Mapping Nearshore Cladophora in Lake Ontario: An Automated Open-Source Sentinel-2 Workflow and Experimental Index

Abstract. Acquiring high-quality Sentinel-2 images is a key first step for nearshore remote-sensing studies, but scene-level cloud metadata can be misleading when the research window is small. This paper presents an open-source, local cloud-aware workflow for two Lake Ontario regions centered on the Olcott (OOL) and Irondequoit (OIR) stations. The workflow first queries Sentinel-2 Level-2A scenes from the Copernicus Data Space STAC catalogue, then downloads only the quicklook image and footprint GeoJSON for each candidate scene. The station ROI is mapped to quicklook image space with a homography transform, cropped, and scored with a local HSV cloud rule. Full spectral bands are downloaded only when the local ROI cloud rate meets the user threshold. Homography improved ROI localization, increasing IoU from 0.2279 to 0.8352 at OIR and from 0.4722 to 0.8706 at OOL. A 500-scene water-land comparison also showed that the mean land cloud rate was about 30 percentage points higher than the lake-water rate. The paper also outlines a controlled tank experiment and retains two candidate Cladophora indices for future validation.