Late-Holocene drought and fire drove a widespread change in forest community composition in eastern North America

Abstract

Several regions of the world have recently experienced climate-induced changes in forest composition, highlighting the need to understand the causes, likelihood, and dynamics of abrupt vegetation change. Although few historical examples of climate-induced forest change exist from recent centuries, particularly in humid regions like the northeastern United States, paleoecological records are rich with examples. For example, pollen records from portions of the northeastern United States indicate that eastern hemlock (Tsuga canadensis) and American beech (Fagus grandifolia) abruptly declined in abundance between 500 and 600 yr BP. Concomitant increases in pine (Pinus spp.) and oak (Quercus spp.) occurred. Hypotheses to explain this change have included cooling during the ‘Little Ice Age’ (LIA), Native American activity, drought, and/or fires. To better understand spatiotemporal patterns of forest change and assess potential causes and dynamics, we synthesized regional pollen records and developed two high-resolution, coupled records of vegetation, fire, and drought from bogs in Maine. Results of our synthesis reveal >70% of regional pollen sites recorded shifts in forest composition during this time period. Bog records revealed that forest composition changed a few decades after the onset of drought and regional fires, consistent with increased recruitment of pine and oak during post-disturbance succession. Vegetation changes persisted until European settlement. Our data demonstrate that widespread, long-lasting forest changes were triggered by decadal-to-multidecadal drought and associated fires, highlighting the potential for abrupt, long-lasting forest changes in response to transient climate and disturbance events, particularly when such events occur against the backdrop of more gradual temperature change.

Keywords

beech drought fire forest dynamics hemlock vegetation change

Introduction

Recent and projected increases in global temperatures, associated changes in regional moisture balance, and changes in the frequency of extreme climatic events have raised concerns about potential rapid ecosystem changes (Kelly and Goulden, 2008; Scheffer et al., 2001; Williams et al., 2013). For example, increased rates of background forest mortality (Van Mantgem et al., 2009) and abrupt mortality events have been documented in a number of regions (Allen et al., 2010). Many tree species grow near their physiological moisture limits (Choat et al., 2012), and modeling experiments suggest that temperature-driven moisture stress will significantly alter forest composition and structure during the next century (Williams et al., 2013). However, few analogues for abrupt climate-driven forest changes exist in the historical record of the past few centuries, and longer-term perspectives can help characterize the spatiotemporal dynamics and potential likelihood of these events.

The paleoecological record is rich with examples of abrupt vegetation changes, although causes are still widely debated (e.g. Allison et al., 1986; Booth et al., 2012a, 2012b; Clark and Royall, 1995; Foster et al., 2006; Shuman et al., 2004, 2009). Much of terrestrial paleoecology has traditionally focused on millennial-scale responses of vegetation to climate change; however, finer temporal resolution is routinely obtainable from many depositional archives, and the response of vegetation to sub-millennial climatic variability has received less attention. Understanding changes at decadal-to-centennial timescales may be especially relevant to forest management by identifying climatic and disturbance thresholds for abrupt forest change (Jackson and Hobbs, 2009; Minckley et al., 2012). Recent developments in age–depth modeling techniques (e.g. Blaauw and Christen, 2011), coupled with high-resolution analyses of both vegetation and climatic history, are poised to significantly contribute to our understanding of the drivers and spatiotemporal dynamics of past abrupt vegetation change, including assessments of temporal synchroneity and lags in response of vegetation to changes in climate and/or disturbance (Booth et al., 2012a; Davis and Botkin, 1985; Liu et al., 2012; Minckley et al., 2012).

Peatland archives are one source of information on past environments, recording information on both the history of upland ecosystems, in the form of pollen and microscopic charcoal, as well as changes in the ecology and hydrology of the peatland itself. Ombrotrophic peatlands, those that derive all nutrients and moisture directly from the atmosphere, are particularly sensitive to climate-induced changes in moisture balance (Booth, 2010; Charman et al., 2009), and a range of techniques have been developed to estimate past surface-moisture conditions on these ecosystems using the peatland paleoenvironmental record (e.g. Amesbury et al., 2012; Booth, 2008). Comparative studies of past surface moisture, upland vegetation, and regional fire history using these archives have proved to be valuable to the identification of linkages among climate, vegetation, and fire history at multidecadal-to-centennial timescales (e.g. Booth et al., 2012a, 2012b; Clifford and Booth, 2013). Critical to this comparative approach is the analysis of climate, vegetation, and disturbance proxies from within the same depositional sequences, so that the relative timing and potential lags in vegetation response can be assessed with minimal chronological uncertainty.

Although numerous examples of abrupt vegetation changes exist in the paleoecological record (Booth et al., 2012a; Foster et al., 1998), only a few of these events have been carefully examined by comparison of vegetation, climate, and disturbance proxies within the same sediment cores (e.g. Booth et al., 2012b). For example, many pollen records from lakes and bogs in northeastern North America record shifts in forest composition between 500 and 600 yr BP; in fact, in New England the vegetation changes at this time were likely the largest and most widespread of the past millennium, except during post-settlement land-clearance (Fuller et al., 1998). Relatively mesic species like Eastern hemlock (Tsuga canadensis), beech (Fagus grandifolia), and sugar maple (Acer saccharum) decreased in abundance at many sites, while pine (Pinus spp.) and oak (Quercus spp.) increased during and after this time period (Fuller et al., 1998; Paquette and Gajewski, 2013). The compositional changes that were initiated at this time persisted until European settlement.

A number of hypotheses have been put forward to explain the forest changes in the Northeast at this time, including Native American agricultural practices, climatic cooling associated with the ‘Little Ice Age’ (LIA), drought, and/or fires – or some combination of all these mechanisms (Booth et al., 2012b; Campbell and McAndrews, 1993; Gajewski, 1987; Paquette and Gajewski, 2013; Shuman et al., 2009). Testing these hypotheses has been challenging, in part because few independent records of climate, fire, and vegetation exist with adequate temporal resolution to assess the relative timing of upland and climatic changes. Furthermore, similar vegetation changes were also observed between 500 and 800 years ago in portions of the central and western Great Lakes region (e.g. Booth et al., 2012b; Campbell and McAndrews, 1993; Gajewski, 1987), but it is unclear whether these changes were synchronous with those observed in the Northeast. The vegetation changes in the Great Lakes region were spatially heterogeneous, but drought and fire intolerant species like beech declined in portions of Michigan and southern Ontario at this time, similar to the patterns in the Northeast (Booth et al., 2012b). Recently, fire and drought have been implicated as likely triggers of these vegetation changes (Booth et al., 2012b), although cooler temperatures of the LIA have also been suggested (Campbell and Campbell, 1994; Campbell and McAndrews, 1993; Gajewski, 1987; Paquette and Gajewski, 2013). To determine whether the vegetation changes in the Great Lakes were synchronous with the changes in the Northeast requires precisely dated records and estimates of chronological uncertainty.

Recently, high-resolution analyses of microscopic charcoal and testate amoebae, a group of moisture-sensitive protists (Booth, 2008, 2010; Mitchell et al., 2008), from three Maine bogs indicated widespread drought and associated fires in southern and central Maine between 500 and 600 yr BP (Clifford and Booth, 2013). In fact, this was the only time in the past 3000 years that concurrent drought and fire were recorded at all three bogs (Clifford and Booth, 2013). Although the widespread drought and fires at this time may have caused the forest changes observed in regional pollen records, testing this hypothesis requires a highly resolved comparison of drought, fire, and vegetation across this time period.

In this paper, we examined the forest transition that occurred between 500 and 600 yr BP in the Northeast, including the potential causes of this abrupt vegetation change and its timing relative to similar changes documented in the Great Lakes region. To do this, we first developed an observation-based hypothesis of the spatial patterns of the vegetation changes in the Northeast using available pollen records. We then tested the hypothesis that abrupt vegetation changes were driven by a combination of drought and fire by developing and examining reconstructions of bog surface-moisture estimated from testate amoebae, fire history inferred from charcoal analysis, and vegetation history inferred through pollen analysis. Each paleoecological proxy was sampled continuously in 1-cm intervals spanning the period of interest, so that the relative timing of events could be directly compared. Pollen-inferred changes in demography may be immediate or lagged in response to multidecadal environmental variability, because pollen-based inferences integrate the effects of mortality and recruitment of new individuals to pollen-producing age classes (e.g. Davis and Botkin, 1985; Jackson and Hobbs, 2009; Minckley et al., 2012; Webb, 1986). Therefore, if drought and fire triggered the changes in forest vegetation between 500 and 600 yr BP, we expected stratigraphic evidence of drought and fire to occur synchronously with, or a few decades prior to, the vegetation changes. Finally, we use our data to compare the timing of the vegetation changes at well-dated sites in the Northeast with those from the Great Lakes region to assess whether the vegetation changes in the two regions were likely synchronous.

Study sites

Sidney Bog (44.39, −69.78) and Saco Bog (43.55, −70.46) are located in central and southern Maine, respectively, and are approximately 100 km apart (Figure 1a). Sidney Bog is located at an elevation of 91 m, while Saco Bog is located at an elevation of 44 m. Regional climate is classified as humid continental (Köppen, Dfb), and annual temperatures (~8°C) and precipitation (119 cm at Saco, 106 cm at Sidney) are similar at the two sites. Both sites were glaciated, but were ice-free by at least 12,000 yr BP (Davis and Jacobson, 1985). However, the upland soils surrounding Sidney Bog comprise mainly poorly sorted glacial till deposits (Maine Geological Survey, 2005), while the uplands surrounding Saco Bog comprise sand and fine grained sediments (Maine Geological Survey, 1999). The upland vegetation near Sidney Bog is currently classified as Laurentian-Acadian Northern Hardwood Forest and Acadian Low-Elevation Spruce–Fir–Hardwood Forest (Gergely and McKerrow, 2013), and upland forests surrounding the bog are dominated by hemlock, beech, sugar maple, and yellow birch (Betula alleghaniensis). Vegetation on the better drained soils surrounding Saco Bog is classified as a mixture of Central Appalachian Oak and Pine Forest and Northern Atlantic Coastal Plain Dry Hardwood Forest. Forests are dominated by several oak species, black birch (Betula lenta), white pine (Pinus strobus), and hemlock (Gergely and McKerrow, 2013). Peatland vegetation found on both bogs was similar, with very scattered tree species, including black spruce (Picea mariana), white pine, and tamarack (Larix laricina), with a typical peatland shrub complex comprising Ericaceae species and a carpet of Sphagnum moss. Saco Bog also contains several stands of Atlantic white cedar (Chamaecyparis thyoides).

Figure 1.

(a) The location of sites used in the regional pollen synthesis and significance of pollen changes between two temporal windows (250–530 yr BP and 550–830 yr BP). Closed circles show sites that had significantly different arboreal pollen percentages between pre-transition and post-transition time periods. The numbers correspond to the site numbers in Table 1. Change in pollen percentages between pre-transition and post-transition time periods for (b) hemlock, (c) beech, (d) pine, and (e) oak. Only sites where >5% arboreal pollen was comprised by each taxon are shown. (f) Boxplots showing the changes in relative pollen percentages among all sites across the two time windows.

Methods

Spatial patterning of forest change

To develop a regional perspective on the forest changes between 500 and 600 yr BP (where present was defined as the year AD 1950), we used pollen records from the Neotoma Database (http://www.neotomadb.org/), including records from Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont, and eastern Canada (Quebec, Nova Scotia, and New Brunswick). Datasets that resided in the database (as of 22 April 2013), and the available pending datasets were used in the analysis. For the datasets that resided in the database, we used sites and age–depth models developed by Blois et al. (2011), whereas for the pending data we used the Neotoma Database provided linear interpolation age–depth model for each site. For a site to be included in the synthesis, it had to contain adequate sampling density, so only sites with >2 pollen samples in both a pre-transition (550–830 yr BP) and post-transition (250–530 yr BP) time period were used. The cutoff date of 250 yr BP was used to avoid the effects of European land-clearance (e.g. Foster et al., 1998; Fuller et al., 1998), and the window of 550–830 yr BP was chosen to be of the same length as the post-transition interval. Although Europeans settled some areas of the region as early as 400 yr BP, the regions around Sidney Bog and Saco Bog did not have permanent settlements until AD mid-1700s (~200 yr BP; Coolidge and Mansfield, 1860). We also visually inspected the pollen records to confirm that the Ambrosia spp. peak was not older than the inferred stratigraphic position of 250 yr BP. A total of 27 sites from the region were included in the analyses. Although many records from the region are not well dated, and therefore, our resulting maps of vegetation change should be interpreted with caution, our intent was to provide an observation-based hypothesis of the spatial pattern of forest change across the time period of interest.

Multiresponse permutation procedure (MRPP) was used to determine if forest community change between the two time periods was statistically different (α = 0.05) at each of the 27 sites. MRPP is a nonparametric test used to determine how closely related groups consisting of multivariate data are to one another (Biondi et al., 1988). In order to standardize the application of MRPP among sites, 26 arboreal taxa were included in the analyses (Table 1). A Bray–Curtis (Sørensen) distance measure was used (Faith et al., 1987). In addition to examining the significance of pollen assemblage changes using MRPP, we also analyzed relative changes in pollen abundance from the pre-transition period to the post-transition period for several of the dominant tree pollen types in most records, including beech, hemlock, pine, and oak.

Table 1.

List of pollen taxa used in community regional analysis.

Taxa name
Abies
Acer
Betula
Carya
Castanea
Cornus
Corylus
Cupressaceae
Fagus
Fraxinus
Juglans
Larix
Liriodendron
Liquidambar
Nyssa
Ostrya
Picea
Pinus
Platanus
Populus
Prunus
Quercus
Taxodium
Tilia
Tsuga
Ulmus

Paleoecological reconstructions

A wide-diameter piston corer (10.2 cm) equipped with a serrated end for cutting through peat was used to collect peat cores from Sidney Bog and Saco Bog (Wright et al., 1984). Field and laboratory methods were also described in Clifford and Booth (2013), as this study used the same cores as this previous work on drought and fire. Peat cores were returned to the laboratory and cut into contiguous 1-cm intervals, and subsamples of 1 cm³ were collected for testate amoebae, pollen, and microscopic charcoal analysis according to Booth et al. (2010). Samples were sieved, and the fraction between 15 and 300 µm was kept for analyses. For testate amoeba analysis, a minimum of 100 tests was typically tallied in each sample and most tests were identified at a magnification of 400×. However, in a few samples, the abundance of tests was quite low (6 samples out of 61 samples between 250 and 550 yr BP), and only 50 tests were counted although this is typically sufficient for transfer function applications (Payne and Mitchell, 2009). To reconstruct fire history and vegetation at each site, charcoal fragments and pollen between 15 and 300 µm were tallied on the same slides as the testate amoebae. A known number of exotic Lycopodium spores were added to each sample so that charcoal accumulation rates (CHAR) and pollen accumulation rates could be calculated. Pollen analysis followed standard procedures (Faegri and Iverson, 1989) and counts were continued until at least 200 arboreal pollen grains were identified and tallied for each sample.

Water-table depths were reconstructed from testate amoeba assemblages using a weighted-averaging transfer function derived from 650 modern samples from North America, including Maine (Booth, 2008). Standard bootstrapping techniques (n = 1000) were used to develop uncertainty estimates for water-table depth reconstructions. Although inferred water-table depths are sometimes detrended to remove millennial-scale patterns, because low frequency changes may be related to lateral bog expansion or other nonclimatic developmental processes (Charman et al., 2006), we present raw water-table depth reconstructions here because we focus on one time period and not the long-term trends. However, detrended versions of these two records can be found in Clifford and Booth (2013).

Timing and dynamics of drought, fire, and forest shifts

We developed age–depth models for both sites using an iterative Bayesian approach using the program Bacon (Blaauw and Christen, 2011). Along with producing a best-fit age–depth model, the method results in multiple possible age–depth models given a set of a priori probability distributions describing the mean peat accumulation rate and ‘memory’, which controls the flexibility of the age–depth model, so that temporal uncertainty at any point in the age–depth model can be estimated. To assess the timing of drought, fire, and vegetation changes during the time window of interest, we assigned thresholds to objectively identify the depth of major events, where the onset of drought in the record was defined as the depth in the core where the change in the inferred water-table depth was greater than the root mean squared error (RMSEP) of the transfer function from the previous sample (i.e. a decrease of water-table depth of approximately >8.5 cm). Fire events were defined as those where the peak in charcoal influx exceeded 1.5 standard deviations above the mean charcoal influx (e.g. Clifford and Booth, 2013). The timing of a decline or increase in a pollen type was defined as the age of the stratigraphic location where that taxon increased or decreased by 50%. To provide a better estimation of the amount of time between vegetation, fire, and drought events in each of the two records, we randomly selected 500 possible age–depth models for each site that were generated using the iterative, Bayesian approach, and used these to develop probabilistic estimates of the amounts of time between the identified stratigraphic locations for drought, fire, and vegetation events.

Results and discussion

Forest community shift between 500 and 600 yr BP

Over 74% of sites (20 sites out of a total of 27 examined) in the region recorded a significant change in pollen assemblages between the two time intervals, highlighting the widespread vegetation changes that were centered on 500–600 yr BP (Table 2). When the responses of the dominant individual taxa (i.e. beech, hemlock, pine, and oak) were examined, 78% of sites recorded a >10% decrease in hemlock pollen, 80% of sites recorded a >10% decrease in beech pollen (Figure 1), and 88% of sites recorded declines of this magnitude in either beech or hemlock. Pine and oak increased at over 70% of sites across the time period. Fuller et al. (1998) noted that drier upland sites in New England that contained high percentages of oak pollen did not undergo major changes in forest composition at this time, and our results are consistent with this observation, as the ~20% of sites that did not record shifts in forest composition (Figure 1) tended to have higher percentages of oak. However, forest changes at this time occurred on both till and sandy soils, as evidenced by the records from Sidney Bog and Saco Bog (Figure 2).

Table 2.

Sites and multiresponse permutation procedure (MRPP) statistics used in the regional analysis of forest community change. Site number corresponds to numbers in Figure 2, and bold p-values are significant at α = 0.05.

Site no.	Site name	Citation	Pre-transition (n)	Post-transition (n)	MRPP	A
1	Aino Pond	Fuller et al. (1998)	10	11	p < 0.001	0.173
2	Balsam Lake	Ibe (1982)	4	4	p < 0.05	0.106
3	Basin Pond	Gajewski (1987)	9	7	p < 0.001	0.156
4	Big Reed Pond Hollow	Schauffler and Jacobson (2002)	3	4	p < 0.01	0.319
5	Conroy Lake	Gajewski (1987)	7	7	p < 0.01	0.183
6	Deep Pond	Parshall and Foster (2002)	5	6	p = 0.10	0.044
7	Duarte Pond	Foster et al. (2002)	9	8	p = 0.09	0.038
8	Ely Lake	Gajewski (1987)	5	4	p < 0.05	0.088
9	Green Pond	Fuller et al. (1998)	9	14	p < 0.01	0.053
10	Harlock Pond	Foster et al. (2002)	4	4	p = 0.48	−0.008
11	Lake Pleasant	Fuller et al. (1998)	5	3	p = 0.52	−0.010
12	Lily Warwick	Fuller et al. (1998)	10	9	p < 0.001	0.413
13	Linsley Pond	Brugman (1978)	9	9	p < 0.01	0.059
14	Little Bolton	Fuller et al. (1998)	8	8	p < 0.05	0.046
15	Little Mirror Lake	Fuller et al. (1998)	9	7	p < 0.01	0.081
16	Mansell Pond	Almquist-Jacobson and Sanger (1995)	4	5	p = 0.81	−0.044
17	North Round Pond	Francis and Foster (2001)	7	5	p < 0.01	0.100
18	Otter Pond	Fuller et al. (1998)	7	8	p < 0.01	0.090
19	Pecker Pond	Francis and Foster (2001)	3	3	p < 0.05	0.230
20	Piermont Marsh	Pederson et al. (2005)	9	9	p < 0.01	0.059
21	Quag Pond	Fuller et al. (1998)	8	7	p < 0.001	0.128
22	Saco Bog	This study	4	8	p < 0.001	0.312
23	Sidney Bog	This study	6	9	p < 0.001	0.307
24	Silver Lake (MA)	Fuller et al. (1998)	22	5	p < 0.05	0.035
25	Silver Lake (PA)	Russell et al. (1993)	4	3	p = 0.09	0.059
26	Snake Pond	Fuller et al. (1998)	3	5	p = 0.09	0.112
27	South Bog	Dieffenbacher-Krall (1996)	3	5	p < 0.05	0.254

Figure 2.

Water-table depth, fire, and dominant pollen changes at (a) Sidney Bog and (b) Saco Bog. A depositional hiatus (peatland fire) at Saco Bog occurred at approximately 600 yr BP, and no record exists between 600 and 1500 yr BP. Gray shading on pollen diagrams represents two-times exaggeration, and gray vertical bars highlight the timing of vegetation transition (500–600 yr BP) in both records.

Pollen records from Sidney Bog and Saco Bog were similar to other records in the region, with large declines in the pollen of more mesic taxa and increases in more drought-tolerant taxa between 500 and 600 yr BP. At Sidney Bog, there was a 50% decline in hemlock and beech pollen, with a commensurate increase in pine and oak. Birch and spruce pollen percentages remained relatively constant during the record (Figures S1 and S3, available online). Similar patterns of vegetation change occurred on the better drained soils surrounding Saco Bog, where hemlock pollen declined and pine pollen increased dramatically. Prior to the vegetation change, forests surrounding Saco Bog were characterized by much more pine and oak than Sidney Bog (Figures S2 and S4, available online), yet significant compositional changes still occurred. However, oak pollen percentages remained relatively stable across the transition and subsequently increased after European settlement. At both sites, the compositional changes between 500 and 600 yr BP persisted until European settlement, with hemlock and beech never regaining their pre-decline level of abundance (Figure 2).

Regional drought and fire between 500 and 600 yr BP

High-resolution hydroclimate records derived from testate amoebae at Sidney Bog and Saco Bog both indicate that a severe and prolonged drought occurred between 500 and 600 yr BP. A similarly timed drought has also been documented at Great Heath Bog located 160 km to the east of Sidney Bog (Clifford and Booth, 2013). During the drought event, all three of these peatland sites recorded a large fire event, the only such time when all three records recorded both drought and fire during the last 3000 years (Clifford and Booth, 2013). Furthermore, at Saco Bog the peatland itself burned at this time, depositing a visible layer of charcoal and likely oxidizing pre-drought peat, resulting in a depositional hiatus between ~580 and ~1500 yr BP.

Temporal dynamics of drought, vegetation, and forest compositional change

The age–depth models for each site suggest variable rates of peat accumulation at both peatlands, with average deposition times at Sidney Bog and Saco Bog of 20 and 7 yr cm⁻¹, respectively, which is within the expected accumulation rate of peatlands in the region (Goring et al., 2012). These accumulation rates provide adequate resolution for estimating the timing of abrupt events, and assessing the relative timing of fire, drought, and vegetation changes within the peat stratigraphy at each site.

The temporal sequence of drought, fire, and vegetation changes centered on 550 yr BP was similar at both Sidney Bog and Saco Bog (Figure 3). Vegetation changes at both sites likely occurred within a few decades following the onset of the drought and after the fire event. This lag is consistent with forest recovery after disturbance, as it would take several decades for newly recruited trees to enter pollen-producing age classes (e.g. Davis and Botkin, 1985; Webb, 1986). At Saco Bog, a peatland fire removed a large portion of the record, making our interpretation of timing a bit more uncertain, but drought was still clearly indicated by very low water-table depths at and immediately after the depositional hiatus. Because of the hiatus, it is not possible to determine the exact timing of when the drought began, but we calculated the ‘drought onset’ in the same manner as at Sidney Bog for consistency, and the vegetation clearly changed rapidly a few centimeters after (i.e. above) the hiatus.

Figure 3.

The timing of fire onset, declines in hemlock and beech pollen, and increases in pine pollen relative to the onset of drought for both Sidney Bog and Saco Bog. Time 0 represents the onset of the drought. The probability distributions of the lags in response of fire and vegetation to the drought were developed using 500 possible age–depth models at each site. For these analyses, the onset of fire and drought were assumed to be synchronous at Saco Bog, where the peatland burned.

Our records provide strong evidence that multidecadal drought and fire were both necessary to trigger the widespread vegetation change. For example, Sidney Bog recorded an earlier prolonged drought at 800 yr BP, but the forest composition did not change at this time. However, there was no evidence of fire associated with this drought event, suggesting that fire was important driver of forest change. Similarly, several earlier fire events were recorded in the Sidney Bog record, and although they were associated with relatively dry conditions, these droughts were not as severe and/or prolonged as the 500–600 yr BP drought.

Several other studies have suggested that temperature changes related to the LIA may have led to the observed abrupt and widespread vegetation changes at the time (e.g. Fuller et al., 1998; Paquette and Gajewski, 2013), and while our new records do not directly reject this hypothesis they do provide strong evidence for drought and fire playing a major role. When the timing of the drought, the associated fire events, and the changes in forest taxa are compared with the timing of the LIA temperature changes from the Northern Hemisphere, our data show that these ecological events also probably occurred prior to the cooling of the LIA (Figure 4). However, temperature reconstructions from the Northern Hemisphere may not be representative of the timing of temperature changes in the Northeast, and unfortunately, no high-resolution, vegetation-independent records of temperature exist for the region. Although temperature-related changes may have influenced post-disturbance recovery patterns, and possibly contributed to the long-term persistence of the compositional changes, our records clearly identify drought and fire as the likely proximate drivers of the abrupt and widespread vegetation change.

Figure 4.

(a) Northern Hemisphere temperature anomaly from Mann and Jones (2003), Moberg et al. (2005), and Ammann and Wahl (2007). (b) Comparison of the timing of drought at two Northeastern sites (Sidney Bog – black, dashed line and Saco Bog – black, solid line) and two sites from the Great Lakes (Minden Bog – gray, solid line and Irwin Smith Bog – gray, dashed line; Booth et al., 2012b). (c) The timing of vegetation changes at two sites in the Northeast and two sites in the Great Lakes both lag the onset of drought, and in the Northeast, where the age–depth models are based on more ¹⁴C dates, this lag was likely a couple decades long.

Forest compositional changes persisted until European settlement (e.g. Figure 2), and secondary succession after the drought and disturbance did not return the system to its pre-disturbance composition. This pattern highlights the potential for transient climate and disturbance events to leave long-term legacies at landscape-to-regional spatial scales. The causes of the persistence of the forest changes are unclear; however, post-fire recruitment would have occurred during unusually dry conditions, favoring species with drought-tolerant seedlings and saplings, and once the forest canopy regenerated, positive feedbacks (e.g. altered microclimate, seed influx) likely maintained a relatively stable forest composition (e.g. Nowacki and Abrams, 2008). Alternatively, cross-scale interactions, such as the combination of multidecadal drought and fire with lower-frequency temperature changes associated with Medieval Climate Anomaly (MCA)-LIA transition, may have influenced post-disturbance forest succession (e.g. Ibanez et al., 2007; Vitasse et al., 2012), favoring the recruitment of pines and oaks over hemlock and beech. The LIA climate of the Northeast may have been particularly favorable for pine and oak species; however, the drought and disturbance events were likely necessary to push the system into a new stable state that persisted until European land-clearance (e.g. Beisner et al., 2003; Overpeck et al., 1990).

Timing of forest changes across eastern North America

The drought and vegetation changes in the Northeast likely occurred several centuries after similar changes occurred in the Great Lakes region (Figure 4). For example, a series of drought and fire events recorded at Minden Bog and Irwin Smith Bog in Lower Michigan were linked to the regional decline of beech populations, and associated expansion of pine and oak, between 700 and 800 yr BP (Booth et al., 2012b). However, comparing the estimated timing of vegetation changes and drought events in the two regions indicates little overlap in age estimates, suggesting that they represent vegetation responses to different events (Figure 4). Droughts and fires during MCA, which extended from the midcontinent to the western USA (Cook et al., 2007), were likely associated with the vegetation changes in the Great Lakes region, whereas the drought and vegetation changes in the Northeast appear to have occurred several centuries later. However, given that both regions experienced major drought-induced forest changes during the past millennium, our records highlight the potential of droughts and associated disturbances to lead to abrupt and persistent ecological change. The role of drought in triggering widespread forest change in humid regions like eastern North America has been underappreciated, likely because severe, multidecadal droughts have not been common during the past century in these regions.

Conclusion

Recent climate-induced forest mortality and rapid compositional change, particularly in semi-arid regions, and projected increases in the frequency of extreme climate-driven vegetation shifts have highlighted the need to better understand both short- and long-term dynamics of such events (Allen et al., 2010; Williams et al., 2013). The widespread changes in forest vegetation that occurred in the Northeast between 500 and 600 yr BP were likely driven by a regional, multidecadal drought and associated fires, similar to the forest changes that occurred in the Great Lakes region several centuries earlier (Booth et al., 2012b). Our coupled records of past hydroclimate, fire, and vegetation indicate that the paleoecological record is well suited to examine the characteristics of abrupt vegetation changes on multidecadal timescales relevant to resource management (e.g. Booth et al., 2012b; Foster et al., 2006; Shuman et al., 2009).

Our records from Maine add to a growing body of evidence indicating that forested ecosystems in humid regions are sensitive to moisture variability and may undergo rapid, widespread, and long-lasting compositional change in response to transient drought and fire events (Pederson et al., 2014). Although the rapid forest changes between 500 and 600 yr BP were likely triggered by transient drought and fire, post-fire succession occurred in the context of more gradual temperature changes associated with the LIA–MCA transition. Ongoing and future ecological changes in the region may exhibit similar characteristics, with disturbances linked to decadal-to-multidecadal drought variability, and subsequent forest recovery and gap dynamics occurring during a time of increasing temperatures. Some regional climate projections suggest that warming temperatures may be associated with increased precipitation and drought variability (Hayhoe et al., 2007), and our records suggest that this potentially increases the likelihood of rapid and widespread vegetation change (e.g. Mohan et al., 2009). Much work is needed to more fully incorporate paleoecological perspectives on abrupt vegetation changes into vegetation models and resource-management decisions (Jackson and Hobbs, 2009; Vegas-Vilarrubia et al., 2011), including efforts to estimate wildfire risk (United States Forest Service, 2010). However, long-term perspectives will continue to improve our understanding of the factors and feedbacks that control forest resilience (Reyer et al., 2015), and when coupled with taxa specific bioclimatic and physiological thresholds of tree mortality, will contribute to the development of necessary and realistic models of forest change under variable and changing climatic conditions.

Footnotes

Acknowledgements

We would like to thank Alex Ireland, Katie LeBoeuf, Michelle Spicer, and Travis Andrews for their helpful discussions regarding this paper. Dan Charman, Tim Daley, and Travis Andrews aided in the field work. Two anonymous reviewers provided thoughtful comments that improved the manuscript.

Funding

Funding was provided by the National Science Foundation Paleo Perspectives on Climate Change (EAR-0902441).

References

Allen

Macalady

Chenchouni

. (2010) A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. Forest Ecology and Management 259: 660–684.

Allison

Moeller

Davis

(1986) Pollen in laminated sediments provides evidence for a mid-Holocene forest pathogen outbreak. Ecology 67: 1101–1105.

Almquist-Jacobson

Sanger

(1995) Holocene climate and vegetation in the Milford drainage basin, Maine, USA, and their implications for human history. Vegetation History and Archaeobotany 4: 211–222.

Amesbury

Barber

Hughes

PDM

(2012) The relationship of fine-resolution, multi-proxy palaeoclimate records to meteorological data at Fågelmossen, Värmland, Sweden and the implications for the debate on climate drivers of the peat-based record. Quaternary International 268: 77–86.

Ammann

Wahl

(2007) The importance of the geophysical context in statistical evaluations of climate reconstruction procedures. Climatic Change 85: 71–88.

Beisner

Haydon

Cuddington

(2003) Alternative stable states in ecology. Frontiers in Ecology and the Environment 1: 376–382.

Biondi

Mielke

Jr Berry

(1988) Data-dependent permutation techniques for the analysis of ecological data. Vegetatio 75: 161–168.

Blaauw

Christen

(2011) Flexible paleoclimate age-depth models using an autoregressive gamma process. Bayesian Analysis 3: 457–474.

Blois

Williams

Grimm

. (2011) A methodological framework for assessing and reducing temporal uncertainty in paleovegetation mapping from late-Quaternary pollen records. Quaternary Science Reviews 30: 1926–1939.

10.

Booth

(2008) Testate amoebae as proxies for mean annual water-table depth in Sphagnum-dominated peatlands of North America. Journal of Quaternary Science 23: 43–57.

11.

Booth

(2010) Testing the climate sensitivity of peat-based paleoclimate reconstructions in mid-continental North America. Quaternary Science Reviews 29: 720–731.

12.

Booth

Lamentowicz

Charman

(2010) Preparation and analysis of testate amoebae in peatland paleoenvironmental studies. Mires & Peat 7: 1–7.

13.

Booth

Brewer

Blaauw

. (2012a) Decomposing the mid-Holocene Tsuga decline in eastern North America. Ecology 93: 1841–1852.

14.

Booth

Jackson

Sousa

. (2012b) Multi-decadal drought and amplified moisture variability drove rapid forest community change in a humid region. Ecology 93: 219–226.

15.

Brugman

(1978) Pollen indicators of land-use change in southern Connecticut. Quaternary Research 9: 349–362.

16.

Campbell

(1994) The impact of late Woodland land use on the forest landscape of southern Ontario. The Great Lakes Geographer 1: 22–29.

17.

Campbell

McAndrews

(1993) Forest disequilibrium caused by rapid Little Ice Age cooling. Nature 366: 336–338.

18.

Charman

Barber

Blaauw

. (2009) Climate drivers for peatland palaeoclimate records. Quaternary Science Reviews 28: 1811–1819.

19.

Charman

Blundell

Chiverrell

. (2006) Compilation of non-annually resolved Holocene proxy climate records: Stacked Holocene peatland palaeo-water table reconstructions from northern Britain. Quaternary Science Reviews 25: 336–350.

20.

Choat

Jansen

Brodribb

. (2012) Global convergence in the vulnerability of forests to drought. Nature 491: 752–756.

21.

Clark

Royall

(1995) Transformation of a northern hardwood forest by aboriginal (Iroquois) fire: Charcoal evidence from Crawford Lake, Ontario, Canada. The Holocene 5: 1–9.

22.

Clifford

Booth

(2013) Increased probability of fire during late Holocene droughts in northern New England. Climatic Change 119: 693–704.

23.

Cook

Seager

Chase

. (2007) North American drought: Reconstructions, causes, and consequences. Earth-Science Reviews 81: 93–134.

24.

Coolidge

Mansfield

(1860) History and Descriptions of New England: Maine. Boston, MA: Austin J. Coolidge, 405 pp.

25.

Davis

Botkin

(1985) Sensitivity of the fossil pollen record to sudden climatic change. Quaternary Research 23: 327–340.

26.

Davis

Jacobson

Jr (1985) Late-glacial and early Holocene landscapes in northern New England and adjacent areas of Canada. Quaternary Research 23: 341–368.

27.

Dieffenbacher-Krall

(1996) Paleo and historical ecology of the Cutler Grasslands, Cutler, Maine (USA): Implications for future management. Natural Areas Journal 16: 3–13.

28.

Faegri

Iverson

(1989) Textbook of Pollen Analysis. 4th Edition. London: Wiley, 329 pp.

29.

Faith

Minchin

Belbin

(1987) Compositional dissimilarity as a robust measure of ecological distance. Vegetation 69: 57–68.

30.

Foster

Motzkin

Slater

(1998) Land-use history as a long-term broad-scale disturbance: Regional forest dynamics in central New England. Ecosystems 1: 96–119.

31.

Foster

Clayden

Orwig

. (2002) Oak, chestnut and fire: Climatic and cultural controls of long-term forest dynamics in New England, USA. Journal of Biogeography 29: 1359–1379.

32.

Foster

Oswald

Faison

. (2006) A climatic driver for abrupt mid-Holocene vegetation dynamics and the hemlock decline in New England. Ecology 87: 2959–2966.

33.

Francis

Foster

(2001) Response of small New England ponds to historic land use. The Holocene 11: 301–312.

34.

Fuller

Foster

McLachlan

. (1998) Impact of human activity on regional forest composition and dynamics in central New England. Ecosystems 1: 76–95.

35.

Gajewski

(1987) Climatic impacts on the vegetation of eastern North America during the past 2000 Years. Vegetatio 68: 179–190.

36.

Gergely

McKerrow

(2013) Land cover-National inventory of vegetation and land use. U.S. Geological Survey Fact Sheet 2013–3085. Available at: https://pubs.er.usgs.gov/publication/fs20133085.

37.

Goring

Williams

Blois

. (2012) Deposition times in the northeastern United States during the Holocene: Establishing valid priors for Bayesian age models. Quaternary Science Reviews 48: 54–60.

38.

Hayhoe

Wake

Huntington . (2007) Past and future changes in climate and hydrological indicators in the US Northeast. Climatic Dynamics 28: 381–407.

39.

Ibanez

Clark

LaDeau

. (2007) Exploiting temporal variability to understand tree recruitment response to climate change. Ecology 77: 163–177.

40.

Ibe

(1982) Quaternary palynology of five lacustrine deposits in the Catskill Mountain region of New York. PhD Dissertation, New York University, New York.

41.

Jackson

Hobbs

(2009) Ecological restoration in the light of historical ecology. Science 325: 567–569.

42.

Kelly

Goulden

(2008) Rapid shifts in plant distribution with recent climate change. Proceedings of the National Academy of Sciences, USA 105: 11823–11826.

43.

Liu

Brewer

Jackson

. (2012) Temporal density of pollen sampling affects age determination of the mid-Holocene hemlock (Tsuga) decline. Quaternary Science Reviews 45: 54–59.

44.

Maine Geological Survey (1999) Surficial geology old orchard beach quadrangle, Maine. Open-File no. 99–94. Augusta, ME: Maine Geological Survey.

45.

Maine Geological Survey (2005) Surficial geology Belgrade quadrangle, Maine. Open-File no. 05–45.

46.

Mann

Jones

(2003) Global surface temperatures over the past two millennia. Geophysical Research Letters 30: 1820.

47.

Minckley

Booth

Jackson

(2012) Response of arboreal pollen abundance to late-Holocene drought events in the Upper Midwest, USA. The Holocene 22: 531–539.

48.

Mitchell

EAD

Charman

Warner

(2008) Testate amoebae analysis in ecological and paleoecological studies of wetlands: Past, present and future. Biodiversity and Conservation 17: 2115–2137.

49.

Moberg

Sonechkin

Holmgren

. (2005) Highly variable Northern Hemisphere temperatures reconstructed from low- and high-resolution proxy data. Nature 433: 613–617.

50.

Mohan

Cox

Iverson

(2009) Composition and carbon dynamics of forests in northeastern North America in a future, warmer world. Journal of Canadian Forest Research 39: 213–230.

51.

Nowacki

Abrams

(2008) The demise of fire and ‘mesophication’ of forests in the eastern United States. BioScience 58: 123–138.

52.

Overpeck

Rind

Goldberg

(1990) Climate-induced changes in forest disturbance and vegetation. Nature 343: 51–53.

53.

Paquette

Gajewski

(2013) Climatic change causes abrupt changes in forest composition, inferred from a high-resolution pollen record, southwestern Quebec, Canada. Quaternary Science Reviews 75: 169–180.

54.

Parshall

Foster

(2002) Fire on the New England landscape: Regional and temporal variation, cultural and environmental controls. Journal of Biogeography 29: 1305–1317.

55.

Payne

Mitchell

EAD

(2009) How many is enough? Determining optimal count totals for ecological and palaeoecological studies of testate amoebae. Journal of Paleolimnology 42: 483–495.

56.

Pederson

Dyer

McEwan

. (2014) The legacy of episodic climatic events in shaping temperate, broadleaf forests. Ecological Monographs 84: 599–620.

57.

Pederson

Peteet

Kurdyla

. (2005) Medieval warming, Little Ice Age, and European impact on the environment during the last millennium in the lower Hudson Valley, New York, USA. Quaternary Research 63: 238–249.

58.

Reyer

CPO

Rammig

Brouwers

. (2015) Forest resilience, tipping points and global processes. Journal of Ecology 103: 1–4.

59.

Russell

EWB

Davis

Anderson

. (1993) Recent centuries of vegetational change in the glaciated North-Eastern United States. Journal of Ecology 81: 647–664.

60.

Schauffler

Jacobson

Jr (2002) Persistence of coastal spruce refugia during the Holocene in northern New England, USA, detected by stand-scale pollen stratigraphies. Journal of Ecology 90: 235–250.

61.

Scheffer

Carpenter

Foley

. (2001) Catastrophic shifts in ecosystems. Nature 413: 591–596.

62.

Shuman

Henderson

Plank

. (2009) Woodland-to-forest transition during prolonged drought in Minnesota after ca. AD 1300. Ecology 90: 2792–2807.

63.

Shuman

Newby

Huang

. (2004) Evidence for the close climatic control of New England vegetation history. Ecology 85: 1297–1310.

64.

United States Forest Service (2010) Northeast Wildfire Risk Assessment. NWRA Streering Committee, Geospatial Working Group, 18 pp. Available at: http://www.na.fs.fed.us/fire/pubs/northeast_wildfire_risk_assess10_lr.pdf.

65.

Van Mantgem

Stephenson

Byrne

. (2009) Widespread increase of tree mortality rates in the western United States. Science 323: 521–524.

66.

Vegas-Vilarrubia

Rull

Montoya

. (2011) Quaternary palaeoecology and nature conservation: A general review with examples from the neotropics. Quaternary Science Reviews 30: 2361–2388.

67.

Vitasse

Hoch

Randin

. (2012) Tree recruitment of European tree species at their current upper elevational limits in the Swiss Alps. Journal of Biogeography 39: 1439–1449.

68.

Webb

III (1986) Is vegetation in equilibrium with climate? How to interpret late-Quaternary pollen data. Vegetatio 67: 75–91.

69.

Williams

Allen

Macalady

. (2013) Temperature as a potent driver of regional forest drought stress and tree mortality. Nature Climate Change 3: 292–297.

70.

Wright

Jr Mann

Glaser

(1984) Piston corers for peat and lake sediments. Ecology 65: 657–659. Augusta, ME: Maine Geological Survey.