Temperature controls on the spatial pattern of tree phenology in China's temperate zone

We used Ulmus pumila leaf unfolding and leaf fall data collected at 46 stations during the 1986–2005 period to construct and validate daily temperature-based spatial phenology models. These models allowed simulation of the 20-year mean and yearly spatial patterns of U. pumila growing season beginning and end dates. This work was undertaken to explore the ecological mechanisms driving tree phenology spatial patterns and examine tree phenology spatial responses to temperature across China's temperate zone. The results show that spatial patterns of daily temperatures within the optimum spring and autumn length periods control spatial patterns of growing season beginning and end dates, respectively. Regarding 20-year mean growing season modeling, mean growing season beginning date correlates negatively with mean daily temperature within the optimum spring length period at the 46 stations. The mean spring spatial phenology model explained 90% of beginning date variance (P < 0.001) with a Root Mean Square Error (RMSE) of 4.6 days. In contrast, mean growing season end date correlates positively with mean daily temperature within the optimum autumn length period at the 46 stations. The mean autumn spatial phenology model explained 82% of end date variance (P < 0.001) with a RMSE of 5.6 days. On average, a spatial shift in mean spring and autumn daily temperatures by 1 °C may induce a spatial shift in mean beginning and end dates by −3.1 days and 2.6 days, respectively. Similarly, a significant negative and positive correlation was detectable between beginning date and spring daily temperature and between end date and autumn daily temperature at the 46 stations for each year, respectively. In general, the explained variances for yearly spatial phenology models are less than those of mean spatial phenology models, whereas the RMSEs of yearly models are greater than those of mean models. On average, a spatial shift in spring and autumn daily temperatures by 1 °C in a year may induce a spatial shift in beginning and end dates between −4.28 days and −2.75 days and between 2.17 days and 3.16 days in the year, respectively. Moreover, both mean and yearly spatial phenology models perform satisfactorily in predicting beginning and end dates of the U. pumila growing season at external stations. Further analysis showed that the negative spatial response of yearly beginning date to spring daily temperature was stronger in warmer years than in colder years. This finding suggests that climate warming in late winter and spring may enhance sensitivity of the growing season's spatial response due to the relationship of beginning date to temperature.

► We define an optimum length period during which mean temperature affects the spatial pattern of tree phenology most remarkably.
► We establish temperature based spatial phenology models for simulating and predicting the average and annual growing season beginning and end dates of Ulmus pumila.
► The higher the spring LP temperature at a location, the earlier the growing season beginning.
► The higher the autumn LP temperature at a location, the later the growing season end.
► Seasonal climate warming will increase sensitivity of spatial response of the growing season beginning to LP temperature.