Preliminary Results on Blue Carbon Content Mapping in Coastal Waters of the Arabian Gulf Using Satellite-Based Modeling Approach

Seagrass beds play an important role in carbon storage, as well as an effective removal of CO2 from the “biosphere-atmosphere” system, which plays a significant role in the amelioration of climate change impacts. The present study focused on carbon content modeling and mapping in submerged aquatic vegetation (SAV) including seagrass and algae species in coastal water areas (-0.5 to -7 m depth) of the Arabian Gulf using field sampling, laboratory analyses and satellite-based approach. The used Landsat-OLI image, acquired simultaneously with field sampling, was preprocessed rigorously to remove all radiometric and geometric anomalies. The modeling is applied on 55 samples collected from the seabed including biomass of the canopy (above-ground biomass: AGB), roots and rhizomes (below-ground biomass: BGB) and sediments (within a depth 15 cm). Each sampling location was automatically labeled and recorded using a “Deep-Blue-Pro” underwater digital camera equipped with a sub-meter GPS survey and connected to a laptop computer allowing a real time viewing of seabed. The carbon content was analyzed in the laboratory considering separately the AGB, BGB, and sediments in each sample, and then converted and calculated at the OLI pixel size level. A linear regression at 95% confidence interval was applied between the obtained carbon content at the pixel level and: a) calibrated reflectances in the VNIR spectral bands, b) water spectral indices, and c) percentage of SAV cover (PSAVC) fraction determined by a semi-empirical relation developed previously. The results revealed that the best fit (R2 = 0.87) was obtained between PSAVC and the total carbon content derived from the biomass (AGB and BGB) and sediments. The cross-validation of derived carbon map based on the proposed modeling approach showed a good correlation (R2 = 0.74) and an acceptable RMSE of ±0.4 kg/m-. However, despite the limited number of sampling points, the uncertainly of carbon content scaling up at the pixel level, and the errors propagation in modeling process, these preliminary results are encouraging for carbon content modeling and mapping over a large area in shallow water based on optical remote sensing.