Modeling canopy openness and understory gap patterns based on image analysis and mapped tree data

Abstract

Ecological relationships beneath a forest canopy are related spatially to the pattern of canopy gaps and sunlight penetration. Methods to characterize and predict canopy light patterns from easily gathered site inventory data are not readily available. We developed a model to estimate the proportion and distribution of canopy openings visible from the under-story, in lieu of hemispherical photography or field instrumentation. An automated procedure for constructing vertical wide-angle views of forest canopies using standard computer-aided design (CAD) software was applied to canopy structure data collected on individual mapped trees in old-growth plots. Vertical hemispherical photos of actual canopies were paired with constructed CAD views of the same point to evaluate the correspondence between the two. Using image analysis software, we assessed total canopy openness (CO) and largest gap size (GAP) on 324 image pairs from seven different plots. With the exception of one plot, a single quadratic model form fit the remaining observations, resulting in r 2 values over 50% between CAD and photo images. We could not adequately model a plot with heavy non-tree understory vegetation. We also compared contour maps of CO gradients from the CAD images and estimates from the regression model to the photographic images to evaluate whether spatial distributions of canopy openings were correctly captured by the constructed model approach. Visual observations of the gradient map show peaks and valleys in canopy openness that visually match openness on photos at corresponding locations in the plot. Lastly, we found a relatively stable correspondence in the distribution of gap sizes between photographic and CAD images. The models presented here may be applied to known or simulated patterns of tree data to derive a spatially-explicit estimation of gap patterns without the need for corresponding photography or instrumentation.