5 Advances in Mountain Hydrology: Connecting Cryosphere, Surface, and Subsurface Processes
We aim to highlight studies addressing the spatial and temporal complexity of mountain hydrology across diverse climatic, geological, and ecological contexts.
Sunday, 19 October 2025
8:00 AM - 12:00 PM
HBGCC, 213AB
Forecast Smarter, Not Harder: Improving Regression-Based Streamflow Prediction in Snow-Dominated Basins with Random Forest and Water Year Clustering
Paper number: 5-13
Program: Technical Sessions
Time: 11:05 AM – 11:20 AM
Authors
-
Eric SmallUniversity of Colorado Boulder, Boulder, Colorado
Abstract
Streamflow forecasts in snow-dominated basins are often limited by spatial and temporal variability in snow accumulation and melt processes, especially as climate change alters precipitation phase and snowpack dynamics. This study presents a data-driven forecasting framework that integrates machine learning-based predictor selection with cluster-based regression modeling to improve seasonal streamflow forecasts in snow-dominated basins.
A suite of snow and climate metrics were selected to capture temperature variability, precipitation characteristics, snow water equivalent, and elevation-dependent processes across 130 western US mountain basins from the CAMELS dataset. Random forest models were then used to evaluate the relative importance of each predictor in explaining April-July streamflow, providing a basin-specific, ranked set of candidate variables for regression forecasting.
To account for interannual climate variability, k-means clustering was applied to group water years based on winter and spring snow and temperature conditions. Separate regression models were trained within each cluster using the top ranked predictors identified in the random forest analysis. This cluster-informed approach improved model performance relative to traditional regression techniques, yielding lower RMSE and higher R2 in basins where snowpack evolution is complex or transitional.
By combining machine learning with unsupervised classification, this framework provides a flexible and transferable method to enhance streamflow forecasting skill in snow-impacted regions. Results underscore the value of adapting predictor selection and model structure to hydroclimatic variability in a changing climate.
Geological Society of America Abstracts with Programs. Vol. 57, No. 6
doi: 10.1130/abs/2025AM-412786
© Copyright 2025 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.