Potential of Sentinel Images to Evaluate Physicochemical Parameters Concentrations in Water Bodies-Application in a Wetlands System in Northern Colombia

This research demonstrated the feasibility of applying Sentinel-2 images to generate empirical models and estimate physicochemical parameters concentration, particularly nutrients in the wetland system called Bajo Sinu wetlands complex, Colombia. Spearman correlations were determined between water quality parameters, which were monitored at 17 points in the wetland on 5 February 2021, with Sentinel-2 images reflectance values from the same monitoring date; the correlations allowed the identification of statistically significant bands in the multiple linear regression algorithm implementation to determine empirical water quality models. The results show significant correlations between the optically active parameters, TSS-Turbidity, which in turn correlated with the optically inactive parameters Turbidity-NO3 and TSS-DO, as well as non-optically active parameters among themselves, TDS-NO3 and TDS-TP; the empirical models presented higher than 74.5% fit (R-2), particularly DO (R-2 = 0.948), NO3 (R-2 = 0.858) and TP (R-2 = 0.779) were the models with the highest fits (R-2). These models allowed us to properly estimate the spatial distribution of nutrient-forming compounds in the wetlands complex. The determinant role played by turbidity in this type of water body is highlighted; it acts as a connecting constituent that makes the estimation of water quality parameters without spectral response through remote sensing feasible. Sentinel-2 images and multiple linear regression algorithms have been shown to be effective in estimating the concentration of water quality parameters without spectral response, such as NO3 and TP in shallow tropical wetlands, due to the processes of transformation, interaction and dependence between the different environmental variables in aquatic ecosystems.

Automated summary

This research demonstrates the feasibility of using Sentinel-2 images to generate empirical models for estimating physicochemical parameters, particularly nutrients, in the Bajo Sinu wetlands complex in Colombia. By correlating water quality parameters monitored at 17 points in the wetland with reflectance values from Sentinel-2 images, statistically significant bands were identified for the implementation of empirical water quality models using multiple linear regression algorithms. The results show significant correlations between optically active and non-optically active parameters, and the empirical models achieved high fits for DO, NO3, and TP.