Objective Under the global warming, the study of phenology is critical in carbon cycle of the terrestrial ecosystem. Phenology is sensitive to climate change, therefore climatic warming is one of the main factors impacting spring and autumn phenological phases. Simulating spring phenological phase had got some similar conclusions, while the leaf coloring date still needs to be explored. This study will contribute to develop phenology model and the study on phenological phase.
Method In this study, we used 10 species in 10 sites to examine the phenological data (leaf bud opening date, leaf out date, first flowering date and leaf coloring date) and meteorological data, and evaluated the accuracy and applicability of spring warming (SW) model. The changing trends of temperature and phenology were analyzed. In addition, the error between the simulated and observed phenological information from the biological characteristics of plants was explored. Based on the observed phenological and meteorological data, the parameters of SW model were estimated. Internal and cross checks were used to assess the validity of SW model.
Result During the past 50 years, temperature of all sites showed an upward trend. Except for Luoyang site in Henan Province of northern China and Dezhou site in Shandong Province of eastern China, spring phenological phase advanced significantly in other sites. Especially, the first flowering date of Lagerstroemia indica in Tai’an of Shandong Province advanced most significantly at a speed of −4.96 d per decade. In Beijing, the leaf bud opening date of Ginkgo biloba advanced least significantly by −0.72 d per decade. In autumn, the leaf coloring date delayed, and it delayed at a speed of 0.12 to 0.49 d per decade in Beijing region. In Qinhuangdao, Hebei Province of northern China, the leaf coloring date of Lagerstroemia indica delayed by 1.05 d per decade. Model performance was assessed according to the root mean square error (RMSE) and the coefficient of determination (R2). SW model performed better in simulating spring phenological phase than simulating autumn phenological phase, simulating arbor better than shrubs. Among trees, the simulation on leaf out date of Populus canadensis was most accurate, with R2 of 0.958 and RMSE of 3.5 d. Among shrubs, the simulation effect on first flowering date of Syringa oblata was best, with R2 of 0.942 and RMSE of 3.6 d. Compared with simulation on spring phenological phase, the simulation on leaf coloring date in autumn had a large deviation, with R2 only of 0.030 to 0.574.
Conclusion In the past 50 years, with the increase of temperature, the spring phenological phase in most sites showed an advanced trend, whereas the autumn phenological phase showed a delayed trend, and there were some differences among species and sites. SW model is suitable for simulation on phenological phase of different life forms, and the difference of simulation effect is not significant. The simulation effects of SW model on varied phenological phases were different, among which, the simulation on leaf out date and first flowering date was most accurate, followed by leaf bud opening date, and the accuracy of simulating leaf coloring date was least. Therefore, the SW model considering only the temperature factor can not truly simulate the autumn phenological phase, and the model should be improved by coupling the factors such as photoperiod and precipitation to improve the accuracy of SW model simulation.