Barr Lake chlorophyll-a and Sentinel-2 spectral index dataset for harmful algal bloom detection (2019–2025)

This repository contains the dataset and analysis scripts used to evaluate Sentinel-2 spectral indices for harmful algal bloom (HAB) detection in Barr Lake, Colorado, USA (2019–2025).

The dataset integrates in-situ chlorophyll-a (Chl-a) measurements with spectral indices derived from Sentinel-2 Level-2A surface reflectance imagery. Each observation represents a field sampling event paired with the closest valid Sentinel-2 acquisition (±5 days) following cloud and quality filtering.

Field samples were collected at four sites grouped into nearshore and open-water zones to account for spatial heterogeneity. Chlorophyll-a was measured using U.S. Environmental Protection Agency (EPA) Method 445.0. Satellite data were processed in Google Earth Engine, including cloud masking, water pixel identification, and spectral index computation.

The dataset includes:
- Processed analysis dataset (in-situ and satellite-derived variables)
- Python scripts for statistical analysis and figure generation
- Google Earth Engine script for spectral index extraction
- Documentation describing dataset structure and variables

Three spectral indices are evaluated:
• Normalized Difference Chlorophyll Index (NDCI)
• Floating Algae Index (FAI)
• Absorption-Band Difference Index (ABDI)

The objective of this dataset is to assess how spectral index performance varies under optically complex conditions. The underlying hypothesis is that index performance differs in sensitivity and classification accuracy depending on optical complexity and spatial heterogeneity, particularly between nearshore and open-water environments.

Bloom conditions are defined as Chl-a > 50 µg L⁻¹, with values > 400 µg L⁻¹ capped and log-transformed for regression analyses.

The data show strong spatial variability, with consistently higher Chl-a concentrations in nearshore areas. Index performance differs across conditions, with NDCI providing the most balanced classification performance, ABDI showing higher sensitivity, and FAI higher specificity.

Version 2 updates include removal of unused threshold categories (30 and 70 µg L⁻¹), updated Python and Google Earth Engine scripts, and improved documentation.

This repository supports reproducibility of the analyses and can be used for evaluating spectral index performance, HAB classification, and remote sensing-based water quality monitoring in optically complex inland waters.