STREAM Dataset

This HydroShare resource contains STREAM, a harmonized collection of high-frequency water quality observations from river monitoring stations across the United States. The resource also includes site metadata, delineated watershed boundaries, watershed characteristics, historical meteorology, annual land use and land cover, anthropogenic inputs, streamflow, and grab-sample water quality data collected at the sensor locations.

Continuous water quality measurements are provided as hourly averages. The supporting datasets make it possible to connect sensor observations with watershed setting, weather, land cover, human activity, and discharge.

Resource Contents

Working with the Data

• Use STREAM_ID as the common station identifier when joining the datasets.
• Use DateTime to align time-varying records within a station.
• Check time_zone in the metadata before interpreting timestamps.
• Review the quality flag paired with each water quality or streamflow value.
• Review flag_catchment_shapefile before using watershed attributes.
• Keep all component files for each watershed shapefile together when extracting or moving files from 03_shapefiles.zip.
• Review hydro_impacts when selecting sites for analyses that may be sensitive to dams or coastal influence.
• Keep source periods in mind. Static BasinATLAS attributes were compiled from datasets representing different years and time periods.

Quality flags

Methods Summary

Site screening and technical validation

A site and parameter were included only when at least one year contained one month or more of observations. This removes very short sensor deployments and highly limited time series.

Sites intersecting HydroLAKES were screened out, and sensor locations were manually reviewed to remove sites in lakes, bays, and artificial channels such as canals. Sites with negative discharge in more than 1% of available records were also excluded to reduce strong tidal influence.

Potential hydrologic impacts are recorded in hydro_impacts rather than used as automatic exclusions:

• Dam: the site is approximately 10 km or less downstream of a dam or related structure identified in the National Inventory of Dams.
• Coast: the site is approximately 1 km or less upstream of the U.S. ocean coastline.

These flags allow users to include or remove potentially affected sites based on the needs of an analysis.

Time-series QA/QC

Quality flags supplied by the original data providers were retained. STREAM applied an additional conservative screening step intended to remove only clear anomalies while preserving unusual but plausible observations.

Values outside the following sensor operating bounds were marked erroneous and removed during processing.

The limits were informed by operating ranges in commonly used sensor manuals, including the YSI EXO2, Sea-Bird SUNA V2, and Sea-Bird HydroCycle-PO4 manuals. A broad turbidity upper limit was used to avoid removing extreme readings that could still be real.

Watershed delineation

Watersheds were delineated using the NHDPlus river network, with the method depending on the source of the monitoring site.

1. USGS sites were first delineated from the USGS site identifier with the nhdplusTools R package.
2. When a USGS site identifier failed or returned an invalid result, the associated NHDPlus COMID was used.
3. Non-USGS sites were spatially linked to an NHDPlus flowline, assigned a COMID, and delineated from that stream segment.
4. When the NHDPlus result could not provide a reasonable watershed, an alternative delineation from Mghydro was evaluated.

All final watershed polygons were projected to the CONUS Albers Equal Area Conic coordinate reference system, EPSG:5070. Topological errors were corrected before watershed area was calculated in square kilometers.

Watershed QA/QC

Delineated watershed area was compared with the drainage area reported by the original monitoring network when that information was available. Sites with a relative difference greater than 25% were flagged for review.

Review included inspection of the watershed boundary, site position, NHDPlus flowline, and assigned COMID. Extra attention was given to large rivers, confluences, reservoirs, braided channels, and modified waterways. Common causes of disagreement included coordinate uncertainty, flowline snapping errors, different hydrographic representations, and small location shifts on large river systems.

The metadata field flag_catchment_shapefile summarizes this check:

• Safe: delineated and reported watershed areas differ by less than 25%.
• Caution: the difference is greater than 25%, or no source drainage area was available for comparison.

Watershed shapefiles

The archive 03_shapefiles.zip contains the final delineated watershed boundaries for individual STREAM monitoring sites. These polygons represent the upstream contributing drainage area used to calculate watershed area and derive the catchment-scale datasets included in this resource.

Use the shapefiles together with the following fields in 01_metadata.zip:

The shapefiles use the CONUS Albers Equal Area Conic coordinate reference system (EPSG:5070). Sites marked Caution were retained so users can decide whether the boundary is appropriate for a particular analysis.

BasinATLAS attributes

Static watershed characteristics were derived by intersecting each delineated watershed with BasinATLAS Level 12 polygons in EPSG:5070.

• Continuous variables, such as climate, topography, soils, and land cover, were calculated as area-weighted watershed averages.
• Categorical variables were assigned using the class with the largest cumulative area weight.
• Variables intended to describe conditions at the watershed outlet were taken from the BasinATLAS polygon containing the monitoring site's pour point.
• Upstream-aggregated BasinATLAS fields were not used because the STREAM watershed polygons already represent each site's full contributing drainage area.

The resulting attributes were joined to the other datasets using the unique site identifier.

Variable Reference

1. Site metadata

The metadata archive provides the information needed to identify stations, connect them with the original data source, determine which datasets are available, and evaluate watershed delineation quality.

2. Continuous water quality

All continuous water quality variables were aggregated from their native sampling intervals to hourly averages. A source quality flag is provided for each corresponding parameter where available.

3. Static catchment variables

Static watershed attributes come from BasinATLAS v10 unless otherwise noted. The source datasets span different years and represent a mix of historical averages, modeled values, and the most recent conditions available when BasinATLAS was assembled.

Variable names are shown as documented in the source, including {} placeholders used in the BasinATLAS naming convention.

Hydrology and surface water

Topography and climate

Land cover and ecosystems

Soils and geology

Population and human influence

River network

For full source citations, native spatial resolutions, source years, and class definitions, see Variables-Description-06-30-2026.pdf and the BasinATLAS catalog.

4. Continuous historical meteorology

Meteorological variables come from the Office of Water Prediction's Analysis of Record for Calibration, version 1.1. Native data are hourly at 30 arc-second resolution. STREAM provides hourly values lumped across each watershed.

Source: Fall, G., et al. (2023), The Office of Water Prediction's Analysis of Record for Calibration, Version 1.1.

5. Dynamic historical land use and land cover

Land use and land cover variables come from the Annual National Land Cover Database. Native data have 30 m spatial resolution and annual temporal resolution. STREAM values are annual watershed fractions. For each station and year, the fractions across all classes sum to 1.

Source: U.S. Geological Survey (2024), Annual NLCD Collection 1 Science Products, version 1.1.

6. Dynamic anthropogenic variables

All variables in this archive are provided as annual watershed averages or watershed fractions.

Key sources:

• Falcone, J. A., Oelsner, G. P., and Bock, A., Estimates of Road Salt Application across the Conterminous United States, 1992-2019.
• National Atmospheric Deposition Program.
• Lu, C., and Tian, H. (2016), Half-degree gridded nitrogen and phosphorus fertilizer use for global agriculture production during 1900-2013.
• Booth, E. G., and Kucharik, C. J. (2026), Historical maps of land use-land cover, crop type, nutrients, and irrigation across CONUS.

7. Streamflow discharge

Streamflow was obtained from the original sources that supplied the corresponding water quality datasets and aggregated to hourly resolution.

8. Grab-sample water quality

Grab samples were collected at the same locations as the high-frequency sensor gauges. These measurements can support comparisons between sensor signals and laboratory or discrete-sample concentrations.

Documentation Notes

The two PDF files included with this resource remain the authoritative source for detailed variable provenance and processing decisions. This README condenses those documents so users can understand the resource structure and begin working with the data without moving between multiple tables.