Improving Groundwater Imputation through Iterative Refinement Using Spatial and Temporal Correlations from In Situ Data with Machine Learning

Obtaining and managing groundwater data is difficult as it common for time series datasets representing levels at wells to have large gaps of missing data. To address this issue, many methods been developed infill or impute the We present a method improving imputation through an iterative refinement model (IRM) machine learning framework that works on any aquifer dataset where each well has complete record can be mixture measured input values. This approach corrects imputed values by using both in situ observations from nearby wells. relied idea similar experience environment (e.g., climate pumping patterns) exhibit changes levels. Based idea, we revisited every “re-imputed” (i.e., had previously imputed) similar, repeated process predetermined number iterations—updating synchronously. Using IRM conjuncture with satellite-based provided better generated could provide valuable insight into behavior, even when limited no were available individual applied our Beryl-Enterprise Utah, gaps. found patterns related agricultural drawdown long-term drying, potential evidence multiple unknown aquifers.