Deep Residual Network and Water Cloud Model-Based Soil Moisture Retrieval Using Satellite Images

Soil moisture is a critical part of the link between land carbon cycle and surface-ground water motion. It also is vital in energy exchange between land/atmosphere; furthermore, soil moisture is an important determinant process for plant growth and productivity. By making the technique use satellite imagery, it is able to predict soil moisture content more accurately rather than using vegetation indices like the "Wide Dynamic Range Vegetation Index (WDRVI), Simple Ratio (SR), or Green Leaf Area Index (GLAI)". A Deep Residual Network (DRN)-based "water cloud model "for soil moisture" forecasting is generated according to the vegetation indices. Additionally, soil moisture is derived using vegetation indices such as the "Wide Dynamic Range Vegetation Index (WDRVI), Simple Ratio (SR), and Green Leaf Area Index (GLAI)". Performance of the developed DRNbased "water cloud model "was also evaluated in terms of RMSE and estimation error. The objective of this research work is to come up with a new modeling approach of soil moisture retrieval by improving "water cloud model "with satellite images and vegetation indices in one hand, and consequently to verify its performance through the application of Deep Residual Network (DRN)-based "water cloud model "on another hand. According to the experimental results, the suggested DRN-based "water cloud model "performed better than the others, with the smallest estimate error and Root Mean Square Error (RMSE) values of 0.523 and 0.69, respectively. The developed method's estimation error is lower than that of the Semi-Empirical Water Cloud Model, Deep Multi Model Fusion Network, Genetic Algorithm (GA), Convolution Neural Network and Regression (CNNR), and Particle Swarm Optimization (PSO), respectively, at 94.14%, 65.32%, 91.63%, 63.12%, and 91.10%. The developed scheme's RMSE is lower than that of the CNNR, PSO, Genetic Algorithm GA, Deep Multi Model Fusion Network, Semi Empirical Water Cloud Model, and 93.52%, 90.33%, 93.11%, and 89.72%, respectively.