Evidence for satellite observed changes in the relative influence of climate indicators on autumn phenology over the Northern Hemisphere

Recent climate changes have elicited diverse phenological shifts in the ecosystems of the Northern Hemisphere (NH), with changes in the end growth of season (EOS) varying considerably with time. Due to the anthropogenic forcing (e.g. anthropogenic radiative forcing, et al.), both the climate conditions and EOS greatly changed since 2000, but the relationship between the two remains unclear. To better understand the responses of EOS to climate change with time, we used Global Inventory Modeling and Mapping Studies (GIMMS) derived normalized difference vegetation index (NDVI) and temperature, precipitation, and cloud cover during 1982–2000 and 2001–2015 separately, aiming to identify and quantify the spatiotemporal distributions of drivers of EOS across different biomes over the NH (>30°). Considering separately these two distinct periods, we found that beginning in the 21st century both the proportion of positive significant pixels (increased by 9.2%) as well as the sensitivity (1.5 ± 4.3 day °C⁻¹ vs. 2.7 ± 7.9 day °C⁻¹) of the temperature vs. EOS relationship had increased. Precipitation affected EOS over a slightly increase area during these two periods, but the sensitivity decreased. Conversely, we found that the sensitivity of EOS to light availability increased for most of biomes, whereas the areas influenced increased by 7.9%. With respect to the sign of relationships between EOS and its drivers, we found that for high latitude regions (i.e., boreal forests and tundra), precipitation advanced EOS during 2001–2015 but delayed EOS during 1982–2000. Although precipitation has significant correlations for a larger area than temperature, its sensitivity was lower. Our results imply that the relative magnitude and sign of drivers of EOS has changed substantially over time and space. Improved representation of these dynamic changes will advance understanding of future climate change on plant phenology by ecosystem models.