Temporal patterns of vegetation phenology and their responses to climate change in mid-latitude grasslands of the Northern Hemisphere

Grassland ecosystem is greatly sensitive to regional and global climate changes. In this study, the start (SOS) and end (EOS) date of growing season were extracted from NDVI data (1981 2014) across the mid-latitude (30°N 55°N) grasslands of Northern Hemisphere. We first validated their accuracy by ground observed phenological data and phenological metrics derived from gross primary production (GPP) data. And then, main climatic factors influencing the temporal patterns of SOS/EOS were explored by means of gridded meteorological data and partial correlation analysis. Based on the results of above statistical analysis, the similarities and differences of spring and autumn phenological responses to climate change among North American grasslands, Mid-West Asian grasslands, and Mongolian grasslands were analyzed. The main results and conclusions are as follows. First, a significant positive correlation was found between SOS/EOS and observed green-up/brown-off date (P<0.05) and GPP-based SOS/EOS (P<0.05), which means remote sensed SOS/EOS can reflect temporal dynamics of terrestrial vegetation phenology. Second, SOS in Mid-West Asian grasslands showed a significant advancing trend (0.22 days/year, P<0.01), whereas the trend of SOS in North American grasslands and Mongolian grasslands was not significant. EOS in North American grasslands (0.31 dyas/year, P<0.01) and Mongolian grasslands (0.09 days/year, P<0.05) both presented a significant delaying trend, but the trend of EOS in Mid-West Asian grasslands was not significant. Furthermore, the correlation analysis of SOS/EOS inter-annual fluctuations and hydrothermal factors showed that a significant negative correlation was found between SOS and the pre-season temperature in 41.6% of pixels (P<0.05), while a significant negative/positive correlation was detected between SOS and pre-season rainfall/snowfall in 14.6%/19.0% of pixels (P<0.05). EOS was significantly positively correlated with pre-season rainfall in 34.5% of pixels (P<0.05), and significantly negatively/positively correlated with pre-season temperature in 12.1%/11.9% of pixels (P<0.05). This indicates that the fluctuations of SOS and EOS are mainly affected by pre-season temperature and pre-season rainfall.