Hyperspectral determination of eutrophication for a water supply source via genetic algorithm–partial least squares (GA–PLS) modeling

• Published in: The Science of the total environment Vol. 426; pp. 220 - 232
• Main Authors: Song, Kaishan, Li, Lin, Tedesco, Lenore P., Li, Shuai, Clercin, Nicolas A., Hall, Bob E., Li, Zuchuan, Shi, Kun
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier B.V 01.06.2012; Elsevier
• Subjects: Airborne sensing; Algorithms; Applied sciences; Chlorophyll - analysis; Chlorophyll-a; Cyanobacteria; Drinking water and swimming-pool water. Desalination; Environmental Monitoring - methods; Eutrophication; Exact sciences and technology; Fresh Water - chemistry; GA–PLS; Hyperspectral; Indiana; Least-Squares Analysis; Mathematical models; Models, Chemical; Phycocyanin; Phycocyanin - analysis; Pollution; Polycarbonates; Spectra; Water Pollutants - analysis; Water Pollution - statistics & numerical data; Water quality; Water supplies; Water Supply - statistics & numerical data; Water treatment and pollution; Indiana; United States; North America; America; USA, Indiana, Indianapolis; Phycocyanin; Chlorophyll-a; Hyperspectral; GA–PLS; Hyperspectral imagery; Urban area; Tap water; Modeling; Remote sensing; Eutrophication; Statistical method; PLS regression; Genetic algorithm; GA-PLS; Bloom; Cyanobacteria; Water quality; Bacteria; Water pollution
• ISSN: 0048-9697; 1879-1026
• DOI: 10.1016/j.scitotenv.2012.03.058

Morse Reservoir (MR), a major source of the water supply for the Indianapolis metropolitan region, is now experiencing nuisance cyanobacterial blooms. These blooms cause water quality degradation, as well as reducing the aesthetic quality of water by producing toxins, scums, and foul odors. Hyperspectral remote sensing data from both in situ and airborne AISA measurements were applied to GA–PLS by relating the spectral signal with measured water eutrophication parameters, e.g., chlorophyll-a (Chl-a), phycocyanin (PC), total suspended matter (TSM), and Secchi disk depth (SDD). Our results indicate that GA–PLS relating field sensor acquired spectral reflectance to the above-mentioned four parameters yielded low root mean square error between measured and estimated Chl-a (RMSE=10.4; Range (R): 1.8–215.8μg/L), PC (RMSE=18.6; R: 1.4–371.0μg/L), TSM (RMSE=3.8; R: 3.6–81.4mg/L), SDD (RMSE=5.8; R: 25–135cm) for MR. The GA–PLS model also yielded high performance with AISA image spectra, and the RMSEs were 12.1μg/L, 25.3μg/L, 5.9mg/L and 5.7cm, respectively for Chl-a, PC, TSM, and SDD. Four water quality parameters were mapped with GA–PLS using AISA hyperspectral image. Based on these results, in situ and airborne hyperspectral remote sensors can provide both quantitative and qualitative information on the distribution and concentration of cyanobacteria, suspended matter, and transparency in MR. ►We estimate chlorophyll-a, phycocyanin, and total suspended matter using hyperspectral data. ►The hybrid GA–PLS model is an effective approach for water quality estimation. ►The GA–PLS coupled with hyperspectral data is an effective tool for water resource managers.