Groundwater level prediction using hybrid ML: Bridging the gap between nature-based solutions and nature-inspired algorithms

Effective groundwater management requires accurate prediction of GroundWater Level (GWL) fluctuations. This study proposes six novel regression algorithms by integrating a Fuzzy Inference System (FIS) with six Nature-inspired Algorithms (NiA) to enhance GWL prediction accuracy. The proposed algorithms contribute to several Nature-based Solutions (NbS) goals, including improving water security by ensuring that groundwater resources are used sustainably and helping to ensure that people have access to clean and safe water. In this study, we coupled FIS with Invasive Weed Optimization (IWO), Ant Colony Optimization (ACO), Teaching-Learning-Based Optimization (TLBO), Differential Evolution (DE), Harmony Search (HS), and Weevil Damage Optimization Algorithm (WDOA). We used precipitation, relative humidity, and groundwater level lag as potential input features to predict the groundwater level. We found that the Fuzzy-IWO-GWL model accurately predicts GWL fluctuations, achieving a high correlation coefficient (R = 0.89), low normalized root mean square error (nRMSE = 0.18), and minimal bias (bias = 0.08). A comparative analysis involving eleven benchmark algorithms (consisting of standalone and deep learning algorithms) reveals the superior performance of the proposed algorithm. This study highlights the potential of Nature-Based Solutions and nature-inspired algorithms in groundwater management applications, providing valuable insights for policymakers and stakeholders involved in ensuring groundwater sustainability.