Chlorophyll Estimation on Hypso-1 Using Ensemble Machine Learning.

This work compares different chlorophyll inversion techniques based on features created with surface reflectance R rs . Band-difference and ratio, as well as more complex biomass descriptors such as TBVI and TBM, can be used to correlate chlorophyll with changes in the spectra. The 6SV1 model will be used to retrieve R rs from HYPSO-1 spectral images through atmospheric correction so that relevant chlorophyll descriptors can be fine-tuned for the hyperpsectral camera used. Based on the findings of this work, an ensemble regression model with optimal descriptors can estimate biomass concentrations in the ocean better than a multivariate linear and polynomial regression. The potential to improve biomass concentration estimates from surface reflectance has been demonstrated in the study.