2 Advances in characterizing groundwater, surface water, and their interactions

This session focuses on characterizing groundwater, surface water, and their interactions through fieldwork, modeling, and data synthesis. We welcome submissions addressing flow, transport, water quality, and water levels under both typical and extreme conditions.

Sunday, 22 March 2026
8:00 AM - 12:00 PM
CCC, Room 25

From viable to valuable: Iterative advancements in national stream temperature modeling

Paper number: 2-4
Program: Technical Sessions
Time: 9:05 AM – 9:25 AM

Authors

Abstract

The U.S. Geological Survey recently developed a machine-learning model to predict daily minimum, mean, and maximum water temperature for 1979 to 2021on more than 50,000 stream reaches across the conterminous U.S. (CONUS). This model was designed to support spatially extensive water availability assessments. To expand the value of the model and make it performant for riverine fisheries management, local water assessment, public recreation, and more, advancements are needed, with high priority on: (1) the predictive accuracy in reaches with strongly groundwater-influenced temperature regimes and (2) the spatial resolution of the stream network.

To improve predictive performance in reaches with strongly groundwater-influenced temperature regimes, we are strategically adding input features to enhance the limited set in the initial model. Specifically, we are identifying catchment attributes that are correlated with quantitative metrics of stream thermal regime as indicated by the relationship between air and water temperature. Potential features include indicators of (1) subsurface properties, such as soil texture or bedrock type, (2) drainage network properties, such as stream density, (3) hydrologic condition, such as mean depth to water, and (4) anthropogenic influences, such as irrigation and impervious cover. We will discuss prioritized features to add and preliminary effects on model performance.

To refine the spatial resolution from 50,000 stream reaches to more than 2,600,000 reaches across CONUS, we developed multi-scale downscaling methods that leverage both the national model and additional local data to predict daily water temperature across local watersheds. We tested the methods in eight watersheds (500 to 1000 km2) across CONUS with different hydrologic settings and under realistic data availability scenarios. Especially when local data were scarce, multi-scale methods generated more accurate predictions than purely national or purely local methods.

By taking an iterative and testing approach to model development, with stable core datasets and a baseline viable model to which we can compare changes, we are able to quantify the improvements over iterations. This approach provides a clear record of progress and a greater ability to identify thresholds where major breakthroughs may be required to continue improving.

Geological Society of America Abstracts with Programs. Vol. 58, No. 2

doi: 10.1130/abs/2026NE-411323

© Copyright 2026 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.