Assessment of water quality using hyperspectral and multispectral data for the characterization of eutrophication in Lake Villarrica.
Loading...
Date
2026
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
El lago Villarrica, situado en el sur de Chile, es un recurso vital de agua dulce cuyo estado ecológico requiere una evaluación continua. El control de la calidad de sus aguas es esencial para detectar procesos de eutrofización. La clorofilaa (Chl-a) es un indicador clave de la biomasa de fitoplancton y su estimación mediante sensores satelitales permite un control eficiente y a gran escala. En este estudio se comparó el rendimiento de diferentes modelos empíricos basados en datos de reflectancia obtenidos a partir de imágenes satelitales corregidas atmosféricamente utilizando el software ACOLITE, calibrados con mediciones in-situ de Chl-a recogidas durante las estaciones de primavera y verano entre 2014 y 2024. Para cada sensor, se seleccionó la mejor combinación de bandas espectrales y se generaron modelos utilizando un procedimiento de bootstrapping con 1000 iteraciones para obtener coeficientes de regresión robustos; los modelos finales se definieron utilizando la mediana de estos coeficientes. El modelo con mejor rendimiento para Landsat-8 y 9 se basó en una combinación de bandas azul y roja (R2= 0.79, RMSE = 2.1 μg ·L−1, MAE = 1.2 μg ·L−1, n = 74), mientras que para Sentinel-2, el modelo óptimo utilizó las bandas azul y verde (R2= 0.75, RMSE = 0.8 μg ·L−1, MAE = 0.72 μg ·L−1, n = 112).En general, los resultados obtenidos mediante teledetección revelan un aumento gradual de los niveles de Chl-a durante la última década. Esta tendencia podría estar relacionada tanto con el calentamiento climático como con el aumento de las presiones antropogénicas, lo que refuerza la necesidad de contar con sistemas de vigilancia continua basados en observaciones satelitales. Ampliar la base de datos in-situ de Chl-a, aumentar la disponibilidad de imágenes satelitales e incorporar mediciones de reflectancia in-situ y cámaras multiespectrales es esencial para mejorar la precisión de los modelos y superar los retos que plantean la cobertura nubosa y los aerosoles.
Lake Villarrica, located in southern Chile, is a vital freshwater resource whose ecological status requires continuous evaluation. Monitoring its water quality is essential for detecting eutrophication processes. Chlorophyll-a (Chl-a) is a key indicator of phytoplankton biomass and estimating it using satellite sensors enables efficient and large-scale monitoring. In this study, the performance of different empirical models based on reflectance data obtained from atmospherically corrected satellite images using ACOLITE software, calibrated with in-situ Chl-a measurements collected during the spring and summer seasons between 2014 and 2024, was compared. For each sensor, the best combination of spectral bands was selected, and models were generated using a bootstrapping procedure with 1000 iterations to obtain robust regression coefficients; the final models were defined using the median of these coefficients. The best-performing model for Landsat-8 and 9 was based on a combination of blue and red bands (R2= 0.79, RMSE = 2.1 μg ·L−1, MAE = 1.2 μg ·L−1, n = 74), while for Sentinel-2, the optimal model used the blue and green bands (R2= 0.75, RMSE = 0.8 μg ·L−1, MAE = 0.72 μg ·L−1, n = 112). In general, the results obtained through remote sensing reveal a gradual increase in Chl-a levels over the last decade. This trend could be associated with both climate warming and increasing anthropogenic pressures, reinforcing the need for continuous monitoring systems based on satellite observations. Expanding the in-situ Chl-a database, increasing satellite image availability, and incorporating insitu reflectance measurements and multispectral cameras are essential to enhance model accuracy and overcome challenges posed by cloud cover and aerosols.
Lake Villarrica, located in southern Chile, is a vital freshwater resource whose ecological status requires continuous evaluation. Monitoring its water quality is essential for detecting eutrophication processes. Chlorophyll-a (Chl-a) is a key indicator of phytoplankton biomass and estimating it using satellite sensors enables efficient and large-scale monitoring. In this study, the performance of different empirical models based on reflectance data obtained from atmospherically corrected satellite images using ACOLITE software, calibrated with in-situ Chl-a measurements collected during the spring and summer seasons between 2014 and 2024, was compared. For each sensor, the best combination of spectral bands was selected, and models were generated using a bootstrapping procedure with 1000 iterations to obtain robust regression coefficients; the final models were defined using the median of these coefficients. The best-performing model for Landsat-8 and 9 was based on a combination of blue and red bands (R2= 0.79, RMSE = 2.1 μg ·L−1, MAE = 1.2 μg ·L−1, n = 74), while for Sentinel-2, the optimal model used the blue and green bands (R2= 0.75, RMSE = 0.8 μg ·L−1, MAE = 0.72 μg ·L−1, n = 112). In general, the results obtained through remote sensing reveal a gradual increase in Chl-a levels over the last decade. This trend could be associated with both climate warming and increasing anthropogenic pressures, reinforcing the need for continuous monitoring systems based on satellite observations. Expanding the in-situ Chl-a database, increasing satellite image availability, and incorporating insitu reflectance measurements and multispectral cameras are essential to enhance model accuracy and overcome challenges posed by cloud cover and aerosols.
Description
Tesis presentada para optar al grado de Magíster en Geofísica.
Keywords
Water quality, Eutrophication, Lakes, Biomass, Phytoplankton