Despite the recognition of high structural heterogeneity in old-growth forests, it is still unclear whether intermingling of diversified neighborhood patches or the amalgamation of different homogeneous patches causes complexity at the stand level. Measures of forest structure, such as the diameter distribution, are typically quantified at the stand-level and gloss over the contribution of individual trees to sub-stand heterogeneity, which reflects the small-scale endogenous processes that shape structure at the neighborhood patch scale. We propose an objective, floating, focal-tree based approach that flexibly scales across natural tree neighborhoods to quantify diameter distributions and investigate the sub-stand structural heterogeneity. Our central hypothesis was that individual trees contribute to multiple, different, and typically heterogeneous diameter distributions such that structural complexity of this mixed-conifer old-growth forest in western Oregon with a stand-level rotated sigmoid diameter distribution could be attributed to the intermingling of diversified neighborhood patches across scales. Tree-centered spatial tessalation using Delaunay triangulation delineated neighborhoods by connecting trees to their natural neighbors across several scales. After identifying the neighborhood of each tree in the stand and constructing its diameter distribution, agglomerative clustering of neighborhoods was used to identify the most prevalent sub-stand diameter distribution types (DDTs) across five scales (5th order neighbors).
Results/Conclusions
At the smallest neighborhood scale, each tree was connected to an average of 6 neighbors in 75 m2 of convex hull area, which increased to 118 neighbors covering 2542 m2 at the largest scale. Agglomerative clustering identified ten distinct sub-stand DDTs across all scales, which accounted for 50% of the variation in Euclidean distances. At the smallest tree neighborhood scale, these ten DDTs were characterized by four different distribution shapes (negative exponential, concave, rotated sigmoid, and unimodal) and varied considerably in the abundance of trees in different size classes and therefore in species composition (Douglas-fir composing the larger trees and western hemlock the smaller). On average, each individual tree contributed to >3 different DDTs at any given scale. Small-scale tree neighborhoods were tremendously heterogeneous, such that the stand-level diameter distribution represented an aggregation across extremely disparate sub-stand distributions. As expected, these distributions exhibited spatial smoothing as small, heterogeneous tree neighborhoods were expanded into larger neighborhoods with more similar average composition, which converged on a rotated-sigmoid shape. Nonetheless, as scale expanded, small heterogeneous tree neighborhoods expanded into larger heterogeneous neighborhoods, indicating that the structure of this mixed-conifer old-growth transcends scale (i.e., intermingles diversified neighborhood patches rather than aggregating different homogeneous patches).