Faculty of Forestry, University of British Columbia, 2424 Main Mall, Vancouver, BC, Canada, V6T 1Z4.
Pacific Forestry Centre, Canadian Forest Service, Natural Resources Canada, 506 West Burnside Road, Victoria, BC, Canada, V8Z 1M5.
Glob Chang Biol. 2017 Apr;23(4):1691-1710. doi: 10.1111/gcb.13428. Epub 2016 Sep 14.
Accounting for water stress-induced tree mortality in forest productivity models remains a challenge due to uncertainty in stress tolerance of tree populations. In this study, logistic regression models were developed to assess species-specific relationships between probability of mortality (P ) and drought, drawing on 8.1 million observations of change in vital status (m) of individual trees across North America. Drought was defined by standardized (relative) values of soil water content (W ) and reference evapotranspiration (ET ) at each field plot. The models additionally tested for interactions between the water-balance variables, aridity class of the site (AC), and estimated tree height (h). Considering drought improved model performance in 95 (80) per cent of the 64 tested species during calibration (cross-validation). On average, sensitivity to relative drought increased with site AC (i.e. aridity). Interaction between water-balance variables and estimated tree height indicated that drought sensitivity commonly decreased during early height development and increased during late height development, which may reflect expansion of the root system and decreasing whole-plant, leaf-specific hydraulic conductance, respectively. Across North America, predictions suggested that changes in the water balance caused mortality to increase from 1.1% yr in 1951 to 2.0% yr in 2014 (a net change of 0.9 ± 0.3% yr ). Interannual variation in mortality also increased, driven by increasingly severe droughts in 1988, 1998, 2006, 2007 and 2012. With strong confidence, this study indicates that water stress is a common cause of tree mortality. With weak-to-moderate confidence, this study strengthens previous claims attributing positive trends in mortality to increasing levels of water stress. This 'learn-as-we-go' approach - defined by sampling rare drought events as they continue to intensify - will help to constrain the hydraulic limits of dominant tree species and the viability of boreal and temperate forest biomes under continued climate change.
由于树木种群对胁迫的耐受不确定性，将水胁迫引起的树木死亡纳入森林生产力模型仍然具有挑战性。本研究利用北美 810 万株个体树木生命状态变化的观测值，采用逻辑回归模型来评估物种特异性死亡率（P）与干旱之间的关系。干旱由每个野外样地的土壤水含量（W）和参考蒸散量（ET）的标准化（相对）值定义。模型还检验了水分平衡变量、地点干旱度（AC）和估计树木高度（h）之间的相互作用。考虑到干旱，在 64 个测试物种中，有 95%（80%）在定标（交叉验证）期间提高了模型性能。平均而言，相对干旱的敏感性随着地点 AC（即干旱程度）的增加而增加。水分平衡变量和估计树木高度之间的相互作用表明，干旱敏感性通常在早期高度发育期间降低，在后期高度发育期间增加，这可能分别反映了根系的扩展和整个植物、叶片特定水力传导率的降低。在整个北美，预测表明，由于 1951 年至 2014 年期间水平衡的变化，死亡率从 1.1% yr增加到 2.0% yr（净变化为 0.9±0.3% yr）。死亡率的年际变化也增加了，这是由 1988 年、1998 年、2006 年、2007 年和 2012 年越来越严重的干旱驱动的。本研究具有很强的可信度，表明水胁迫是树木死亡的一个常见原因。本研究具有较弱到中等的可信度，这加强了先前归因于死亡率上升的观点，即与不断增加的水胁迫水平有关。这种“边学边做”的方法——在干旱事件继续加剧的情况下对其进行抽样——将有助于限制主要树种的水力极限以及在持续气候变化下北方和温带森林生物群落的生存能力。
