Faculty of Forestry and Environmental Management, University of New Brunswick, New Brunswick, Canada.
Department of Forestry, Natural Resources Faculty, University of Tehran, Karaj, Iran.
PLoS One. 2015 Apr 7;10(4):e0121172. doi: 10.1371/journal.pone.0121172. eCollection 2015.
Mapping landscape variation in tree species richness (SR) is essential to the long term management and conservation of forest ecosystems. The current study examines the prospect of mapping field assessments of SR in a high-elevation, deciduous forest in northern Iran as a function of 16 biophysical variables representative of the area's unique physiography, including topography and coastal placement, biophysical environment, and forests. Basic to this study is the development of moderate-resolution biophysical surfaces and associated plot-estimates for 202 permanent sampling plots. The biophysical variables include: (i) three topographic variables generated directly from the area's digital terrain model; (ii) four ecophysiologically-relevant variables derived from process models or from first principles; and (iii) seven variables of Landsat-8-acquired surface reflectance and two, of surface radiance. With symbolic regression, it was shown that only four of the 16 variables were needed to explain 85% of observed plot-level variation in SR (i.e., wind velocity, surface reflectance of blue light, and topographic wetness indices representative of soil water content), yielding mean-absolute and root-mean-squared error of 0.50 and 0.78, respectively. Overall, localised calculations of wind velocity and surface reflectance of blue light explained about 63% of observed variation in SR, with wind velocity accounting for 51% of that variation. The remaining 22% was explained by linear combinations of soil-water-related topographic indices and associated thresholds. In general, SR and diversity tended to be greatest for plots dominated by Carpinus betulus (involving ≥ 33% of all trees in a plot), than by Fagus orientalis (median difference of one species). This study provides a significant step towards describing landscape variation in SR as a function of modelled and satellite-based information and symbolic regression. Methods in this study are sufficiently general to be applicable to the characterisation of SR in other forested regions of the world, providing plot-scale data are available for model generation.
绘制物种丰富度（SR）的景观变化图对于长期管理和保护森林生态系统至关重要。本研究考察了在伊朗北部高海拔落叶林中，作为该地区独特地貌、包括地形和沿海位置、生物物理环境和森林特征的代表的 16 个生物物理变量，对实地评估 SR 的前景进行制图。本研究的基础是开发中等分辨率的生物物理表面和相关的 202 个永久采样点的图块估计。生物物理变量包括：（i）直接从该地区数字地形模型生成的三个地形变量；（ii）四个来自过程模型或第一原理的生态生理相关变量；以及（iii）七个由 Landsat-8 获得的表面反射率变量和两个表面辐射率变量。通过符号回归，结果表明，在 16 个变量中，只有 4 个变量可以解释 SR 观测值的 85%（即风速、蓝光表面反射率和代表土壤含水量的地形湿润指数），平均绝对误差和均方根误差分别为 0.50 和 0.78。总的来说，局部计算风速和蓝光表面反射率解释了 SR 观测值的 63%左右的变化，其中风速占了 51%的变化。其余的 22%由与土壤水分相关的地形指数及其相关阈值的线性组合解释。一般来说，以 Carpinus betulus 为主导的地块（占一个地块中所有树木的 33%以上）的 SR 和多样性最大，其次是 Fagus orientalis（中位数差异为一个物种）。本研究为描述 SR 的景观变化提供了重要的一步，作为模型化和卫星信息和符号回归的函数。本研究中的方法具有足够的通用性，可应用于世界上其他森林地区的 SR 特征描述，只要有模型生成所需的图块数据即可。
