What can Sentinel-2 time series tell us about the phenology of dominant deciduous tree species in Poland?

Abstract Vegetation phenophases serve as indicators of long-term climate-driven environmental changes, impacting ecosystem functions like the carbon cycle and plant–animal interactions. Satellite-based phenology assessments, particularly with high-spatial and temporal resolution Sentinel-2 data, enable large-scale monitoring where field observations are scarce. This study modeled the start of the season (SOS) for deciduous species in Poland using Sentinel-2 data (2018–2023). We tested four vegetation indices [MTCI (MERIS Terrestrial Chlorophyll Index), EVI (Enhanced Vegetation Index), NDMI (Normalized Difference Moisture Index), VARI (Visual Atmospheric Resistance Index)] for their ability to correctly capture SOS in silver birch stands by comparing Sentinel-2 based SOS dates with ground observations, which were available only for this species. We then used the best-performing index to map SOS for black alder, pedunculate and sessile oak, and common beech. Results showed that Sentinel-2 indices effectively track SOS patterns at national scale, though different indices yielded varying SOS dates: EVI and NDMI estimates were similar, MTCI indicated later SOS with high variability, and VARI showed the earliest dates. Correlations with ground data were moderate (coefficients >0.5) and varied yearly. EVI provided the most stable results, making it most suitable for tracking other species. The earliest SOS dates appeared in 2018 and the latest in 2021, with birch generally showing the earliest SOS and oak the latest. Spatially, SOS started earlier in western lowlands and later in the Baltic moraine hills, northeast, and mountainous areas. Annual shifts likely reflected air temperature variability in spring, as shown in a preliminary growing degree days analysis. Our findings enhance understanding of forest phenology and demonstrate remote sensing’s potential for tracking large-scale seasonal changes. Uncertainties resulting from limited and variable ground observations, spatial mismatches between field and satellite estimations, and possible gaps in Sentinel-2 coverage highlight the challenges of validating SOS estimates; nonetheless, Sentinel-2 enables the detection of meaningful regional and species-level phenological patterns. Expanding the Sentinel-2 archives will further enhance this capability. Future studies should explore the relationship between SOS and temperature metrics, drawing on high-resolution gridded meteorological data to capture spatial patterns in spring temperature conditions and support spatial analysis of SOS–temperature relationships.