SCHELDT - Airborne Hyperspectral Potential for Coastal Biogeochemistry of the Scheldt Estuary and Plume

Context and objectives

Estuaries are obligate pathways for the transfer of dissolved and particulate material from the continent to the marine system. The tidal regime of some estuaries leads to an increased residence time of the freshwater in the estuarine mixing zone and pronounced changes in the speciation of elements. European estuaries are subject to intense anthropogenic disturbance reflected in elevated loading of detrital organic matter, which induce high respiration rates and the production of large quantities of dissolved CO2.
In the present days, researches on the functioning of estuarine and coastal ecosystems are based on highly time consuming, costly sea campaigns and laboratory analyses. Although optical spaceborne remote sensing already proved useful in such coastal ecosystems studies, hyperspectroscopy opened a new dimension by allowing improved distinction of various biogeochemical compounds through characteristic spectral signature identification.

Project outcome

Expected scientific results

Ratio Multiple Regression and well as First Derivative approach proved very convenient for first statistical exploration of large hyperspectral databases.
Some results obtained are very encouraging in terms of correlation coefficient (r²), range and also distribution (i.e. CDOM).
In case of limited number of ground truth stations, attention must be paid to the general distribution range and pattern established in order to avoid inconsistent correlations.

CASI data recorded over water seems affected by :
a)“sun glim” and cross track systematic artefacts as revealed clearly by some parameter distribution (i.e. DIC)
b)Bathymetry effects (i.e. DOC)

Combining our previous empirical approach with (a) physical models including fine bathymetry effect and (b) classical hyperspectral algorithms making use of known isolated spectral signatures (from literature and laboratory spectral measurements on water samples) would also allow to improve prediction of in situ parameters from remote sensing hyperspectral data.