Comparison of five canopy cover estimation techniques in the western Oregon Cascades

Abstract

Estimates of forest canopy cover are widely used in forest research and management, yet methods used to quantify canopy cover and the estimates they provide vary greatly. Four commonly used ground-based techniques for estimating overstory cover – line-intercept, spherical densiometer, moosehorn, and hemispherical photography – and cover estimates generated from crown radii parameters of the western Cascades variant of the Forest Vegetation Simulator (FVS) were compared in five Douglas-fir/western hemlock structure types in western Oregon. Differences in cover estimates among the ground-based methods were not related to stand-structure type (p = 0.33). As expected, estimates of cover increased and stand-level variability decreased with increasing angle of view among techniques. However, the moosehorn provided the most conservative estimates of vertical-projection overstory cover. Regression equations are provided to permit conversion among canopy cover estimates made with the four ground-based techniques. These equations also provide a means for integrating cover data from studies that use different techniques, thus aiding in the ability to conduct synthetic research. Ground-based measures are recommended for specific objectives. Because the FVS-estimated cover levels were consistently lower and more variable than most of the ground-based estimates (by up to 44, 17% on average), ground-based measures of canopy cover may be preferable when accuracy is an important objective.

Introduction

Estimates of canopy cover are widely used in forest research and management, including in the Pacific Northwest Region of the USA (PNW). Regulations for certain regional wildlife species require maintenance of certain levels of canopy cover (e.g., Weiss et al., 1991, Verner et al., 1992). Canopy cover is often used as a criterion for classifying stand structure (e.g., Wisdom et al., 2000, Azuma and Hanson, 2002), and as a surrogate for shade when monitoring stream temperatures (e.g., OWEB, 1999). In addition, cover estimates are used to estimate penetration of light to the understory (e.g., Canham et al., 1990, Lieffers et al., 1999, Englund et al., 2000).

Despite the importance of quantitative estimates of canopy cover, there is no standard measurement method. Instead, estimates are derived with a wide variety of ground-based techniques. Commonly used ground-based methods include ocular estimates, the moosehorn (Robinson, 1947), spherical densiometers (concave and convex; Lemmon, 1956), the densitometer (Stumpf, 1993), hemispherical photography (Evans and Coombe, 1959), point counts, and the line-intercept method (Canfield, 1941, O’Brien, 1989). Less commonly cited ground-based methods include stem and crown mapping, the vertical tube (Johansson, 1985), and the gimbal sight (Walters and Soos, 1962). Additionally, predictive relationships between tree size and canopy cover derived from empirical measures are used in stand or tree growth models to estimate vertically projected canopy cover, such as Forest Vegetation Simulator (FVS; Donnelly and Johnson, 1997), ORGANON (Hann, 2003), and certain forest-gap models (e.g., Garman et al., 2003). Where direct measures of canopy cover cannot be acquired, estimates of cover can be derived by applying these predictive relationships to ground-based measures of tree sizes.

Estimates of percent cover vary among ground-based methods, primarily due to differences in the angle of view from zenith captured (e.g., Bunnell and Vales, 1990, Applegate, 2000). Larger angles of view result in greater estimates of canopy cover because canopy gaps visually “close” as the angle of view is lowered from directly overhead towards the horizon (Kirchoff and Schoen, 1987, Bunnell and Vales, 1990). For a given amount of vertical canopy cover, we would expect that estimates from narrow- and wide-angle techniques would be more similar in single-layer stands than in multi-layer stands.

Conceptually, “canopy cover” is the vertical projection of plant foliage onto a horizontal surface. In practice, measurements of “canopy cover” assess either foliage, foliage plus stems, or canopy perimeters, and may do so with instruments with a variety of angles of view. While a few researchers have distinguished between vertically projected cover and cover measured with wider angles of view (e.g., crown completeness, Bunnell et al., 1985; angular cover, Nuttle, 1997), there is a general tendency for overstory cover measured with different angles of view to be referred to as “canopy cover”. As a result, different techniques with different angles of view are estimating cover values for different meanings of canopy cover. In our comparison of techniques, we refer to “canopy cover” in this broader sense that encompasses different angles of view, and distinguish it from the term “vertical canopy cover”. Only as the angle of view of canopy reduces to zero, only measuring the area directly overhead, does angular canopy cover become equivalent to vertical canopy cover. The line-intercept method, with a theoretical zero width, is therefore expected to provide the least-biased, most accurate estimates of vertical canopy cover. It is also the most directly comparable measure to line-intercepts used to estimate cover from remote sensing imagery (O’Brien, 1989).

Predictive models of canopy cover are constrained by several factors. The range of conditions in which parameterization data are sampled limits model applications to similar conditions. Of greater importance is the influence of the type of method used to collect the ground-based measures for model development. Given the variability in measures with different canopy-cover estimation methods, predictive models are generally no better than the ground-based methods, having the same error and limitations as the methods used to generate the model-parameterization data.

Previous studies comparing field and/or modeling methods for estimating canopy cover have demonstrated important differences among methods (e.g., Bunnell and Vales, 1990, Ganey and Block, 1994, Cook et al., 1995, Applegate, 2000, Englund et al., 2000). However, we are unaware of previous studies comparing the line-intercept method with other methods in multiple structure types. Given that the line-intercept method is commonly used (e.g., Azuma and Hanson, 2002, Fiala, 2003), and is expected to offer the most reliable estimates of vertical canopy cover, further study is warranted to determine the relationship of other commonly used cover techniques relative to the line-intercept method among stand-structure types. With the lack of standardized methods, a means for comparing cover measures recorded among techniques is also desired for integration of multiple datasets. This is especially important given that different techniques are measuring alternative definitions of canopy cover that are inaccurately used interchangeably. Recognizing the large amounts of time that are often required to collect detailed canopy cover data, it is increasingly common for managers to rely solely on modeled estimates of canopy cover. However, the relation of these estimates to what is observed on the ground has not been well explored. Therefore, it is also of interest to compare modeled cover with ground-based estimates.

The objectives of this study were to: (1) compare estimates of canopy cover among the ground-based line-intercept, hemispherical photography, moosehorn, and convex spherical densiometer methods; (2) compare the variability in cover estimates obtained by these techniques; (3) create regression equations to facilitate comparisons of estimates made among these methods; and (4) compare ground-based estimates with canopy cover estimates generated by the FVS equations. The FVS was chosen for evaluation in this study because of the availability of tree attributes in our data, the diversity of stand-structure types in our study, and the prevalence of FVS and its extensions in use in current forest research and management (e.g., Christensen et al., 2002, Hummel et al., 2002). The methods were compared in five Pseudotsuga menziesii (Mirb.) Franco/Tsuga heterophylla (Raf.) Sarg. (Douglas-fir/western hemlock) structure types in the western Oregon Cascade Range.