Modeling Multispectral UAV-Obtained Data for Monitoring Cyanobacterial Blooms in Reservoirs

Abstract

This study evaluated imagery acquired by an unmanned aerial vehicle (UAV) equipped with red-green-blue (RGB) and red-green-near-infrared (RGN) sensors for rapid detection and monitoring of cyanobacterial blooms. Water samples were collected at 26 sites across four contrasting reservoirs in trophic status to quantify pigment concentrations, phytoplankton density and biovolume, and cyanobacterial density and biovolume, while concurrent UAV surveys were conducted. Aerial imagery was processed to extract spectral-waveband values and calculate vegetation indices. RGB-derived variables correlated more strongly with measured responses than RGN-derived variables, with the red and green wavebands, Excess Green (ExG), and Green Leaf Index (GLI) showing the strongest relationships. Multiple linear regression yielded significant predictive models for chlorophyll-a (R2 = 0.98, limit of detection [LOD] ≈ 3.5 μg L–1), phycocyanin (R2 = 0.78, LOD ≈ 20 μg L–1), phytoplankton density (R2 = 0.84, LOD ≈ 6 organisms mL–1), cyanobacterial density (R2 = 0.75, LOD ≈ 24 organisms mL–1), phytoplankton biovolume (R2 = 0.78, LOD ≈ 97.7 mm3 L–1), and cyanobacterial biovolume (R2 = 0.83, LOD ≈ 69 mm3 L–1). Applying these models to the orthomosaics enabled spatial mapping of pigment concentrations and cyanobacterial biomass, demonstrating the potential of UAV-based imaging for rapid bloom monitoring and water quality management.