Caravan-Qual (lite): A global scale integration of water quality observations into a large sample hydrology dataset

Caravan-Qual is an open access dataset that brings water quality to the research paradigm of large sample hydrology (LSH), integrating daily water quality data from 100 constituents with catchment attributes, meteorological forcing and co-located streamflow observations.

The dataset has been published in Scientific Data. The associated manuscript can be found here.

We envisage the dataset to facilitate research into topics including:

Spatio-temporal analysis of river water quality dynamics at regional to global scales.

Investigation of the relationships between (constituent-specific) water quality responses and hydrological, meteorological and catchment characteristics.

The development and evaluation of process-based, hybrid and data-driven water quality models across diverse hydrological and climatic conditions.

Key features

~96 million water quality observations from 150,000+ monitoring stations located worldwide, harmonised across multiple databases

Linked to streamflow observations from 26,000+ gauges, harmonised across multiple databases

Daily meteorological forcing data (ERA5-Land)

Comprehensive catchment and stream attributes (HydroATLAS and GEOGLOWSv2)

Provided in both .zarr (1980-01-01 to 2025-09-30 only) and .csv format.

Access notes

All water quality data (with combined streamflow observations) are  provided as .csv files (in wqms-csvs.zip) and in .zarr format.   

Please note that, due to data storage restrictions, only monthly weather data is stored on Zenodo (in Caravan-Qual_monthly_weather.zarr.zip) and is stored seperately to the daily water quality data and catchment attributes (in Caravan-Qual_lite.zarr.zip).

The 'full' Caravan-Qual dataset (i.e., including daily weather data) and the auxilliary data required for extending the dataset can be accessed here.

Channel log

11 December 2025: Version 0.1 (beta version release).

16 March 2026: Version 1.0 (version of the official paper release).

~26 million water quality observations added, including global pharmaceutical data (Wilkinson et al., 2021), an Iranian water quality dataset (Zarei et al., 2025) and updated the GEMStat data to the February 2026 version.

Streamflow data from CAMELS-NZ (Bushra et al., 2025) added.

Limit flags (e.g. "<") and detection limits associated with each observation are now preserved.

Observations detected as outliers (physical or statistical) are flagged, opposed to removed.

Figures for spatio-temporal patterns per constituent are made avaliable (Caravan-Qual_figures.zip)

Full license information, including download links, for each dataset included in Caravan-Qual is provided (Caravan-Qual_licenses.zip).

2nd May 2026: Associated manuscript published in Scientific Data (https://doi.org/10.1038/s41597-026-07352-7)

The most up-to-date developments are documented on Caravan-Qual's GitHub page. The original Caravan paper can be accessed here, while developments and community extensions are documented on Caravan's GitHub page.