Applying a calibrated primary production model to the Scheldt Estuary using Envisat-MERIS images

Abstract

Using variations of the BPI model by Cole & Cloern (1987), primary production in the Scheldt Estuary’s two contrasting regions (Oosterschelde and Westerschelde) was first modelled using in-situ values and calibrated for. Primary production was successfully modelled using in-situ values, returning r2 values of 0.75 and 0.89 for the Oosterschelde and Westerschelde respectively using only the simple BPI model. Later, in-situ values of individual variables were compared with MERIS-derived estimates. Results show that retrieval of these variables works better for the Oosterschelde than the Westerschelde. Using the calibrated variations of the BPI model, these were applied onto the MERIS satellite imagery. Primary production was more successfully modelled in the Oosterschelde than the Westerschelde, with r2 values of 0.816 compared to 0.2 respectively for the simple BPI model alone. Monthly composite images during March-September 2011 revealed reasonable simulation of trends in GPP values for the Westerschelde, however, there were consistently missing pixel values for Kd in the Oosterschelde, resulting in the inability to model GPP in the Oosterschelde. This error appears to only affect composite images, as the same does not occur for exact date matchups, suggesting that this might be a result of a processing issue. This pilot project into remote sensing of the Scheldt Estuary demonstrates that the possibility of using satellite imagery to rapidly model primary production in coastal systems similar to the Oosterschelde remains promising. In addition, MERIS product algorithms needs to be validated and trained against datasets of regions with high turbidity and SPM concentrations, if improvements of modelling the Westerschelde are to be made.