The role of fire during climate change in an eastern decidious forest at Devil's Bathtub, New York.
Much speculation has been devoted to understanding the role of fire under past climate regimes and during times of rapid climate change (Graham et al. 1990, Walker 1991, Rizzo and Wiken 1992, Clark 1993, Wotton and Flannigan 1993, Stocks 1989). The mediation of climate-driven vegetation change by fire results from fire-fuel interactions. Effects suggested to date can be classed in two categories: (1) fire modifies the timing of changes that would occur anyway (i.e., in the absence of fire); and (2) fire modifies the actual outcome of climate change. Historic data (Abrams and Scott 1989), simulation models (Overpeck et al. 1990), and the paleo record suggest that disturbance, including fire, can either accelerate (Amundson and Wright 1979, Clark 1990) or delay (Tolonen 1987, Bradshaw and Zackrisson 1990) the effects of climate change, depending on initial state and the direction of climate change. Paleo data further suggest that fire can actually modify the ecological consequences of climate change (Grimm 1983), producing "surprises," i.e., outcomes not expected on the basis of climate change alone (Tsukada and Sugita 1982, Cwynar 1987, Clark 1993). Vegetation changes brought on by fire-climate interactions probably occur at decade to century scales and so are difficult to observe directly.
The understanding of how fire mediates vegetation response to climate change suffers from fire records that are poorly resolved in time. Sediment analyses provide relatively continuous records of fire and vegetation change at a single location, thereby holding edaphic and topographic factors constant. They can span numerous periods of climate change, thus offering temporal perspective. Unfortunately, the methods used to reconstruct fire histories generally do not resolve individual fires, and so do not permit resolution of fire intervals at the time scales relevant to plant response (e.g., maturation age, thinning rates, senescence). Without these estimates of individual fires, we cannot determine whether vegetation changes that occur at decade time scales lead change in fire regime or vice versa. Comparisons of this sort could help us to establish whether fire might be a proximate cause for vegetation change or whether fire regime instead responds to changes in fuels that attend vegetation change.
Here, we use a method of charcoal analysis that resolves individual fires in lake sediment to determine whether or not fire may have mediated climate effects on forest composition over the last 10 400 yr in an eastern deciduous forest. Previous work showed that fire was common in xeric forests close to the prairie-forest border before fire suppression (Clark 1990), and that fire regimes changed abruptly with changing agricultural practices during the Holocene in Europe (Clark et al. 1989). There have yet to be full-Holocene analyses of this type from North America. We applied our approach to an eastern deciduous forest to explore (1) whether changes in fire regime led dramatic vegetation changes of the early Holocene; and (2) whether fire might be responsible for maintenance of diversity in mixed deciduous forests, as has been hypothesized on the basis of historical records, life history, and physiology (Abrams 1990, 1992). Evidence used to examine (1) includes timing of change in pollen and charcoal records at times of transition. Evidence for question (2) includes indication of fire occurrence at time scales necessary to affect diversity, i.e., decades to centuries.
We selected a small lake with varved (annually laminated) sediments for our analysis, to maximize the spatial (Prentice 1985) and temporal (Clark 1988a) resolution possible from pollen and charcoal reconstructions. Devil's Bathtub (43 [degrees] 00 [minutes] N, 77 [degrees] 34 [minutes] W; elevation 198 m) is a small (0.7 ha), steep-walled meromictic kettle pond within the Mendon Ponds County Park, south of Rochester, western New York [ILLUSTRATION FOR FIGURE 1 OMITTED]. There are no streams flowing in or out of the lake. The site is located in the till plains of the Erie-Ontario Lowland, [approximately equal to]20 km north of the Allegheny Plateau.
Regional vegetation and climate history
Deglaciation of the Erie-Ontario Lowland was completed between 14000 and 12000 yr BP (Teller 1987). Pollen and plant macrofossil records show herbaceous tundra on the Allegheny Plateau as late as 12 600 yr BP. Boreal taxa Picea, Pinus, and Abies expanded across the Plateau before 10000 yr BP, followed by temperate deciduous trees and Tsuga (Spear and Miller 1976). Forest change during this interval is explained as a consequence of rising summer temperatures (declining Picea, increasing Pinus), rising temperatures in both seasons (increasing Quercus), and increasing moisture availability (decreasing nonarboreal taxa, increasing Abies, Betula). Later arrivals in the northeastern U.S. (Carya, Ulmus, Fagus, and Tsuga), from 9000 to 7000 yr BP, suggest continued increases in precipitation, temperatures, or both (Prentice et al. 1991), or migration lag (Davis 1989). The early- to mid-Holocene occurrence of Pinus strobus, Tsuga, and Betula alleghaniensis at elevations higher than today in the Adirondack and New England mountains suggests that temperatures may have been warmer in the Northeast before 5000 yr BP (Davis et al. 1980, Jackson 1989). Rapid decline in Tsuga pollen after 5000 yr BP is characteristic of most pollen records in the eastern U.S., possibly due to a pathogen (Davis 1981b). Increased Picea pollen after 2000 yr BP suggests decreasing temperatures (Davis 1983, Gajewski 1987, 1988, Whitehead and Jackson 1990).
According to original land survey records, the most widely distributed tree species on the Erie-Ontario Lowland and Allegheny Plateau were Fagus grandifolia and Acer saccharum (Gordon 1940, Seischab 1990). On the Lowland, Tilia americana, Ulmus, and Fraxinus americana were the most common associates of Fagus grandifolia and Acer saccharum, with oaks common on droughty sites. Devil's Bathtub is located within an area with oak forest in the uplands, surrounding lowland open-water and marsh communities. On the Allegheny Plateau, Quercus alba, Pinus strobus, Quercus velutina/rubra, Tsuga canadensis, Castanea dentata, Ulmus spp., and Carya spp. were the next most common types after Fagus grandifolia and Acer saccharum (Seischab 1990).
Iroquois occupied the region before European settlement, but evidence for their effects on vegetation is meager. The Land Surveys of the late 18th century follow the collapse of Iroquois and neighboring Algonquian linguistic groups due to disease and warfare (White 1991, Clark 1995). Seneca maize agriculture is indicated in the Phelps and Gorham tract by 13 fields noted in the Lowland and three fields on the Plateau. Gordon (1940) identifies an abandoned Iroquois village dominated by Pinus strobus on the nearby Holland purchase [ILLUSTRATION FOR FIGURE 1 OMITTED], and Day (1953) reports several firsthand accounts of Iroquois burning in central New York. The apparent low incidence of fire reported by surveyors of the Phelps and Gorham purchase was interpreted to indicate the infrequent use of burning as a manipulative tool by the native Seneca (Seischab 1990). Windthrow was much more commonly recorded in the surveys from central New York (Seischab 1990) and northwestern Pennsylvania (Lutz 1930) than was fire.
The catchment is dominated today by Quercus spp. in the overstory and abundant Acer saccharum in the understory. Outside this small catchment are several plantations of Pinus resinosa established in the 1930s.
We analyzed charcoal, sedimentation rate, and pollen to determine (1) potential consequences of fire for Holocene vegetation change, and (2) environmental factors that could influence the charcoal record. The rate of sediment accumulation responds to lake productivity, transport of allochthonous material to a lake, and redistribution of material within the basin. Varve thickness can provide information on climate history (Dean et al. 1984, Clark 1990), and can suggest when high charcoal accumulation rates might result from bulk sediment accumulation rather than from charcoal production on the surrounding landscape. Because we measure charcoal accumulation and varve thickness independently, we can use the relationship between them to assess charcoal source. If charcoal accumulation is strongly influenced by bulk sediment delivery, then there will be a positive correlation between charcoal and bulk sediment accumulation. If charcoal accumulation is controlled by sources outside the lake (independent of sediment inputs), then there will be no relationship.
Sediment charcoal records respond to fire occurrence at local (100 m) to regional ([10.sup.5] m) spatial scales (Clark and Royall 1994), because samples contain a range of size classes and, thus, residence times in the atmosphere. Low-frequency changes in charcoal abundance emphasize the regional scale. Most particles in sediment samples are small (diameter 5-10 [[micro]meter]) and have residence times of days to weeks, with settling and wet deposition being the principal removal mechanisms (Muller 1984). Residence times exceeding days imply transport of [greater than or equal to]100 km (Andreae 1983). Settling velocities for such particles are generally [less than]1 mm/s (Clark et al. 1996). The low-frequency component of charcoal series emphasizes the regional scale; although individual burns are highly stochastic, the average area burned for a whole region depends on regional climate (Stocks and Street 1983). This low frequency integrates the background source over areas much broader than the scales describing individual burns (Tolonen 1983, Patterson et al. 1987, Clark 1988b, Clark and Royall 1994). "Background" charcoal is taken here to be this average accumulation over extended intervals.
The high-frequency (in this context, annual) component of charcoal series can record individual fires sufficiently close to the lake to produce "peaks" in the charcoal series. Settling is the overwhelming removal process for large (50-1000 [[micro]meter] diameter) particles, most of which have settling velocities of 1-100 cm/s and short atmospheric lifetimes (Sandberg et al. 1979, Muller 1984, Radke et al. 1991; J. S. Clark et al., unpublished manuscript). In the only study of transport from an experimental burn that includes large particles, J. S. Clark et al. (unpublished manuscript) found an order of magnitude decline in particle deposition at the burn edge, and relatively even deposition out to a distance of 100 m. The same study showed that particle diameters were lower in sediments when overall fluxes were low, supporting the interpretation that low values also represent longer distance transport. High-frequency data contain these large peaks associated with individual burns (Clark 1990). Although most large particles settle rapidly, some remain in the atmosphere sufficiently long to be transported over large regions (Radke et al. 1991, Garstang et al. 1996; J. S. Clark et al., unpublished manuscript).
Identification of the peak magnitude that might indicate local fire is implicit in many charcoal studies, but rarely calibrated. This magnitude is critical, however, because it determines the fire frequency that is calculated from the series. Identification of scales of variability in charcoal data narrows the range of possible factors that could be responsible for an observed series of sediment accumulation rates. To differentiate among scales of variability, we examined sediment accumulation at Devil's Bathtub in time and frequency domains, using power spectra that show variance as a function of temporal scale.
To evaluate whether or not there is a clear separation between "peaks" and "background" values, we calculated a sensitivity index that shows how the mean interval between charcoal peaks varies with the threshold value used to define peaks. "Fires" identified by application of an index were compared individually with varve series to determine the extent to which they might result from increased bulk sediment supply to the center of the lake. An alternative method made use of previous analysis in northwestern Minnesota, where comparisons of sediment charcoal in three lakes with fire scars on Pinus resinosa trees were used to identify an index constituting the charcoal accumulation corresponding to local (catchment) fire (Clark 1990). Although comparability to Devil's Bathtub is limited by differences in vegetation, fire behavior, and hydrology (see Discussion), cautious use of the method provides insight. The Minnesota series are the only existing sites where annual charcoal records have been compared with mapped fires, so they represent a valuable, albeit limited, baseline for interpretation of sediment charcoal.
We constructed distributions of intervals between peaks for portions of core having clear separation of peak and background levels. The distribution of intervals between fires represents a model-independent indicator of biotic control over fire occurrence, e.g., accumulation of fuels (Clark 1989). Changing fire hazard with time since the last fire can be interpreted from such distributions. Biotic control over this risk is represented by significant departures from the time-in-dependent exponential distribution. The stochastic component of the model reflects the inherently unpredictable nature of ignitions.
We analyzed sediment pollen to determine changes in Holocene vegetation that attend changes observed in the charcoal record and Holocene climatic change (see Regional vegetation and climate history). Sediment pollen integrates vegetation pattern at a range of spatial scales, biased toward taxa that produce large quantities of wind-dispersed pollen. In order to assess the vegetation types and climate settings to which past fire regimes are most likely to apply, we searched a modern (surface-sample) pollen data set for analogs to fossil spectra. Modern regions having pollen assemblages most similar to fossil spectra are most likely to have similar vegetation and climates. Maps of closest analogs to fossil spectra were used to make this assessment.
Sediments were recovered from the deepest location in Devil's Bathtub. The sediment surface was sounded from the ice surface to produce a bathymetric map [ILLUSTRATION FOR FIGURE 1 OMITTED]. Overlapping 1-m sections were recovered from 14.12 to 18.89 m using piston (Wright et al. 1984) and freeze corers operated from the ice surface. A freeze core filled with dry ice and ethanol (Swain 1973) was used to obtain the upper 1.3 m of sediment, including the sediment interface at 14.12 m.
Stratigraphy was described for reference with thin sections. Many laminations had characteristic thicknesses, structure, and color that made them readily identifiable. Frozen sediments were described, and all laminations were measured at subfreezing temperatures before sampling to determine potential losses during processing. Ten samples were taken for 14C dating.
Charcoal quantification and varve measurements were completed on petrographic thin sections (Merkt 1971, Clark 1988a). Frozen and piston core sediments were cut into segments 5 cm long x 2 cm wide. Each section overlapped 1 cm with the stratigraphically adjacent ones to insure continuity of counts. Samples were dehydrated with acetone, impregnated with epoxy resin, and sectioned using standard petrographic methods. Varves were identified and marked directly on thin sections. Lamination thicknesses were measured using a stereoscope mounted on a dendrochronograph. Although annual laminations are discernible throughout most of the Devil's Bathtub core, short sections exist in which laminations are not well differentiated. These segments do not appear to be massive events in thin section; they contain structure, but not sufficiently clear to permit confident measurements. Lamination age estimates for these gaps were estimated from mean thicknesses of 10 laminations on either side of the gap. Because these estimates are subject to errors resulting from changes in sedimentation rate, we did not attempt to interpret deposition rates or charcoal accumulation across these intervals. The most obvious error would result from episodic deposition. This error affects the time scale, but it does not affect charcoal accumulation estimates from sediments adjacent to gaps, because they are independent of varve measurements and counts.
Annual records of charcoal accumulation are presented as indices representing estimates of the fraction of the sediment surface covered by charcoal. Charcoal identification, measurement, and quantification follow methods of Clark (1988a, 1990). Only completely opaque and angular particles [greater than]60 [[micro]meter] in length were tallied at 64x on a dissecting microscope. Using stereology relationships, the summed linear measure of charcoal along a varve ("varve length") is representative of charcoal area per sediment area,
C = charcoal length/varve length = charcoal area/sediment area/yr,
the second equality being the stereology relationship (Clark 1988a). Years are in the denominator, because each count is taken from a single varve.
Samples were obtained from frozen and piston cores for pollen analysis. Samples 0.5 [cm.sup.3] were removed using a brass piston sampler. Samples were prepared and counted according to standard methods of Faegri and Iversen (1975). Haploxylon and diploxylon Pinus types are termed "Pinus strobus" and "Pinus banksiana-type," respectively, as no attempt was made to separate pollen of P. banksiana, P. resinosa, or other diploxylon types that might contribute pollen. Pinus pollen grains that could not be identified to type were apportioned to the types according to the relative portions of identified types in the spectrum.
Quantitative treatment and interpretation
Smoothing and power spectra for varve thickness and charcoal. - Trends in background charcoal were represented by low-pass (30-yr window) filtered series, smoothed using a Fourier transform. We calculated power spectra for varve-thickness and charcoal series over portions of the core containing high variability: the late Holocene for varves and the early Holocene for charcoal. A three-point smoothed power spectrum is presented with a 95% confidence interval (Clark 1990). The confidence interval is wider than for an unsmoothed spectrum, accommodating a dependence among adjacent points introduced by the smoothing procedure.
Threshold indices. - Histograms of raw indices, C, emphasize the occurrence of nearby fires. Clark (1990) compared these indices from sediments of three Minnesota lakes with fire scars on nearby Pinus resinosa trees to estimate a parameter C[prime], values above which correspond to local fire. In that study, the average background for three different lakes ranged from 42 to 68 [mm.sup.2][cm.sup.-2][yr.sup.-1] above the background level. We examined the distribution of peaks in the charcoal series corresponding to the parameter C[prime] obtained from the Minnesota series.
A sensitivity coefficient is used to identify evidence for the peaks that would distinguish a nearby fire from the low values representative of nonfire years. Assume that background values are relatively constant and of much lower magnitude than the large peaks that can occur when a fire burns to the lake edge (Clark 1990; J. S. Clark et al., unpublished manuscript). Under this assumption, there is a range of relatively rare values, intermediate between background and peaks, that are unlikely to occur. Within this intermediate range, the mean interval (in years) between peak values is relatively insensitive to the index used to calculate it. We calculate here a simple coefficient that is minimized in such a range, if it indeed exists. Let [Mu] be the mean waiting time in years between charcoal values that exceed C[prime] (i.e., "peaks"). The dimensionless coefficient [s.sub.[Mu]C[prime]], the sensitivity of the mean waiting time [Mu] to charcoal index C[prime], is defined as
[s.sub.[Mu]C[prime]] = [Delta][Mu]/[Mu] / [Delta]C[prime]/C[prime].
It is the proportionate change in mean waiting time for a proportionate change in the index that is used to define a "fire." If there is an identifiable threshold charcoal amount corresponding to local fire, then this index will be positive and decreasing below that threshold, near zero at the threshold, and positive and increasing above. Such a trend would reflect a clustering of charcoal values from nonfire years at low values, and values for fire years well above this range. Plots of charcoal distributions, together with [s.sub.[Mu]C[prime]], should indicate whether or not such a separation exists between peak and background years.
Distribution of fire intervals. - Distributions of waiting times between charcoal peaks were fitted to portions of the charcoal series having clear separation between peaks and background levels, using maximum-likelihood estimators (Clark 1989). We tested the hypothesis that fire probability increases with time since the last fire using a likelihood ratio test of the maximum likelihood parameter estimates against an exponential model.
Identification of modern pollen analogs. - We calculated chord distances between fossil spectra and surface [TABULAR DATA FOR TABLE 1 OMITTED] samples (Overpeck et al. 1985) contained in the North American Pollen Data Base to identify geographic distributions of modern analogs. We first calculated the mean for fossil spectra spanning 1000-yr intervals of the core. Chord distances were calculated from each of these spectra to each modern pollen spectrum. Taxa used for these comparisons were Pinus, Abies, Picea, Acer rubrum, Acer saccharum, Betula, Ostryal Carpinus, Carya, Corylus, Fagus, Fraxinus, Populus, Quercus, Tsuga, and Ulmus. Maps of modern spectra having the lowest chord distances were prepared for each interval.
Pollen indicators, lamination counts, and 14C chronologies (Table 1) were used to construct an age/depth model for the core [ILLUSTRATION FOR FIGURE 2 OMITTED]. We did not feel that confident lamination measurements could be obtained near the surface of the core, so a floating varve chronology was tied to the Tsuga pollen decline, a widespread event that appears synchronous across eastern North America at 4634 [+ or -] 304 14C yr BP, mean [+ or -] 1 SD (Webb 1982). This chronology includes short sections where we did not feel confident lamination counts could be made. The portions of the core where estimates were used are too small to be observed in Fig. 2 (with the exception of a 4.5-cm missing segment centered on 7000 yr BP); we include them in the charcoal diagrams that are shown at larger scales. Agreement between the age/depth model obtained from lamination thickness measurements and the 14C depth model provides strong evidence that laminations are annual, albeit interrupted in places.
Several pollen assemblage zones are defined for purposes of discussion. Only dominant pollen types important for the fire history interpretation are included here [ILLUSTRATION FOR FIGURE 3 OMITTED]. The full data set is available from the NOAA North American Pollen Data Base.
Picea/Pinus banksiana-type, DB-4:1885-1840 cm; 10400 to 9200 yr BP. - Picea is abundant throughout this zone, with Pinus banksiana-type increasing from 10 100 yr BP up to the top of the zone. DB-4a is the earlier (10 400 to 10 100 yr BP) Picea-dominated phase with herbaceous taxa, Artemisia, Gramineae, and Cyperaceae; DB-4b is the later phase, with higher percentages of P. banksiana-type (10 100 to 9200 yr BP). Other arboreal taxa include Ostrya/Carpinus, Ulmus, Quercus, Fraxinus, and Tsuga.
Pinus strobus, DB-3: 1840-1803 cm: 9200-8300 yr BP. - Pinus strobus increases with declines of Picea and Pinus banksiana-type between 9200 and 8300 yr BP. The beginning of the zone is located at the Picea decline, followed by maximum Betula, Populus, Abies, and Corylus values. Picea declines more rapidly than does Pinus banksiana-type.
Mixed hardwoods, DB-2: 1803-1460 cm; 8300 yr BP to AD 1800. - An abrupt decline of Pinus strobus and disappearance of Populus pollen attends the rise of several hardwood taxa, notably Fagus. Also increasing in this zone are deciduous taxa Quercus, Carpinus/Ostrya, and Acer. Tsuga declines after 5000 yr BP and then rises steadily after 4000 yr BP. Populus reappears with the Tsuga decline. Picea and Castanea increase after 3000 yr BP.
Agricultural indicators, DB-1: 1460-1412 cm; AD 1800-1990. - Agricultural indicators Gramineae, Ambrosia, Rumex, and Plantago increase as several arboreal types decline, particularly Tsuga, Fagus, and Acer saccharum.
Surface samples least dissimilar (having the lowest chord distances) to fossil spectra are presented as maps for each of the pollen zones, together with 200-yr portions of the charcoal series [ILLUSTRATION FOR FIGURE 4 OMITTED] that are representative of that pollen zone in terms of peak frequency and average accumulation.
DB-4a. - Few modern samples fall within chord distances [less than]0.25 [ILLUSTRATION FOR FIGURE 4B OMITTED]. Closest analogs are near James Bay, a region where Picea pollen is especially abundant, but Pinus is less abundant than in regions to the south and west (Anderson et al. 1991). Mean January and July temperatures here are [approximately equal to]-20 [degrees] and 15 [degrees] C, respectively, and mean annual precipitation is 800 mm.
DB-4b. - The region of closest analogs is expanded to much of western Ontario, Manitoba, and Northern Saskatchewan [ILLUSTRATION FOR FIGURE 4C OMITTED]. Less similar samples extend into the Northwest Territories and eastward into Quebec, Labrador, and the northeastern USA. Analogs west of Hudson Bay reflect the increased Pinus of zone DB-4b. Those east of Hudson Bay may result from increases in Betula and the appearance of Abies. Mean temperatures span ranges of -30 [degrees] to 10 [degrees] for January and 10 [degrees] to 20 [degrees] for July. Precipitation ranges from 300 to 1000 mm (Anderson et al. 1991).
DB-3. - Close modern analogs are found in the western Great Lakes region dominated by pines, particularly northeastern Minnesota, Wisconsin, and northern Michigan [ILLUSTRATION FOR FIGURE 4D OMITTED]. Mean January and July temperatures are - 10 [degrees] and 20 [degrees] C, respectively, and mean annual precipitation is 700-900 mm (Anderson et al. 1991).
DB-2. - Relatively few modern samples are close analogs for mid-Holocene samples [ILLUSTRATION FOR FIGURE 4E OMITTED]. Chord distances [less than]0.30 are scattered and lie within mixed forest that today includes eastern deciduous taxa, northern hardwoods, Pinus strobus, and Tsuga, largely depending on substrate. Mean temperatures in January are -7 [degrees], and in July, 20 [degrees]. Mean annual precipitation is 900-1100 mm (Anderson et al. 1991).
Graphs of varve thickness include several large peaks from 6600 to 6000 yr BP in DB-4, and after 5000 yr BP in DB-2 [ILLUSTRATION FOR FIGURE 2C OMITTED]. The thick varves at 10 130 yr BP consist of silt and organic debris. Varve thicknesses increase in variability after 5000 yr BP, continuing up to the point at which they can no longer be confidently counted after 2500 yr BP. The power spectrum for this portion of the varve series shows high variability at many scales, but variance is not continuously distributed across all scales [ILLUSTRATION FOR FIGURE 5 OMITTED]. The lack of correlation between charcoal accumulation and sediment accumulation [ILLUSTRATION FOR FIGURE 6 OMITTED] indicates that charcoal supply is independent of sedimentation processes within the lake.
The high (annual) resolution of the record, together with its long duration (8000 yr), means that we could not display the entire record in a way that shows the high frequency variability. We therefore present it in two sections [ILLUSTRATION FOR FIGURES 7 AND 8 OMITTED], together with important pollen types for reference.
Several gaps occur in the charcoal series; the most extensive is a 4.6-cm missing segment at an intersection of core sections centered at 7000 yr BP [ILLUSTRATION FOR FIGURE 7 OMITTED]. Shorter gaps are the intervals where confident varve measurements could not be made: 9500, 7900, 4050, 3700, 3000, and 2600 yr BP [ILLUSTRATION FOR FIGURES 7 AND 8 OMITTED]. Because we did not confidently identify laminations in the upper part of the core, we do not present charcoal indices for the sediments younger than 2500 yr BP.
General trends are emphasized by filtered series, i.e., decadal rather than year-to-year variability [ILLUSTRATION FOR FIGURES 7 AND 8 OMITTED]. Charcoal increases to a maximum in DB-4b, declines at the beginning of DB-3 (8800 yr BP), and declines further at the beginning of DB-2 (8000 yr BP) [ILLUSTRATION FOR FIGURE 7 OMITTED]. Both of these declines are rapid, with changes occurring over several years' time. Values are low throughout the mixed hardwood zone DB-2, rising somewhat following the Tsuga decline after 5000 yr BP, and again after 3500 yr BP [ILLUSTRATION FOR FIGURE 8 OMITTED].
Estimation of local fire
Distributions of charcoal accumulation rates in zone DB-4 suggest that "local" fires might best be approximated by accumulation rates near C[prime] = 50-60 [mm.sup.2][center dot][cm.sup.2][center dot][yr.sup.-1]. These high values might represent burns that come to within 1-100 m of the lake edge (see Methods). Lower values produce a homogeneous distribution [ILLUSTRATION FOR FIGURE 9 OMITTED], as observed at other sites in eastern North America (Clark et al. 1996). Only one value falls between 40 and 60 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1]. Above 60 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1] are several large values, constituting [less than]2% of the total values. Sensitivity coefficients fall to minimum values within this range of 40 to 60 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1] and subsequently rise [ILLUSTRATION FOR FIGURE 9 OMITTED]. A paucity of intervals in this range is consistent with the interpretation that "background" levels may be relatively distinct from fire years. In contrast, the distribution of values in DB-3 shows no clear region of minimal sensitivity [ILLUSTRATION FOR FIGURE 9 OMITTED], all but one value falling below 60 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1]. On the basis of sensitivity coefficients, we suggest that local fire is probably represented by C[prime] values [greater than]60 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1] in zone DB-4. Coefficients suggest that there is no such value in DB-3 [ILLUSTRATION FOR FIGURE 4D OMITTED], suggesting that a clear separation between fire and nonfire years may not have existed at this time, or that local fires did not occur.
Two lines of evidence support the interpretation that local fire is represented by values [greater than]60 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1] in zone DB-4. Application of threshold values from previous studies in Minnesota, where fire years were found to be 50 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1] (i.e., C[prime] [approximately equal to]100 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1]) above background levels [ILLUSTRATION FOR FIGURE 4A OMITTED], gives values of 53 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1] (DB-4a) and 58 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1] (DB-4b). These values produce clear separation between peaks and background in zones DB-4a [ILLUSTRATION FOR FIGURE 4B OMITTED] and DB-4b [ILLUSTRATION FOR FIGURE 4C OMITTED]. The mean fire interval [Mu] predicted by values of C[prime] = 55 is 88.0 yr for DB-4a and 89.8 yr for DB-4b (Table 2).
A second line of support for this interpretation comes [TABULAR DATA FOR TABLE 2 OMITTED] from power spectra of the two zones [ILLUSTRATION FOR FIGURE 10 OMITTED]. DB-4 displays peaks at 60 and 90 yr [ILLUSTRATION FOR FIGURE 10B OMITTED], close to the mean intervals of 80-90 yr, the interval implied by a C[prime] value of 55. It is not unexpected to recover the mean interval from spectra, provided events tend to be clustered at regular intervals. The DB-3 portion of the series shows no significant peaks [ILLUSTRATION FOR FIGURE 10A OMITTED], also consistent with lack of a clear threshold C[prime] [ILLUSTRATION FOR FIGURE 9A OMITTED].
Distributions of fire intervals in the early Holocene
Distributions of fitted intervals for the DB-4 zone show modal values near 80 yr, and intervals broadly distributed over 50 to 200 yr [ILLUSTRATION FOR FIGURE 11 OMITTED]. Using the threshold C[prime] = 55, we obtained a good fit for a Weibull distribution that suggests increasing fire risk with time since the last fire (Table 2). A likelihood ratio test led us to reject the hypothesis that Weibull parameter c = 1 (P = 0.0387).
Interpretation of fire and vegetation
Pollen and charcoal of the early Holocene. - Clear transitions in both the background and in distributions of peaks attend changes in vegetation types of the early Holocene. Before the Pinus banksiana-type expansion at 10000 yr BP, background charcoal is low, but large peaks indicate three fires in 300 yr [ILLUSTRATION FOR FIGURE 12 OMITTED]. Large individual peaks would seem consistent with stand-replacement crown fires of the sort common in modern Picea forests, but such values might also represent surface fires; peaks of similar magnitude characterize the surface fires of presettlement northwestern Minnesota [ILLUSTRATION FOR FIGURE 4A, B OMITTED]. The large peak at 10 370 yr BP is attended by three years of high sediment accumulation [ILLUSTRATION FOR FIGURE 12 OMITTED]. The increase in charcoal far exceeds the increase in sediment accumulation, indicating that the charcoal peak did not simply result from secondary deposition of particles already present in lake sediments. No other charcoal peaks in the early Holocene are attended by increased sedimentation rate. The transition to increased Pinus banksiana (DB-4b) is marked by 25 yr of thick varves at 10 170 yr BP, followed by an increase of background charcoal and continued occurrence of peak values [ILLUSTRATION FOR FIGURE 5 OMITTED]. Layers contain silt and macrophyte fragments, suggesting secondary deposition of littoral sediments.
The period of transitory increase in Abies, Populus, and Betula before Pinus strobus gains dominance is marked by steadily declining background charcoal, but possibly frequent local fire, represented by peaks approaching threshold values of 30-50 [mm.sup.2][center dot][cm.sup.-2][center dot][yr.sup.-1] [ILLUSTRATION FOR FIGURE 13 OMITTED]. A large peak at 8880 yr BP is followed by an abrupt decline in background charcoal, at a time when Abies, Populus, and Betula decrease and Pinus strobus reaches maximum values [ILLUSTRATION FOR FIGURE 7 OMITTED]. Background charcoal remains relatively low throughout the period of Pinus strobus dominance, until another large peak at 8080 yr BP. Following this event, it appears that Pinus strobus is replaced by hardwoods (particularly Fagus), and background charcoal declines abruptly to low values. Thus, the time of Pinus strobus dominance may have begun and ended with fire, but local and regional fire appears less important than during previous times [ILLUSTRATION FOR FIGURE 4D OMITTED].
Background charcoal and varves of the mixed hardwood zone. - Despite a lack of charcoal peaks that could confidently be attributed to local fire [ILLUSTRATION FOR FIGURE 4E OMITTED], the mixed hardwood zone DB-2 shows some clear trends in background charcoal and in varve thicknesses [ILLUSTRATION FOR FIGURE 8 OMITTED]. Most notable is the increase and subsequent decline of background charcoal that follows the Tsuga decline from 5000 to 4500 yr BP. Increases at 5000 yr BP in pollen of taxa that may respond to fire, Betula and Populus, followed by modest increases in Quercus [ILLUSTRATION FOR FIGURE 3 OMITTED], may result from some increase in fire occurrence at a regional scale. Lack of clear peaks suggests that fires may not have occurred locally.
Individual thick varves and larger average varve thicknesses characterize the late Holocene, after 5000 yr BP, especially following the decline in charcoal [ILLUSTRATION FOR FIGURE 8 OMITTED]. Varves are highly irregular, with thick ([greater than]2 mm) individual varves occurring at 47.8 [+ or -] 46.5 yr intervals (mean [+ or -] 1 SD). The power spectrum does not contain clearly differentiated peaks [ILLUSTRATION FOR FIGURE 5 OMITTED], suggesting that sedimentation rates in the late Holocene may have become sensitive to a range of different processes operative at different scales. Because varves in our core did not extend to the surface, we were unable to document fire regimes of the immediate and post-European settlement.
This first annually resolved record of charcoal accumulation spanning the Holocene provides insights into the role of fire before, during, and since the establishment of the mixed hardwood forests of central New York. This interval experienced transition from Late-Glacial to Interglacial climate and vegetation assemblages that included open spruce woodland, spruce/pine, white pine, and mixed hardwood. We found that a single site can experience vastly different fire importances at different times [ILLUSTRATION FOR FIGURE 4 OMITTED], and that those importances can shift abruptly [ILLUSTRATION FOR FIGURES 7, 8 OMITTED]. We observed clear changes in both the background and peak charcoal accumulations as climate and vegetation changed. Peak charcoal values of the magnitude that today accompany local fire [ILLUSTRATION FOR FIGURE 4A OMITTED] (Clark 1990) are associated at Devil's Bathtub with pollen data that suggest fire-prone vegetation [ILLUSTRATION FOR FIGURE 4B, C OMITTED]. Background charcoal values (Table 2) change as pollen assemblages change [ILLUSTRATION FOR FIGURE 7 OMITTED], consistent with modern relationships in eastern North America (Clark and Royall 1994).
The role of fire during vegetation transitions can be difficult to detect in time series data, because feedbacks between vegetation (fuel) and fire probably determine how vegetation responds to climate change. Fire could play a role at times of transition, be they climatically driven or otherwise, by opening the preexisting stand, changing light and nutrient availability, and stimulating germination of some species. Climate change might alter ignition frequency and/or fuel moisture, with fire being the proximate cause for vegetation change. Alternatively, fire regime could simply respond to changing fuels as vegetation responds to changing climate. The source of sediment might shift with changes in climate and catchment vegetation, thereby affecting accumulation of charcoal in sediments.
Interpretation of the role of fire across transitions among vegetation types must consider the different scales at which charcoal and pollen sense landscape pattern. Although fire is partly controlled by regional climate, giving coherent geographic trends in charcoal accumulation (Clark and Royall 1994), the record of large particles is local (Clark 1990; J. S. Clark et al., unpublished manuscript). Large charcoal particles quantified by thin-section methods have short atmospheric residence times (Clark 1988b, Radke et al. 1991), and record fire mostly around a lake catchment. Pollen data record vegetation over a broader region (Prentice 1985, Jackson 1990), even those from a lake as small as Devil's Bathtub. The principal pollen transitions here are characteristic of other sites in the Northeast (Davis et al. 1980, Davis and Jacobson 1985, Webb 1988, Jackson and Whitehead 1991). To facilitate comparison of pollen and charcoal records, we have analyzed the smallest lake we could identify having varved sediments in mixed deciduous forest. Although individual fires cannot be confidently tied to changes in pollen percentages because the data document different scales, we found that the most dramatic vegetation changes recorded in sediment pollen were attended by local fire and by changes in "background" charcoal abundance [ILLUSTRATION FOR FIGURE 7 OMITTED]. In some of the cases we will interpret, the data suggest changes in fuel structure as climate changed, resulting in modified fire behavior. In other cases, the cause is unclear. Our interpretation avoids speculation on the precise arrival time of different taxa, because increased pollen percentages do not distinguish between a taxon approaching a lake as a wave front and population growth in place. Nor do we interpret "wiggles" in pollen curves as evidence of responses to fire, because of the different scales represented by charcoal and pollen. We focus instead on the differences in charcoal accumulation among pollen assemblage zones, and on the ways in which charcoal accumulation changes between zones.
The Pinus banksiana expansion (10 200 yr BP)
The Pinus banksiana expansion is an exceptional vegetation change for this site: it shows no change in the distribution of fire intervals, despite large increases in background charcoal. A climatic shift to conditions favorable for fire that might have precipitated Pinus banksiana expansion is not supported by the charcoal record, because we did not observe increased fire frequency. Thick varves at the transition [ILLUSTRATION FOR FIGURE 12 OMITTED] suggest severe local disturbance, but there is no evidence for increased fire. The possibility remains that fire frequency changed on a regional scale but did not increase at Devil's Bathtub. Analysis of this possibility requires additional sites.
An alternative interpretation of increasing background charcoal after 10 200 yr BP is that climate (e.g., increased temperature) or migration lag explains expansion of Pinus banksiana, which resulted in higher particulate emissions per fire without an overall increase in fire frequency ([ILLUSTRATION FOR FIGURE 12 OMITTED], Table 2). Declining herb pollen [ILLUSTRATION FOR FIGURE 3 OMITTED] suggests a transition from open spruce woodland to a more closed-canopy Pinus banksiana/Picea forest with crown fire at 50-200 yr, as observed today in northern Ontario and Manitoba (Lynham and Stocks 1991), regions consistent with modern analog comparisons for this interval [ILLUSTRATION FOR FIGURE 4C OMITTED]. The increase in background charcoal could reflect higher fuel loads, higher burn efficiencies, higher emission factors (mass of particulate release per mass of fuel consumed), and/or a change in transport and deposition processes. Fuel loads are among the more obvious candidates, because the decline in herb taxa suggests that the canopy closed with the expansion of Pinus banksiana. Crown fires are more likely when canopies are homogenous (Payette et al. 1989, Timoney and Wein 1991), and an understory of Picea can facilitate crowning in stands of mature Pinus banksiana (Stocks 1989). Burn efficiencies and emissions also may have changed. Background charcoal should respond to higher emissions, because there would be greater delivery from "nonlocal" fires (Clark and Royall 1994). Unfortunately, the few airborne estimates of emissions from boreal forests (Radke et al. 1991, Cofer et al. 1990; J. S. Clark et al., unpublished manuscript) do not include comparisons from Pinus banksiana vs. Picea stands. The increase in fuel loads and potential for crowning in the closed canopies following Pinus banksiana expansion lead us to believe that climate was responsible for observed changes in vegetation, which in turn affected fire behavior.
Support for the potential influence of Pinus banksiana on fire regime comes from the distribution of fire intervals [ILLUSTRATION FOR FIGURE 11 OMITTED], strikingly similar to those characteristic of some modern boreal forests. Intervals broadly distributed at 50 to 200 yr compare with similar distributions for parts of Alberta (Johnson 1992), Ontario (Lynham and Stocks 1991), Quebec (Payette et al. 1989, Bergeron and Brisson 1990), and Labrador (Foster and King 1986), except where precipitation is especially high and fire breaks are abundant (Foster 1983). These regions are all represented in the suite of closest modern analogs to the pollen spectra of this interval [ILLUSTRATION FOR FIGURE 4C OMITTED].
The transition to Pinus strobus (8000 to 10000 yr BP)
Diminished fire frequency does not appear to explain the sequence of pollen changes leading to Pinus strobus expansion, unless the true Pinus strobus expansion was delayed 1000 yr after the beginning of increasing pollen abundance. The transition from Picea/Pinus banksiana to Pinus strobus spans a gradual decline in background charcoal, followed by abrupt decline at 8000 yr BP. Local fires may have occurred regularly during the 9500 to 8800 yr BP increase of Pinus strobus and decline of Picea, particularly at the beginning of the decline (9440 yr BP) and its end (8870 yr BP) [ILLUSTRATION FOR FIGURE 13 OMITTED]. Declines in background charcoal and in the magnitude of charcoal peaks during the Pinus strobus increase and Picea/Pinus banksiana decline might suggest that Pinus strobus (in addition to several other taxa) was able to invade while fires were still frequent, and that the emissions gradually declined as it became dominant. The interpretation that Pinus strobus invaded a landscape of frequent fire and subsequently modified that fire regime is dependent on the interpretation of the timing of that expansion. However, it is also possible that Pinus strobus arrived later.
The Picea decline coincides with the transient rise of early-successional hardwood taxa Betula and Populus and, to a lesser extent, Abies. Macrofossil studies in the Adirondacks show that Betula pollen of the time was produced there by Betula papyrifera (Jackson and Whitehead 1991). The transient assemblage containing the less fire-tolerant Abies is reminiscent of the eastern boreal forest of Quebec and Labrador, where fires occur less frequently than in drier parts of western Ontario and Manitoba (Foster 1985, Flannigan and Harrington 1988, Payette et al. 1989). Abies is a "late-successional" species today, being abundant late in stand development in eastern Canada, where fire is less frequent than in the Midwest (Flannigan and Harrington 1988). It is not clear whether the modern geographical abundance distribution of Abies results from poor seed production and dispersal (a disadvantage when burns are large), low seedling survival, and/or higher fire sensitivity (Rowe and Scotter 1973). The sedimentary record at Devil's Bathtub shows the rise of Abies, together with Betula and Populus, at a time when fire may have been frequent, although background levels were declining [ILLUSTRATION FOR FIGURE 7 OMITTED].
Alternatively, Pinus strobus may not have arrived near Devil's Bathtub until after 9000 yr BP. The extent to which fire may explain the transition to Pinus strobus is complicated by the fact that Pinus pollen is extremely well dispersed, and may therefore record transitions at a regional scale. A shift from Pinus banksiana-type pollen to Pinus strobus is witnessed elsewhere in the Northeast in the early Holocene (Whitehead 1979, Davis et al. 1980, Spear et al. 1994). Jackson and Whitehead (1991) used macrofossil data to argue that pollen of Pinus strobus might lead the local appearance of trees by 1000 yr. If this were the case at Devil's Bathtub, the "arrival" of P. strobus may indeed have occurred near the fire at 8800 yr BP, 1000 yr after the pollen percentage began to increase, and coincident with the decline of Picea pollen and abrupt decline in charcoal. If P. strobus arrival did not occur until after the large peaks in the charcoal series ceased at 8800 yr BP, then it is possible that frequent (possibly crown) fire in P. banksiana delayed its expansion, as postulated for Picea abies in parts of Sweden (Bradshaw and Zackrisson 1990). The change in fire regime at 8800 yr BP may simply reflect changing fuels, with the rapid decline of Picea and Pinus banksiana and expansion of deciduous taxa, Abies, and Pinus strobus. A shift to surface fires or a decrease in fire occurrence might explain the sudden decline in charcoal abundance at 8800 yr BP.
It is important to note that the close-interval sampling does not resolve the "arrival time" of a widely dispersed pollen taxon such as Pinus. Broad dispersal reduces temporal resolution of changing tree distributions. We did not attempt closer interval sampling of pollen, because we recognized that "arrival times" could not be precisely determined. It is possible that macrofossils would help to better resolve this arrival.
The interpretation of fire in Pinus strobus (8800 to 8100 yr BP)
We do not find a clear relationship between fire intervals and Pinus strobus dominance. Assemblages may not have burned as frequently as did the preceding Picea/Pinus banksiana stands. Decreased fire importance is consistent not only with model predictions of a more positive water balance and rising temperatures (Kutzbach and Guetter 1986) and presettlement analog comparisons suggesting stands where crown fire is uncommon (Clark et al. 1996, Clark and Royall 1996) [ILLUSTRATION FOR FIGURE 4D OMITTED], but also with decreased fuels that would have attended a switch from the crown fire regime. It is possible that the high background charcoal (relative to subsequent hardwood assemblages) may represent low-emission surface fires. If understory fuels were modest, low-intensity fires may not have been recorded as individual events in our data. Because Pinus strobus spans a broad range of climate and vegetation types in modern forests, there is not a particular fire regime common to Pinus strobus. We are unaware of emissions estimates from fires in closed-canopy Pinus strobus stands that might aid interpretation of our results. The interpretation that low-emission surface fires dominated at this time remains tentative, because surface fires in stands of other pine species can produce large amounts of charcoal [ILLUSTRATION FOR FIGURE 4A OMITTED].
The hardwood expansion (8100 yr BP)
The early Holocene expansion of hardwoods throughout the Northeast may be a direct response to rising temperatures and precipitation (Prentice et al. 1991), as well as to decreased fire. The timing of charcoal and pollen changes suggests a connection between fire regime and vegetation change. The transition from Pinus strobus to Fagus occurred after a fire at 8100 yr BP [ILLUSTRATION FOR FIGURE 7 OMITTED] that marked an abrupt shift to the lower average charcoal values that characterize the hardwood zone [ILLUSTRATION FOR FIGURE 4E OMITTED]. The large difference between average charcoal before and after this peak suggests that a new pattern of fire and fuels began with a single year's burn near 8100 yr BP [ILLUSTRATION FOR FIGURE 7 OMITTED]. The time scale is more consistent with local events (e.g., fire) than with regional climate change. In addition, the localized dispersal of Fagus pollen (Prentice 1985) makes the spatial resolution of charcoal and pollen comparable. If surface fires occurred in Pinus strobus forests before 8000 yr BP, then they may have delayed Fagus expansion. Alternatively, inflammable Fagus-dominated stands may have been expanding regionally during previous centuries when Pinus, Betula, and Populus were declining; the distinct local expansion of Fagus after 8000 yr BP greatly reduced probability of fire immediately around the lake.
Interpretation of fire in mixed hardwoods
Is fire necessary to explain the widespread abundance of oak in eastern North American pollen spectra of presettlement time? Life history and physiology of the oak species that dominate in our region (Quercus rubra, Q. velutina) suggest that fire might be needed to ensure their long-term persistence in the face of competition from more shade-tolerant taxa (reviews by Crow 1988, Abrams 1992). Our failure to find clear evidence of fire during the late Holocene raises the possibility that oak forests might be maintained in the absence of fire. This result is important, because fire has been identified as one of the sources of variability that might be required to permit coexistence of a large number of species that compete in similar ways for a few limiting resources. It is possible that, although Devil's Bathtub is in an area of sandy soils with abundant oaks, the overall fire frequency was less than that of similar xeric sites not embedded within a matrix of northern hardwood forest that dominant the bulk of the Erie-Ontario Lowland [ILLUSTRATION FOR FIGURE 1 OMITTED]. Land survey records that suggest little evidence of fire in this area (Seischab 1990) are supported by the charcoal evidence that fires may not have occurred.
Given the lack of emissions estimates from surface fires in deciduous forests, the possibility remains that fires occurred in the late Holocene and failed to leave a record in sediments. Against this argument are the clear charcoal peaks associated with individual surface fires in Pinus resinosa forests [ILLUSTRATION FOR FIGURE 4A OMITTED]. Sporadic crowning is possible in some of the burns represented in Figure 4a, but they must have been primarily surface fires, because an extensive Pinus resinosa canopy persisted for centuries while these burns were occurring. Large charcoal peaks associated with slash-and-burn agriculture in southern Ontario (Clark and Royall 1995b) and southwestern Germany (Clark et al. 1989) indicate that high charcoal values can be common in mesic deciduous forests. It is possible that factors other than surface fires contributed to charcoal values in these cultural settings (Clark 1995), but, in both of these examples, charcoal accumulation remained substantially higher during occupation phases than that present before occupation. If fires were occurring in the late Holocene forests near Devil's Bathtub, then they contributed far less charcoal to the sediments than observed in these previous studies. Emission estimates are needed to aid interpretation of this period.
Potential effects of recent climate and cultural change on fire regimes are difficult to interpret. Pollen data suggest modest vegetation changes in northeastern North America over the last 2000 yr, consistent with a cooler climate with a possible increase in moisture availability (Davis et al. 1980, Gajewski 1987, Jackson and Whitehead 1991). Although our charcoal data do not span this time of changing climate and rise of Iroquois maize agriculture, the abundances of dominant pollen types, including Quercus, do not change appreciably from the top of our charcoal series to the rise of European agriculture. It is, therefore, plausible that the limited fire importance suggested in the Land Survey records applies to the forest composition present in our late Holocene charcoal data. We do not observe increased fire frequency at other northeastern sites during the last 2000 yr (Clark and Royall 1995a, 1996), except near an Iroquois settlement at Crawford Lake, Ontario for AD 1350-1650 (Clark and Royall 1995b).
Fire is not the only potentially important disturbance in the late Holocene mixed hardwood forests. Varve thicknesses are controlled in part by climate effects on lake processes, the delivery of dissolved and particulate matter from catchments, and winter density currents that move littoral sediments to deeper areas. We cannot differentiate among these sources, but the data suggest that disturbance within the catchment may have increased since the mid-Holocene [ILLUSTRATION FOR FIGURE 8 OMITTED]. Thin sections show that many layers contain leaf fragments and other debris, suggesting that climate variability may have increased; such variability may have driven quasi-cyclic changes in forest composition. The time scales describing that variability [ILLUSTRATION FOR FIGURE 5 OMITTED] include those comparable with life histories of the dominant tree taxa. The trend toward a more positive water balance in recent millennia may have been attended by greater annual to decadal variability in climate.
These results raise two questions regarding the role of disturbance in eastern deciduous forests. First is the possibility that periodic fire does not explain the presence of Quercus taxa on some sites of intermediate to mesic moisture status. Quercus pollen is abundant in presettlement pollen spectra. If fire were necessary for Quercus persistence, we would expect to see at least some evidence for fires over a period spanning millennia. However, more sites are needed to better assess the regional occurrence of fire in presettlement time. Second, if not fire, then maybe windthrow and/or ice storms contribute to oak importance. Our varve data cannot be taken as conclusive evidence for catchment disturbance, and windthrow or ice storms are not generally identified as important mechanisms for Quercus regeneration (Crow 1988).
Tsuga decline (4800 yr BP)
The catastrophic decline of Tsuga pollen over eastern North America after 5000 yr BP suggests an episodic event, e.g., a pathogen (Davis 1981b, Allison et al. 1986). Our annual-resolution charcoal data show increased background charcoal, suggesting that regional, but probably not local, fire may have followed the Tsuga decline. Possible regional fires may reflect increased fuels following large-scale mortality, but low Tsuga percentages even before the decline suggest that it may not have been a dominant taxon locally.
Disturbance regimes and life history
Long-term disturbance data such as these permit scrutiny of widely held, but rarely tested, notions that the relationship between life history and the distribution of intervals between disturbances explain whether or not a species can persist. If disturbance is to serve as an explanation for existence of a species at a site (e.g., Heinselman 1973, Huston 1979, Sousa 1984), it must occur on time scales that match those describing life history activities (maturation times, thinning rates, longevities) of the dependent species (Clark 1991). If the density of waiting times between fires is of the right temporal scale, fire could explain persistence of the species on the landscape. If the density describes a time scale incompatible with life history, it cannot be invoked as an explanation. In the case of fires, short intervals could extirpate a population before it achieved reproductive maturity. Long intervals could lead to invasion of species not adapted to disturbance.
The charcoal data of the Picea/Pinus banksiana assemblage of the early Holocene provide an opportunity to address the relationship between disturbance and life history time scales. The suggestion that fire probability increases with stand age, together with the observed increase in background charcoal as Pinus banksiana increased, are both consistent with the interpretation that Pinus banksiana altered the fire regime, due to its effects on fuel loads. The time scale represented by the density of intervals [ILLUSTRATION FOR FIGURE 11 OMITTED] and by the power spectrum of charcoal data [ILLUSTRATION FOR FIGURE 10 OMITTED] is consistent with the life histories of the dominant species (Rowe and Scotter 1973). The increasing probability of fire with time since the last fire suggests that the regrowing vegetation is partly responsible for this fire regime. This internal (biotic) control is implied both by the fact that the mode of the distribution of fire intervals is displaced from zero [ILLUSTRATION FOR FIGURE 11 OMITTED] and by peaks in the power spectra at 60 and 100 yr [ILLUSTRATION FOR FIGURE 10 OMITTED]. The correspondence between scales of fire frequency and jack pine life history supports the notion that disturbance can be an important control over species composition.
Fire regimes changed at each of the Holocene vegetation transitions at Devil's Bathtub, New York. Together with pollen data, fossil charcoal data show times when the fire regime appears to have changed as a consequence of vegetation change. One example is a transition to crown fire with the expansion of Pinus banksiana, which appears to have invaded in the absence of any change in fire frequency. On the basis of consistent fire intervals across the transition and increased background charcoal, it is likely that increased temperatures resulted in the expansion of Pinus banksiana, which, in turn, modified fire regimes in a way that permitted long-term maintenance of populations. The abrupt decline in charcoal with Fagus expansion at 8100 yr BP is further evidence for the fire regime responding to a change in fuels.
Because of the uncertainty over the timing of Pinus strobus expansion, we entertain two alternative explanations of the role of fire. If the Pinus strobus pollen curve represents population expansion at the spatial scale representative of charcoal data, then it is likely that factors other than fire directly controlled its expansion; Pinus strobus invaded the fire regime of Pinus banksiana. Rising moisture levels that may explain the increase of Abies also may have facilitated the expansion of Pinus strobus. The attendant decrease in fire importance may have resulted from the change in crown and ladder fuels as Pinus strobus and hardwoods replaced Pinus banksiana and Picea. Alternatively, increasing moisture availability may have reduced the importance of fire (for any of several reasons), which had previously retarded expansion of Pinus strobus. A combination of the two alternatives is also plausible and cannot be excluded on the basis of our data.
Fire regimes of the Late-Glacial support the notion that scales of disturbances must match those of species' life history if apparently disturbance-dependent species are to persist. Despite potentially different climate and stand structure in the Late-Glacial, the fire intervals of that time are consistent with modern ones in which Picea and Pinus banksiana occur. Fire does not appear to explain the existence of mixed hardwood forests during the Holocene at our site, although wind or ice storms may.
Our results indicate several important areas for future research. Better emissions estimates from different fire types in different modern forests are critical for interpretation of fossil charcoal data. Clear examples here include the Pinus strobus zone of the early Holocene and mixed hardwoods. Lack of fire during the bulk of the Holocene indicates that further study is needed to understand the role of disturbance in eastern deciduous forests. A larger number of sites, preferably from a range of soil and climate types, is needed to obtain a perspective on the importance of fire in the presettlement period.
For discussion and/or reviews of the manuscript, we thank Marc Abrams, Dave Foster, Steve Jackson, and Norton Miller. Catherine Royall helped core Devil's Bathtub. This research was supported by NSF grant BSR-9107272.
Abrams, M. D. 1990. Adaptations and responses to drought in Quercus species of North America. Tree Physiology 7: 227-238.
-----. 1992. Fire and the development of oak forests. BioScience 42:346-353.
Abrams, M. D., and M. L. Scott. 1989. Disturbance-mediated accelerated succession in two Michigan forest types. Forest Science 35:42-49.
Allison, T. D., R. E. Moeller, and M. B. Davis. 1986. Pollen in laminated sediments provides evidence for a mid-Holocene forest pathogen outbreak. Ecology 67:1101-1105.
Amundson, D. C., and H. E. Wright. 1979. Forest changes in Minnesota at the end of the Pleistocene. Ecological Monographs 49:1-16.
Anderson, P. M., P. J. Bartlein, L. B. Brubaker, K. Gajewski, and J. Ritchie. 1991. Vegetation-pollen-climate relationships for the Arcto-Boreal region of North America and Greenland. Journal of Biogeography 18:565-582.
Andreae, M. O. 1983. Soot carbon and excess fine potassium: long range transport of combustion-derived aerosol. Science 220:1148-1151.
Bergeron, Y., and J. Brisson. 1990. Fire regime in red pine stands at the northern limit of the species' range. Ecology 71:1352-1364.
Bradshaw, R. H. W., and O. Zackrisson. 1990. A two thousand-year history of a northern Swedish boreal forest stand. Journal of Vegetation Science 1:519-528.
Clark, J. S. 1988a. Charcoal-stratigraphic analysis on petrographic thin sections: recent fire history in northwest Minnesota. Quaternary Research 30:67-80.
-----. 1988b. Particle motion and the theory of charcoal analysis: source area, transport, deposition, and sampling. Quaternary Research 30:81-91.
-----. 1989. Ecological disturbance as a renewal process: theory and application to fire history. Oikos 56: 17-30.
-----. 1990. Fire and climate change during the last 750 years in northwestern Minnesota. Ecological Monographs 60:135-159.
-----. 1991. Disturbance and tree life history on the shifting mosaic landscape. Ecology 72:1102-1118.
-----. 1993. Paleoecological perspectives on modeling broad-scale responses to global change. Pages 315-332 in P. Kareiva, J. G. Kingsolver, and R. B. Huey, editors. Biotic interactions and global change. Sinauer, Sunderland, Massachussetts, USA.
-----. 1995. Climate and Indian effects on southern Ontario forests. The Holocene 5:371-379.
Clark, J. S., T. C. Hussey, and P. D. Royall. 1996. Presettlement analogs for Quaternary fire regimes in eastern North America. Journal of Paleolimnology 16:79-96.
Clark, J. S., J. Merkt, and H. Muller. 1989. Post-Glacial fire, vegetation, and human history of the northern Alpine forelands, southwestern Germany. Journal of Ecology 77:897-925.
Clark, J. S., and P. D. Royall. 1994. Pre-industrial particulate emissions and carbon sequestration from biomass burning in North America. Biogeochemistry 23:1-17.
Clark, J. S., and P. D. Royall. 1995a. Particle-size evidence for source areas of charcoal accumulation in Late Holocene sediments of eastern North American lakes. Quaternary Research 43:80-89.
Clark, J. S., and P. D. Royall. 1995b. Transformation of a northern hardwood forest by aboriginal (Iroquois) fire: charcoal evidence from Crawford Lake, Ontario, Canada. The Holocene 5: 1-9.
Clark, J. S., and P. D. Royall. 1996. Local and regional sediment charcoal evidence for fire regimes in presettlement northeastern North America. Journal of Ecology, in press.
Cofer, W. R., J. S. Levine, E. L. Winstead, and B. J. Stocks. 1990. Gaseous emissions from Canadian boreal forest fires. Atmospheric Environment Part A-General Topics 24:1653-1659.
Crow, T. R. 1988. Reproductive mode and mechanisms for self-replacement of northern red oak (Quercus rubra) - a review. Forest Science 34:19-40.
Cwynar, L. C. 1987. Fire and the forest history of the North Cascade Range. Ecology 68:791-802.
Davis, M. B. 1981a. Quaternary history and the stability of forest communities. Pages 132-153 in D. C. West, H. H. Shugart, and D. B. Botkin, editors. Forest succession: concepts and application. Springer-Verlag, New York, New York, USA.
-----. 1981b. Outbreaks of forest pathogens in Quaternary history. Proceedings of the Fourth International Palynological Conference, Luchnow, India 3:216-227.
----. 1983. Holocene vegetational history of the eastern United States. Pages 166-181 in J. H. E. Wright, editor. Late quaternary environments of the United States. University of Minnesota Press, Minneapolis, Minnesota, USA.
-----. 1989. Lags in vegetation response to greenhouse warming. Climatic Change 15:75-82.
Davis, M. B., R. W. Spear, and L. C. K. Shane. 1980. Holocene climate of New England. Quaternary Research 14:240-250.
Davis, R. B., and G. L. Jacobson. 1985. Late Glacial and early Holocene landscapes in northern New England and adjacent areas of Canada. Quaternary Research 23:341-368.
Day, G. M. 1953. The Indian as an ecological factor in the northeastern forest. Ecology 34:329-346.
Dean, W. E., J. P. Bradbury, R. Y. Anderson, and C. W. Barnosky. 1984. The variability of Holocene climate change: evidence from varved lake sediments. Science 226:1191-1194.
Faegri, K., and J. Iversen. 1975. Textbook of pollen analysis. Blackwell Scientific, Oxford, England.
Flannigan, M. D., and J. B. Harrington. 1988. A study of the relation of meteorological variables to monthly provincial area burned by wildfire in Canada (1953-1980). Journal of Applied Meteorology 27:441-452.
Foster, D. R. 1983. The history and pattern of fire in the boreal forest of southeastern Labrador. Canadian Journal of Botany 61:2459-2471.
-----. 1985. Vegetation development following fire in Picea mariana (black spruce) - Pleurozium forests of southeastern Labrador, Canada. Journal of Ecology 73:517-534.
Foster, D. R., and G. A. King. 1986. Vegetation pattern and diversity in S.E. Labrador, Canada: Betula papyrifera (birch) forest development in relation to fire history and physiography. Journal of Ecology 74:465-483.
Gajewski, K. 1987. Climatic changes on the vegetation of eastern North America during the past 2000 years. Vegetatio 68:179-190.
-----. 1988. Late Holocene climate changes in eastern North America estimated from pollen data. Quaternary Research 29:255-262.
Garstang, M., P. D. Tyson, H. Cachier, and L. Radke. 1996. Atmospheric transports of particulate and gaseous products by fires. In J. S. Clark, H. Cachier, B. Stocks, and J. G. Goldammer, editors. Sediment records of biomass burning and global change. Springer Verlag, New York, New York, USA, in press.
Gordon, R. B. 1940. The primeval forest types of southwestern New York. New York State Bulletin 321:1-97.
Graham, R. L., M. G. Turner, and V. H. Dale. 1990. How increasing C[O.sup.2] and climate change affect forests. Bioscience 40:575-587.
Grimm, E. C. 1983. Chronology and dynamics of vegetation change in the prairie-woodland region of southern Minnesota, USA. New Phytologist 93:311-350.
Heinselman, M. L. 1973. Fire in the virgin forest of the Boundary Waters Canoe Area, Minnesota. Quaternary Research 3:329-382.
Huston, M. 1979. A general hypothesis of species diversity. American Naturalist 113:81-101.
Jackson, S. T. 1989. Postglacial vegetational changes along an elevational gradient in the Adirondack Mountains (New York). Biological Survey Museum Bulletin 465.
-----. 1990. Pollen source area and representation in small lakes of the northeastern United States. Review of Palaeobotany and Palynology 63:53-76.
Jackson, S. T., and D. R. Whitehead. 1991. Holocene vegetation patterns in the Adirondack Mountains. Ecology 72: 641-653.
Johnson, E. A. 1992. Fire and vegetation dynamics. Cambridge University Press, Cambridge, England.
Kutzbach, J. E., and P. J. Guetter. 1986. The influence of changing orbital parameters and surface boundary conditions on climate simulations for the past 18 000 years. Journal of Atmospheric Sciences 43:1726-1759.
Lutz, H. J. 1930. Original forest composition in northwestern Pennsylvania as indicated by early land survey notes. Journal of Forestry 28:27-38.
Lynham, T. J., and B. J. Stocks. 1991. The natural fire regime of an unprotected section of the boreal forest in Canada. Pages 99-108 in Proceedings of the 17th Tall Timbers Fire Ecology Conference. Tall Timbers Research Station, Tallahassee, Florida.
Merkt, J. 1971. Zuverlassige Auszahlungen yon Jahresschichten in Seesedimenten mit Hilfe von Gross-Dunnschliffen. Archiv fur Hydrobiologie 69:145-154.
Muller, J. 1984. Atmospheric residence time of carbonaceous particles and particulate PAH-compounds. The Science of the Total Environment 36:339-346.
Overpeck, J. T., I. C. Prentice, and I. T. Webb. 1985. Quantitative interpretation of fossil pollen spectra: dissimilarity coefficients and the method of modern analogs. Quaternary Research 23:87-108.
Overpeck, J. T., D. Rind, and R. Goldberg. 1990. Climate-induced changes in forest disturbance and vegetation. Nature 343:51-53.
Patterson, W. A., K. J. Edwards, and D. J. Maguire. 1987. Microscopic charcoal as a fossil indicator of fire. Quaternary Science Reviews 6:3-23.
Payette, S., C. Morneau, L. Sirois, and M. Desponts. 1989. Recent fire history of the northern Quebec biomes. Ecology 70:656-673.
Prentice, I. C. 1985. Pollen representation, source area, and basin size: toward a unified theory of pollen analysis. Quaternary Research 23:76-86.
Prentice, I. C., P. J. Bartlein, and T. Webb III. 1991. Vegetation and climate change in eastern North America since the last glacial maximum. Ecology 72:2038-2056.
Radke, L. F., D. A. Hegg, P. V. Hobbs, J. D. Nance, J. H. Lyons, K. K. Laursen, R. E. Weiss, P. J. Riggan, and D. E. Ward. 1991. Particulate and trace gas emissions from large biomass fires in North America. Pages 209-224 in J. S. Levine, editor. Global biomass burning: atmospheric, climatic, and biospheric implications. Massachusetts Institute of Technology Press, Cambridge, Massachusetts, USA.
Rizzo, B., and E. Wiken. 1992. Assessing the sensitivity of Canada's ecosystems to climatic change. Climatic Change 21:37-55.
Rowe, J. S., and G. W. Scotter. 1973. Fire in boreal forest. Quaternary Research 3:444-464.
Sandberg, D. V., J. M. Pierovich, D. G. Fox, and E. W. Ross. 1979. Effects of fire on air. United States Department of Agriculture Forest Service General Technical Report WO-9, Washington, DC.
Seischab, F. K. 1990. Presettlement forests of the Phelps and Gorham Purchase in western New York. Bulletin of the Torrey Botanical Club 117:27-38.
Sousa, W. P. 1984. The role of disturbance in natural communities. Annual Review of Ecology and Systematics 15:353-391.
Spear, R. W., M. B. Davis, and L. C. K. Shane. 1994. Late quaternary history of low- and mid-elevation vegetation in the White Mountains, New Hampshire. Ecological Monographs 64:85-109.
Spear, R. W., and N. G. Miller. 1976. A radiocarbon-dated pollen diagram form the Allegheny Plateau of New York State. Journal of the Arnold Arboretum 57:369-403.
Stocks, B. J. 1989. Fire behavior in mature jack pine. Canadian Journal of Forest Research 19:783-790.
Stocks, B. J., and R. B. Street. 1983. Forest fire weather and wildfire occurrence in the boreal forest of northwestern Ontario. Pages 249-265 in R.W. Wein, R. R. Riewe, and I. R. Methven, editors. Resources and dynamics of the Boreal Zone. Association of Canadian Universities Northern Studies, Ottawa, Ontario, Canada.
Swain, A. M. 1973. A history of fire and vegetation in northeastern Minnesota as recorded in lake sediment. Quaternary Research 3:383-396.
Teller, J. T. 1987. Postglacial lakes and the southern margin of the Laurentide Ice Sheet. Pages 39-70 in W. F. Ruddiman and J. H. E. Wright, editors. North America and adjacent oceans during the last deglaciation. The geology of North America. Geological Society of America, Boulder, Colorado, USA.
Timoney, K. P., and R. W. Wein. 1991. The areal pattern of burned tree vegetation in the subarctic region of northwestern Canada. Arctic 44:223-230.
Tolonen, K. 1983. The postglacial fire record. Pages 21-44 in R. W. Wein and D. A. Maclean, editors. The role of fire in northern circumpolar ecosystems. Wiley, Chichester, UK.
Tolonen, M. 1987. Primary and secondary forest successions reflected in a late Holocene pollen sequence from an upland mire near the southwest coast of Finland. Annales Botanici Fennici 24:97-109.
Tsukada, M., and S. Sugita. 1982. Late Quaternary dynamics of pollen influx at Mineral Lake, Washington. Botanical Magazine (Tokyo) 95:401-418.
Walker, B. H. 1991. Ecological consequences of atmospheric and climate change. Climatic Change 19:301-316.
Webb, T. III. 1982. Temporal resolution in Holocene pollen data. Proceedings of the Third North American Paleotological Convention 2:569-572.
-----. 1988. Eastern North America. Pages 385-414 in B. Huntley and T. Webb III, editors. Vegetation history. Kluwet Academic, Dordrecht, The Netherlands.
White, R. 1991. The middle ground: Indians, Empires, and Republics in the Great Lakes Region, 1650-1815. Cambridge University Press. Cambridge, England.
Whitehead, D. R. 1979. Late-Glacial and postglacial vegetational history of the Berkshires, western Massachusetts. Quaternary Research 12:333-357.
Whitehead, D. R., and S. T. Jackson. 1990. The regional vegetational history of the High Peaks (Adirondacks Mountains), New York. New York State Museum Bulletin Number 478, Albany, New York, USA.
Wotton, B. M., and M. D. Flannigan. 1993. Length of the fire season in a changing climate. Forestry Chronicle 69:187-192.
Wright, H. E., D. H. Mann, and P. H. Glaser. 1984. Piston corers for peat and lake sediments. Ecology 65:657-659.
|Printer friendly Cite/link Email Feedback|
|Author:||Clark, James S.; Royall, P. Daniel; Chumbley, Craig|
|Date:||Oct 1, 1996|
|Previous Article:||Tundra plant uptake of amino acid and NH4+ nitrogen in situ: plants compete well for amino acid N1.|
|Next Article:||Channel narrowing and vegetation development following a great plains flood.|