Publications

Explainable artificial intelligence can help interpret deep learning models of environmental processes, but the reliability of these explanations remains difficult to assess. Models with similar predictive skill may produce different explanations, and agreement across methods does not necessarily indicate accurate recovery of the underlying relationships. We propose using the Canadian Fire Weather Index System as a controlled benchmark to investigate the sources of explanation uncertainty and the role of correlated feature proxies. Its known mathematical structure provides a reference for assessing whether explanations recover the dependencies used to calculate the index. We will examine how explanation variability changes across training runs, model designs, interpretation methods, and input dependence. The study will test whether training randomness dominates explanation uncertainty and whether stable explanations accurately reflect the reference calculation. The findings will inform how predictive performance, explanation reproducibility, and fidelity should be evaluated when interpreting deep learning models in environmental applications.

Growing water scarcity increases the need to understand how water markets can improve allocation across competing uses. This study develops a market simulation framework to assess the economic welfare implications of water trading in the western United States. The framework combines information on water rights, water use, and market prices with economic demand relationships to estimate potential gains from transfers between agricultural and urban users. Alternative scenarios examine how trading opportunities and water availability shape these gains and their distribution across regions. Sensitivity analysis evaluates how uncertainty in input data and behavioral parameters affects estimated outcomes. The analysis provides a basis for distinguishing opportunities for beneficial reallocation from outcomes that depend strongly on modeling assumptions. By linking regional water conditions to economic incentives, the project supports assessment of market design and highlights the importance of considering both aggregate efficiency and the distribution of benefits when evaluating alternative water management strategies.

Explainable Artificial Intelligence (XAI) methods are increasingly used to understand the predictions made by machine learning (ML) models in fields such as hydrology. However, there is a lack of clear guidance about which methods are most appropriate for a given analysis. These complexities result in uncertainties about explanations. We illustrate these uncertainties using a synthetic ground-truth benchmark and a perfect model experiment for attributing water scarcity across the Western United States. Identifying drivers of water scarcity is critical for water-resource management but complicated by interactions between meteorological variability and human-system uses. We train an ML emulator on simulated unmet demand from a process-based hydrological model, then use eight XAI methods to interpret its predictions. We identify two distinct sources of uncertainty: explanation uncertainty from different XAI methods and numerical instability from explanation sensitivity to training seed. These findings show that XAI methods should be selected intentionally and explanations tested for uncertainty before being relied upon.

Ensuring continued access to freshwater is an essential part of managing consumptive use, withdrawals and diversions in the Great Lakes system. This is primarily accomplished through the Great Lakes Compact Agreement, administered by the Great Lakes and St. Lawrence Governors and Premiers’ Regional Body and Compact Council. We used the latest findings from the Large Lake Statistical Water Balance Model to better understand historical trends in precipitation, evaporation, runoff and inter-lake flow in support of this Cumulative Impact Assessment.

Groundwater systems are intrinsically heterogeneous with dynamic spatio-temporal patterns, which cause significant challenges in quantifying and mapping their complex processes. However, accurate forecasting of regional groundwater contamination is commonly needed to identify its spatio-temporal dynamics, which helps the public anticipate the timing and severity of potential groundwater quality issues and could serve as an early warning system. This study focuses on modeling a plume of 1,4-dioxane originating from the Gelman site beneath the city of Ann Arbor, Michigan. It proposes a novel methodology that accounts for the spatially and temporally irregular and uncertain nature of groundwater contamination data to analyze historical trends in dioxane concentration and predict its transport: 1. A random forest interpolation model was deployed to fill or extend gaps in fragmented time series across all monitoring wells; 2. The Mann-Kendall test was applied to evaluate trends in dioxane concentrations at various wells; 3. An automated time series machine learning (AutoTS) package was used to forecast future concentrations; and 4. An R-based Shiny web application was designed to allow visualization and quantification of dioxane contamination analytical data. This research introduces a novel framework for filling spatial and temporal data sampling gaps in groundwater contamination to offer an effective and promising way to predict future plume concentration and spatial distribution.