ChinaCoast-DIN-DIP v1.0: a seamless monthly dataset of surface dissolved inorganic nitrogen and phosphorus at 4 km resolution in the coastal waters of China (2003–2022)

ChinaCoast-DIN-DIP v1.0 is a gridded dataset of monthly surface dissolved inorganic nitrogen and dissolved inorganic phosphorus in the coastal waters of China. It covers January 2003 to December 2022 at 4 km resolution, spanning 17° N to 41° N and 106° E to 127° E from the Bohai Sea to the northern South China Sea, including the Changjiang and Pearl River estuaries. Method. DIN and DIP were retrieved jointly with FTT-SG, a feature-tokenizer Transformer with a spectral branch and target-specific gating. The model was trained on paired in situ observations from 2017 to 2022. Its inputs were MODIS-Aqua remote sensing reflectance and sea surface temperature, GLORYS12V1 sea surface salinity and mixed layer depth, and GEBCO bathymetry. Gaps caused by clouds and missing satellite data were filled with DINEOF, giving spatially continuous monthly fields. Accuracy. The model achieved R² of 0.668 for DIN and 0.674 for DIP. Validation against independent in situ data from four coastal regions gives same-month correlations of 0.69 to 0.91. The typical uncertainty is a factor of 1.96 for DIN and 1.79 for DIP. Data format. The dataset is provided as a single NetCDF file, ChinaCoast_DIN_DIP_v1.0_monthly_4km_2003_2022.nc, on a regular latitude and longitude grid with 240 monthly time steps. Concentrations are given in mg L⁻¹, with DIN expressed as nitrogen and DIP as phosphorus. Variables: DIN, DIP: monthly concentrations (time, lat, lon)gapfill_flag: 0 no data, 1 direct retrieval, 2 DINEOF gap filledland_mask: 0 ocean, 1 landbathymetry: water depth from GEBCO (m)distance_to_coast_km: distance to the nearest coastline (km)valid_retrieval_months: number of months with complete model input (of 240) Contact. Jin Qi, Zhejiang University, [e-mail]