Oral Session 4 | Tuesday, October 4, 15:00–15:20 | Abstract 633
Improving phytoplankton classification from hyperspectral measurements taking the SNR into account
The many bands and the high spectral resolution of hyperspectral sensors such as PRISMA, DESIS or EnMAP appear very promising for phytoplankton classification, but their increased sensor noise compared to multispectral sensors imposes limitations on threshold concentrations and the number of phytoplankton groups that can be distinguished. An analytic equation for a spectral weighting function (w) of the sensor bands is presented which optimizes the retrieval of phytoplankton groups from hyperspectral data. The function w depends on the reflectance differences (dR) induced by variable phytoplankton type and concentration, and on the signal-to-noise ratio (SNR) of the measurement. Extensive simulations covering wide concentration ranges of water constituents and major phytoplankton groups have been made to derive typical spectra of dR. Examples of w are presented based on these simulated dR spectra and on measured SNR spectra from hyperspectral satellite sensors. The improvement for phytoplankton classification is demonstrated for simulated measurements and for some hyperspectral images from PRISMA and DESIS.
Peter Gege, German Aerospace Center (DLR), 0000-0003-0939-5267