| Soil moisture refers to the humidity in the soil of the cultivated layer of crops that are suitable for the growth and development of plants.It plays an important role in climate change,energy balance,and the global water cycle.Monitoring soil moisture can better grasp crop growth,predict crop yield,and other agricultural applications,and directly affect crop growth and development.It is a characterization factor of soil moisture.Nong’an County is located in the black soil region of northeast China,with a fertile black soil layer and high organic matter content.It is also a supply base for commercial grains such as corn and soybean,but with low perennial precipitation and low water resource utilization.Against the background of severe degradation of black soil resources,the study combined Sentinel-1SAR data and Sentinel-2 optical data from the same period as data,using improved water cloud models and neural networks,Using the measured soil moisture data as validation data,the inversion study of soil moisture in the farmland area of Nong’an County was conducted.The main conclusions of this paper are as follows:(1)In vegetation covered areas,surface vegetation has a high degree of attenuation of microwave backscatter signals.The study found that the average influence value on the backscatter coefficient of VV and VH polarized images using NDVI as input parameters increased by 0.57 d B and 0.92 d B,respectively,compared to NDMI as input parameters after improved water cloud model calculation;Through comparison,it can be seen that the NDVI index is more suitable as an input parameter for the water cloud model.After calculating the improved water cloud model using NDVI as a parameter,the VH and VV polarization R~2 of the soil backscatter coefficients are 0.39 and 0.50,respectively;Compared with the fitting of the total backscatter coefficient,it increases by 0.22 and 0.21,respectively.Therefore,the soil backscatter coefficient calculated by the improved water cloud model can better represent the changes in soil moisture.(2)The GA-BP neural network with feature selection is more suitable for large-scale soil moisture retrieval.Using the improved water cloud model to obtain relevant microwave remote sensing images and optical remote sensing characteristic data after removing the impact of vegetation,as input parameters of the neural network model,a neural network model for retrieving soil moisture was constructed.From the prediction results,it can be observed that the optimal prediction results obtained by linear fitting were fitted with the measured water content,with a regression correlation of R~2=0.60,and a root mean square error of RMSE=2.29.Among the schemes with the highest accuracy in the neural network group,the bias(bias)is 0.0915,the correlation(R~2)is0.74,and the root mean square error(RMSE)is 1.49.However,after feature selection and dimensionality reduction,the prediction accuracy of the GA-BP neural network model decreases slightly.The bias(bias)of the validation results increases by 0.0167,the correlation(R~2)decreases by 0.03,and the root mean square error(RMSE)increases by 0.19,but at the same time,the model operation time,the space occupation,and network complexity are greatly reduced,which can meet the high inversion accuracy requirements while improving the inversion speed and reducing data redundancy.(3)The validation of different soil moisture products shows that the spatial distribution of soil moisture retrieved using this research method is highly reliable.Utilizing the SMAP/Sentinel soil moisture product map in the study area,the inversion results in this paper were verified,and the correlation result was R~2=0.58.According to the soil moisture level map of the research area obtained from the final inversion,38%of the surface soil moisture in the agricultural cultivated area of Nong’an County is between 9%and 15%,belonging to a mild drought state,and 61%of the surface soil moisture is between 15%and 20%,belonging to a suitable state of soil moisture.The regional distribution of farmland soil moisture at various levels in the study area is consistent with the actual situation. |
PDF Full Text Request