Multi-decade biomass dynamics in an old-growth hemlock-northern hardwood forest, Michigan, USA

Estimating living biomass and CWD inputs

Above-ground living biomass was estimated for all trees >12.5 cm dbh (to conform with the 5 inch minimum dbh recorded in 1962 and 1967), using published dbh-based allometric equations. Equations were selected from Jenkins et al. (2004) on criteria including: location of development, favoring equations developed for the upper Great Lakes region; size range, favoring relationships based on samples including large trees; and size of sample used to develop relationships (see Table S1 for equations used and original sources).

It is not possible to assess either precision or accuracy of resulting estimates, but significant error and bias are probably inescapable. For some species, equations were not available for the upper Great Lakes region. No available allometric equations were fitted using trees larger than about 70 cm dbh. For all major canopy species in the current data-set, individuals of >70 cm dbh are relatively common, with maximum sizes around 100 cm dbh, and large trees account for substantial fractions of total biomass. Where multiple appropriate equations were available, I compared resulting estimates; estimates for trees up to 50–60 cm dbh were generally similar among equations but, for larger stems, they diverged substantially—by as much as 30% for a 90 cm stem.

Inputs to the CWD pool were estimated using equations for bole biomass only, applied to trees that died during the interval between two measurements. In most cases, CWD input estimates were based on dbh from the last measurement before the tree’s death, so any growth after that measurement and before death would be small (annual diameter increments for canopy trees in this study average around 1.0–1.5 mm/yr depending on species and interval, and growth rates are typically reduced in the years immediately preceding death (Woods, 2000)). However, for the 22-year interval between 1967 and 1989, I used diameter measurements of standing dead trees taken in 1989 where loss of bark and decay were not significant. As much as 30% of above-ground living biomass, in large trees, can be in branches, twigs, and foliage. However, only the very largest branches would contribute significantly to CWD as measured here, and allometric equations typically do not allow differentiation of branches by size. The effects of using dbh for up to several years prior to death and of using bole biomass only will result in some systematic underestimation of CWD input rates. CWD inputs are presented as annualized rates, evenly distributed within each interval between measurements.

Estimating CWD volume and mass

Volume of individual CWD segments was estimated by treating them as regular conic frusta; the height (length) of the frustum was derived from the mapped polar coordinates of the end-points (using three-dimensional coordinates where appropriate). For CWD segments lying only partially within the plot, I calculated the length within the plot perimeter and estimated diameter at that point assuming linear taper between measured ends; volume was calculated for the portion within the plot only. Stumps were also treated as frusta (or multiple frusta where flare led to sectioned measurements). Snag volume was estimated using similar approaches unless appropriate dbh-based allometric equations for bole volume were available (Table S1). CWD volume was converted to dry biomass using taxon-specific relationships between density and decay class from Liu et al. (2006), which uses decay classes nearly identical to those used here (for unidentifiable fragments in decay classes 3 to 5, values for ‘all species’ were used).

Plot groupings

For comparisons across stands of different history and composition, I used plot groupings based on location, habitat, and canopy composition dominance (see Woods, 2000). The Acer group includes plots in two localities with fine-textured, high-cation soils and near-complete hardwood dominance and with maximum ages >300 years. The Tsuga group includes two plots in a single stand with high hemlock dominance and apparent maximum age of about 300 years (K Woods, 1993, unpublished increment core data). Six plots on level, coarse, deep sandy soils, with upper canopy dominated by hardwoods but significant subcanopy and codominant Tsuga component were initially assigned to a ‘mixed flat’ group and three plots in areas of shallower soils on moderate slopes and with mixed hardwood-Tsuga canopy were treated as a ‘mixed slope’. However, all results and trends were very similar for these two groups, and they are pooled here as a single ‘mixed’ plot group. Finally, three plots in the 1830 burn area (slight to moderate slopes, and strong Tsuga dominance) were assigned to the ‘burn’ group.

Statistical approaches

Changes in plot-level characteristics over the full study period were assessed using paired t-tests comparing starting and ending values with appropriate transformations. Differences among compositional plot groupings were assessed using non-parametric Kruskal–Wallis tests with post hoc Mann–Whitney pair-wise tests with Bonferroni corrections.

Coarse woody debris

Coarse woody debris pools in 2007, including snags, averaged 151 m³/ha (range among plot groups, 119–178 m³/ha) and 46.0 Mg/ha (38–52 Mg/ha) (Table 2, Fig. 7 and Table S3). CWD densities were significantly higher in Acer-dominated stands, but were not otherwise different among plot groups (p > 0.05, Kruskal–Wallis test, post hoc Mann–Whitney pair-wise tests with Bonferroni correction). The proportion of CWD in the first two decay classes was substantially lower in the Acer group (31% by mass) compared to other groups (47–61%), and highest for the ‘burn’ group. Snag densities ranged from 9.9/ha (‘Acer’ group) to 98.8/ha (‘burn’) with an overall average of 27.3/ha. Snags constituted 10–20% of total CWD biomass for mixed and hardwood dominated plot groups, but 49% for the ‘Tsuga’ group and 99% for the burn group (Table S3).

Annualized inputs to the CWD pool (total stem mass of dying trees divided by length of measurement interval) averaged 1.9–3.2 Mg/ha/yr, depending on plot group, for the entire 47 year period (Fig. 8). Fluctuations among measurement intervals can be attributed to the effects of occasional large-tree mortality in relatively small sample areas. However, both mixed and Tsuga- dominated fire group showed large increases in CWD input in the second half of the study period, from 1989 to 2009, compared to 1962–1989 (Fig. 7 and Table 3). Rates of CWD input nearly quadrupled over the last two decades, compared to earlier decades, for the ‘burn’ plot group.

Average annual CWD input was 13% (by mass) of total CWD in decay classes 1 and 2 in 2007. For plot groups, corresponding values ranged from 6% (‘burn’ group) to 20% (‘Acer’ group). These proportions are likely somewhat underestimated because input estimates do not incorporate large branches that were included in the CWD pool as measured in 2007.

Coarse woody debris

Overall, 2007 CWD volume (averaging 151 m³/ha for all plots) and biomass (46 Mg/ha), with about 25% of these values in standing snags, are within the range of earlier findings for old-growth forests of the upper Great Lakes and northeastern forest regions (Hura & Crow; Tyrrell & Crow, 1994; Goodburn & Lorimer, 1998; McGee, Leopold & Nyland, 1999; Angers et al., 2005; D’Amato, Orwig & Foster, 2008). Conifer CWD generally shows lower decay rates than hardwood CWD of comparable size (Harmon et al., 1986). However, studies comparing total CWD pools between hardwood and conifer-dominated forests have shown inconsistent results. Most of these, however, have focused on successional or boreal forests (Lee et al., 1997; Pedlar et al., 2002). Here, the ‘Tsuga’ plot group had relatively low CWD densities compared to other groups in this study, while the strongly hardwood-dominated ‘Acer’ group had very high values, suggesting substantially higher recent input rates for hardwood stands. Tyrrell & Crow (1994) show CWD increasing with age in Tsuga stands; values here are in the highest range for their study (suggesting stand ages >300 yr). CWD densities for the Acer group are among the highest published for cool-temperate deciduous forests. These pools are dominated by wood in the most decayed classes, possibly suggesting results of a significant, landscape-scale disturbance several decades ago; this may be consistent with the sharp decrease in biomass observed for these plots between 1962 and 1967.

Inputs of CWD from 1962–2009 averaged 2.5 Mg/ha/yr over all plots—equivalent to 5.5% of the 2007 total CWD pool and 11.7% of the pool of decay classes 1 and 2. If CWD pools are relatively stable or equilibrial, this implies an average residence time of around two decades overall and one decade in the relatively intact classes 1 and 2. Few studies have attempted to estimate residence time for CWD in hemlock-northern hardwood forests, and none have used long-term monitoring of mortality to assess input rates, so it is difficult to draw comparisons. However, existing estimates of residence time vary substantially. Zielonka (2006) estimates residence times of 13 and 24 yrs in classes 1 and 2 in conifer-dominated forests in Poland— substantially longer than turnover times suggested here under steady-state assumptions. However, estimates from old-growth hardwood forest in Ontario suggest a half-life for ‘down woody debris’ (exclusive of snags) of ca. 20 yr, with mean age of wood in classes 1 and 2 of three and five years (Vanderwel et al., 2008), and CWD in a logged, successional forest in New Hampshire decreased in mass by 90% in 20 years (Arthur, Tritton & Fahey, 1993). Given variations in approach (inclusion of snag mass or not, differences in decay class definitions) and coarseness of estimates, these decay rates could be compatible with current CWD pools being in approximately steady-state.

Nearly all previous studies of CWD dynamics are based on one-time measurements and space-for-time substitutions; consequently conclusions about trends over time are strongly assumption-laden. Direct measurement of trends in CWD input, based on long-term permanent-plots with multiple remeasurements, permits an unusual direct assessment of trends. Most strikingly, overall per-year CWD input increased by nearly half between the first 27 years of the study and the last 20 years (from 2.1 to 3.1 Mg/ha/yr), suggesting non-equilibrial status of biomass and carbon dynamics at decadal scales. Such trends might be anticipated in cohort-structured stands where one or more cohorts are reaching senescence; consistent with this interpretation, overall canopy-tree mortality rate increased from 0.6%/yr for 1962–1989 to 0.9%/yr from 1989–2009. These increases were most evident in the youngest stands—the ‘burn’ group—with an increase in CWD input from 0.8 to 3.0 Mg/ha/yr for these two periods, and in the mixed plots (increase 2.1 to 3.9 Mg/ha/yr).

Over all plots, size distribution of stems contributing to total mortality and CWD input did not show the increasing dominance of large stems seen in the living biomass pool (Fig. 4B), but size distributions of CWD input do suggest an increased importance of smaller to mid-size canopy trees (ca. 25–50 cm dbh) in the last two decades. This may be due to enhanced mortality of suppressed sub-canopy trees and sub-dominant canopy trees.

The ‘Acer’ group, with the highest CWD inputs in 1962–1989, and the ‘Tsuga’ group showed less change; if these two groups are in approximate steady-state, implied residence time of CWD is about 16 years overall and 5 years in decay classes 1 and 2 for the hardwood-dominated ‘Acer’ plots, and about 22 years and 10 years for the ‘Tsuga’ group. This is in keeping with general expectations of more rapid decay of hardwood CWD and, for Acer plots, consistent with estimates from Liu et al. (2006) and Gough et al. (2007), who estimate a decomposition rate-constant of 0.09/yr, implying residence time of 11–13 yrs (Liu et al., 2006), for forests of similar composition. Arthur, Tritton & Fahey (1993) estimate 90% loss of CWD biomass in 23 yr for hardwoods in New Hampshire; this would be a more advanced state of decay than in the classes incorporated in this study. However, calculations based on steady-state assumptions should be approached cautiously; 47 years is much less than half of canopy turnover time estimates for similar forests (Frelich & Graumlich, 1994; Dahir & Lorimer, 1996).

However, several factors make these interpretations very tentative. First, input estimates here may be low because of exclusion of branch biomass for dying trees. Second, changes in size distribution towards increasing dominance of very large stems suggest that CWD inputs will tend to shift to larger-diameter materials, with correspondingly longer residence-times. Finally, over longer periods, canopy-tree mortality may be dominated by pulses associated with more severe disturbance (Woods, 2004; Woods, 2007; Frelich & Lorimer, 1991). No such events have occurred during the study period or, as suggested by size and age structures (Woods, 2000), in several decades prior to its initiation. Thus long-term average inputs are likely to be greater than those observed here. All of these factors suggest that estimates of recent CWD input for the study plots may be an underestimate of longer-term, future inputs. If so, CWD pools may be expected to increase.

Sources of error and uncertainty

While the long-term data-set suppporting these analyses supports unusually detailed insight into carbon/biomass dynamics and relationships between tree growth and mortality and CWD pools in old-growth forests, some important uncertainties and potential errors remain. Two of these are likely most important.

First there are large uncertainties connected with biomass estimation for very large trees (>70 cm dbh) that constitute an increasing proportion (up to 30%) of total above-ground living biomass. Until allometric relationships for large stems become available, it will be difficult to reduce this uncertainty. All estimates of biomass in late-successional forests are subject to the same potential errors.

Second, even though the study period of 47 years is unusually long, canopy-tree mortality rates of <1%/yr indicate that it remains brief in the innate time-scale for dynamics in these forests. Observed dynamics represent less than half of a single generation of canopy trees. Further, multiple studies have suggested that dynamics in similar forests are strongly influenced by major disturbance episodes with return times measured in centuries (Woods, 2004; Woods, 2007; Frelich & Lorimer, 1991). Such events would have major consequences for CWD dynamics, but none have occurred at this site during the study period.