Trained monthly natural flow models and calculated natural flow anomalies at reference and biologically-paired USGS streamgages across the conterminous United States

This data release contains the outputs of a national-scale natural flow model and subsequent calculation of natural flow anomalies (NFA) at reference USGS streamgages and USGS streamgages linked to biological sampling sites compiled from USGS National Water Quality Assessment (NAWQA) and EPA Wadeable Streams Assessment (WSA), Environmental Monitoring and Assessment Program (EMAP), and National Rivers and Streams Assessment (NRSA) programs. This data release was produced by the CONUS Ecological Flows project in support of national-scale ecological flow analyses.

The workflow has two main components. First, a random-forest model predicts the long-term average monthly reference natural flow conditions expected for a gage (the expected, or E), using natural landscape and climate predictor variables from the NHDPlus version 2 catchment characteristics dataset. Second, natural flow predictions are compared to observed monthly flows (the O) to determine the departure from long-term reference conditions (the O/E ratio), also called the Natural Flow Anomaly (NFA). NFA values are centered on 1 and are direct proportional values related to natural normal flows, with values less than 1 reflecting deficit conditions (lower flows than expected under natural conditions) and values greater than 1 representing surplus conditions (greater flows than expected). NFA at the monthly scale can be driven by climate variability, landcover change, water use, or other water-management factors which affect the availability of water for a given month.

Data included in this release are:

Site location and gage pairing information for reference gages and biologically-linked gages (CONUS_Monthly_Ref_Sites_v3COMID_coord_fix.csv and all_sites_locations_gagesV5.csv respectively)

Catchment splitting results used to adjust drainage areas (final_refgage_prop_area.csv, final_biogage_prop_area.csv, and final_biosample_prop_area.csv)

Trained monthly random forest model objects to produce new predictions of monthly natural flow expectations (trained_ref_mean_e_model_MON.rds, where MON is a 3 digit month abbreviation).

Model performance metrics (performance_metrics.csv)

The resulting monthly NFA timeseries for all biologically-paired streamgages and associated biological sampling sites (biosites_monthly_mean_NFA.parquet). The value applies to both gage and sampling site, due to the matching technique, and that NFA is proportional to expected flows.