Nagamoto et al. (2026) Widespread drought-driven declines in streamflows and water quality in the Upper Colorado River Basin during 1998-2022

Identification

• Journal: Communications Earth & Environment
• Year: 2026
• Date: 2026-09-14
• Authors: Emily Nagamoto, Mohammed Ombadi, Fabio Ciulla, Jared Willard, Rosemary W. H. Carroll, Charuleka Varadharajan
• DOI: 10.1038/s43247-026-03890-5

Research Groups

• Earth & Environmental Sciences Area, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
• Department of Geography, University of Colorado Boulder, Boulder, CO, USA
• Department of Climate and Space Sciences and Engineering, University of Michigan, Ann Arbor, MI, USA
• Computing Sciences Area, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
• Artificial Intelligence (AI), Learning, and Intelligent Systems Group, National Laboratory of the Rockies, Golden, CO, USA
• Division of Hydrologic Science, Desert Research Institute, Reno, NV, USA

Short Summary

This study investigates the cascading impacts of meteorological droughts on streamflow, water temperature, and specific conductance in the Upper Colorado River Basin from 1998-2022 using a data-driven framework. It finds that severe droughts reduced streamflows by 15% while increasing water temperature by 9% and specific conductance by 5%, with catchment attributes significantly influencing spatial variability in vulnerability and resilience.

Objective

• To quantify the magnitude of changes in streamflow (Q), water temperature (WT), and specific conductance (SC) in response to severe meteorological droughts relative to adjacent normal years in the Upper Colorado River Basin (UCRB).
• To evaluate catchment recovery following drought episodes.
• To identify catchment attributes that influence the spatial variability in responses to drought.

Study Configuration

• Spatial Scale: Upper Colorado River Basin (UCRB), encompassing headwater catchments in Colorado, Wyoming, Utah, and New Mexico, with elevations ranging from 1200 to 4300 meters. The study included 202 sites for streamflow, 35 for water temperature, and 23 for specific conductance.
• Temporal Scale: Water years 1998 to 2022. Specific severe meteorological drought episodes analyzed were 2001–2002, 2012, 2018, and 2020–2021.

Methodology and Data

• Models used:
  • DIRE (Drought Impacts and Resilience Evaluation) data-driven framework.
  • Deep learning model (Long Short-Term Memory - LSTM) for predicting water temperature in unmonitored basins.
  • XGBoost machine learning regression model for identifying catchment attributes influencing streamflow changes.
  • Statistical methods: Spearman correlations, attribute z-scores, Mann-Kendall trend test, Box-Cox transform, regression analysis for precipitation-runoff relationships.
  • Standardized Precipitation Evapotranspiration Index (SPEI) for meteorological drought identification.
• Data sources:
  • Streamflow (Q; cubic meters per second per square kilometer), water temperature (WT; degrees Celsius), and specific conductance (SC; microsiemens per centimeter at 25 degrees Celsius) from the USGS National Water Information System (NWIS) database.
  • Meteorological data (precipitation, air temperature) from the PRISM (Parameter-elevation Regressions on Independent Slopes Model) dataset at 4 km resolution via Google Earth Engine (GEE).
  • 250 catchment characteristics (natural and human factors) from the USGS GAGES-II (Geospatial Attributes of Gages for Evaluating Streamflow) dataset.
  • Reservoir storage level data for 43 reservoirs from the US Bureau of Reclamation website.
  • Land cover characteristics from the Multi-Resolution Land Characteristics (MRLC) Consortium Web Viewer tool (NLCD).

Main Results

• Severe meteorological droughts (2001–2002, 2012, 2018, 2020–2021) in the UCRB caused significant impacts on water quantity and quality.
• Annual streamflows (Q) declined by a median of 15% across all sites during drought years relative to normal adjacent years. Peak flows (95th percentile) decreased by 47%, and low flows (5th percentile) by 15%.
• Water temperature (WT) increased by a median of 9% across most sites during drought. Peak and low annual WT increased by approximately 2–5%.
• Specific conductance (SC) increased by a median of 5% across most sites during drought. Low SC (5th percentile) increased by 44%, while peak SC (95th percentile) slightly declined by 2%.
• The greatest impacts (declines in Q, increases in WT and SC) occurred during spring and summer months.
• Streamflow and SC generally recovered to pre-drought baselines within 1–10 years and 1–6 years, respectively, for droughts up to 2018. However, successive droughts since 2018 led to prolonged impacts on Q and SC, with many sites not recovering by 2022.
• WT impacts were transient, with most catchments recovering within 1–2 years.
• Catchment attributes associated with greater streamflow declines during drought included higher freshwater withdrawals, deciduous forest cover, higher mean annual air temperatures, finer soils (silt or clay), and longer subsurface residence times.
• Factors enhancing streamflow resilience included higher subsurface flows, greater soil water retention and infiltration capacity (e.g., soil permeability, sand content), northern aspect, and June precipitation.
• WT increases were positively correlated with higher elevation, slopes, precipitation, and riparian woody wetlands/forests, and negatively correlated with agricultural/shrub lands and finer soil textures.
• The presence of open water bodies and reservoirs mitigated SC increases.
• Precipitation-runoff relationships did not significantly change at most sites during drought years relative to normal years, and most sites showed no significant long-term trends in Q, runoff efficiency, WT, or SC between 1998 and 2022.

Contributions

• This study is the first to comprehensively assess post-drought recovery for streamflow, water temperature, and specific conductance in the Upper Colorado River Basin.
• It introduces DIRE (Drought Impacts and Resilience Evaluation), a novel, robust, modular, and flexible data-driven framework that integrates machine learning with other computational approaches to evaluate drought impacts, recovery, and resilience.
• The research provides critical insights into how both natural and anthropogenic catchment attributes influence drought vulnerability and resilience in mountainous regions, demonstrating that responses vary spatially even under uniform meteorological drought intensities.
• It employs advanced methods such as network analysis for clustering redundant catchment attributes and explainable artificial intelligence (XAI) for interpreting high-dimensional data and identifying key drivers of change.

Funding

• iNAIADS Early Career Research Program award (U.S. Department of Energy (DOE), Office of Science, Office of Biological and Environmental Research, Berkeley Lab Contract Number DE-AC02-05CH11231)
• U.S. Department of Energy, Office of Science, Office of Workforce Development for Teachers and Scientists (WDTS) under the Science Undergraduate Laboratory Internship (SULI) program
• Watershed Function Science Focus Area (Office of Biological and Environmental Research, Berkeley Lab Contract Number DE-AC02-05CH11231)
• National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of Science User Facility (Lawrence Berkeley National Laboratory, Contract No. DE-AC02-05CH11231)

Citation

@article{Nagamoto2026Widespread,
  author = {Nagamoto, Emily and Ombadi, Mohammed and Ciulla, Fabio and Willard, Jared and Carroll, Rosemary  W. H. and Varadharajan, Charuleka},
  title = {Widespread drought-driven declines in streamflows and water quality in the Upper Colorado River Basin during 1998-2022},
  journal = {Communications Earth & Environment},
  year = {2026},
  doi = {10.1038/s43247-026-03890-5},
  url = {https://doi.org/10.1038/s43247-026-03890-5}
}