The companion manuscript is published in AGU Water Resources Research: Episodic salinization of Midwestern and Northeastern US rivers by road salt (Platt and Dugan, 2025). The companion code is captured in a data pipeline published on Zenodo at https://doi.org/10.5281/zenodo.17095176.

Data and file descriptions

There are 3 data files included in this data publication. The first two represent snapshots of publicly available data and the third represents the specific conductance regime assigned to each site based on a time series analysis (see the companion manuscript Platt and Dugan 2025 for more information). Ultimately, the static attributes and site classification's were used as model inputs to a classification random forest to determine which stream attributes may be associated with which specific conductance regime.

1. 1_specific_conductance.csv contains a time series of continuous specific conductance measurements across 437 sites from the US Geological Survey National Water Information System (USGS NWIS). The dates range from January 2, 2000 to May 1, 2025.
2. 2_static_attributes.csv contains 17 different stream attributes matched to each of the 437 USGS sites gathered from a different sources, including the US Environmental Protection Agency's National Hydrography Dataset Plus (US EPA NHDPlus), National Oceanic and Atmospheric Administration National Water Model (NOAA NWM), as well as Zell and Sanford, 2020 (DOI: 10.1029/2019WR026724) and Falcone et al., 2023 (DOI: 10.5066/P96IX385). More information about each of these sources and exactly what was accessed is available in the companion paper, Platt and Dugan 2025. Also, see the table further below for definitions and units for each of the attributes.
3. 3_site_classification.csv contains the classification model output stating whether a site's specific conductance patterns were "Episodic" or "Not episodic".

How to use the data here

The data in this resource represents a snapshot of data downloaded and processed from public resources (USGS National Water Information System and ScienceBase) in May 2025. If you want to be able to recreate the results from the companion manuscript, these data should be used as inputs to the reproducible data pipeline built in R, available at https://github.com/lindsayplatt/episodic-river-salinization-model. Details on how to do this are provided below. If you want to just build the data pipeline using the most recent data available from the raw inputs, see more in the GitHub repository README file.

By using the files in this data release, you can skip phases 1-3 in the companion pipeline (1_Download, 2_Prepare, 3_Filter).

1. By using data from 1_specific_conductance.csv instead of p3_ts_sc_qualified in 4_EpisodicSalinization.R, you can rerun code used to classify a site as "Episodic" or "Not episodic". You can compare these outcomes to the data in 3_site_classification.csv.
2. By using data from 2_static_attributes.csv instead of p3_static_attributes in 5_DefineCharacteristics.R, you can rerun code used to evaluate which characteristics are more associated with episodic vs not episodic through the random forest modeling code.

Information on units

In 1_specific_conductance.csv, specific conductance are in microSiemens per centimeter at 25 degrees Celcius (uS/cm at 25 degC).

Static attributes in 2_static_attributes.csv are described in this table (if the source is not explicitly stated, it is from the US EPA NHDPlus: