A Machine Learning Approach for Mapping Chlorophyll Fluorescence at Inland Wetlands

Wetlands are a critical component of the landscape for climate mitigation, adaptation, biodiversity, and human health prosperity. Keeping an eye on wetland vegetation is crucial due to it playing major role in planet’s carbon cycle ecosystem management. By measuring chlorophyll fluorescence (ChF) emitted by plants, we can get precise understanding current state photosynthetic activity. In this study, applied Extreme Gradient Boost (XGBoost) algorithm map ChF Biebrza Valley, which has unique Europe peatlands, as well highly diversified flora fauna. Our results revealed advantages using set classifiers derived from EO Sentinel-2 (S-2) satellite image mosaics accurately spatio-temporal distribution terrestrial landscape. The validation proved that XGBoost quite accurate estimating with good determination 0.71 least bias 0.012. precision measurements reliant upon determining optimal S-2 overpass time, influenced developmental stage plants at various points during growing season. Finally, model performance indicated biophysical factors characterized greenness- leaf-pigment-related spectral indices. However, utilizing indices based extended periods remote sensing data better capture land phenology features improve accuracy mapping fluorescence.