Projects

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.

Water markets can support drought adaptation by reallocating scarce water across users, sectors, and locations, but institutional and physical barriers may limit their effectiveness. This project examines water markets across twelve western U.S. states by combining empirical analysis of transaction records with economic simulations of agriculture-to-urban water transfers. Transactions are linked to drought, hydrologic, climatic, demographic, and agricultural conditions to evaluate how market participation, transfer activity, prices, and transaction values respond to water scarcity. A complementary simulation framework integrates water rights, water use, and demand relationships to estimate potential welfare gains and their distribution across regions and sectors under alternative trading and water availability scenarios. Sensitivity analysis assesses how uncertainty in data and behavioral parameters influences these estimates. Together, these approaches examine whether markets respond to drought as expected, who benefits or bears costs, and how market design and trading barriers shape the potential of water trading as an adaptation strategy.

Water shortages across the western United States are driven both by natural factors, such as temperature and precipitation, and by human factors, such as population density, irrigation efficiency, cropland, and water-use intensity. This project uses the University of New Hampshire Water Balance Model (WBM) to quantify how each of these drivers contributes to water shortage. Because full sensitivity analyses of a process-based model are computationally impractical, I developed a deep learning emulator that captures WBM behavior with high fidelity and far greater efficiency, and then applied multiple explainable artificial intelligence (XAI) methods to quantify each input’s contribution. Different XAI methods can disagree even for the same model and data, so I combined their results into consensus-strength maps that show where the methods agree on the main drivers, and quantified how unstable and uncertain these attributions are.

The Laurentian Great Lakes and St. Lawrence River basin comprises the largest freshwater system on Earth, containing about one-fifth of the world’s surface fresh water. However, the Great Lakes basin has recently experienced rapid shifts between highs and lows in some water balance components. To resolve the regional water budget in the Great Lakes over an extended historical period, and provide insight moving forward in our changing climate, we applied the Large Lake Statistical Water Balance Model (L2SWBM) to quantify uncertainty and reconcile discrepancies between model- and measurement-based estimates of each water balance component from various datasets. The L2SWBM contextualizes and reduces uncertainty while closing the water balance over consecutive historical periods. The model assimilates multiple datasets for each hydrologic component (i.e., over-lake precipitation, over-lake evaporation, runoff, connecting channel flows, and inter-basin diversions) and runs millions of iterations to reconstruct potential historical water budgets. Using observed and modeled water balance data across the historical record, the L2SWBM iteratively solves for coefficient values and estimates a range of reasonable uncertainties in individual components that remain faithful to the water balance. After applying the L2SWBM, uncertainty in the Great Lakes water balance was significantly reduced by approximately 10–40%, depending on original estimates.

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, including the following key components:

Groundwater plays a critical role in the water balance; however, the groundwater component of the hydrologic cycle is frequently overlooked at basin scales because it is difficult to observe and quantify. We address this problem through a novel framework that combines existing hydrological models and datasets with groundwater flux estimates across Earth’s largest system of lakes, the Laurentian Great Lakes. Beyond serving as a template for combining surface water and groundwater data and models, the Great Lakes are a compelling test case: they recently transitioned from a period of water scarcity (water levels were persistently below average from 1998 through 2013) to extreme water abundance (all-time high water levels were set in 2017 and 2019). Throughout this transition, we know of no comprehensive data record or modeling system that explicitly links changes in observed land surface hydrology with critical subsurface groundwater processes. In addition to using our framework to combine existing surface water and groundwater datasets and model simulations, we will demonstrate the impact of groundwater representation in existing lake physics models that serve as the basis for understanding lake evaporation, circulation patterns, and pollutant fate and transport dynamics.

The Laurentian Great Lakes represent the largest system of lakes on Earth and contain roughly 20% of the world’s fresh surface water. Our team leads research aimed at understanding short- and long-term changes in Great Lakes water level variability, including pathways through which climate change impacts the major components of the Great Lakes water balance. This research extends into other Great Lakes regional projects, including model and dataset development for water quantity and quality management.