Jaeger et al. (2026) Lost in translation: Reconciling different streamflow permanence data products

Identification

• Journal: Journal of Environmental Management
• Year: 2026
• Date: 2026-03-20
• Authors: Kristin L. Jaeger, Susan A. Wherry, Malia H. Scott, Audrey Martinez, Roy Sando, Evan Thaler
• DOI: 10.1016/j.jenvman.2026.129219

Research Groups

• U.S. Geological Survey, Washington Water Science Center, Tacoma, WA, USA
• U.S. Geological Survey, Oregon Water Science Center, Portland, OR, USA
• U.S. Geological Survey, Wyoming-Montana Water Science Center, Helena, MT, USA
• U.S. Geological Survey, Idaho Water Science Center, Boise, ID, USA
• College of Forestry, Oregon State University, Corvallis, OR, USA

Short Summary

This study develops a framework to reconcile and evaluate two streamflow permanence datasets (NHDPlus HR and PROSPER model output) for the Pacific Northwest, finding 68% agreement regionally and identifying reliability patterns to inform land and water management decisions.

Objective

• To provide a framework for reconciling different streamflow permanence data products (NHDPlus HR and PROSPER) to determine if existing information is sufficient for streamflow class determination or if further field verification is required.

Study Configuration

• Spatial Scale: Hydrologic Unit Code (HUC) 17 Pacific Northwest Region, U.S., encompassing Washington, most of Oregon and Idaho, and parts of Montana and Wyoming. The study area includes three EPA Level II ecoregions: Marine West Coast Forest, Cold Deserts, and Western Cordillera.
• Temporal Scale:
  • PROSPER model predictions: Annual streamflow permanence for years 2004–2016.
  • NHDPlus HR classifications: Based on historical mapping largely occurring between 1955 and 1990, with an average range of 1950–1990 for flowline map dates.
  • Climate data (scPDSI): Monthly self-calibrating Palmer Drought Severity Index averaged over water years (October 1 of previous year to September 30 of current year).

Methodology and Data

• Models used:
  • PRObability of Streamflow PERmanence (PROSPER) model (random forest model, version 2.1 update).
  • Logistic regression (to assess climate bias in NHDPlus HR).
• Data sources:
  • National Hydrography Dataset Plus High Resolution (NHDPlus HR) hydrographic classification (10-meter spatial resolution).
  • PROSPER model output (streamflow permanence probabilities at 30-meter spatial resolution).
  • Monthly self-calibrating Palmer Drought Severity Index (scPDSI) (4-kilometer gridded data, processed as flow-conditioned parameter grids at 10-meter resolution).
  • Historical field observations (used for NHDPlus HR and to train PROSPER).
  • Covariates for PROSPER: total annual precipitation, annual mean daily minimum air temperature, percent forest cover, monthly mean evapotranspiration for August, and baseflow index.

Main Results

• Regional Agreement: The NHDPlus HR and PROSPER datasets agree on streamflow permanence classification for an average of 68% of flowline lengths within HUC8 watersheds.
• Class-Specific Agreement: Agreement is higher for nonperennial classifications (78% of flowline lengths) compared to perennial classifications (46% of flowline lengths).
• Geographic Patterns: High agreement is observed in areas dominated by nonperennial streams, such as the Cold Deserts ecoregion and parts of the Western Cordillera. Lower agreement is found in the western portions of the study region (western Oregon and Washington).
• PROSPER Reliability:
  • Nonperennial classifications in the Cold Deserts ecoregion are generally reliable (94% high confidence, 91% nonperennial).
  • PROSPER predictions are considered less reliable for high-elevation mountain regions (e.g., Olympic Mountains, north Cascade Range) and larger streams/rivers (drainage area greater than 100 square kilometers).
  • Less confident PROSPER classifications (streamflow permanence probability between 0.4 and 0.6) are more frequent at higher elevations (approximately 1500 meters), smaller drainage areas (peaking at 0.1 square kilometers), and moderate annual precipitation (500–800 millimeters).
  • 9.4% of stream-reach locations have covariate values outside the calibration range for one of the top five most important variables, with precipitation being the most common (7.7%).
• NHDPlus HR Reliability:
  • A slight but statistically significant positive climate bias was identified, indicating that perennial NHDPlus HR classifications were more likely associated with climatically wet periods (positive scPDSI values).
  • 73% of flowlines by length in HUC8 watersheds correspond to normal climate conditions at the time of mapping and are thus interpreted as reliable classifications.
  • "Unexpected" class-climate pairs (e.g., perennial classification during a climatically dry year) are considered "more reliable," while "expected" pairs (e.g., perennial during a climatically wet year) are "less reliable."

Contributions

• Provides a reproducible and flexible two-level framework for reconciling disparate streamflow permanence data products (PROSPER probabilities and NHDPlus HR classifications).
• Offers a systematic decision procedure to help land and water managers determine when existing data is sufficient for streamflow classification or when field verification is needed, potentially leading to cost savings.
• Evaluates the agreement between PROSPER and NHDPlus HR, and assesses the reliability of each dataset based on confidence metrics, model suitability, and climatic context.
• Highlights the strengths and weaknesses of both PROSPER and NHDPlus HR across diverse physioclimatic conditions within the Pacific Northwest.
• Integrates multiple data sources (modeled probabilities, categorical classifications, climate indices, and the potential for field observations) into a unified decision support tool.

Funding

• USGS Water Mission Area Water Budget Program

Citation

@article{Jaeger2026Lost,
  author = {Jaeger, Kristin L. and Wherry, Susan A. and Scott, Malia H. and Martinez, Audrey and Sando, Roy and Thaler, Evan},
  title = {Lost in translation: Reconciling different streamflow permanence data products},
  journal = {Journal of Environmental Management},
  year = {2026},
  doi = {10.1016/j.jenvman.2026.129219},
  url = {https://doi.org/10.1016/j.jenvman.2026.129219}
}