中国农科院资划所李召良团队在PNAS姊妹期刊最新研究成果！

在国家自然科学基金创新研究群体项目“农业遥感机理与方法”（41921001）支持下，中国农业科学院农业资源与农业区划研究所李召良团队在PNAS姊妹期刊PNAS Nexus上发表了论文《Asymmetric temperature effect on leaf senescence and its control on ecosystem productivity》。项目负责人李召良研究员为论文通讯作者，何磊博士为论文第一作者。本研究是项目组去年关于春季展叶非对称响应研究工作（He et al., PNAS Nexus, 2023，https://doi.org/10.1093/pnasnexus/pgad308）的延续，两篇论文系统地回答了植被春、秋季物候对自然升温、降温的非对称响应问题。（PNAS Nexus | 北方植被生态系统展叶期对自然升温和降温的非对称响应）

摘要

由于太平洋年代际振荡和西伯利亚高压增强，北半球在2004至2018年期间出现了大范围的秋季降温现象；然而，降温对叶片衰老和植被生产力的影响仍然不甚明确，这阻碍了我们对植被如何适应复杂气候变化的理解。本研究利用欧洲自1950s以来的地面秋季物候观测时序记录以及北半球中高纬地区1989至2018年的遥感绿度数据，发现北方森林的叶片衰老日期（leaf senescence date, LSD）对秋季升温的响应比对降温的响应更大。LSD对温度的非对称响应，归因于升温条件下植被在冬季之前获取资源的需求增加、霜冻风险降低以及水分供应需求变化。LSD非对称响应可能会进一步导致秋季植被生产力对升温和降温的非对称响应，而当前的动态全球植被模型还难以准确地表征这种复杂性。这一新的发现和认知，对于改进植被和气候模型，提高未来碳循环和气候预测的准确性有着至关重要的作用。     

LSD对自然升温和降温的响应

在树种尺度和植物群落尺度上，研究表明LSD对秋季自然升温的响应更大。基于PEP725数据库中4个欧洲优势树种的叶片衰老日期数据（BBCH code 94），岭回归和线性回归的结果均表明，3个树种（Betula pendula Roth, Fagus sylvatica L., Quercus robur L.）对自然升温的响应更大（Fig 1B–C）。基于NDVI估算的LSD数据的结果，同样发现2004–2018年北半球中高纬地区常绿针叶林、落叶阔叶林和混交林的LSD对秋季升温响应更大（Fig 2C–D）。同时，在时间尺度上的分析也支持这一结论，即在1989–2003年、2004–2018年期间发生升温和降温转变的森林，也表现出对秋季升温的响应更大（Fig 2E–F）。    

Fig. 1. European distribution of study sites and analyses of leaf senescence date responses to natural autumn warming and cooling inB. pendula Roth, F. sylvatica L., Q. robur L., and A. hippocastanum L. A) Location of the long-term study site records for the four European temperate tree species. B) Ridge regression analyses of leaf senescence date (LSD) responses. C) Multiple linear regression analyses of LSD responses. Records of warming and cooling were selected based on change in temperature and partial correlation analysis at P < 0.05 and differences in LSD responses between warming and cooling conditions were analyzed using Student's t-test at P < 0.05. Boxplots show median (horizontal line) and mean (cross) data within the 25–75th percentiles; ***P < 0.001, **P < 0.01, and *P < 0.05. 

Fig. 2. Grid cells of remotely sensed data in which there was autumn warming and cooling during the period (2004–2018) and analyses of forest biome leaf senescence date responses. A) Location of autumn warming and cooling grid cells. B) Trends in biome leaf senescence date (LSD). C) Location of forest biome types that experienced warming and cooling, based on the MODIS MCD12C1 IGBP land cover product. D) Ridge regression analysis of biome LSD sensitivity to temperature, controlling for effects of precipitation and radiation. E) Location of grid cells in which there were shifts in warming and cooling between the period 1989–2003 and 2004–2018. F) Ridge regression analysis of biome LSD sensitivity to temperature. Differences in LSD responses between warming and cooling conditions were analyzed using Student's t-test at P < 0.05. Boxplots show median (horizontal line) and mean (cross) data within the 25–75th percentiles; ***P < 0.001, **P < 0.01, and *P < 0.05.

Fig. 3. Mechanisms of asymmetric leaf senescence date responses to autumn warming and cooling. A) Sensitivities of leaf senescence date (LSD) to cold degree days (CDD) under autumn warming and cooling with contrasting CDD base temperatures of 5, 10, and 15°C (CDD5, CDD10, and CDD15, respectively). B) Linear regression analysis of relations between LSD and CDD10 under autumn warming and cooling conditions; differences in regression slopes tested by covariance analysis at P < 0.001. C) Standard deviation in preseason temperature (Preseason Tstd) under autumn warming and cooling conditions.D) Mean frost risk index under autumn warming and cooling conditions. E) Linear regression analysis of relations between LSD sensitivity to temperature (ridge sensitivity absolute values) and frost risk index. F) Mean Standardized Precipitation-Evapotranspiration Index (SPEI) under autumn warming and cooling conditions. Differences between autumn warming and cooling conditions (A, C, D, F) were tested using a linear mixed method with random intercepts among forest biomes. Boxplots show median (horizontal line) and mean (cross) data within the 25–75th percentiles; ***P < 0.001, **P < 0.01, and *P < 0.05.

基于kNDVI和多套GPP产品数据，研究发现秋季植被生产力对秋季升温的响应相比对降温的响应更大，这种不对称性与LSD对秋季升温的响应更大有关。然而，当前动态全球植被模型只能部分且有限地捕捉到秋季生产力对升温和降温响应的差异。未来如果能将非对称LSD响应的详细特征纳入全球植被模型中，则有可能提升过程模型对秋季升温和降温对生产力非对称影响的表现，从而提高季节性植被动态预测的准确性。

Fig. 4. Responses and drivers of autumn productivity to warming and cooling. A) Sensitivities of autumn productivity to temperature under warming and cooling, based on kernel NDVI (kNDVI), Global Land Surface Satellite (GLASS) gross primary productivity (GPP) (GLASS-GPP), two-leaf light use efficiency modeled (TLLUE-GPP), and near-infrared reflectance of vegetation-based GPP (NIRv-GPP) data sets; kNDVI scaling factor is 0.01. B) Linear regression analysis of the relations between autumn kNDVI sensitivities to temperature [S(kNDVI|T)] and LSD sensitivities to temperature [S(LSD|T)]. C) Partial least squares structural equation modelling analysis of S(LSD|T) and climate factor effects on mean S(kNDVI|T); MAP: mean annual precipitation; MAT: mean annual temperature; SWD: downward solar radiation; VPD: vapor pressure deficit. D) Mean sensitivities of autumn GPP to temperature under warming and cooling in 16 DGVMs in the TRENDY-v11 project (See Methods); sensitivity differences among models were tested using Student's t-test at P < 0.05 and the bars indicate standard errors. ***P < 0.001, **P < 0.01.

 https://doi.org/10.1093/pnasnexus/pgae477