Analysis of sampling methods for coarse woody debris

https://doi.org/10.1016/j.foreco.2004.03.042Get rights and content

Abstract

Coarse woody debris (CWD) is an important structural component of forest ecosystems and plays a key role in ecosystem functioning. It includes whole fallen trees and branches, pieces of fragmented wood, stumps, and standing dead trees (snags). A number of sampling approaches have been used to quantify CWD but there has been little systematic analysis of alternative approaches in terms of sampling efficiency. This study analysed sampling methods for CWD in three Australian forest types: eucalypt-dominated woodlands in south central Queensland, eucalypt open forest (dry sclerophyll forest) in south coastal New South Wales and eucalypt tall open forest in southern Tasmania. All downed wood (greater than 10 cm in woodland and 15 cm in other forest types) was measured and mapped on several large plots (1 ha) at each site. Alternative sampling approaches were compared using a computer-based system. These included varying length and arrangement of lines for the line intersect method and varying sizes of fixed-area plots.

Variability of CWD volume in forests is relatively high and achieving estimates with a precision (coefficient of variation) less than 50% of the mean requires a high sampling intensity. In analysis of the line intersect method, precision of volume estimates increased as total transect length increased. In denser forests, transects of at least 100 m are required to achieve a precision of less than 100% of the mean. A single long transect resulted in lower coefficient of variation than a number of short transects and is generally more cost effective. There was no improvement in precision by arranging the total transect length in an ‘L’ shape, a triangle or a square.

In woodland vegetation (20–50% canopy cover) the amount of CWD is relatively low and variability is high. In these situations fixed-area plots are preferred. 20m×20 m plots give estimates with a mean coefficient of variation less than 100% in the more dense forest structures but well over 100% in woodlands. Measuring all pieces on a 50m×50 m plot in woodlands gave estimates with a precision of ±21–73% of the mean.

Introduction

Coarse woody debris (CWD) plays a role in many aspects of ecosystem functioning, both in aquatic systems and on land, including habitat for wildlife and fungi, nursery site for seedling establishment, nutrient cycling, and soil stability. CWD may include whole fallen trees, fallen branches, pieces of fragmented wood, stumps, and standing dead trees (snags). This study focuses on sampling strategies for forest floor CWD (i.e. excludes stumps and snags), and is referred to as CWD throughout this paper.

CWD variables of primary interest are total volume and mass. Volume and mass in different piece size classes or decay classes (e.g. Sollins et al., 1987, Stewart and Burrows, 1994, Pyle and Brown, 1999) are also of interest. Volume is required to determine mass, with sub-samples usually taken to assess density.

Many sampling designs for quantifying CWD volume have been used (e.g. Brown, 1971, Delisle et al., 1988, Stewart and Burrows, 1994, Gove et al., 1999, McKenzie et al., 2000, USDA Forest Service, 2001), but there has been little systematic analysis of these alternative approaches. Most sampling designs for CWD in Australia have been adopted from overseas studies and may not be applicable to all Australian forest conditions. The most widely used method for sampling CWD is the line-intersect method (Warren and Olsen, 1964, van Wagner, 1968) where diameter of CWD is measured at the point of intersection along a transect of a given length but no width. The design options associated with the line intersect method are transect length, transect layout, and number of replicates.

As the frequency of CWD generally increases with decreasing diameter size, some CWD methodologies determine the length of transect (or length of a transect section) to be sampled by the diameter of woody debris, particularly when fine woody debris is included (e.g. Delisle et al., 1988, USDA Forest Service, 2001). In these methods, transects are divided into sections, and measurements on these sections correspond to diameter size classes, i.e. all woody debris is measured on the first section of transect, and then for each subsequent section, the smallest diameter class is disregarded. This ensures that over-sampling does not occur for the small-sized wood pieces, and that a sufficient number of the largest-sized pieces are sampled.

Transects are often arranged in different orientations at a site to reduce potential for orientation bias. Non-random orientation of CWD can occur due to harvesting practices (e.g. cable harvesting), windthrow, flooding, or steep slopes. CWD will be underestimated if a single sample line is orientated parallel to the wood pieces, and will be overestimated if a transect is orientated perpendicular to the wood pieces.

Many transect layouts have been adopted, including an equilateral triangle (e.g. Delisle et al., 1988), three transects radiating from a common point (e.g. Waddell, 2002), a square, an ‘L’ shape, a single line (Bell et al., 1996), and variations of these. The British Columbia Ministry of Forests use a triangle with 30 m sides for determining fuel loading prior to a prescribed burn, while an ‘L’ shape with two 24 m lines is used in their Vegetation Resources Inventory (Marshall et al., 2000). A transect layout recommended in the sampling protocols by McKenzie et al. (2000) for quantifying carbon in Australian ecosystems, is a variation on the ‘L’ arrangement. McKenzie et al. recommend sampling downed CWD >2.5 cm in diameter using two 10 m transects arranged at right angles to each other, when >10 pieces of CWD occur in a 25m×25 m plot. Starting from a random point in the plot, the first 10 m transect is laid out in a random direction. If the plot boundary is intersected before the full 10 m, the second transect is started at right angles to the first. This is continued (turning at right angles whenever the plot boundary is intersected) until a total of 10 m in both directions is reached.

There is no advantage in using one transect arrangement over another if CWD pieces are orientated at random (Bell et al., 1996). Therefore, at sites with randomly orientated pieces there is no apparent benefit in using a methodology with a complex transect arrangement if a single line transect will be quicker to implement and will provide similar results. If orientation bias of wood pieces is present, the effect can be greatly reduced by running individual sample lines in more than one direction and averaging the results (van Wagner, 1968). Howard and Ward (1972, cited in Nemec and Davis, 2002) recommended using randomly orientated straight line to reduce bias in situations where non-random patterns of CWD are likely to arise.

The alternative to the line intersect method is a full census of CWD on a fixed-area plot (quadrat). Each piece is measured for length and diameter or other variables of interest. McKenzie et al. (2000) recommend that when there are ≤10 pieces of CWD within a 25m×25 m quadrat the dimensions of all pieces within the quadrat are measured. Snags and stumps are usually measured on plots or along a transect with a given width.

This study investigated alternative methods for quantifying CWD in Australian forest conditions with a focus on developing methodologies for carbon inventory purposes. Native forests in Australia are highly diverse, including tropical and cool temperate rainforests, tall (>50 m in height) and medium (20–50 m) open eucalypt forests, and more open and shorter stature woodland forests. A wide range of management practices are used in different forest types, resulting in a large array of forest conditions. Therefore, quantifying CWD is a challenging task.

The primary aim of this study was to compare alternative approaches to measurement of CWD in terms of the precision of the estimate (coefficient of variation), the spatial extent and pattern of sampling, and to develop recommendations for appropriate sampling designs across a broad range of Eucalyptus dominated forest types.

Section snippets

Study areas

Three geographical locations were used for this study: a woodland site in south central Queensland (Injune), an open forest site in south coastal NSW (Kioloa), and a tall open forest site in Tasmania (Warra) (Fig. 1). At all sites forests were dominated by Eucalyptus, characteristic of approximately 80% of Australian native forests. These locations represented the range of Eucalyptus forest structures across Australia.

In each study area plots were selected with a variety of forest ages and

Methods

The analytical approach involved a full survey and mapping of all CWD on 1 ha plots in each of the three locations. This was used as the basis of a computer simulation analysis of alternative sampling strategies for CWD. There is frequently large spatial variation in the amount of CWD within a forest stand, and a sample plot of 1 ha (100m×100 m) allowed assessment of within site variability normally observed in these ecosystems.

Results

CWD volumes calculated directly from plot measurements within each forest type ranged from 7 to 40 m3 ha−1 in woodland forest (mean of 19 m3 ha−1), 31 to 194 m3 ha−1 in open forest (mean of 100 m3 ha−1), and 744 to 1615 m3 ha−1 in tall open forest (mean of 1198 m3 ha−1) (Table 1). At the woodland site live tree basal area ranged from 8 to 16 m2 ha−1, and average top canopy height ranged from 14 to 16 m. In the open forest basal area ranged from 19 to 47 m2 ha−1, with an average top height of 18–36 m. In tall

Line intersect sampling

At all locations and plots, CV decreased as amount of CWD increased for a given transect length. As transect length increased across all plots, the range of volume estimates decreased and therefore precision of the estimate increased. Sampling effort is increased by longer transects, as more CWD pieces are intersected for a given density of CWD. The longer each individual line transect is, the smaller the variability will be among lines (Marshall et al., 2000). When adopting the line intersect

Conclusions

The focus of this study has been on examining the accuracy of the line intersect method for quantifying CWD, as line intersect sampling strategies are as diverse as the forests in which the methods are implemented. Estimates with a desired level of accuracy can be achieved in most forest conditions, providing transect length, number of replicates, and arrangement of transects are appropriate for that forest condition. The line intersect method is efficient, giving a high degree of precision for

Acknowledgements

The Queensland Environment Protection Agency, and Alex Lee of the Bureau of Rural Sciences Canberra provided additional plot data for Injune. Stuart Davey, Jenet Austin and Kimberley van Niel provided data from their unpublished theses (Australian National University) for the Kioloa study region. Forestry Tasmania provided additional data from CFI plots in Tasmania. Thanks to the many people from the Bureau of Rural Sciences, Forestry Tasmania, and the Queensland Environmental Protection Agency

References (25)

  • J.W Hazard et al.

    Simulation studies on line intersect sampling of forest residue. Part II

    For. Sci.

    (1986)
  • J.E Hickey et al.

    Fire history of the tall wet eucalypt forests of the Warra ecological research site, Tasmania

    Aust. For.

    (1999)
  • Cited by (0)

    View full text