Asymmetric temperature effect on leaf senescence and its control on ecosystem productivity

Lei He, Jian Wang, Josep Peñuelas, Constantin M. Zohner, Thomas W. Crowther, Yongshuo H. Fu, Wenxin Zhang, Jingfeng Xiao, Zhihua Liu, Xufeng Wang, Jiahao Li, Xiaojun Li, Shouzhang Peng, Yaowen Xie, Jian‐Sheng Ye, Chenghu Zhou, Zhao-Liang LiPNAS 2024 · DOI

TL;DR

This study reveals that autumn leaf senescence in northern forests is more sensitive to warming than to cooling temperatures. This asymmetry, driven by factors like resource acquisition and frost risk, has significant implications for how we model plant productivity and carbon cycles, as current models often fail to capture these non-linear responses.

What Problem Does This Paper Address?

While the impact of rising global temperatures on ecosystems is a major research focus, the effects of natural cooling trends, particularly in autumn, on vegetation processes have been largely overlooked. This is a critical gap because autumn leaf senescence is a key regulator of carbon and nutrient cycling, and its timing directly influences plant productivity and overwintering survival. Understanding how vegetation responds to both warming and cooling is essential for accurately predicting how ecosystems will adapt to complex climate change scenarios and for improving the reliability of climate models. The paper specifically addresses the question of whether autumn leaf senescence dates (LSD) respond differently to warming versus cooling temperatures and what mechanisms drive these potential asymmetries.

Methods

The researchers employed a multi-faceted approach combining in situ phenological observations with satellite-derived data.

  • In situ Phenological Data: They utilized a large dataset of over 36,000 leaf senescence records from 11,138 European sites, spanning back to the 1950s, for four dominant temperate tree species. This allowed for species-level analysis of LSD responses.
  • Satellite-Derived Greenness Data: Thirty years (1989-2018) of satellite-derived Normalized Difference Vegetation Index (NDVI) data were used to analyze biome-level LSD responses across northern forests.
  • Statistical Analysis: Ridge regression and multiple linear regression analyses were central to quantifying the sensitivities of LSD to warming and cooling. They also developed a "frost risk index" and used the Standardized Precipitation-Evapotranspiration Index (SPEI) to assess the influence of climatic factors.
  • Process-Oriented Modeling: To understand the implications for ecosystem productivity, they analyzed plant productivity data (kernel NDVI and gross primary productivity from multiple models) and employed Partial Least-Squares Structural Equation Modeling (PLS-SEM) and random forest modeling. They also incorporated data from 16 Dynamic Global Vegetation Models (DGVMs).

Datasets

  • Pan European Phenology (PEP) database (PEP725): A collection of over 36,000 in situ phenological time series from 11,138 European sites, dating back to the 1950s, for four temperate tree species (Aesculus hippocastanum L., Betula pendula Roth, Fagus sylvatica L., and Quercus robur L.).
  • PKU GIMMS NDVI: Bi-weekly NDVI data at 1/12° spatial resolution from 1989 to 2018, extended to 2022, used for estimating LSD.
  • MODIS MCD12C1 IGBP land cover product: Used to identify forest biome types.
  • European gridded observational (E-OBS v27.0e) climate data: Daily data at 0.1° spatial resolution (temperature, precipitation, radiation) for species-scale analysis.
  • Multi-Source Weather (MSWX) climate data: Daily data at 0.1° spatial resolution for biome-scale analysis.
  • Monthly SPEI dataset (three-month scale): Used to calculate soil water availability.
  • Plant productivity data:
    • kNDVI: Calculated from monthly MODIS MOD13C2 v061 product.
  • GPP datasets: GLASS GPP (monthly), TLLUE modeled GPP (monthly), and NIRv-based GPP (monthly).
  • Process-oriented modeled GPP: Data from 16 DGVMs participating in the TRENDY-v11 project.

Key Insights and Ideas

  • Asymmetry is Key: The most striking finding is the pronounced asymmetry in leaf senescence response to temperature. Autumn warming leads to a significant delay in leaf senescence, while autumn cooling causes a less pronounced advance. This suggests that vegetation's "strategy" for dealing with temperature changes in autumn is not symmetrical.
  • Why the Asymmetry? The paper proposes and tests several hypotheses. The stronger response to warming is linked to plants needing to maximize resource acquisition before winter, a greater need under warmer conditions. Conversely, cooler conditions might trigger more conservative responses due to increased frost risk and potentially altered water availability.
  • Beyond Simple Reversal: This asymmetry means that simply inverting the effects of warming to predict cooling impacts is inaccurate. This has direct consequences for how we model ecosystem processes.
  • LSD as a Driver of Productivity: The study demonstrates a strong link between the asymmetry in leaf senescence and the asymmetry in autumn plant productivity. This highlights leaf senescence as a crucial intermediary controlling how much carbon plants can assimilate in the autumn.
  • Model Limitations: Current process-oriented models often fail to capture this observed asymmetry in leaf senescence and its subsequent impact on plant productivity, indicating a need for model improvements.

Key Results

  • Quantified Asymmetry: A 1°C increase in autumn temperature caused a delay in leaf senescence of 7.5 ± 0.2 days, while a 1°C decrease led to an advance of only 3.3 ± 0.1 days (P < 0.001).
  • Species-Level Agreement: This asymmetric response was observed across four dominant temperate tree species.
  • Biome-Level Confirmation: Satellite data confirmed this asymmetry across evergreen needleleaf, deciduous broadleaf, and mixed forests.
  • Mechanistic Support:
    • LSD sensitivities to cold degree days (CDD) were significantly larger under warming than cooling.
  • Cooling sites exhibited higher temperature variability and greater frost risk compared to warming sites.
  • Water stress levels were higher at cooling sites.
  • Productivity Link: Autumn productivity sensitivities to temperature were positively correlated with LSD sensitivities, with LSD accounting for 46% of the spatial variability in autumn productivity sensitivities.
  • Model Agreement: 11 out of 16 DGVMs showed larger autumn productivity responses to warming than cooling, though this was not always statistically significant at the aggregate level.

Limitations and Future Work

  • Spatial Separation of Warming/Cooling Sites: The authors acknowledge that in their spatial analysis, warming and cooling sites were spatially separated, which could introduce some uncertainty. They addressed this by conducting a temporal analysis on grid cells that shifted between warming and cooling.
  • Model Representation: While DGVMs showed some evidence of asymmetric responses, the study suggests that current models only partially capture these observed differences, highlighting the need to incorporate more detailed characterizations of asymmetric LSD responses.
  • Future Work: The paper emphasizes the need for more comprehensive studies to understand the complex interplay between temperature, water availability, and their combined effects on leaf senescence. It also calls for the integration of these asymmetric effects into vegetation and climate models to improve the accuracy of future carbon cycle and climate projections.

Summary updated

Abstract

Widespread autumn cooling occurred in the northern hemisphere (NH) during the period 2004-2018, primarily due to the strengthening of the Pacific Decadal Oscillation and Siberian High. Yet, while there has been considerable focus on the warming impacts, the effects of natural cooling on autumn leaf senescence and plant productivity have been largely overlooked. This gap in knowledge hinders our understanding of how vegetation adapts and acclimates to complex climate change. In this study, we utilize over 36,000 in situ phenological time series from 11,138 European sites dating back to the 1950s, and 30 years of satellite greenness data (1989-2018), to demonstrate that leaf senescence dates (LSD) in northern forests responded more strongly to warming than to cooling in autumn. Specifically, a 1 °C increase in temperature caused 7.5 ± 0.2 days' delay in LSD, whereas a 1 °C decrease led to an advance of LSD with 3.3 ± 0.1 days (<i>P</i> < 0.001). This asymmetry in temperature effects on LSD is attributed to greater preoverwintering plant-resource acquisition requirements, lower frost risk, and greater water availability under warming than cooling conditions. These differential LSD responses highlight the nonlinear impact of temperature on autumn plant productivity, which current process-oriented models fail to accurately capture. Our findings emphasize the need to account for the asymmetric effects of warming and cooling on leaf senescence in model projections and in understanding vegetation-climate feedback mechanisms.
BibTeX
@inproceedings{he2024asymmetric,
  author = {Lei He and Jian Wang and Josep Peñuelas and Constantin M. Zohner and Thomas W. Crowther and Yongshuo H. Fu and Wenxin Zhang and Jingfeng Xiao and Zhihua Liu and Xufeng Wang and Jiahao Li and Xiaojun Li and Shouzhang Peng and Yaowen Xie and Jian‐Sheng Ye and Chenghu Zhou and Zhao-Liang Li},
  title = {{Asymmetric temperature effect on leaf senescence and its control on ecosystem productivity}},
  booktitle = {PNAS},
  year = {2024},
  doi = {10.1093/pnasnexus/pgae477},
}
Figure