Monitoring water transparency of a hypertrophic lake (the Albufera of València) using multitemporal Sentinel-2 satellite images

espanolDesde los anos 70, la Albufera de Valencia es un lago hipertrofico, y a pesar de los esfuerzos para revertir el sistema a un estado de aguas claras, como el tratamiento de aguas residuales y la construccion de filtros verdes, estos no han dado los resultados esperados, clasificandose aun como malo segun la implementacion de la Directiva Marco del Agua en la legislacion espanola. Actualmente el lago requiere un monitoreo constante y la transparencia del agua, como profundidad del disco Secchi (SDD), es un parametro clave para el seguimiento de la calidad del agua. La teledeteccion ofrece ventajas sustanciales sobre los metodos tradicionales de monitoreo, siendo una herramienta optima para el monitoreo continuo del estado de calidad de las aguas superficiales. El objetivo es calibrar y validar un algoritmo para la estimacion de la SDD a partir de Sentinel-2 (A y B), con el sensor MSI, multiespectral (13 bandas) de 404 nm a 2200 nm, una resolucion espacial de 10, 20 y 60 m y una frecuencia temporal de 5 dias (revisita en el Ecuador), valores impensables en imagenes de libre acceso. De las 81 imagenes tomadas por el satelite entre 2016 y 2017, solo se pudieron utilizar 40 imagenes debido a la presencia de nubes. Se utilizo el software SNAP 5 para el procesamiento, utilizando la herramienta Sen2Cor para la correccion atmosferica y el algoritmo desarrollado para estimar la SDD del lago. Para la validacion con muestras de campo, se llevaron a cabo 20 campanas de muestreo, tomando 114 medidas de la SDD. Tambien se midio la concentracion de clorofila a en cada punto de muestreo y se recogieron datos hidrologicos, de precipitacion y viento. La calibracion muestra su robustez con un R2 de 0.673 utilizando 79 muestras. Los resultados de la validacion utilizando 35 muestras son muy buenos, con un RMSE de 0.06 m, mostrando la precision del algoritmo. De la interpretacion de los mapas tematicos, extrajimos que la variacion temporal de la SDD sigue un patron bimodal anual, donde el aumento de la SDD viene determinada por un aumento significativo de la renovacion del agua. El algoritmo desarrollado para estimar la SDD a partir de imagenes S2 es preciso y apropiado para su uso dentro de un protocolo para monitorizar el estado ecologico del lago EnglishThe Albufera of Valencia has been a hypertrophic lake since the 1970s. Extensive efforts to revert the system to a clear water state, such as wastewater treatment and green filters construction, have not yielded the desired results; Albufera is still qualified as “bad” according to the Spanish Water Framework Directive implementation. Currently, the lake requires constant monitoring, and water transparency, measured by Secchi disc depth (SDD), is a key parameter for evaluating water quality. Remote sensing offers substantial advantages over traditional monitoring methods such as SDD because it allows the quality of the surface waters to be continuously monitored. This work aimed to calibrate and validate an algorithm for SDD retrieval from Sentinel-2 (S2) (A and B) satellites with multispectral instrument (MSI) sensors (13 bands) from 404 nm to 2200 nm, spatial resolutions of 10, 20 and 60 m and a temporal frequency of 5 days (revisit at the equator)-values previously unattainable from open access images. The study was carried out with images from 2016 and 2017; only 40 of the 81 images of the Albufera captured by the S2 satellites could be used, mainly due to the presence of clouds. Once the images were downloaded, they were processed using SNAP 5 software. Images were then atmospherically corrected using the Sen2Cor tool, and the lake’s SDD was estimated using the developed algorithm. The estimated SDD data were validated against field samples; a total of 20 sampling campaigns were carried out to measure the SDD, and 114 samples were taken. Chlorophyll a concentrations from each sample point were also measured to allow for better data interpretation; hydrological, precipitation and wind data were also collected. The algorithm model’s calibration showed its robustness with an R2 of 0.673 using 79 samples. Validation of the algorithm’s accuracy using 35 samples produced a low root mean squared error of 0.06 m, indicating a perfect fit between the predicted and observed data. Interpretation of thematic maps showed that SDD temporal variations follow an annual bimodal pattern where the increase of SDD is determined by a significant increase in water renewal. The retrieval algorithm to estimate the SDD from S2 satellite images is accurate and appropriate to use within a protocol whose main purpose is to monitor the ecological status of the Albufera of Valencia.