Abstract
We present a standardized pan-watershed dataset of annual chlorophyll-a concentration (Chl-a) in 27,313 lakes (≥ 10 ha) draining into Lake Winnipeg, Canada, spanning 1984-2023. Lake polygons from HydroLAKES were integrated with Landsat 5/7/8 Collection 2 imagery processed in Google Earth Engine (GEE) using a reproducible workflow that (1) filters July-October scenes (peak phytoplankton biomass season), (2) masks non-water pixels from each scene, (3) converts Landsat digital numbers to surface reflectance values, (4) applies a cross-sensor Chl-a retrieval model calibrated against in-situ samples, (5) calculates the spatial mean of Chl-a in each lake for each scene, and (6) calculates the median value of all spatial-mean values per lake per year. Outputs include per-lake annual Chl-a provided as both natural-log and back-transformed Chl-a (µg L⁻¹) plus annual trophic state classes delivered in an Excel workbook and two geodatabases for mapping. The accompanying annotated GEE and R codes, input lake boundaries, and documentation enable transparent reuse and straightforward adaptation to other regions, time periods, or sensors. This resource fills a critical monitoring gap for an agriculturally influenced, bloom-prone continental watershed and supports research and management by establishing productivity baselines, detecting departures from historical conditions, and assessing bloom timing at scales relevant to decision-making. All data, inputs, and code are openly available via Zenodo.