Chin J Plant Ecol ›› 2022, Vol. 46 ›› Issue (7): 753-765.DOI: 10.17521/cjpe.2021.0254

• Research Articles • Previous Articles     Next Articles

Process-based simulation of autumn phenology of trees and the regional differentiation attribution in northern China

CHEN Yi-Zhu, LANG Wei-Guang, CHEN Xiao-Qiu()   

  1. College of Urban and Environmental Sciences, Laboratory for Earth Surface Processes of the Ministry of Education, Peking University, Beijing 100871, China
  • Received:2021-07-07 Accepted:2021-09-27 Online:2022-07-20 Published:2021-12-16
  • Contact: CHEN Xiao-Qiu
  • Supported by:
    National Natural Science Foundation of China(41771049);National Natural Science Foundation of China(41471033)


Aims Revealing occurrence mechanisms of autumn phenology in temperate deciduous trees is of vital importance for improving estimation accuracies of ecosystem carbon sequestration and vegetation productivity. This study aimed to uncover the mechanisms of leaf senescence in response to environmental changes and simulation accuracies of autumn phenology through process-based models, and further investigate impacts of water conditions on leaf senescence mechanisms and simulation accuracy of autumn phenology.

Methods We used the low temperature and photoperiod multiplicative model (TPM) to fit the first leaf coloration dates and leaf fall end dates of six tree species at more than 90 stations across the temperate zone of northern China from 1981 to 2014. The TPM contains two sub-models, i.e., photoperiod-initiated leaf senescence model (TPMp) and temperature-initiated leaf senescence model. We evaluated simulation accuracies and their spatiotemporal variations of station-species specific optimum models, and analyzed regional differentiation of proportions of two sub-models among optimum models and its spatial dependence on arid-humid gradient.

Important findings (1) Photoperiod shortening plays more important roles in initiating leaf senescence than temperature decrease. The simulated average root mean square errors of optimum models for first leaf coloration and leaf fall end dates are 6.9 d and 6.0 d, respectively. Proportions of significantly positive correlations between simulated and observed time series are 71.4% (first leaf coloration) and 83.6% (leaf fall end), respectively. (2) Simulated regional mean absolute errors and multi-year mean absolute errors of optimum models for first leaf coloration date and leaf fall end date are less than 2.4 d. However, amplitudes of temporal and spatial variations in simulated phenological dates are usually smaller than those in observed phenological dates, which are closely related to the high temporal variability of autumn phenological occurrence date. (3) Water conditions affect the selection of leaf senescence initiation pathways to a certain extent. This is mainly manifested in that proportions of TPMp among optimum models for first leaf coloration dates in arid and semi-arid regions are higher than those in humid and semi-humid regions, while simulation accuracies of optimum models in humid and semi-humid regions are higher than those in arid and semi-arid regions.

Key words: autumn phenology, simulation, low temperature and photoperiod multiplicative model (TPM), spatiotemporal characteristics of simulation effect, moisture dependence, temperate deciduous trees