Soil Moisture Retrieval Using Sail Squirrel Search Optimization-based Deep Convolutional Neural Network with Sentinel-1 Images

Abstract

Soil Moisture (SM) is an environmental descriptor, which acts as the affiliation between the atmosphere and the earth’s surface. Various SM retrieval methods are developed to abolish the influence of vegetation cover attenuation, surface roughness, and scattering to find an association among SM and backscatter coefficient. To understand the relationship between various vegetation parameters and backscatter coefficient poses a great challenge in SM retrieval. Hence, an efficacious SM retrieval method is afforded using the proposed Sail Squirrel Search Optimization-based Deep Convolutional Neural Network (SSSO-based Deep CNN). Here, the proposed SSSO is derived by concatenating the Sail Fish Optimization (SFO) with Squirrel Search Algorithm (SSA). The Deep CNN performs the process of SM retrieval using vegetation indices. The fitness measure of the proposed optimization enables to find the best solution to update the weights of the classifier for increasing the efficiency of the retrieval mechanism. By training Deep CNN with the proposed optimization, the soil moisture of an area is effectively retrieved. However, the proposed SSSO-based Deep CNN obtained minimal estimation error and minimal RMSE of 0.550 and 0.726 using sentinel-1 data, respectively.