Bannari et al. 2023. “A Simple Modeling-Approach For Mapping Submerged Aquatic Vegetation in Shallow Coastal Waters Based on Sentinel-MSI Data and Field Truth

Bannari, Abderrazak, Thamer-Salim Ali, and Asma Abahussain. 2023. “A Simple Modeling-Approach For Mapping Submerged Aquatic Vegetation in Shallow Coastal Waters Based on Sentinel-MSI Data and Field Truth.” Pasadena, California.

Abstract
Within the framework of the present study, semi-empirical models based on vegetation indices were evaluated, validated, and compared for mapping the percentage of submerged aquatic vegetation cover (PSAVC) in shallow water (0.5 to 20 m depth) located at the southwest of the Arabian Gulf. As well as, the potential of the blue and green bands instead of the red was evaluated to discriminate the PSAVC density classes. Sentinel-MSI image acquired almost simultaneously with field truth measurements was pre-proceed and used. A field sampling survey was conducted for the validation procedure, in which 55 sampling locations with variable biomass cover were selected. Likewise, the examined models’ efficiency and accuracy analysis (p ˂ 0.05) were carried out using field survey measurements. The results revealed that regardless of the integrated bands, the semi-empirical models integrating the indices WDVI, VARI, NDWI, NDVI, and WAVI are not conclusive, yielding an insignificant coefficient of determination (R2 ≤ 0.31) and a high root mean square error (0.30 ≤ RMSE ≤ 0.43). While the best performances are accomplished by the models incorporating WEVI, Diff(B-G), and WTDVI indices using the blue and green bands. These three indices achieve the best fits (0.89 ≤ R2) and an overall low RMSE of 0.10. This accuracy is equal or a bit less than that provided by the in situ sampling quadrant testifying to the ability of this simple semi-empirical approach for mapping PSAVC in shallow water using Sentinel-MSI data.


Fig. 4. Derived PSAVC map using the prediction equation based on the WTDVI index thresholding.