Holocene dynamics of an eastern hemlock ( Tsuga canadensis ) forest site at the northern range of the species limit

Abstract

Eastern hemlock (Tsuga canadensis) is a shade-tolerant tree of the temperate conifer-hardwood forests of northeastern North America whose northern limit of distribution coincides with the St Lawrence River around Québec City (Canada). We have analyzed the structure and dynamics of one of the very few old-growth hemlock stands in this area to evaluate its successional status at the Holocene scale. To document the origin and long-term development of the hemlock site, we have used conventional forest surveys and macrofossil analysis of woody debris and charcoal pieces at the soil surface and buried in the mineral soil. The ‘Rivière-du-Moulin hemlock forest’ is an old-growth forest, at least 1000 years old, whose structure has been rejuvenated by recurrent surface fires killing most plants of the shaded forest floor and facilitating hemlock regeneration. According to the number of fires and the corresponding fire intervals, the hemlock site experienced a sustained fire regime since the mid-Holocene, first in a developmental context of hardwood forests where beech (Fagus), butternut (Juglans), and birch (Betula) were growing, and then for the last 2400–2100 years as conifer forests where hemlock prevailed throughout or during a large part of the period. Our data highlight the influence of fire on the dynamics of hemlock-hardwood stands, a forest ecosystem generally viewed as being controlled by local light and medium canopy-gap disturbances. Soil charcoal analysis of conifer-hardwood forests may be used concurrently with canopy-gap analysis to decipher the influence of stand-scale disturbances and to calculate better forest turnover at several time scales.

Keywords

charcoal fire history gap Holocene old-growth forest woody debris

Introduction

Eastern hemlock (Tsuga canadensis (L.) Carrière) is a long-lived, shade-tolerant conifer tree inhabiting the cool temperate zone of eastern North America, from the southern Appalachian Mountains to southeastern Canada (Taylor, 1993). The northern range of the species extends from the Great Lakes area to the St Lawrence valley (Canada) where the last trees are thriving in well-drained sites east of Québec City (Bhiry and Filion, 1996b; Marie-Victorin, 1995). Hemlock is currently a major tree species of the temperate conifer-hardwood forests in northeastern North America (Barbour and Billings, 1999; Nichols, 1935), and often behaves as a late-successional species (Doyon et al., 1998; Mosseler et al., 2003).

Eastern hemlock is adapted to small-scale disturbances created by canopy turnover in old-growth conifer-hardwood forests from the Great Lakes region to New England and eastern Canada (D’Amato et al., 2008; Frelich, 2002; Rankin and Tramer, 2002; Ziegler, 2002), and may invade burned sites (Henry and Swan, 1974; Miles and Smith, 1960) and achieve dominance in absence of large-scale disturbances (Woods, 2000). Most species of the conifer-hardwood forests including hemlock are adapted to acidic and relatively dry to mesic soil conditions (Goerlich and Nyland, 2000; Rogers, 1978), although hemlock and white pine (Pinus strobus L.) are among the few species relatively sensitive to drought stress (Abrams et al., 2000). The regional distribution of eastern hemlock varies greatly, the species being more abundant in the east than in the west where it is largely controlled by precipitation loads. Reduced temperatures in summer limit the distribution of eastern hemlock to southern Québec where the species has been exploited in the 19th century for firewood (Simard and Bouchard, 1996), lumber, and leather-tanning purposes (Marchand and Filion, 2014).

Among the diversity of tree species forming the main forest biomes of eastern North America, eastern hemlock stands as an iconic example of the dramatic biogeographical changes that occurred during the Holocene. It is the only tree species which has experienced a major decline of its populations across the species range (Davis, 1981). The origin of the decline has been the focus of several studies since the first mention of the event (Deevey, 1939). The coincidental abundance of insect larvae during the decline as recorded in lake sediments (Anderson et al., 1986), and the direct field evidence of several insect attacks killing large hemlock trees (Bhiry and Filion, 1996a, 1996b) are presently the sole reliable data on the likely causes of the event and supporting the biotic hypothesis. According to pollen data, the ongoing post mid-Holocene decline recovery shows lower densities of hemlock populations across the species range (Davis, 1981). Hemlock was much more abundant before European settlement (Mosseler et al., 2003; Rogers, 1978; Whitney, 1990, 1996). In southern Québec, forest stands composed of eastern hemlock as a dominant or co-dominant canopy tree are confined to small well-drained sites, particularly along river courses, and typically scattered across the agro-forestry landscape because of extensive logging and land colonization over the last 150 years (Marchand and Filion, 2014).

Given the post-settlement context of large-scale environmental disturbances caused by land colonization and forestry practices, natural hemlock forests along the St Lawrence valley are rare ecosystems in need of protection and conservation. This prompted the Québec Government to create in 1975, the first ecological reserve of a national network now comprising 72 reserves and represented by the old-growth hemlock forest of the Rivière-du-Moulin Ecological Reserve. This forest is considered one of the few remaining representative types of the northern conifer-hardwood forest zone of eastern North America whose ecological regime is presumably controlled by canopy-gap disturbances. Whether the dynamics of the northernmost, disjunct hemlock forests, like the Rivière-du-Moulin forest, is similar to that of most stands distributed across the species range remains open and needs further scrutiny. Are there other factors outside the canopy-gap realm influencing the development and maintenance of northern hemlock forests? An integrated ecological and paleoecological analysis of this northern forest type located in an agricultural environment should give the opportunity to document the long-term development of the forest and to evaluate its resilience in relation to canopy-gap turnover, climate, and environmental factors. Accordingly, in this study, we have focused on the detailed analysis of the hemlock forest with the objectives to document its origin, structure, and development within the canopy-gap realm, and to reconstruct the long-term dynamics of the forest site using macrofossil analysis. To fulfill these objectives, we have used conventional forest stand surveys (age and size structure of extant tree species) and macrofossil analysis of woody debris at the soil surface and charcoal pieces at the soil surface and buried in the mineral soil.

Methods

Study site

The studied hemlock forest (46°38′N, 71°53′, 10.7 ha) is part of the Rivière-du-Moulin Ecological Reserve situated in the Lotbinière county, about 50 km west of Québec City, near the northern range limit of the species (Figure 1). The forest site is located on a fluvio-marine terrace (about 20 m a.s.l.) formed after the recession of the Champlain Sea waters in the early Holocene (Occhietti et al., 2001).

Figure 1.

Location of the study site and northern limit of eastern hemlock (Tsuga canadensis).

The hemlock site stands as a forest composed of tall canopy trees dominated by eastern hemlock with American beech (Fagus grandifolia Ehrh.), white pine, red maple (Acer rubrum L.), yellow birch (Betula alleghaniensis Britton), eastern white cedar (Thuja occidentalis L.), and red oak (Quercus rubra L.) as companion species. Large sugar maple (Acer saccharum Marsh.) trees also occur sporadically in the forest. Balsam fir (Abies balsamea (L.) Mill.) and white spruce (Picea glauca (Moench) Voss) are growing on the forest floor as rare seedlings and saplings. Striped maple (Acer pensylvanicum L.), American beech, and particularly eastern hemlock are distributed in the shrub layer, whereas all the other tree species are absent from the regeneration layer. The forest canopy includes several dead trees in standing position, and scattered, undecomposed, or slightly decayed tree stems are lying on the forest floor. The vegetation cover of the heavily shaded forest floor is patchy as more than 90% of the ground is covered by a dense litter of twigs, branches, and leaves. The remaining cover of the forest floor is composed of small, patchy carpets of the moss Pleurozium schreberi (Brid.) Mitt. along with other companion moss species, as well as large, open clones of the clubmosses Lycopodium dendroideum Michx. and Huperzia lucidula (Michx.) Trevisan, the herb Coptis trifolia (L.) Salisb., and the fern Dryopteris carthusiana (Vill.) H.P. Fuchs.

The hemlock forest is bordered by a young softwood/hardwood stand (hereafter border forest) composed of several tree species including large-toothed aspen (Populus grandidentata Michx.), trembling aspen (Populus tremuloïdes Michx.), paper birch (Betula papyrifera Marsh.), yellow birch, and white pine. According to an air photograph (Q65358-134, Ministère des terres et forêts, Québec) taken in 1965, the border forest established after this date when succeeding to an abandoned farm field.

The soil of the hemlock stand is covered by an organic horizon 12.5 ± 5.7 cm thick subdivided into three sub-horizons, that is, F horizon (5.0 ± 2.6 cm), FH horizon (3.9 ± 2.4 cm), and H horizon (3.7 ± 1.4 cm). pH (water) of the mineral soil of the border forest varies between 3.7 and 4.7 (4.4 for a fossil plowed horizon A_p), and the soil pH of the hemlock stand (B horizon) is 4.8. The soil texture of the border forest is a sandy loam, whereas it is more clayey in the ecotonal zone (between the two forest stands) and the hemlock forest. Except for the soil of the border forest that was disturbed by plowing (A_p horizon), the soil of the hemlock stand corresponds to a well-drained to moderately well-drained podzol (drainage class 3-4; CSSC, 1998).

Sampling

A first sampling survey was done during the fall of 2012 to describe the structure of the hemlock forest and to recover soil charcoal pieces for botanical identification and radiocarbon dating. A rectangular plot (50 m × 20 m) was positioned at about 30 m from the border forest. The diameter at breast height (dbh at 1.3 m above ground) of all tree stems was recorded systematically in the sampling plot. Charcoal sampling included 25 microsites with 22 out of the 25 microsites distributed every 5 m along the upper and lower lines of the sampling plot and the other three microsites located every 25 m at the center of the plot. At each microsite, a 20 cm × 20 cm sample of the organic horizon at the contact with the mineral soil was used for charcoal analysis. The thickness of each sample of organic horizon was measured before being extracted. During sampling, all intrusions of surface charcoal into the mineral soil were avoided. Two mineral soil samples were then taken at each of the microsites using a soil corer, allowing the sampling of 750 cm³ of soil per core. A total of 25 cores of the first 15 cm of the solum (A samples) were sampled, and another 25 soil cores (B samples) were extracted beneath each A sample about 15 cm below the mineral soil surface, for a total of 50 samples of mineral soil. Once in the laboratory, each sample was placed for 12 h in a solution of 2–5% KOH to disperse the soil aggregates. The samples were then cleaned and processed using running water over 2 mm and 4 mm mesh screens. Charcoal pieces were extracted manually under a binocular microscope, dried at room temperature and weighed. All charcoal samples collected were >2 mm long, which indicates that they were formed and deposited in situ (Ohlson and Tryterud, 2000). The charcoal pieces are buried into the mineral soil over time through bioturbation triggered by tree uprooting and other soil disturbance agents (Carcaillet and Talon, 1996; Payette et al., 2012; Talon et al., 2005). Charcoal pieces were examined under an incident light microscope at magnifications of 200×, 500×, and 1000× for botanical identification. They were sectioned in order to observe the transversal, radial, and tangential anatomical planes. The anatomical criteria described by Panshin and De Zeeuw (1980), Inside Wood (2004), and the charcoal reference collection of the Centre d’Études Nordiques (Université Laval, Québec) were used for the identification of the charcoal pieces. Because it was not always possible to identify wood charcoal to genus or species, groups of taxa were created. For example, the Picea/Larix laricina group includes charcoal possessing parenchyma rays containing a transversal resin canal and small cross-field pits (piceoids and cupressoids). Tsuga canadensis was joined to Picea/Larix laricina when transversal tracheids (above and below ray cells) were present, although the pieces were too small or too altered to detect the presence of parenchyma rays containing a transversal resin canal. Abies balsamea, Tsuga canadensis, and Thuja occidentalis do not possess parenchyma rays containing a transversal resin canal (except when traumatic), but hemlock can be differentiated from the other two species by the presence of transversal tracheids. The tangential end wall of the parenchyma/ray cells is nodular because of the presence of punctuations in Abies balsamea and Tsuga canadensis but not in Thuja occidentalis. Therefore, hemlock, fir, and cedar were often grouped in different combinations according to the presence of one or both of the following features: the nodular end wall and the transversal tracheids. The state of preservation, the small size, and the extraneous colored material filling cells of the charcoal pieces are the main factors complicating the botanical identification.

A second field survey was executed in the fall of 2013 based on the analysis of data obtained in 2012. A plot of 130 m × 20 m was installed in the forest complex beginning at the border forest, crossing the ecotonal zone, and ending in the heart of the hemlock forest. The principal aim of the second sampling was to evaluate the spatial extent of recent fire occurrences detected in the 2012 charcoal data of the hemlock forest. After a detailed field survey, no fire scars were observed on all the trees of the hemlock site. The first 30 m of the sample plot was covering the border forest and the remaining 100 m extended from the ecotonal zone to the center of the hemlock forest. The sample plot has been divided in 13 quadrats of 10 m × 20 m each. The dbh of all the trees distributed in the quadrats was measured for size structure, which varied between 22 cm and 82 cm. Woody debris (stem diameter > 10 cm) on the forest floor were sampled and identified in the laboratory for botanical composition. A total of 22 dominant trees were cored with a Pressler probe at 30 cm above ground for age determination. In order to identify and evaluate the spatial distribution of surface fires, four monoliths (20 cm × 20 cm) of the organic horizon above the mineral solum were sampled for charcoal extraction. The monoliths were recovered from quadrats 5, 6, 11, and 12 (hereafter Q5, Q6, Q11, and Q12), respectively. In the laboratory, each monolith was partitioned in slices 2 cm thick (except 1-cm slices in Q11). The organic slices were screened and processed the same way as with the 2012 soil samples. Four and nine charcoal pieces of Q11 and Q5, respectively, were botanically identified and radiocarbon-dated using the same procedure as in 2012.

Charcoal dating and fire history

The fire history was reconstructed with the aid of ¹⁴C dating (AMS method: accelerator mass spectrometry) of wood charcoal selected according to species/genus status and to position in the 2012 plot and the 2013 monoliths. Charcoal pieces from the organic samples belong to the organic compartment at the soil surface and charcoal pieces from the mineral solum to the mineral compartment. Each charcoal piece selected was pretreated to CO₂ in the ¹⁴C laboratory of the Centre d’Études Nordiques and dated at the Keck Carbon Cycle AMS facility (University of California, Irvine, CA, USA). The program Calib 7.02 (Stuiver et al., 2013) and the database IntCal13.14C (Reimer et al., 2013) and UWSingleYEAR98 (for dates <71 yr BP) (Stuiver et al., 1998) were used to calibrate the ¹⁴C dates in calendar years. For each radiocarbon date (± 2 standard deviations), intervals of calendar (hereafter cal.) ages with their associated probabilities were computed. Because of the large excursions of atmospheric ¹⁴C through time, probability distributions within the 2-sigma ranges vary greatly in particular during the recent periods of maximum and minimum variations of solar activity (Solanki et al., 2004), resulting in the partitioning of the probability distribution over time for a given date. The mean year of the interval having the highest probability was retained as the cal. yr BP of a given charcoal sample. Sixty-four years were added to the age of the sample in order to show all the fire events in years before 2014. The charcoal dates with overlapping probability distributions of cal. yr were considered as originating from the same fire. To estimate the date of a fire event, we averaged the mean calendar ages of the charcoal pieces originating from the event. These dated events were then used to estimate a fire interval, that is, the number of years that separates two fire events.

An accumulation curve (De Lafontaine and Payette, 2012) was constructed to evaluate whether the number of fires dated corresponds to the number of all the fires that actually occurred in the site. We have used this type of analysis to measure the sampling effort required to properly represent a given population (Soberón and Llorente, 1993). The accumulation curve relates the number of fires identified according to the number of charcoal pieces dated. Most of the fires that theoretically occurred at the site have been detected when the accumulation curve develops into an asymptote. The calculation of the accumulation curve was realized with the accumresult() function using the random method of the BiodiversityR package in the R software (Kindt and Coe, 2005; R Development Core Team, 2011). This function plots 100 accumulation curves by randomly resampling the ¹⁴C dates of all the charcoal pieces of the hemlock stand. The mean accumulation curve is then calculated from these 100 curves and plotted. An indication of the number of fire that could have been detected with further sampling is estimated by fitting a non-linear regression to the average curve using the nls function of the R program with F(n) = F(max) (1 − e^kn). F(n) is the number of fires detected, n the number of dated charcoal pieces, F(max) the maximum number of fires that could have been detected (asymptote), and k the constant (Frégeau, 2013).

Results

Stand composition and structure

The size structure of the hemlock forest (2012 sampling plot, Figure 2) shows the neat dominance of eastern hemlock with trees 50–64 cm dbh and the scattered occurrence of red maple, red oak, and striped maple. The sampled plot (1000 m²) harbors 25 hemlock, 11 red maple, and 2 red oak stems > 20 cm dbh. The advance regeneration of hemlock is relatively dense but scattered in the other parts of the forest stand. The stem size distribution of hemlock shows a reverse J-shape pattern (corresponding to the negative exponential model) suggesting equilibrium conditions between regeneration, growth, and mortality through senescence, although saplings are less represented. Given the current demographic conditions of hemlock, the species will be able to maintain its dominance and regeneration if present site factors remain constant. Although red maple is relatively well established in the stand, the species does not produce seedlings and saplings at this moment.

Figure 2.

Stem-size structure of the eastern hemlock forest of the Rivière-du-Moulin Ecological Reserve. Eastern hemlock: black bars; red maple: gray bars. Note the natural logarithm scale of the number of tree stems.

The distribution of all the tree stems per species from the border forest to the hemlock forest highlights the nature of the canopy of all the forest sites (Figure 3a). The border forest occupies the first three quadrats and the ecotonal zone the fourth quadrat. The hemlock forest is divided in two sections, with the first section from quadrat 5 to quadrat 8 composed of eastern hemlock and white pine and the second section from quadrat 9 to quadrat 13 dominated by eastern hemlock accompanied by red maple. Striped maple is mainly distributed in the second section of the hemlock stand. All the trees of the border forest are relatively small (mean dbh: <25 cm), except for a few white pine and large-toothed aspen stems 30 cm large (Figure 4). The highest number of tree species is in the border forest where several small dead trees in standing position, including white pines, birches, and aspens, are widespread. The border forest is a young secondary stand established rapidly after the abandonment of a farm field a few years after 1965. The abandonment of the farm field is probably related to the creation of the Ecological Reserve in 1975. White pine trees are the tallest and largest trees (15–33 m high) of the section of the hemlock forest nearest to the ecotonal zone. Hemlock trees of the hemlock-red maple section reach a size similar to white pine trees, and it is in this section where red maple and striped maple have the largest size.

Figure 3.

Number of tree species (a) living and (b) dead along the transect from the border forest to the hemlock forest. Q1–Q3: border forest; Q4: ecotonal zone; Q5–Q13: hemlock forest. See Supplementary material Table 1 (available online) for data on dead stems and stumps on the ground.

Figure 4.

Mean, minimum, and maximum diameters of tree species along the transect from the border forest to the hemlock forest. Q1–Q3: border forest, left of dotted line; Q4: ecotonal zone, between dotted lines; Q5–Q13: hemlock forest. The squares indicate the average dbh of all the trees present in each quadrat (see Figure 3 for number of trees per quadrat). The error bars show the maximum and minimum values of dbh.

Most dead trees in standing position and leaning on the ground are predominantly hemlock, and followed by a rather small number of red maple, white pine, and striped maple (Figure 3b; Supplementary material Table 1, available online). A greater diversity of conifer trees (in particular balsam fir and eastern white cedar) occurs in the woody debris compartment than in the living wood compartment. A number of stumps of eastern white cedar, eastern hemlock, balsam fir, and white pine are distributed on the forest floor and indicative of selective logging before the creation of the ecological reserve.

The botanical assemblage of the dead wood population is, in general, similar to the present composition of living trees. All the woody debris are well preserved and no decayed woods were recorded along the transect. Several dead trees in standing position are slowly decomposing without falling on the ground. The number of blowdowns in the forest is smaller than the number of standing dead trees. However, the forest floor in several other sections of the forest is composed of a succession of pits and hollows probably caused by old blowdowns. Also, a large part of the forest ground near the river (Rivière du Moulin) is made of open and closed depressions formed by flooding events in the distant past, at a time when the river floodplain was slightly above the St Lawrence river level after the recession of the Champlain Sea waters.

All the trees growing in the border forest (Q1–Q3) were ⩽40 years old, as also confirmed by the fact there were no trees growing in the site in 1965 when the air photo was taken. Two red maple trees located in the ecotonal zone (Q4) were at least 75 years old. The minimum age of most trees in the hemlock stand (Q5–Q11) were 95–155 years. However, the oldest tree was a 220-year-old hemlock from Q7. It is noteworthy that the oldest hemlocks are located in another section of the forest where several trees are 400–500 years old (Delwaide and Filion, 1999).

Charcoal data

The distribution, number, and mass of charcoal sampled in the 2012 plot vary greatly from one microsite to the other. More than 337 charcoal pieces (including 258 pieces > 2.5 mg) were extracted from the organic compartment in 15 out of the 25 microsites (Table 1). The mineral compartment contained 34 charcoal pieces, that is, 10 times less than in the organic compartment. The mass of the charcoal pieces of the mineral compartment was similar to that of the organic compartment. The botanical composition of the charcoal pieces included about the same number of gymnosperms (59 charcoal pieces; at least 3–4 taxa: Tsuga, Picea/Larix, Abies/Tsuga/Thuja) and angiosperms (58 charcoal species; at least 4 taxa: Betula, Fagus, Juglans, and 1–2 undetermined broadleaf species) (Table 1). Most charcoal pieces of the mineral soil are angiosperms.

Table 1.

Characteristics and botanical identification of the charcoal pieces extracted from the surface and the mineral compartments of the 2012 sampling plot.

Compartment	Surface	Mineral A	Mineral B
n	337	19	15
n ⩾ 2.5 mg	258	18	13
Mean mass (mg)	8.11 (±17.92)	9.01 (±10.26)
Betula sp.	4	0	1
Fagus grandifolia	27	9	2
Broadleaf (except Fagus grandifolia)	2	0	0
cf. Juglans sp.	0	0	2
Undetermined broadleaf	25	5	5
Tsuga canadensis	23	0	0
Picea sp./Larix laricina	2	0	0
Picea sp./Larix laricina/Tsuga canadensis	15	0	0
Abies balsamea/Thuja occidentalis/Tsuga canadensis	3	0	0
Conifer (except Pinus sp.)	6	0	1
Undetermined conifer	10	1	2
Undetermined (xylem)	24	3	2
Undetermined (bark)	192	1	0
Others	4	0	0

At least 192 charcoal pieces of the organic compartment are composed of undifferentiated parenchyma corresponding to bark and stem pith, whereas only one charcoal of these types was found in the mineral compartment (Table 1). The genus/species taxa identified at the soil surface are all representative of the current composition of the hemlock forest, in particular beech and hemlock. Some charcoal pieces of the broadleaf group contain spiral thickenings in the vessel elements most likely corresponding to sugar maple or red maple, but it has not been possible to identify further because of the advanced state of wood decay. Butternut (Juglans) charcoal was recorded in the mineral compartment although the species is absent from the site presently, a situation in contrast to white pine where no charcoal pieces of the species were found in both soil compartments.

Several charcoal pieces were extracted from the four monoliths of the 2013 transect (Figure 5). Hemlock/balsam fir and soft maple charcoal were found in Q5. Hemlock and undetermined conifer charcoal were present in Q11. Sugar maple charcoal and birch and bark charcoal were recorded in Q6 and Q12, respectively.

Figure 5.

Distribution of charcoal pieces in Q5, Q6, Q11, and Q12 monoliths. Charcoal sampling in 2 cm slices in Q5, Q6, and Q12, and 1 cm slices in Q11. Calibrated dates of dated charcoal pieces shown next to bars and different fires identified by letter between brackets. Mineral horizon shown by crosses.

Fire history

According to the 40 ¹⁴C dates of charcoal pieces, 21 fire events were identified from 5830 cal. yr BP to Present (Tables 2 and 3). Based on the accumulation curve of all the charcoal dates, an estimated maximum of 30 fire events could have occurred at the site since the mid-Holocene (Figure 6b) (pooled ¹⁴C-dated charcoal samples of 2013 monoliths and 2012 plot). More fires occurred during the last 1000 and 400 years, the latter period corresponding to European (French) settlement in the study region. The fire intervals per time slice also varied greatly (Figure 6a; Table 3). The mean fire interval calculated for the Holocene period was about three times longer than that of the last 1000 years. It is probable that the fire record before 1000 cal. yr BP is incomplete as suggested by the number of estimated fires deduced from the accumulation curve.

Table 2.

¹⁴C yr BP and calibrated age and botanical identification of charcoal pieces and number of fires recorded in the eastern hemlock site of the Rivière-du-Moulin Ecological Reserve.

No. University of California (UCIAMS)	Quadrat (Q) or Microsite (MP)	Location^a	Taxa identification	¹⁴C Age	Calibrated year^b	Probability	2-sigma interval (cal. yr BP)	n fire events cumul.
131004	MP7	S	Picea sp./Larix laricina	0 ± 15	60	0.99	58–58	21
137222	Q5	3–4 cm	Abies balsamea/Tsuga canadensis	55 ± 25	120	0.51	96–140	20
137776	Q5	5–6 cm	Acer ‘soft’	120 ± 15	165	0.6	123–208
137774	Q5	9–10 cm	Acer ‘soft’	120 ± 15	165	0.6	123–208
137221	Q5	3–4 cm	Abies balsamea/Tsuga canadensis	100 ± 20	150	0.73	91–204
131018	MP23	S	Abies balsamea/Tsuga canadensis	110 ± 15	165	0,61	121–205
137777	Q5	5–6 cm	Abies balsamea/Tsuga canadensis/Picea sp.	110 ± 15	165	0.6	121–205
137220	Q5	7–8 cm	Abies balsamea/Tsuga canadensis cf. Abies	105 ± 20	160	0.59	116–205
137772	Q5	7–8 cm	Acer ‘soft’	105 ± 15	160	0.6	119–204
137775	Q5	9–10 cm	Abies balsamea/Tsuga canadensis	100 ± 15	160	0.58	117–202

133252	MP3	S	Conifer	170 ± 15	260	0.55	232–283	19

131005	MP13	S	Tsuga canadensis	255 ± 15	360	0.96	349–373	18

133247	MP14	S	Indeterminate	280 ± 20	465	0.49	438–493	17
133250	MP19	S	Deciduous	285 ± 15	475	0.5	455–491
131023	MP4	S	Tsuga canadensis	330 ± 20	470	0.79	411–526
133249	MP10	S	Tsuga canadensis	295 ± 15	465	0.67	438–493
137778	Q11	6 cm	Conifer	610 ± 15	680	0.8	644–715	16
137773	Q11	7 cm	Tsuga canadensis	610 ± 15	680	0.8	644–715
137218	Q11	9 cm	Abies balsamea/Tsuga canadensis	705 ± 25	730	0.92	713–751	15
150983	MP 21	A	No xylem, bark or parenchyma tissues	775 ± 20	2280	1.00	740–791

133248	MP17	S	Conifer (except Pinus, Thuya and Taxus)	880 ± 15	830	0.87	799–862	14

137219	Q11	8 cm	Tsuga canadensis	1005 ± 20	1000	0.98	973–1026	13

133246	MP19	B	Conifer (except Pinus, Thuya and Taxus)	1780 ± 20	1775	0.48	1749–1804	12

133253	MP14	A	Indeterminate	1815 ± 20	1825	1	1769–1882	11
131006	MP6	A	Conifer	1870 ± 15	1885	0.86	1834–1934

131027	MP19	B	Conifer	2020 ± 20	2025	0.96	1985–2068	10
131010	MP22	B	Conifer	2040 ± 15	2060	1	2009–2114

131007	MP14	A	Indeterminate	2185 ± 15	2330	0.68	2217–2341	9
150440	MP14	A	No xylem, bark or parenchyma tissues	2215 ± 20	2280	0.86	2294–2370

131003	MP1	S	Betula sp.	2285 ± 20	2390	0.88	2370–2413	8

137771	Q5	11–12 cm	Betula sp.	2740 ± 15	2890	1	2848–2928	7

131009	MP7	B	Betula sp.	2835 ± 15	2995	0.98	2942–3044	6

131017	MP2	A	Fagus grandifolia	3315 ± 20	3600	1	3541–3656	5
131021	MP2	S	Fagus grandifolia	3340 ± 20	3660	0.88	3618–3701
131008	MP2	B	Fagus grandifolia	3370 ± 20	3670	0.91	3634–3707
131016	MP9	A	Deciduous	3740 ± 15	4180	0.8	4142–4214	4
133254	MP9	B	Juglans sp.	4635 ± 20	5480	0.8	5445–5513	3
131024	MP9	B	Juglans sp.	4720 ± 20	5420	0.42	5390–5450
133251	MP9	S	Deciduous	4835 ± 20	5655	0.77	5644–5670	2
150436	MP9	B	Deciduous (cf. with spiral thickenings)	4985 ± 20	5830	1	5718–5810	1

S = surface compartment, depth according to the soil surface (cm) is indicated when sample came from a sliced monolith, A = upper mineral core, B = lower mineral core

Mean cal. yr of interval with highest probability + 64 years, in number of years before 2014

Table 3.

Number of fire events and mean fire interval (±SD) from mid-Holocene to Present.

Period (cal. yr BP)	n fires	Mean interval (years ± SD)
5830–present	21	288 ± 307
5830–2400	8	491 ± 401
2400–present	13	179 ± 180
1000–present	9	118 ± 48
400–present	4	100 ± 3

Figure 6.

(a) Cumulative number of fire recorded during the studied period and (b) accumulation curve of the number of fire events according to the number of radiocarbon-dated charcoal pieces. In (a), dots show fires represented by conifer charcoal (black), hardwood charcoal (white), or a mix of conifer and/or hardwood and/or undetermined charcoal (gray). In (b), dots indicate the mean of 100 random iterations ± 1 standard deviation. F(n) = F(max) (1 − e^kn), where F(n) is the number fires detected, n the number of dated charcoal, F(max) the maximum number of fires that could have been detected (asymptote), and k the constant.

Q5 charcoal belongs to two different fire events, the most recent charcoal piece being 160 years old and the oldest, a birch charcoal, about 2900 years old. The eight charcoal pieces from Q5, each one located in organic slices 2 cm thick and stratigraphically positioned from 3–4 cm to 9–10 cm from the soil surface are of the same age. This situation corresponds to charcoal pieces coming at least from the two wood stems (3 Acer pieces and 5 Abies pieces), which probably remained in a living position after the fire event. The burned stems have been progressively buried during the ongoing vertical development of the organic horizon since the fire event. The recent fire event of Q5 has not reached Q11 where three fire events were recorded at different depths in the organic horizon. The fire at Q5 was recorded in the 2012 plot. Given that direct evidence for the passage of several fires has been recorded both in the monoliths (2013) and in the quadrat (2012), the spatial distribution of the recent fires across the hemlock site was somewhat patchy.

The botanical assemblage of the dated charcoal only includes angiosperms from 5830 to 2300 cal. yr BP. In contrast, gymnosperms are dominating the site at least since about 2000 cal. yr BP (Table 2). It has not been possible to identify the trees killed by the 2300 cal. BP fire. The oldest charcoal pieces identified in the site belong to beech, birch, walnut wood, and unidentified hardwood showing spiral thickenings in the vessel elements.

Discussion

The hemlock stand of the Rivière-du-Moulin Ecological Reserve is one of the rare forests to occur at the northeastern edge of the species range in eastern Canada. Also, the forest is located near the St Lawrence River in the heart of one of the oldest agricultural regions of North America, with the very first seigneuries established during the 17th century (Bureau, 1968; Courville, 1996). The reconstruction of the Holocene history of terrestrial sites based on macrocharcoal analysis shows that fire has been a common and relatively frequent disturbance having impacted the hemlock stand, a forest ecosystem considered ill-adapted to this type of disturbance. It is generally assumed that eastern hemlock is adapted to frequent and light disturbances associated with blowdowns but not to stand-replacing disturbances (D’Amato and Orwig, 2008), in particular fire disturbance.

The current demographic status of eastern hemlock seems at equilibrium according to the uneven-size distribution of the population from seedlings and saplings to mature trees. Despite this distribution, the abundance of standing dead trees and the small number of rotten or poorly decomposed logs suggest a forest currently in a growing state of development. This condition contrasts with the other tree species, including red maple, beech, and red oak, which are not presently producing a regeneration layer. Despite the well-balanced size structure of hemlock, the age of all the trees cored is less than 250 years and the lack of decayed wood at the ground surface suggest that this section of the forest is relatively young. There exists a relationship between the age of several trees of the forest and the age of the most recent fires which have affected the ground layer vegetation. Because of the small number of cored trees and undervaluation of the age of each tree, it has not been possible to calculate the numerical importance of the post-fire cohorts. Needless to say, however, that hemlock > 400 years old are currently growing in the northern part of the ecological reserve, including a tree that was 459 years old in the 1980s (Delwaide and Filion, 1999) and still alive today (at least 500 years old).

The most frequent disturbances which affected recently the hemlock stand are all related to agricultural practices, selective cutting, and fire. The border forest is composed of pioneer tree species that invaded an abandoned farm field (as shown by the plowed A_p horizon) several decades ago (after 1965). The age of most trees inhabiting the ecotonal zone and the nearby hemlock stand is about 75–100 and ⩽200 years, respectively. It is possible that the ecotonal zone is several hundred years old, and maintained as such since the initial period of land colonization during the second half of the 17th century (Courville, 1996). Also, the scattered stumps distributed across the forest suggest the superficial impact of selective cutting in the recent past which did not significantly influence the size structure of the forest. Finally, even after a general appraisal of the present physiognomy of the forest, it would not come easily to our mind that the most common disturbance in the stand is fire.

Indeed, the evidence for the influence of fire in the hemlock site is abundant and direct, thanks to the charcoal pieces retrieved from the organic and the mineral soil compartments. The charcoal assemblages of the hemlock site allowed the reconstruction of the fire history at least since the second part of the Holocene. No charcoal predating the mid-Holocene was found in the site. Although no data are available on the rate of postglacial uplift in the region, and given the present low-elevated terrace where the forest is located, the topographical position of the site before 5800 cal. yr BP was probably at about the level of the St Lawrence River, thus explaining the fossil channels of the site, with water saturated soils less prone to fire ignition and spread. A total of 21 fires were recorded in the site over the last 5800 years, possibly more when considering the number of estimated fires based on the accumulation curve. The fire frequency and the mean fire interval progressively increased and decreased, respectively, from the mid-Holocene to Present.

All the charcoal pieces dated from 5830 to 2400 yr BP were broadleaf tree species (including beech and butternut) belonging to the northern hardwood forest of eastern North America (Barbour and Billings, 1999). During this period, eight fires impacted the site, with a mean fire interval shorter than most plant ecologists would assumed for this type of forest (Frelich and Lorimer, 1991; Whitney, 1986). According to the charcoal assemblages, the fire regime changed after 2400 cal. yr, most likely in line with the dominance of conifer tree species. Fire frequency increased after 2400 cal. yr, and the time elapsed between each successive fire decreased dramatically. The overall trend in the fire frequency may be explained, at least in part, by the possibly incomplete detection of fires in the site before the mid-Holocene. In contrast, an increase in the number of fires was recorded during the last 1000 years, a situation never experienced before in the site. However, caution must be exercised here because of the probability of unrecorded fires before 1000 cal. yr BP because of sampling depth (the number of ¹⁴C-dated samples). According to the accumulation curve of fire events, there exists a probability of the occurrence of a limited number of fires between 5800 and 1000 cal. BP not recorded in our dataset.

The passage of the most recent fires (ca. 1955), not clearly evidenced by the current size structure of the stand, was detected in the stratigraphy of the organic monoliths. The abundance of bark charcoal in the organic compartment testifies for the incidence of light fires having spread over the forest floor, likely burning the external part of the stems (bark, small branches, and twigs of the lower part of the stems) without killing trees, which supports the contention that the branch and twig charcoal ¹⁴C dates are close to the fire dates. Several mature trees survived the surface fires probably because of a thick-bark. The forest which burned intermittently was made of tall canopy trees like eastern hemlock able to create a deeply shadowed undergrowth floor devoid of dense vegetation, thus favoring the passage of low-severity surface fires. Among the sparse plant cover of the forest floor, eastern hemlock thrives well as it is among the best adapted tree species to deep shade conditions. Surficial fires generally burn seedlings and saplings as well as decaying wood on the soil surface without damaging mature trees which, similar to eastern white pine, take advantage of the free forest floor for seed dispersal and germination. The spruce (or larch) charcoal that permitted the detection of a fire that occurred ~60 years ago most likely came from the burning of a sapling. The fact that this fire is represented by this sole piece of charcoal also testifies for the low severity of this fire event. The repetitive surficial fire events having impacted the forest highlight the ability of eastern hemlock to behave like thick-bark white pine trees to survive light fires and to regenerate on a cleared and blackened forest floor.

The hemlock forest established in the site long after the dramatic decline of the species between 5700 and 4250 cal. yr BP (Bhiry and Filion, 1996a, 1996b). The study site was not occupied by hemlock before the decline probably because it was forming a floodplain too humid for mesophilous forest growth. It is not possible to confirm unequivocally the arrival of eastern hemlock at the site prior to the last 1000 years. However, there exists a high probability for the species to have established at the site before the last millennium according to the increased pollen representation of the species in the study area between 3000 and 800 yr BP (Filion et al., 2009). The latter period has been marked by the very small pollen representation if not the absence of eastern white cedar and also the very small representation of balsam fir pollen. Then, it is probable that the conifer charcoal from this period and identified to the Tsuga/Abies/Thuja group is genuine Tsuga charcoal. Many charcoal pieces identified as Tsuga/Abies/Thuja located in the same sampling site were indeed Tsuga pieces. A large part of these pieces may originate from the same tree, but they were too small or altered to be identified to the species level. To our knowledge, there exists several balsam fir stands but only a small number of eastern white cedar stands in the study area, whereas small hemlock stands are currently distributed nearby in the St Lawrence lowlands. This suggests that the hemlock forest developed progressively during the last two millennia with changing abundance of the species according to fire disturbance and other allogenic factors possibly associated with climate variability (Anderson et al., 1986; Filion et al., 2009; Fuller, 1998; Lafontaine-Boyer and Gajewski, 2014; Parshall, 2002).

The hemlock forest, established at least 1000 years ago, has been affected by light surface fires which killed only a small proportion of the canopy trees without modifying the basic size structure of the forest. Although less likely, one cannot totally exclude the occurrence of stand-killing fires during the last 1000 years. Indeed, this condition may happen, although rarely, as documented in a recent case, in the Great Lakes region, of the impact of a stand-replacing fire on a Wisconsin hemlock forest (Murray et al., 2012). Seen from this perspective, the development of the hemlock forest influenced by light recurrent fires proceeded according to a mode similar to that of pine forests composed of tall, thick-bark, self-thinned canopy trees (Keeley and Zedler, 1998). Light surface fires generally occur in stands with cleared forest floor that facilitates the passage of fires and the regeneration of the dominant canopy trees. In this respect, it is probable that the majority of the fires which occurred over the last 400 years were man-made given the fact that the hemlock forest thrived in an agricultural landscape since the very first years of the French settlement, and then to nowadays where fire has been used repetitively for land clearing and grazing (Bureau, 1968).

Conclusion

The hemlock forest of the Rivière-du-Moulin Ecological Reserve is an ancient forest, at least 1000 years old, whose structure has been rejuvenated by recurrent surface fires killing most plants of the shaded forest floor and facilitating hemlock regeneration. Hemlock is a shade-tolerant tree species able to survive the surface fires and then to produce an abundant seedling and sapling bank. According to the number of fires and the corresponding fire intervals, the hemlock site experienced a sustained fire regime since the mid-Holocene, first in a developmental context of hardwood forests where beech, butternut, and birch were growing, and then for the last 2400–2100 years as conifer forests where eastern hemlock trees prevailed throughout or during a large part of the period.

Our data highlight the influence of fire on the dynamics of this eastern hemlock-hardwood forest, a forest ecosystem generally viewed as being controlled by local light and medium canopy-gap disturbances. The overall disturbance regime of hemlock-hardwood forests at the regional scale is generally attributed to local gap-dynamics of canopy trees (D’Amato et al., 2008; Frelich, 2002; Frelich and Lorimer, 1991). Because most studies on the dynamics of the conifer-hardwood forest in eastern North America mainly focused on the classical gap-dynamics syndrome, future studies on these forests should take advantage of the terrestrial charcoal analysis of forest soils to decipher the respective influence of all the possible stand-scale disturbances, including fire and blowdown disturbances, in order to calculate realistic statistics of forest turnover at different time scales.

Footnotes

Acknowledgements

We are most grateful to Ann Delwaide, Pierre Grondin, Jason Laflamme, and Francis Saint-Amour for help in the field, and to Bianca Bédard, Barbara Godbout, Joanie Tremblay, Marie-Hélène Tremblay, and Benoît Filion for assistance in the laboratory. Special thanks to the ‘Direction du patrimoine écologique et des parcs’ of the Québec Department of Durable Development, Environment and Parks for giving us the permission to work at the Rivière-du-Moulin Ecological Reserve. We would like to extend our appreciation to two anonymous reviewers for their insightful comments on an earlier draft of the manuscript.

Funding

This project has been financially supported by the Direction of Forest Surveys (‘Direction des inventaires forestiers’) of the Department of Forest, Wildlife and Parks, Québec Government.

References

Abrams

van de Gevel

Dodson

. (2000) The dendroecology and climatic impacts for old-growth white pine and hemlock on the extreme slopes of the Berkshire Hills, Massachusetts, U.S.A. Canadian Journal of Botany 78: 851–861.

Anderson

Davis

Miller

. (1986) History of late- and post-glacial vegetation and disturbance around Upper South Branch Pond, northern Maine. Canadian Journal of Botany 64: 1977–1986.

Barbour

Billings

(eds) (1999) North American Terrestrial Vegetation. 2nd Edition. New York: Cambridge University Press.

Bhiry

Filion

(1996a) Characterization of the soil hydromorphic conditions in a paludified dunefield during the Mid-Holocene hemlock decline near Québec City, Québec. Quaternary Research 46: 281–297.

Bhiry

Filion

(1996b) Mid-Holocene hemlock decline in eastern North America linked with phytophagous insect activity. Quaternary Research 45: 312–320.

Bureau

(1968) Les solitudes de Lotbinière. Master in Geography, Université Laval.

Carcaillet

Talon

(1996) Stratigraphie et datations de charbons de bois dans les Alpes: quelques aspects taphonomiques. Géographie physique et Quaternaire 50: 233–244.

Courville

(1996) Population et territoire. Atlas historique du Québec. Québec City: Presses de l’Université Laval.

CSSC (1998) Canadian System of Soil Classification, 3rd Edition. Agriculture and Agri-Food Canada. Ottawa: NRC Press.

10.

D’Amato

Orwig

(2008) Stand and landscape level disturbance dynamics in old-growth forests in western Massachusetts. Ecological Monographs 78: 507–522.

11.

D’Amato

Orwig

Foster

(2008) The influence of successional processes and disturbance on the structure of Tsuga canadensis forests. Ecological Applications 18: 1182–1199.

12.

Davis

(1981) Outbreaks of forest pathogens in forest history. In: Proceedings of the IV International Palynological Conference (1976-77), vol. 3. Lucknow: Birbal Sahni Institute of Paleobotany, pp. 216–227.

13.

Deevey

Jr (1939) Studies on Connecticut lake sediments. I. A postglacial climatic chronology for southern New England. American Journal of Science 237: 691–724.

14.

De Lafontaine

Payette

(2012) Long-term fire and forest history of subalpine balsam fir (Abies balsamea) and white spruce (Picea glauca) stands in eastern Canada inferred from soil charcoal analysis. The Holocene 22: 191–201.

15.

Delwaide

Filion

(1999) Dendroséries du pin blanc (Pinus strobus L.) et de la pruche de l’est (Tsuga canadensis L. [Carr.]) dans la région de Québec. Géographie physique et Quaternaire 53: 265–275.

16.

Doyon

Bouchard

Gagnon

(1998) Tree productivity and successional status in Québec northern hardwoods. Ecoscience 5: 222–231.

17.

Filion

Lavoie

Querrec

(2009) The natural environment of the Québec City region during the Holocene. Post-Medieval Archaeology 43: 13–29.

18.

Frégeau

(2013) Dynamique de la pessière à mousses au nord du Lac Saint-Jean (Québec). MSc Thesis, Université Laval.

19.

Frelich

(2002) Forest Dynamics and Disturbance Regimes. Cambridge: Cambridge University Press.

20.

Frelich

Lorimer

(1991) Natural disturbance regimes in hemlock-hardwood forests of the upper Great Lakes region. Ecological Monographs 61: 145–164.

21.

Fuller

(1998) Ecological impact of the Mid-Holocene hemlock decline in southern Ontario, Canada. Ecology 79: 2337–2351.

22.

Goerlich

Nyland

(2000) Natural regeneration of eastern hemlock: A review. In: McManus

(ed.) Proceedings of the Symposium on Sustainable Management of Hemlock Ecosystems in Eastern North America (General Technical Report NE-267). Durham, NH: USDA Forest Service, pp. 14–22.

23.

Henry

Swan

JMA

(1974) Reconstructing forest history from live and dead plant Material. An approach to the study of forest succession in southwest New Hampshire. Ecology 55: 772–783.

24.

Inside Wood (2004) Present. Available at: http://insidewood.lib.ncsu.edu/search.

25.

Keeley

Zedler

(1998) Evolution of life histories in Pinus. In: Richardson

(ed.) Ecology and Biogeography of Pinus. New York: Cambridge University Press, pp. 219–250.

26.

Kindt

Coe

(2005) Tree Diversity Analysis: A Manual and Software for Common Statistical Methods for Ecological and Biodiversity Studies. Nairobi: World Agroforestry Centre (ICRAF).

27.

Lafontaine-Boyer

Gajewski

(2014) Vegetation dynamics in relation to late Holocene climate variability and disturbance, Outaouais, Québec, Canada. The Holocene 24: 1515–1526.

28.

Marchand

Filion

(2014) A dendrochronological analysis of eastern hemlock and white pine in relation to logging in La Mauricie National Park (Québec, Canada). The Forestry Chronicle 90: 351–360.

29.

Marie-Victorin

(1995) Flore laurentienne. 3rd Edition. Montréal: Presses de l’Université de Montréal.

30.

Miles

Smith

(1960) A study of the origin of hemlock forests in southwestern Nova Scotia. The Forestry Chronicle 36: 375–392.

31.

Mosseler

Lynds

Major

(2003) Old-growth forests of the Acadian forest region. Environmental Reviews 11: S47–S77.

32.

Murray

Holmes

Webster

. (2012) Post-disturbance plant community dynamics following a rare natural-origin fire in a Tsuga canadensis forest. PLoS ONE 7: e43867.

33.

Nichols

(1935) The hemlock-white pine-northern hardwood region of eastern North America. Ecology 16: 403–422.

34.

Occhietti

Chartier

Hillaire-Marcel

. (2001) Paléoenvironnements de la mer de Champlain dans la région de Québec, entre 11 300 et 9750 BP: Le site de Saint-Nicolas. Géographie physique et Quaternaire 55: 23–46.

35.

Ohlson

Tryterud

(2000) Interpretation of the charcoal record in forest soils: Forest fires and their production and deposition of macroscopic charcoal. The Holocene 10: 519–525.

36.

Panshin

De Zeeuw

(1980) Textbook of Wood Technology. 4th Edition. New York: McGraw-Hill.

37.

Parshall

(2002) Late Holocene stand-scale invasion by hemlock (Tsuga canadensis) at its western range limit. Ecology 83: 1386–1398.

38.

Payette

Delwaide

Schaffhauser

. (2012) Calculating long-term fire frequency at the stand scale from charcoal data. Ecosphere 3: 59. Available at: https://dx-doi-org.web.bisu.edu.cn/10.1890/ES12-00026.1.

39.

Rankin

Tramer

(2002) Understory succession and the gap regeneration cycle in a Tsuga canadensis forest. Canadian Journal of Forest Research 32: 16–23.

40.

R Development Core Team (2011) R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing.

41.

Reimer

Bard

Bayliss

. (2013) IntCal13 and Marine13 radiocarbon age calibration curves 0-50000 years cal. B.P. Radiocarbon 55: 1869–1887.

42.

Rogers

(1978) Forests dominated by hemlock (Tsuga canadensis): Distribution as related to site and post-settlement history. Canadian Journal of Botany 56: 843–854.

43.

Simard

Bouchard

(1996) The precolonial 19th century forest of the upper St. Lawrence region of Quebec: A record of its exploitation and transformation through notary deeds of wood sales. Canadian Journal of Forest Research 26: 1670–1676.

44.

Soberón

Llorente

(1993) The use of species accumulation functions for the prediction of species richness. Conservation Biology 7: 480–488.

45.

Solanki

Usoskin

Kromer

. (2004) Unusual activity of the sun during recent decades compared to the previous 11,000 years. Nature 431: 1084–1087.

46.

Stuiver

Reimer

Braziunas

(1998) High-precision radiocarbon age calibration for terrestrial and marine samples. Radiocarbon 40: 1127–1151.

47.

Stuiver

Reimer

(2013) CALIB 7.02. Available at: http://radiocarbon.pa.qub.ac.uk/calib/.

48.

Talon

Payette

Filion

. (2005) Reconstruction of the long-term fire history of an old-growth deciduous forest in southern Québec, Canada, from charred wood in mineral soils. Quaternary Research 64: 36–43.

49.

Taylor

(1993) Tsuga canadensis. In: Flora of North America Editorial Committee (ed.) Flora of North America, vol. 2. New York: Oxford University Press, p. 364.

50.

Whitney

(1986) Relation of Michigan’s pre-settlement pine forests to substrate and disturbance ecology. Ecology 67: 1548–1559.

51.

Whitney

(1990) The history and status of the hemlock-hardwood forests of the Allegheny Plateau. Journal of Ecology 78: 443–458.

52.

Whitney

(1996) From Coastal Wilderness to Fruited Plain: A History of Environmental Change in Temperate North America from 1500 to Present. New York: Cambridge University Press.

53.

Woods

(2000) Long-term change and spatial pattern in a late-successional hemlock-northern hardwood forest. Journal of Ecology 88: 267–282.

54.

Ziegler

(2002) Disturbance regimes of hemlock-dominated old-growth forests in northern New York, U.S.A. Canadian Journal of Forest Research 32: 2106–2115.