Spatio-temporal analysis of suspended sediments in two estuaries in Atlantic Canada with remote sensing

This study evaluates suspended sediment concentration (SSC) in two estuaries in New Brunswick, Canada (Bouctouche and Cocagne). A linear relationship was first established between turbidity and SSC (R2 = 0.97 and 0.94). Water-leaving reflectance was retrieved from Sentinel-2A and 2B imagery using the ACOLITE atmospheric correction. Turbidity was then estimated from satellite images using in situ turbidity data and a semi-empirical algorithm applied to the red-edge band (704 nm), selected based on a combined evaluation of statistical performance and physical consistency of retrieval patterns. This model yielded R2 values of 0.40 in Bouctouche and 0.63 in Cocagne. Based on imagery from 2016 to 2024, seasonal patterns were detected in both estuaries, with higher SSC observed in April and May, likely due to increased river flow and snowmelt. The analysis also indicates that Bouctouche exhibits higher SSC values than Cocagne. A generalized linear model (GLM) was fitted using images from 2016 to 2023 where top-of-atmosphere SWIR (1600 nm) reflectance is below 0.02, minimizing sun glint and adjacency effects. The GLM predicts median SSC in each estuary using wind data, tidal levels and freshwater inflow, achieving R2 values of 0.55 in Bouctouche and 0.54 in Cocagne in fivefold cross-validation. In both estuaries, higher SSC was associated with increased river discharge, stronger winds and ebb tides. Additionally, in Cocagne, lower tidal levels were also linked to elevated SSC. Finally, the study investigates longitudinal SSC variations along the estuarine channels. A decreasing gradient in SSC from upstream to downstream was observed in 67% of Bouctouche images and 82% of Cocagne images. These results suggest that sediment inputs originate both from the watershed and upstream river flow.