A 1-km Gridded Decadal (2016–2024) Nationwide Inland Water Quality Dataset for South Korea

Overview This dataset provides a decadal-scale (2016–2024) spatiotemporal reconstruction of nationwide inland water quality across South Korea at a 1 km gridded resolution. The dataset was generated using a multi-layer perceptron (MLP)-based deep learning framework that integrates Harmonized Landsat and Sentinel-2 (HLS) S30 surface reflectance with ERA5-Land hydrometeorological reanalysis data. Data Characteristics Spatial Resolution: 1 km (Aggregated from native 30 m predictions) Temporal Coverage: 2016 – 2024 Spatial Scope: Nationwide, South Korea (Five major river basins: Han, Geum, Nakdong, Yeongsan, and Seomjin) Parameters Included: Water Temperature (WT, °C) Suspended Solids (SS, mg/L) Chlorophyll-a (Chl-a, mg/m³) Dissolved Oxygen (DO, mg/L) Total Nitrogen (TN, mg/L) Total Phosphorus (TP, mg/L) File Structure The dataset is provided in GeoTIFF format, categorized by water quality parameters. Each file includes pixel-wise mean concentrations and associated uncertainty metrics derived from Monte Carlo dropout inference. Methodology The reconstruction utilizes the MLP model trained on over 55,000 in-situ matchups. It accounts for both bio-optical signals from satellites and thermodynamic/hydrological drivers from reanalysis data. For technical details, please refer to the associated manuscript submitted to Earth System Science Data (ESSD). Related Resources Manuscript: "Decadal-scale Spatiotemporal Trends of Nationwide Water Quality using Satellite-Reanalysis Integrated Deep Learning" (Submitted to Earth System Science Data). The DOI will be linked here upon official publication.