Predicting water quality variability in a Mediterranean hypereutrophic monomictic reservoir using Sentinel-2 MSI: the importance of model functional form

Anthropogenic eutrophication is a pressing global environmental problem that threatens the ecological functions of many inland freshwaters and diminishes their abilities to meet their designated uses. Water authorities worldwide are being pressed to manage the negative consequences of harmful algal blooms (HABs) based largely on data collected from conventional monitoring programs that lack the needed spatio-temporal resolution for effective lake/reservoir management. This study assesses the potential of using Sentinel 2 MSI to predict and assess the spatio-temporal variability in the water quality of the Qaraoun Reservoir, a poorly-monitored Mediterranean hypereutrophic monomictic reservoir that is subject to extensive HABs during the growing season. The performance and transferability of water quality models previously calibrated based on Landsat 7 and 8 surface reflectance to predict Chlorophyll-a (Chl-a), total suspended solids (TSS), Secchi Disk Depth (SDD), and Phycocyanin (PC) levels in the reservoir are first assessed. The results showed poor transferability between Landsat and Sentinel 2, with all models experiencing a significant drop in their predictive skill. Sentinel 2 specific models were then developed for the reservoir based on 153 water quality samples collected over two years. Different model functional forms were then tested, including multiple linear regressions (MR), multivariate adaptive regression splines (MARS), and support vector regressions (SVR). Our results showed that for Chl-a, the MARS model outperformed MR and SVR, with an R² of 60%. Meanwhile, the SVR-based models outperformed their MR and MARS counterparts for TSS, SDD and PC (R² = 59%, 94%, and 72% respectively).