Short-term vegetation loss versus decadal degradation of grasslands in the Caucasus based on Cumulative Endmember Fractions

Quantifying drought effects in Central European grasslands through regression-based unmixing of intra-annual Sentinel-2 time series

“…This study demonstrates the opportunities provided by the Sentinel-2 mission for retrieving fractional cover time series for grassland ecosystems. Our findings from a Central European region complement similar studies based in other regions or on other sensors (e.g., Guerschman et al, 2009;Lewińska et al, 2020). The trajectory of PV fractional cover represented the expected seasonal phenology well within individual years, with highest growth rates, i.e., increase in PV fractions in late spring to early summer, and growth cessation, i.e., decrease in PV fractions, in late summer to autumn with decreasing photoperiod and beginning cold acclimation at temperatures of 5-10 • C (Wingler and Hennessy, 2016).…”

“…To ensure consistent and comparable PV, NPV, and soil cover estimates through time, we selected a global set of library spectra representative of the entire time series based on the triangular NDVI/ SWIR ratio feature space from 4 years of Sentinel-2 data. We followed the common approach to choose one representative spectral signature for each cover class (e.g., Lewińska et al, 2020;Röder et al, 2008) from the vertices of the feature space (Guerschman et al, 2009;G. Wang et al, 2019a).…”

“…However, compared to Sentinel-2, Landsat has a considerably lower observation frequency. Sparse time series require modified approaches such as pooling observations from several years (Melaas et al, 2013;Senf et al, 2017) or annual cumulative fractional cover estimates (Lewińska et al, 2020). Detailed intra-annual analyses of grassland time series are thus limited to recent years with Sentinel-2 data.…”

“…Mapping time series of photosynthetic vegetation (PV), nonphotosynthetic vegetation (NPV), and soil fractional cover constitutes a universally applicable framework for characterizing the state and dynamics of land cover in a quantitative way. With regard to grasslands, fractions of PV, NPV, and soil provide comprehensive measures of the essential ground cover components in interpretable quantitative units (Lewińska et al, 2020). The temporal interactions between these components support a better understanding of processes related to management and environmental conditions such as drought (Guerschman et al, 2020).…”

“…Using linear unmixing, Asner and Heidebrecht (2002) and Guerschman et al (2009) assessed the ability of multi-and hyperspectral sensors to estimate PV, NPV, and soil fractions in desert grass-and shrublands, and savanna ecosystems, respectively. Lewińska et al (2020) used SMA to analyze short-and long-term changes and trends of vegetation cover in the Caucasus with multispectral Landsat data. Similarly, Numata et al (2007) found that cover fractions derived from a linear mixing model were suitable for monitoring pasture conditions in Brazil.…”