Mapping vegetation phenology in the high Arctic is critical for understanding ecosystem responses to rapid warming, but persistent cloud cover and low solar angles pose substantial challenges for optical satellite monitoring. We evaluated whether Landsat 8 can successfully map the start of growing season (SOS) at 78°N in Svalbard by developing spectral-based cloud detection algorithms tailored to Arctic conditions. From 17 images during the 2014 growing season, we generated cloud-free NDVI time series and extracted SOS dates at 30-m resolution across Adventdalen (Advent Valley). Validation against six field plots equipped with phenocameras demonstrated strong agreement (R² = 0.884, RMSE = 4.1 days, NSE = 0.869). The resulting SOS map revealed systematic spatial patterns ranging from late May to early August, with elevation explaining 6.6% of variance as the strongest single predictor. Combined topographic models (elevation, slope, distance from sea) explained 14.6% of variance, indicating that unmeasured factors particularly aspect, snow distribution, and vegetation composition exert dominant controls on phenological timing. These results confirm the viability of Landsat-based phenology mapping in the high Arctic. Critically, by establishing a validated processing framework this study provides the methodological foundation for application to the historical Landsat archive (Landsat 4/5 TM and Landsat 7 ETM+), subject to cross-sensor radiometric calibration. This opens unprecedented opportunities for reconstructing 30-m resolution phenological time series extending back to 1984, enabling quantification of four decades of vegetation responses to Arctic warming in regions experiencing some of Earth's most rapid climate change.
Stendardi et al. (Wed,) studied this question.