Elsevier

Forest Ecology and Management

Volume 272, 15 May 2012, Pages 26-34
Forest Ecology and Management

A comparison of selected parametric and imputation methods for estimating snag density and snag quality attributes

https://doi.org/10.1016/j.foreco.2011.06.041Get rights and content

Abstract

Snags (standing dead trees) are an essential structural component of forests. Because wildlife use of snags depends on size and decay stage, snag density estimation without any information about snag quality attributes is of little value for wildlife management decision makers. Little work has been done to develop models that allow multivariate estimation of snag density by snag quality class. Using climate, topography, Landsat TM data, stand age and forest type collected for 2356 forested Forest Inventory and Analysis plots in western Washington and western Oregon, we evaluated two multivariate techniques for their abilities to estimate density of snags by three decay classes. The density of live trees and snags in three decay classes (D1: recently dead, little decay; D2: decay, without top, some branches and bark missing; D3: extensive decay, missing bark and most branches) with diameter at breast height (DBH)  12.7 cm was estimated using a nonparametric random forest nearest neighbor imputation technique (RF) and a parametric two-stage model (QPORD), for which the number of trees per hectare was estimated with a Quasipoisson model in the first stage and the probability of belonging to a tree status class (live, D1, D2, D3) was estimated with an ordinal regression model in the second stage. The presence of large snags with DBH  50 cm was predicted using a logistic regression and RF imputation. Because of the more homogenous conditions on private forest lands, snag density by decay class was predicted with higher accuracies on private forest lands than on public lands, while presence of large snags was more accurately predicted on public lands, owing to the higher prevalence of large snags on public lands. RF outperformed the QPORD model in terms of percent accurate predictions, while QPORD provided smaller root mean square errors in predicting snag density by decay class. The logistic regression model achieved more accurate presence/absence classification of large snags than the RF imputation approach. Adjusting the decision threshold to account for unequal size for presence and absence classes is more straightforward for the logistic regression than for the RF imputation approach. Overall, model accuracies were poor in this study, which can be attributed to the poor predictive quality of the explanatory variables and the large range of forest types and geographic conditions observed in the data.

Highlights

► Snag density by decay class predictions had higher accuracy on private lands. ► Presence/absence of large snags was more difficult to model on private lands. ► Large snag presence/absence classification more accurate with logistic regression. ► Decision thresholds should account for unequal sizes in presence/absence classes.

Introduction

Detailed information about the density of standing dead trees (snags) and their quality attributes (e.g., size, decay class) are not only important for carbon storage and fire effects but also essential for managing biodiversity and wildlife habitat. Many wildlife species depend on snags occurring across the landscape in a variety of sizes and decay classes for nesting, roosting, denning, foraging, or shelter (Bull et al., 1997). The potential use of snags by wildlife species depends on snag characteristics such as tree species, diameter, height, decay stage, and proximity to other snags and live trees (Bull et al., 1997).

Size and decay stage are two prominent characteristics of dead wood that influence habitat suitability for individual wildlife species. Many species prefer large snags with diameter at breast height (DBH) greater than 50 cm for nesting, roosting, and foraging (Marcot et al., 2010, Bull, 2002). The number of species provided with suitable habitat, reproductive output of cavity-nesters, security of nest and roost sites from predators and weather, and the longevity of available habitat all increase with snag size (Hagar, 2007). In addition to snag size, decay stage is considered one of the most important attributes that affects snag use by wildlife species associated with dead wood (Vaillancourt et al., 2008) and vertebrates in general (Harmon et al., 1986). The decay stage of snags influences the ability of cavity-using species to excavate nest sites (Lundquist and Mariani, 1991), and the extent of colonization by arthropods, an important food resource for many species of vertebrate wildlife.

Forest management effects on snag density have been well documented (Graves et al., 2000). Snag density tends to be low in areas of intensive timber harvest and increased human access (Wisdom and Bate, 2008), while large snags are more abundant in unharvested stands (Marcot et al., 2010). Snag density also differs by ownership, ecoregion, and forest type. For Coastal Oregon, Kennedy et al., 2008, Ohmann et al., 2007 reported more snags on public lands than on private lands. In Oregon and Washington, Ohmann and Waddell (2002) found lowest snag densities in dry habitats east of the Cascade crest of Oregon and Washington and greatest snag densities at high elevations.

To successfully conserve forest structure and biodiversity and manage forests for broad ecological goals including wildlife, information about snag density along with snag quality attributes is necessary. While research has focused on modeling snag density ignoring snag quality attributes (e.g., Frescino et al., 2001, Eskelson et al., 2009a, Pierce et al., 2009), few studies have attempted to model snag density by decay class (Bater et al., 2009) or snag presence by size class (Martinuzzi et al., 2009).

The aim of this study was to (1) apply nearest neighbor imputation and a two-stage parametric model to predict snag density by decay class; (2) predict the presence/absence of large snags (DBH  50 cm) in western Oregon and Washington with logistic regression and nearest neighbor imputation; and (3) evaluate the performance of the different approaches on private and public lands in western Washington and western Oregon.

Section snippets

Data

Data from forested Forest Inventory and Analysis (FIA) plots, located in western Oregon (OR) and western Washington (WA) and collected between 2001 and 2008, were used in this study. A detailed description of the FIA inventory data is available in Bechtold and Patterson (2005). Only plots (n = 2356) for which all four subplots belonged to the same condition class were used in this study, which allows using plot-level variables as explanatory variables. The number of live trees per hectare (LTPH)

Snag density by decay stage

For both states and both ownership types, CANOPY, STDAGE, TM7, WET and the forest type indicator variables were significant explanatory variables in the QP model for estimating TPH. Other significant variables that were selected in three of the four QP models were R54, R57, and TM5. For the OR private land model, no climate variables were significant, while the other three models included both temperature and precipitation variables among the significant explanatory variables (Table 5).

Snag density by decay stage

The results of our study indicate that STDAGE was the most important explanatory variable in all QP and ORD models, since TPH tends to decrease with stand age and the proportion of DTPH is larger in old forests than young and pole-sapling stands (Bater et al., 2009) and hence, increases with STDAGE. CANOPY, however, was only significant in the QP models, but did not explain much variability in the ORD models for estimating the proportion of LTPH and DTPH by decay class. This is due to CANOPY

Acknowledgments

We thank Dr. Janet Ohmann, Travis Woolley, and two anonymous reviewers for their helpful comments on an earlier version of the manuscript. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the US Government.

References (32)

  • E.L. Bull et al.

    Trees and Logs Important to Wildlife in the Interior Columbia River Basin. Gen. Tech. Rep. PNW-GTR-391

    (1997)
  • J.J. Chen et al.

    Decision threshold adjustment in class prediction

    SAR QAR Environ. Res.

    (2006)
  • Christensen, R.H.B., 2010. Ordinal–regression models for ordinal data. R package version 2010.07-23....
  • J. Cohen

    A coefficient of agreement of nominal scales

    Educ. Psychol. Meas.

    (1960)
  • N.L. Crookston et al.

    YaImpute: an R package for kNN imputation

    J. Stat. Softw.

    (2008)
  • B.N.I. Eskelson et al.

    Estimating cavity tree and snag abundance using negative binomial regression models and nearest neighbor imputation methods

    Can. J. For. Res.

    (2009)
  • Cited by (0)

    1

    Tel.: +1 541 737 8549; fax: +1 541 737 4316.

    2

    Tel.: +1 541 758 8815; fax: +1 541 758 8806.

    View full text