Key points are not available for this paper at this time.
Monitoring vegetation is essential in Earth Observation (EO) due to its link with the global carbon cycle, playing a crucial role in ecosystem management. The fluorescence of chlorophyll (ChF) is a reliable indicator of plants' photosynthetic activity and growth, especially when they are experiencing unfavourable conditions, particularly in terrestrial wetlands. These wetlands are integral components of the landscape, contributing significantly to climate mitigation, adaptation, biodiversity, and the well-being of both the environment and humanity. We conducted a research study using the XGBoost machine learning algorithm to map the chlorophyll fluorescence parameter Fv/Fm in the Biebrza River Valley, which is known for its marshes, peatlands, and diverse flora and fauna. Our study highlights the benefits of using ensemble classifiers derived from EO Sentinel-2 satellite imagery for accurately mapping Fv/Fm across terrestrial landscapes under the Ramsar Convention at Narew River Valley (Poland) and Čepkeliai Marsh (Lithuania). The XGBoost algorithm provides an accurate estimate of ChF with a robust determination coefficient of 0.747 and minimal bias at 0.013, as validated using in situ data. The precision of Fv/Fm chlorophyll fluorescence parameter estimation from remote sensing sensors depends on the growth stage, emphasizing the importance of identifying the optimal overpass time for S-2 observations. Our study found that biophysical factors, as denoted by spectral indices related to greenness and leaf pigments, were highly impactful variables among the top classifiers. However, incorporating soil, vegetation and meteorological indicators from remote sensing data could further increase the accuracy of chlorophyll fluorescence mapping.
Bartold et al. (Wed,) studied this question.