Author(s): Mohammad Zeynoddin; Silvio Jose Gumiere; Hossein Bonakdari

Linked Author(s):

Keywords: Machine Learning; SOILPARAM; Water Table Depth; ELM; Feature Selection

Abstract: The SOILPARAM application, developed using Google Earth Engine (GEE), offers a practical solution for accessing hydroclimatic data, enhancing the accuracy of water table depth (WTD) predictions. This study examines the use of the Extreme Learning Machine (ELM) for modeling WTD based on data retrieved through SOILPARAM, including meteorological data, soil moisture, and temperature. Feature importance was analyzed using Local Interpretable Model-Agnostic Explanations (LIME) to refine predictions, identifying key variables such as precipitation and soil moisture that most influence WTD. Results show that ELM achieved an R-value of 0.843 before feature analysis and an RMSE of 6.316 cm. Post-LIME, although the correlation decreased slightly, the prediction errors were further minimized. The findings illustrate the real-world application of SOILPARAM in improving WTD prediction accuracy supporting water resource management efforts by simplifying data access and modeling processes.

DOI:

Year: 2024