Illuminating groundwater flow modeling uncertainty through spatial discretization and complexity exploration

Abstract

Uncertainty in groundwater modeling presents a significant challenge, originating from various sources. This groundbreaking study aims to quantitatively assess uncertainties arising from spatial discretization and complexity dynamics. The research focuses on the Najafabad Aquifer in Esfahan, Iran, as a compelling case study. Five distinct conceptual models were developed, with parameter counts of 16 (model 1), 20 (model 2), 22 (model 3), and 26 (model 4 and 5), and subjected to a consistent spatial discretization of 500 m. Additionally, two alternative models with spatial discretizations of 250 m (model 1a) and 1000 m (model 1 b) were introduced based on the least complex model with 16 parameters. The study comprehensively examines groundwater uncertainty by manipulating spatial discretization while considering complexity dynamics. Model Muse facilitates simulation, and UCODE is utilized for calibration using observed hydraulic head data. Uncertainties are explored using Bayesian model-averaging (BMA) and model selection criteria. Comparing probabilities of the initial five models reveals increasing uncertainty with a greater number of parameters (KIC in model 1: 99.25%, model 2: 0.41%, model 3: 0.34%, model 4 and 5: 0%). Investigation of seven alternative models highlights the dominant influence of coarser spatial discretization on groundwater modeling uncertainty. Remarkably, despite the lowest complexity in model 1 with probability of 99.25%, the model with coarse spatial discretization (model 1b) exhibits the zero probability (KIC in model 1a: 93.42%, model 1: 6.53%, model 1b: 0%, model 2: 0.03%, model 3: 0.02%, model 4 and 5: 0%.). Thus, considering optimal parameter count and spatial discretization size is crucial in conceptual model development. This study pushes the boundaries of understanding the intricate relationship between spatial discretization, complexity, and groundwater modeling uncertainty. Findings hold significant implications for improving model accuracy and decision-making in hydrogeological studies.