Overstory Mortality Varies with Level and Pattern of Green-tree Retention

Charles B. Halpern1 and Juraj Halaj2 1College of Forest Resources Box 352100 University of Washington Seattle, WA 98195-2100 chalpern@u.washington.edu 2Cascadien, Inc. 1903 NW Lantana Drive Corvallis, OR 97330-10160

Introduction

With growing concern over the loss and fragmentation of old-growth forests in the Pacific Northwest, green-tree or structural retention harvest has replaced the historic practice of clearcut logging on federal lands. By retaining greater structural complexity at the time of harvest, it is thought that green-tree retention will ensure greater persistence and more rapid recovery of the species and ecosystems processes that characterize older forests.

In retaining live trees through timber harvest, managers have the ability to vary two important elements of overstory structure: level of retention (the number or proportion of trees or basal area) and the spatial pattern in which trees are retained. In simplest form, trees can be retained in groups (aggregates) or as individuals dispersed across the harvest unit. Franklin et al. (1997) discuss many of the silvicultural and ecological benefits and tradeoffs implicit in these approaches. However, variable retention harvests naturally lead to greater potential for wind-damage or windthrow of isolated or newly exposed trees, thus compromising these potential benefits. In this study, we examine how level and pattern of green-tree retention affect subsequent rates of tree mortality. We pose the following questions:

• Do overall rates of mortality vary with level or pattern of overstory retention?
• Are patterns of mortality similar among the primary species (Douglas-fir and western hemlock) or canopy classes of trees (co-dominant, intermediate, and suppressed)?
• Does the size distribution of dead trees differ from that of live trees and do these relationships vary among treatments?
• Do the principal types of mortality (standing with crown, broken stem, uprooted) vary with level and pattern of retention?

Methods

Our results are based on permanent plot observations of mortality over 2-3 yr period after treatment implementation. The full experimental design consists of six, 13-ha retention treatments, including a control (see Experimental Design), replicated at each of the six locations or blocks (see Study Areas). Here we examine five of the treatments: the control (100% retention) as reference point for “natural” rates of mortality, and four that we analyze as a fully balanced, two-factor design to contrast retention level (15%, the minimum required by the Northwest Forest Plan vs. 40%), and spatial pattern (1-ha aggregates vs. uniformly dispersed trees).

Within each treatment unit, a series of circular, 0.04 ha permanent sample plots were established along a systematic grid (40-m spacing). In the control treatments, a total of 32 plots were established at alternate grid points. In the dispersed retention treatments, where tree densities were considerably lower, 63-64 plots were established (i.e. all grid points). In the aggregated retention treatments, plots were established at all grid points within the aggregates (10 and 25 plots in the 15 and 40% retention treatments, respectively). In each plot, all trees >5 cm dbh were tagged, identified to species, measured for diameter, and assigned to a canopy class: suppressed, intermediate, or co-dominant (very few trees were considered dominants/emergents). Mortality was assessed for 3 yr after harvest (1999-2001) in three blocks, and for 2 yr (2000, 2001) in the remaining three blocks.

Two-way ANOVA models (degrees of freedom: block = 5, level = 1, pattern = 1, level x pattern = 1, error = 15) were used to test for effects of harvest treatments on cumulative mortality (proportion of dead trees). Separate tests were conducted for all trees, dominant tree species, principal canopy classes, and principal forms of mortality (standing with crown, stem breakage, and uprooting). Kolmogorov-Smirnov (KS) two-sample tests were used to compare the diameter distributions of live and dead trees (summed across blocks) within each of four treatment classes (aggregated, dispersed, 15% retention, and 40% retention).

Results

Overall rates of mortality

• Annual rates of mortality tended to decline with time, although individual blocks showed considerable variation in temporal trends and cumulative mortality (Table 1).

• Cumulative mortality was significantly greater both at lower levels of retention (15%; P = 0.004) and in dispersed (vs. aggregated) treatments (P = 0.0098).

Primary tree species

• For Douglas-fir, the most common species (55% of tagged stems), mortality was significantly greater both at lower retention (P = 0.015) and in dispersed treatments (P = 0.038).
• In contrast, western hemlock (19% of tagged stems) did not show a significant response to level or pattern of retention.

Overstory canopy classes

• Among co-dominant trees, mortality was significantly greater at lower levels of retention (P = 0.005), but only marginally greater in dispersed treatments (P = 0.093).
• Among intermediate-sized trees, mortality was extremely variable and did not differ among treatments.
• Among suppressed trees, mortality was significantly greater in dispersed treatments (P = 0.007).

Size distributions of live and dead trees

• The size distributions of live and dead trees differed in all but aggregated treatments, with smaller-diameter trees comprising a larger percentage of the population of dead trees than of live trees.

Forms of mortality

• Treatments influenced, in part, the form of mortality. This was not the case for trees that died standing with crown (no significant main effects).
• However, for trees with broken stems (wind-snap), rates of mortality were marginally greater at lower levels of retention (P = 0.0594).
• For uprooted (wind-thrown) trees, we observed a significant interaction (P = 0.0036) between level and pattern of retention, with significantly greater mortality in dispersed than in aggregated treatments at 15 but not 40% retention.

Discussion

Our study provides broad-based experimental evidence of the potential for post-harvest mortality of overstory trees following variable retention harvests. Two to three years of observation indicate that both level of retention (proportion of original basal area) and its spatial arrangement (trees in 1-ha aggregates vs. fully dispersed) can have significant effects on cumulative mortality and on its distribution among canopy classes and diameter classes of trees. At 15% retention -- the minimum required on federal “matrix” lands within the range of the northern spotted owl -- cumulative mortality was two to three times greater than at 40% retention or in control treatments (where “natural” or background levels of mortality averaged ~1.1% per year). Retaining trees in 1-ha patches greatly reduced the potential for mortality (particularly at lower retention): mortality rates in aggregated treatments were 50% of those in dispersed treatments.

Co-dominant trees showed a significant response to level of retention with mortality more than five times greater in 15% dispersed than in 40% aggregated or dispersed treatments. This outcome is particularly relevant to management, because it is typically the larger, more vigorous trees that are selected for retention in dispersed settings. Cumulative mortality in excess of 7% over 2-3 yr represents a considerable loss of overstory cover in treatments in which initial densities were already low.

Suppressed trees showed a predictably different response: significantly greater mortality in dispersed than in aggregated treatments. In the former, mortality may have been induced by greater levels of logging damage (see Damage to overstory trees). In the latter, suppressed trees remained protected within the undisturbed aggregates. These patterns are consistent with our assessment of the diameter distributions of live and dead trees. In aggregated treatments, these were similar, but in dispersed treatments, smaller trees comprised a larger percentage of the population of dead trees. These smaller trees are more likely to be the less-common, shade-tolerant species, thus there may be direct or indirect consequences for biological diversity beyond the more obvious effects on stem density and canopy structure.

Our analyses of the “forms” of mortality suggest that level and pattern of retention can differentially affect the input of coarse woody debris to these forests. Among the primary forms of mortality, the proportion of dead stems resulting in snags (“standing with crown” and “broken stem”) did not differ among treatments. However, windthrow ("uprooted trees") was more common at lower retention, particularly in dispersed treatments. Thus, while snag production may be comparable among treatments, inputs of fresh downed wood are likely to be higher where residual trees are dispersed at low levels of retention.

It has been observed that most wind-induced mortality occurs within the first 5-6 yr after timber harvest. In this study, 3 yr of observations confirm that rates of mortality have generally declined with time. However, the potential for mortality remains high: recent sampling at Watson Falls revealed significant wind-induced mortality in the 15% dispersed treatment during early 2003.

Although the potential for overstory mortality is only one of several factors that managers consider in designing variable retention harvests, our results clearly highlight (1) the short-term stability of forest aggregates of 1-ha in size, within which annual rates of mortality are comparable to uncut forests, and (2) the susceptibility to wind damage of dispersed treatments supporting minimal levels of retention. Future measurements of these sites will provide valuable information on the ecological consequences of these effects.