Dataset supporting: A Sentinel-2-based ecological indicator framework for probabilistic chlorophyll-a exceedance mapping in a small inland reservoir
Description
This repository contains the cleaned analytical dataset supporting the associated manuscript on Sentinel-2-based probabilistic chlorophyll-a exceedance screening at Barr Lake, Colorado, USA. The workbook contains 70 quality-controlled site-date matchups from 51 sampling dates during July–September 2019–2025. Each record includes laboratory-measured chlorophyll a, concurrent in situ pH, Sentinel-2 Level-2A Normalized Difference Chlorophyll Index (NDCI), ERA5-Land regional daytime wind speed, sampling site and ecological-zone information, coordinates, the selected satellite acquisition date, and signed and absolute satellite–field temporal offsets. The primary outcome, Chla50, equals 1 when observed chlorophyll a is ≥50 µg L⁻¹ and 0 otherwise. Chla50 is an elevated-phytoplankton-biomass screening outcome; it is not a direct measure of cyanobacterial dominance, cyanotoxin concentration, bloom toxicity, or public-health risk. Eighty-three field observations were initially available; 70 met the prespecified temporal-matchup and valid-data criteria. Row-level records for the 13 pre-analysis exclusions were not archived. Version 2 renames HAB50 to Chla50 without changing its values, removes ancillary environmental fields not used in the manuscript analysis, standardizes site and zone labels and date formatting, adds absolute temporal offset, and updates the README and data dictionary.
Files
Steps to reproduce
1. Collect field chlorophyll-a and pH observations at Barr Lake during July–September 2019–2025. Open-water samples at the Dam and Outlet represented the upper water column; targeted 2025 nearshore samples at the Boat Ramp and Fishing Dock represented accessible near-surface water. 2. Analyze chlorophyll a following U.S. EPA Method 445.0 using 90% acetone extraction and acidification-corrected fluorometry. 3. Retrieve Sentinel-2 Level-2A surface reflectance from COPERNICUS/S2_SR_HARMONIZED in Google Earth Engine and scale band values by 0.0001. 4. Require granule-level CLOUDY_PIXEL_PERCENTAGE <30%; mask SCL classes 3, 8, 9, 10, and 11. Apply MNDWI ≥0 only at open-water sites; do not apply the MNDWI threshold at nearshore sites. 5. Calculate NDCI = (B5 − B4)/(B5 + B4) and extract mean NDCI within a 15 m buffer around each field coordinate. 6. Search acquisitions within ±5 calendar days and rank valid candidates by smallest absolute temporal offset, greater valid-data support, and lower granule cloud percentage. 7. Derive ERA5-Land wind on the field date from the magnitude of the mean 10 m horizontal wind vector after separately averaging u and v components for 10:00–13:00 local time. 8. Retain the 70 observations meeting matchup and valid-data criteria. Code Chla50 = 1 for chlorophyll a ≥50 µg L⁻¹ and 0 otherwise. 9. Fit the prespecified unpenalized binomial GLM using mean-centered NDCI, pH, and wind; preserve sampling-date structure in inference and validation. Evaluate ecological zone separately as a sampling-design sensitivity analysis.
Institutions
- University of Colorado BoulderColorado, Boulder
- Colorado Parks and WildlifeColorado, Denver