Estimation of Surface Soil Moisture Using a Structural Equation Model and an Artificial Neural Network (SEM-ANN) Combined Method

Abstract: Soil moisture is an important parameter affecting environmental processes, such as hydrology, ecology and climate. However, due to the influence of soil type, soil structural conditions, topographic features, vegetation environment and human activities, the distribution of soil water content is characterized by spatial heterogeneity. It is currently also very difficult to accurately monitor soil moisture distribution over larger areas. In order to investigate the direct or indirect influence of various factors on soil moisture for better soil moisture inversion results, structural equation models were used to determine the structural relationships between soil moisture influencing factors and the degree of influence. These models were then transformed into the topology of Artificial neural networks (ANN). Finally, a structural equation model coupled with an artificial neural network was constructed as an SEM-ANN soil moisture inversion model. Results showed that: (1) the most important predictor of spatial variability of soil moisture in the dry season was the temperature vegetation drought index; surface temperature was the most important predictor in the wet season; (2) the SEM-ANN model recorded R2 values of 0.87 and 0.82, RMSE values of 7.68 and 8.63 for the training set (R2 of 0.85 and 0.81), and an RMSE of 8.03 and 7.72 for the validation set in the dry and wet seasons, respectively. These values were higher than those obtained using the ANN model without a SEM structure; (3) No significant differences in soil moisture distribution trends were recorded between the dry and wet season. Low soil moisture content was recorded in the northwest-southeast stripes, having low vegetation cover in the dune area and strong surface evapotranspiration. Results for the interdune beach area recorded a high soil moisture content due to the high vegetation cover and good soil water-holding capacity