Department of Biology, University of Regina, 3737 Wascana Parkway, Regina, Saskatchewan, S4S 0A2, Canada.
Laboratory of Geo-Information Science and Remote Sensing, Wageningen University, P.O. Box 47, 6700 AA, Wageningen, The Netherlands.
Ecol Appl. 2017 Oct;27(7):2128-2141. doi: 10.1002/eap.1596. Epub 2017 Sep 6.
Global environmental change is expected to induce widespread changes in the geographic distribution and biomass of forest communities. Impacts have been projected from both empirical (statistical) and mechanistic (physiology-based) modelling approaches, but there remains an important gap in accurately predicting abundance across species' ranges from spatial variation in individual-level demographic processes. We address this issue by using a cohort-based forest dynamics model (CAIN) to predict spatial variation in the abundance of six plant functional types (PFTs) across the eastern United States. The model simulates tree-level growth, mortality, and recruitment, which we parameterized from data on both individual-level demographic rates and population-level abundance using Bayesian inverse modelling. Across a set of 1° grid cells, we calibrated local growth, mortality, and recruitment rates for each PFT to obtain a close match between predicted age-specific PFT basal area in forest stands and that observed in 46,603 Forest Inventory and Analysis plots. The resulting models produced a strong fit to PFT basal area across the region (R = 0.66-0.87), captured successional changes in PFT composition with stand age, and predicted the overall stem diameter distribution well. The mortality rates needed to accurately predict basal area were consistently higher than observed mortality, possibly because sampling effects led to biased individual-level mortality estimates across spatially heterogeneous plots. Growth and recruitment rates did not show consistent directional changes from observed values. Relative basal area was most strongly influenced by recruitment processes, but the effects of growth and mortality tended to increase as stands matured. Our study illustrates how both top-down (population-level) and bottom-up (individual-level) data can be combined to predict variation in abundance from size, environmental, and competitive effects on tree demography. Evidence for how demographic processes influence variation in abundance, as provided by our model, can help in understanding how these forests may respond to future environmental change.
预计全球环境变化将导致森林群落的地理分布和生物量广泛变化。已经从经验（统计）和机制（基于生理学）建模方法中预测到了这些影响，但从个体水平的人口统计过程的空间变化准确预测物种范围内的丰度仍然存在一个重要差距。我们通过使用基于队列的森林动态模型（CAIN）来解决这个问题，该模型预测了美国东部六个植物功能类型（PFT）的丰度在空间上的变化。该模型模拟了树木级别的生长、死亡和繁殖，我们使用贝叶斯反演建模从个体水平的人口统计率和种群水平的丰度数据对其进行了参数化。在一组 1°网格单元中，我们为每个 PFT 校准了本地的生长、死亡率和繁殖率，以使预测的森林林分特定年龄的 PFT 基底面积与 46603 个森林清查和分析图的观测值相匹配。由此产生的模型对该地区的 PFT 基底面积具有很强的拟合度（R=0.66-0.87），捕捉到了与林龄有关的 PFT 组成的演替变化，并很好地预测了总体茎直径分布。为了准确预测基底面积，所需的死亡率始终高于观测到的死亡率，这可能是因为采样效应导致在空间异质的图中对个体水平的死亡率估计存在偏差。生长和繁殖率没有表现出与观测值一致的方向变化。相对基底面积受繁殖过程的影响最大，但随着林分的成熟，生长和死亡率的影响往往会增加。我们的研究说明了如何将自上而下（种群水平）和自下而上（个体水平）的数据结合起来，从树木种群动态的大小、环境和竞争效应预测丰度的变化。我们的模型提供了关于人口统计过程如何影响丰度变化的证据，可以帮助理解这些森林可能对未来环境变化的反应。
