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

Anthropogenic eutrophication is a global environmental problem threatening the ecological functions of many inland freshwaters and diminishing their abilities to meet their designated uses. Water authorities worldwide are being pressed to improve their abilities to monitor, predict, and manage the incidence of harmful algal blooms (HABs). While most water quality management decisions are still based on conventional monitoring programs that lack the needed spatio-temporal resolution for effective lake/reservoir management, recent advances in remote sensing are providing new opportunities towards better understanding water quality variability in these important freshwater systems. This study assessed the potential of using the Sentinel 2 Multispectral Instrument 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 periods of HABs. The work first evaluated the ability to transfer and recalibrate previously developed reservoir-specific Landsat 7 and 8 water quality models when used with Sentinel 2 data. The results showed poor transferability between Landsat and Sentinel 2, with most models experiencing a significant drop in their predictive skill even after recalibration. Sentinel 2 models were then developed for the reservoir based on 153 water quality samples collected over 2 years. The models explored different functional forms, including multiple linear regressions (MLR), multivariate adaptive regression splines (MARS), random forests (RF), and support vector regressions (SVR). The results showed that the RF models outperformed their MLR, MARS, and SVR counterparts with regard to predicting chlorophyll-a, total suspended solids, Secchi disk depth, and phycocyanin. The coefficient of determination (R2) for the RF models varied between 85% for TSS up to 95% for SDD. Moreover, the study explored the potential of quantifying cyanotoxin concentrations indirectly from the Sentinel 2 MSI imagery by benefiting from the strong relationship between cyanotoxin levels and chlorophyll-a concentrations.

Integration of Remote Sensing and Mexican Water Quality Monitoring System Using an Extreme Learning Machine

Remote Sensing, as a driver for water management decisions, needs further integration with monitoring water quality programs, especially in developing countries. Moreover, usage of remote sensing approaches has not been broadly applied in monitoring routines. Therefore, it is necessary to assess the efficacy of available sensors to complement the often limited field measurements from such programs and build models that support monitoring tasks. Here, we integrate field measurements (2013–2019) from the Mexican national water quality monitoring system (RNMCA) with data from Landsat-8 OLI, Sentinel-3 OLCI, and Sentinel-2 MSI to train an extreme learning machine (ELM), a support vector regression (SVR) and a linear regression (LR) for estimating Chlorophyll-a (Chl-a), Turbidity, Total Suspended Matter (TSM) and Secchi Disk Depth (SDD). Additionally, OLCI Level-2 Products for Chl-a and TSM are compared against the RNMCA data. We observed that OLCI Level-2 Products are poorly correlated with the RNMCA data and it is not feasible to rely only on them to support monitoring operations. However, OLCI atmospherically corrected data is useful to develop accurate models using an ELM, particularly for Turbidity (R² = 0.7). We conclude that remote sensing is useful to support monitoring systems tasks, and its progressive integration will improve the quality of water quality monitoring programs.