CAMELS-KR: Catchment attributes, meteorology, and streamflow for large-sample hydrology in South Korea

CAMELS-KR is the first CAMELS (Catchment Attributes and Meteorological data for Large-sample Studies) dataset developed for South Korea. It provides static catchment attributes, daily hydrological and meteorological time series, and streamflow generated using both a regional Long Short-Term Memory (LSTM) model and a locally calibrated HBV conceptual model for 282 catchments across the country. The static attributes are organized into seven categories: location, topography, climate, soil, land cover, hydrology, and human influence. The time series data include observed daily streamflow and water level, catchment-averaged meteorological forcings, and daily simulated streamflow from the LSTM and HBV models. The dataset is designed to support large-sample hydrology, machine learning–based streamflow prediction, regionalization, hydrological model comparison, and benchmarking studies. Detailed descriptions of the dataset structure, variables, and file organization are provided in the accompanying CAMELS_KR_Data_Description.pdf document. Changelog v1.1 Removed the previous README file and added the CAMELS_KR_Data_Description.pdf document providing detailed information on the dataset. v1.2 Removed “reconstructed” from the dataset title, changing “reconstructed streamflow” to “streamflow” to clarify that observed streamflow is included in the dataset. Added two topological attributes, is_headwater and upstream_basin_ids, to CAMELS_KR_location_attributes.csv. Added a new CAMELS_KR_geology_attributes.csv file containing nine geological and hydrogeological catchment attributes: geo_porosity and geo_log10_permeability derived from GLHYMPS, and pa_perc, mt_perc, va_perc, ss_perc, pb_perc, vb_perc, and sc_perc derived from GLiM. Added the water_surface_level variable to CAMELS_KR_Hydrological_timeseries.csv, representing water-surface elevation relative to mean sea level.