Divergent responses of grassland vegetation phenology to climate change at different scales

Understanding grassland phenology responses to climate change is of crucial importance for revealing regional and species differences in ecosystem processes. By means of thousands of ground observations and the counterparts derived from long-term remote sensing data, the spatiotemporal patterns of grassland phenology and its links to climate changes and biotic factors were investigated over the Northern Hemisphere. In site-species scale, the leaf-out date did not show significant variation trend for 68.6% of the site-species, while the leaf senescence date has significantly delayed for 31% of the site-species during 1982-2011 over China. Sunshine hours, temperature, precipitation, and the leaf-out date primarily explained the variation of leaf senescence for 36.6%, 31.7%, 22.0%, and 9.8% of the investigated site-species, respectively. Sunshine hours were the foremost factor in controlling leaf senescence for 60.1% of graminoid species and temperature for 42.3% of forb species. Specifically, in the Inner Mongolia Grassland, a predominant significant positive correlation between the leaf-senescence date and previous precipitation in 54.6% of site species. In pixel-landscape scale, a significant advance (P < 0.05) of the start of growing season (SOS) was detected in 23.2% of grid cells, while a predominantly and significantly delaying trend (P < 0.05) of the end of growing season (EOS) was identified in 20.5% of grid cells during 1981–2014 over the grasslands in the Northern Hemisphere. They jointly resulted in a primarily significant prolongation trend of growing season length in 22.7% of grid cells. Meanwhile, the time span of SOS/EOS (from the earliest SOS/EOS to the last SOS/EOS) and the growing season length (from SOS to EOS) have extended for the entire study region. For the Inner Mongolia Grassland, SOS was mainly controlled by pre-SOS precipitation with the sensitivity being largest in desert steppe. EOS was closely connected with pre-EOS air temperature in meadow steppe and typical steppe, but more closely related to pre-EOS precipitation in desert steppe. Moreover, a significant negative correlation between EOS and SOS was observed in part of grassland areas, but no significant relationship between NPP and EOS was observed. Aside from correlation analysis, a newly model by incorporating the effect of drought stress (CDDP) into the traditional chilling-degree-days (CDD) model was developed to simulate the leaf senescence. CDDP model was selected as the optimal model for 73% of site species with insufficient water supply in preseason, while CDD model was selected as the optimal model for 18% of site species with a relatively wet but cold preseason. Overall, they highlight the diverse responses in the timing of spring and autumn phenology to preceding temperature and precipitation in different grassland types and their dependence on species, functional-types, and geographical gradients.