Dataset for Climate-driven latitudinal divergence in lake phosphorus dynamics across Canada

This dataset provides annual near-surface total phosphorus (TP) concentrations for Canadian lakes from 1984 to 2023, derived from Landsat surface reflectance imagery and TPNet model retrievals. It includes two data components: Grid_TP, containing the median lake surface TP for each of Canada's 3,329 grids (1° × 0.5°) with latitude (LAT) and longitude (LON) representing each grid's centroid; and Basin_TP, containing the median lake surface TP for each of Canada's 366 HydroBASINS Level-5 watersheds, with LAT and LON representing each watershed's centroid. Each watershed is uniquely identified by HYBAS_ID, which can be used to obtain its full boundary. BASIN_TYPE categorizes watersheds by land use: 1 for urban, 2 for agricultural, 3 for tundra, 4 for boreal forest, and 5 for temperate forest. TP concentrations are expressed in µg/L. The dataset also includes the pre-trained TPNet model (TPNet.pth), developed using the PyTorch framework, along with demo code (TPNet.py) to guide the estimation of near-surface TP from Landsat imagery. For any questions or suggestions regarding the dataset or model, please contact Hongwei Guo at [e-mail].