A Comparative Analysis of SMAP-Derived Soil Moisture Modeling by Optimized Machine Learning Methods: A Case Study of the Quebec Province

Abstract

Many hydrological responses rely on the water content of the soil (WCS). Therefore, in this study, the surface WCS products of the Google Earth Engine Soil Moisture Active Passive (GEE SMAP) were modeled by a support vector machine (SVM), and extreme learning machine (ELM) models optimized by the teacher learning (TLBO) algorithm for Quebec, Canada. The results showed that the ELM model is only able to forecast 23 steps with Correlation Coefficient (R) = 0.8313, Root Mean Square Error (RMSE) = 6.1285, and Mean Absolute Error (MAE) = 5.0021. The SVM model could only estimate the future steps, one step ahead, with R = 0.8406, RMSE = 18.022, and MAE = 17.9941. Both models’ accuracy dropped significantly while forecasting longer periods.

Keywords

• teacher learner
• optimization
• ELM
• SVM
• LSTM
• forecast