water content data to GF-1 soil water content data.(2)Select the mean square error MSD,MPDu,MPDs,MPDu/MSD and MPDs/MSD and other different evaluation indicators to quantitatively evaluate the data of different sensors.There are the following conclusions:(1)There is a high consistency between GF-1 and GF-6soil water content data.The evaluation results of error indicators show that the systematic error between GF-1 and GF-6 soil water content data accounts for the main part.(2)The consistency calculation results of GF-1 and Sentinel-1 soil water content data are low.The calculation results of the error index of GF-1 and Sentinel-1 soil water content data show that the difference between GF-1 and Sentinel-1 soil water content data is Time non-systematic errors account for the main part.(3)Based on the analysis of error characteristics,this study found that the systematic error between GF-1 and GF-6 soil water content data accounted for the main part,so for the systematic error,this paper uses the least squares regression model OLS to convert.After analyzing the difference between GF-1 and Snetinel-1 soil water content data,it is found that the main part of the non-systematic error station,because OLS is only a one-way conversion model,and its assumption is that the converted data is an unbiased calculation,so After comparing the inversion results of optical and microwave soil moisture content,this paper converts the inversion results of microwave soil moisture content and eliminates the nonsystematic errors in the microwave soil moisture content data by selecting Frost filter and 5*5filter window.(4)The sample data was selected and the fusion method proposed in this paper was verified.It was found that the elimination of a series of errors based on regression model conversion and filtering can realize the conversion from GF-6 to GF-1 and from Sentinel-1 to GF-1 conversion;finally selected five measured stations to apply the time series soil water content construction method of this article.The results show that the time series soil water content constructed based on the fusion method of this article has a good relationship with the measured soil water content data.Consistency,high-resolution time-continuous soil water content data suitable for this study area can be constructed. |
PDF Full Text Request