CAMELS: A Dataset of Watershed Properties and Meteorology From the Large Sample Study of the United States

The dataset consists of two aspects: time series and catchment attributes.

Time series datasets

The dataset is basin-scale hydrometeorological forcing data developed by the research team for 671 basins in the U.S. Geological Survey 2009 Hydroclimatic Data Network (HCDN-2009, Lins 2012), which belongs to the contiguous United States basin subset. Model time series output is available for the same time period as the forcing data. USGS streamflow data for all basins are also provided for all available dates between January 1 and December 31, 2014. The research team then implemented the hydrological model and calibration procedures traditionally used by the NWS, the hydrological modeling system based on SNOW-17 and the Sacramento Soil Moisture Accounting (SAC-SMA), and the Shuffled Complex Evolution (SCE) optimization method.

Dataset structure

The basin_timeseries_v1p2_metForcing_obsFlow.zip file contains all the basin forcing data, observed flows, basin metadata, a README file, and basin shape files for all three meteorological products. The three *_modelOutput_*.zip files contain all the model outputs for the various forcing datasets indicated in the link names. Finally, the basin_set_full_res.zip file is a full-resolution basin shape file containing the original basin boundaries from the geospatial structure.

Watershed attribute dataset

The dataset covers the same 671 catchments as the large-sample hydrometeorological dataset introduced by Newman et al. in 2015. For each catchment, the research team described multiple attributes that influence catchment behavior and hydrological processes. Separate datasets characterizing these attributes have been available for some time, but comprehensive multivariate catchment-scale assessments have been difficult until now because these datasets often have different spatial configurations, are stored in different archives, or use different data formats. By creating catchment-scale estimates of these attributes, the research team aims to simplify the assessment of their interrelationships.

Topographic characteristics (e.g., elevation and slope) were taken from Newman et al. (2015). Climate indices (e.g., aridity and frequency of dry days) and hydrological characteristics (e.g., mean annual discharge and baseflow index) were calculated using time series provided by Newman et al. (2015). Soil properties (e.g., porosity and soil depth) were characterized using the STATSGO dataset and the Pelletier et al. (2016) dataset. Vegetation characteristics (e.g., leaf area index and root depth) were inferred using MODIS data. Geological characteristics (e.g., geological class and subsurface porosity) were calculated using the GLiM and GLHYMPS datasets.

A fundamental difference between this dataset and similar datasets is that it provides both quantitative estimates of various catchment properties and involves an assessment of the limitations of the data and methods used to calculate these properties (see Addor et al., 2017). The large number of catchments, combined with the diversity of geophysical features, makes these data well suited for large sample studies and comparative hydrology.