Determining Absorption and Scattering Coefficients by Optimizing Radiative Transfer Modelling with In-Situ Coastal Hyperspectral Reflectance Spectra

Monitoring ocean color is a crucial tool in understanding marine ecosystems and their health, as it provides quantitative information on chlorophyll-a (chla) concentration, absorption by colored dissolved organic matter (CDOM). Ocean color data can be used to characterize phytoplankton pigment groups from the ocean, by using a few color bands (multispectral imaging). Hyperspectral sensors have the potential to distinguish pigment groups in more detail. This capability is necessary for detailed monitoring of local ecosystems and improving predictions of algae blooms. As part of the European Copernicus Programme, the Copernicus Hyperspectral Imaging Mission for the Environment (CHIME) will provide routine hyperspectral observations globally over land and coastal zones in support to European Union policies for the management of natural resources, ecosystem services and societal benefits. CHIME primarily focuses on land applications, however, secondary applications include, among others, detection of floating debris and water quality monitoring in inland and coastal waters.

Radiative transfer modelling (RTM) is a useful tool for determining the expected level of detail in which phytoplankton pigment groups can be distinguished with a new observing system (e.g. satellite mission). The forward RTM simulation uses assumed absorption and scattering properties of phytoplankton in each depth layer to calculate light progression and the reflectance spectrum. Since these properties are difficult to determine and vary with factors such as depth [1] and the community that the pigment groups appear in [2], it can lead to inaccurate model output.

During an observation campaign in September 2024, we sampled at two different locations in Trondheimsfjorden: one located in the middle of the fjord, to compare results with satellite data while minimizing land-mixing, and the other located at the critical Gaula river outlet, which is important for sediment and nutrient transport into the fjord. We measured among others phytoplankton accessory pigments and taxonomy, CDOM, the surface reflectance spectrum, and the downwelling irradiance in the water column. This data allows us to compare the model output to the observation while we tune absorption and scattering properties. An optimization process lets us tune them to minimize the deviation in the output, thereby resulting in estimates of absorption and scattering properties.

The forward RTM exercise was set up to simulate the reflectance of the water surface and the downwelling irradiance in the water column. Measured chla and CDOM concentrations were used as inputs. In the optimization exercise, the sum of squares of the difference between the measured vertical downwelling irradiance profile and the simulated vertical downwelling irradiance profile is minimized, by tuning the absorption and scattering coefficients. For testing, the measured reflectance spectrum and RTM simulated reflectance spectrum were compared.

The final absorption and scattering coefficients were compared to measured absorption properties [3] and simulated absorption and scattering properties [2].  These simulations can better determine the sensitivity of future hyperspectral missions (e.g. CHIME) to distinguishing phytoplankton pigment groups.

[1] Sundarabalan et al., 2013. Journal of Quantitative Spectroscopy and Radiative Transfer. https://doi.org/10.1016/j.jqsrt.2013.01.016
[2] Lain et al., 2023. Scientific Data. https://doi.org/10.1038/s41597-023-02310-z
[3] Johnsen et al., 2007. Journal of Phycology. https://doi.org/10.1111/j.1529-8817.2007.00422.x