An 840-year record of fire and vegetation in a boreal white spruce forest.
Fire and climate are key processes that control boreal forest vegetation dynamics (Shugart et al. 1992). Fire frequency in the boreal forest has been studied using historical and dendrochronological records (Johnson 1992). Postfire vegetation dynamics have been studied using chronosequence methods (e.g., Dix and Swan 1971). Results of chronosequence studies have been criticized because they combine sites with different microenvironments and different ecological histories (Pickett 1989). Indeed, dendrochronological reconstructions of nonboreal tree population dynamics in western North American forests have found that individual sites do not undergo the vegetation dynamic that was suggested by chronosequence studies (Johnson et al. 1994, Fastie 1995). Dendrochronological records have not, however, been able to shed light on whether boreal forest communities have changed in response to past climate changes.
As an alternative, it has been suggested that fine-resolution fossil pollen, charcoal, and sedimentation records from lake sediments can provide long-term, site-specific records of (1) fires from which the mean fire interval can be estimated; (2) short-term ([less than]100 yr) postfire vegetation dynamics; and (3) climate-induced, centennial-scale changes in forest composition (e.g., Swain 1978, Green 1981, Clark et al. 1989, Campbell and McAndrews 1993, Rhodes and Davis 1995). In this study, we use fine-resolution fossil pollen, charcoal, and sediment analysis to reconstruct an 840-yr record of fire and vegetation dynamics in a boreal white spruce forest. The lake contains annually laminated sediments that allow accurate chronological control of the charcoal, pollen, and sediment records. A comparison of the pollen and charcoal record from the last 200 yr with a dendrochronological record of local fires indicates that the pollen and charcoal record will provide evidence of most large, local fires (MacDonald et al. 1991). This conclusion is reinforced by recent modeling of vegetation disturbance and pollen deposition at the site (Sugita et al. 1997).
Previous studies of boreal white spruce forests provide results against which our findings will be compared, and against which the validity of the fossil pollen approach will be assessed. Dendrochronological records from boreal white spruce forests have found fire cycles of 96 yr (71-142 yr, 95% CI) in northern Alberta (Larsen 1997) and 113 yr in central Alaska (Yarie 1981). A chronosequence study of boreal white spruce forests in Alaska found that, after a fire, there was sequential dominance by herbs, shrubs, deciduous trees, and then white spruce (Foote 1983). Chronosequence studies of trees in Ontario and Quebec also found that white spruce dominance tends to follow that of aspen and jack pine (Cogbill 1985, Bergeron and Dubuc 1989). A study of postfire understory vegetation dynamics in Quebec, however, found no evidence for change in abundance of herb or shrub species. (DeGrandpre et al. 1993).
In this paper, we use the pollen, charcoal, and sediment records to identify past fires and then to examine three questions regarding forest dynamics of boreal white spruce. First, will the pollen stratigraphy following different fires exhibit a predictable sequence of peaks from different pollen taxa? Second, what is the mean fire interval and how does this compare to published dendrochronological estimates? Third, have there been long-term changes in the relative abundance of different pollen taxa that suggest changes in relative abundance of different species around the site?
Lake basin characteristics
Rainbow Lake A (RLA) is located in Wood Buffalo National Park (WBNP), Alberta, Canada [ILLUSTRATION FOR FIGURE 1 OMITTED]. The lake has a radius of [approximately]93 m, a surface area of 2.7 ha, and a maximum depth of 10 m. RLA is connected by a small channel, [approximately]2 m deep, 10 m wide, and 110 m long, to an adjacent lake that has a surface area of 4.6 ha and a maximum depth of 14 m [ILLUSTRATION FOR FIGURE 2 OMITTED]. The drainage basin of the two lakes together is [approximately]38 ha. There are no significant inflow or outflow streams associated with the lakes. The morphometry of RLA is typical of lakes that contain annually laminated sediments (Larsen and MacDonald 1993).
The Rainbow Lakes are solutional sinkholes formed in Devonian gypsum (Airphoto Analysis Associates 1979). The bedrock is overlain by well-drained lacustrine sands that overlie a calcareous, Late Pleistocene glacial till (Airphoto Analysis Associates 1979). The soils are degraded eutric brunisols (Airphoto Analysis Associates 1979). The maximum topographic relief around the lakes is [approximately]30 m.
Climate normals (AD 1950-1980) are available from Fort Smith, Northwest Territories, 25 km to the northeast of RLA (Environment Canada 1982). Precipitation occurs throughout the year, but peaks in July; mean annual precipitation is 35 cm. Mean January and July temperatures are -26.8 [degrees] C and 16 [degrees] C, respectively. The extreme minimum and maximum temperatures recorded in Fort Smith between 1950 and 1980, are -53.9 [degrees] C and 35.0 [degrees] C, respectively. The average frost-free period is 72 d, extending between 8 June and 20 August.
The current vegetation around RLA was established after a series of six fires that occurred between AD 1844 and 1943 (MacDonald et al. 1991), with the last extensive fire appearing to have burned in 1862. Picea glauca dominates the forest (botanical nomenclature follows Budd 1979). Populus balsamifera and especially Populus tremuloides, often decadent, are common on ridge tops and south-facing slopes. Alnus incana is frequent along the lakeshore. Betula papyrifera is common on north-facing slopes. Scattered individuals of Picea mariana and Larix laricina are found on moist sites near the shore, whereas scattered pockets of Pinus banksiana occur on well-drained interfluves.
Shrubs of Amelanchier alnifolia, Cornus stolonifera, Rosa acicularis, Salix spp., Shepherdia canadensis, Vaccinium uliginosum, and Viburnum edule are found in the understory, sometimes as decadent stems, but are most common in forest gaps. The ground cover is dominated by a mat of Pleurozium schreberi. Herbs and low shrubs such as Cornus canadensis and Ledum groenlandicum are common in the ground layer, but Equisetum spp. and Gramineae are more common in open-canopy sites where moss cover is poor. In general, forbs, grasses, and shrubs are low in abundance. Emergents such as Typha latifolia, Nymphae tetragonea, and Myriophyllum exalbescens are found within RLA's littoral zone.
In close proximity to RLA are mixed and pure patches of Picea glauca, Pinus banksiana, and Populus tremuloides, along with several small patches of wetland [ILLUSTRATION FOR FIGURE 2 OMITTED]. Extensive stands of Pinus banksiana begin 5 km to the west of RLA. Picea mariana and Sphagnum spp. muskeg that occupy poorly drained glacio-lacustrine deposits dominate the vegetation [approximately]10-20 km in all directions from RLA.
Fire is the main form of vegetation disturbance in the western boreal forest (Johnson 1992). Between 1950 and 1985, [approximately]18 680 [km.sup.2] of the 44 800 [km.sup.2] of Wood Buffalo National Park (WBNP) was burned by forest fires (Larsen and MacDonald 1995). Individual fires varied in size, from lightning strikes that burned individual trees to a maximum observed fire of 1813 [km.sup.2] (WBNP file). Fires of [greater than or equal to]1 [km.sup.2] comprised 14.9% of the fires and were responsible for 99.81% of the area burned. Fires of [less than or equal to]0.01 [km.sup.2] accounted for 58.9% of the fire events and were responsible for 0.01% of the area burned. The mean fire interval in white spruce forests in WBNP is 96 yr (71-142 yr, 95% CI) (Larsen 1997).
Pollen catchment area
Empirical and theoretical analyses of pollen source areas suggest that most of the pollen deposited in the center of a small lake, such as RLA, should come from within a several-kilometer radius of the lake (Prentice 1985, Sugita 1993). based on the Prentice-Sugita models and assuming a homogenous forest, [approximately]60% of the relatively large and heavy Picea cf. glauca pollen should come from an area within a 500-m radius of the lake, and [approximately]50% of the relatively light pollen from Pinus and Betula should come from within a 3-km radius.
Local disturbances should have a much greater influence on the pollen record than more distant disturbances. Sugita et al. (1997) used computer simulations to evaluate the influence that fires with different sizes and proximities would have on the Picea cf. glauca pollen record in RLA. It was found that a 4-ha fire would cause a 12% reduction in the input of P. cf. glauca pollen if it reached the lakeshore, and a 2% reduction if it reached 100 m from the lakeshore. A 100-ha fire would cause a 38% reduction if it reached the lakeshore, and 11% and 3% reductions if it reached 100 m and 500 m, respectively, from the lakeshore. A 2500-ha fire would cause a 60% reduction if was at the lakeshore, and 11% and 5% reductions if it reached 100 m and 500 m, respectively, from the lakeshore.
The pollen record from RLA will, therefore, contain a background of pollen from distant regional sources, as well as a signal from the local vegetation. Because RLA is small, it is likely that the signal from large changes in the local vegetation, such as the destruction of the watershed forest by burning, should be strong enough to be apparent above the background of regional pollen. Thus, we assume that fires that destroy the majority of forest in the watershed of RLA will be represented in the pollen record. We recognize that this record will be biased towards fires of [greater than or equal to]1 [km.sup.2], but such fires account for most of the area burned in northern Alberta; hence, this biased record should be adequate for inferring the mean fire interval. We note that the local vegetation consists of stands with different time intervals since the last fire, and with different microenvironments that favor different species. Because the pollen record from RLA integrates information from these different forest stands, it will contain a mixed signal. It may, therefore, be difficult to determine whether the pollen signal is due to postdisturbance vegetation dynamics in one patch, or to varied postdisturbance dynamics in different patches.
FIELD AND LABORATORY METHODS
Rainbow Lake A was cored on 24 April 1986, using a modified Livingstone piston sampler (Wright et al. 1984). The upper 100 cm of sediments, along with [approximately]5 cm of the water column, were collected using a clear plastic tube. The water was drained and the core was transported to the laboratory in a vertical position. As a result of draining, the core compacted to 77 cm. A second core, spanning from 80 to 180 cm below the sediment-water interface, was collected adjacent to the first core.
The plastic tube core was extruded by driving a piston upwards using a scissor jack. A 40x Nikon stereomicroscope was fixed in a horizontal position next to the core to identify and count sediment laminae. The laminae were composed of couplets of light-colored layers dominated by calcite crystals, and dark-colored layers composed primarily of fine organics that contained abundant diatom frustules (as observed under a light microscope at 400x). This pattern conforms to the biogenic annual laminations discussed in O'Sullivan (1983).
Disks containing five couplets of light and dark laminae, representing five years of sedimentation, were sliced from the core using a thin, rigid piece of acetate. The thickness of each 5-yr sample was measured prior to slicing. A total of 129 disks, containing a total of 645 laminae couplets, were obtained from the upper core. The second core was placed flat on a bench beneath a stereomicroscope and was sampled in five-couplet disks sliced vertically from the core. We sampled a total of 207 sample disks, containing 1035 couplets of laminae, from the top 80 cm of the second core.
The stratigraphies from the two overlapping core sections were correlated on the basis of measured depths and slotting of loss on ignition (LOI) and calcium concentration (in milligrams per kilogram) stratigraphies. LOI is indicative of the percentage of organic matter in a sample (Dean 1974). The concentration of calcium was determined using short-lived, short-decay, thermal instrumental neutron activation analysis (for details, see MacDonald et al. 1991). The top and bottom cores exhibited statistically significant slotting of the LOI (r = 0.56, n = 29, P [less than] 0.001) and calcium concentration records (r = 0.86, n = 29, P [less than] 0.001) [ILLUSTRATION FOR FIGURE 3 OMITTED].
The chronology created by the stratigraphic slotting of the cores was tested by radiocarbon dating a small piece of wood found [approximately]90 cm below the sediment-water interface in the second core. The slotted stratigraphies suggest that this sample should have been deposited between AD 1640 and 1645. A radiocarbon age of 460 [+ or -] 71-yr BP (mean [+ or -] 2 SD) was obtained for this fossil (AMS Date TO-768). The corrected radiocarbon date corresponds to the calendar years between AD 1409 and 1484 (Stuiver and Reimer 1993). The discrepancy between the radiocarbon date and the date suggested by the slotted stratigraphies may be due to erosion and redeposition of the wood, or to imprecise measurement of the depth at which the second core was taken. Therefore the laminations are believed to be annual. The most recent 840 couplets of sediment were used for further analysis.
Pollen and charcoal counting
A sediment subsample 1 [cm.sup.2] in surface area was removed from each of the 5-yr sediment disks and was prepared for pollen and charcoal analysis following standard methods (Faegri et al. 1989). A known quantity of pre-acetolyzed Lycopodium spores was added to the samples prior to processing, to allow the calculation of pollen accumulation rates (PARs) and charcoal accumulation rates (CHARs) (Stockmarr 1971). Between 259 and 354 pollen grains from terrestrial plants were counted from each sample. Identifiable spores and the pollen from nonterrestrial plants were also counted. Palynomorphs were identified using a reference collection and an illustrated key (McAndrews et al. 1973). Picea cf. glauca and Picea cf. mariana pollen were differentiated using the criteria outlined by Hansen and Engstrom (1985).
Microscopic pieces of charcoal on the pollen slides were measured using an ocular grid (MacDonald et al. 1991). The charcoal pieces were placed into the following size classes: 75-374 [Mu][m.sup.2], 375-679 [Mu][m.sup.2], 680-1434 [Mu][m.sup.2], 1435-2199 [Mu][m.sup.2], and [greater than]2200 [Mu][m.sup.2]. CHAR, expressed in square micrometers per square centimeter per year, was calculated by multiplying the midpoint of each size class with the number of fragments observed in each size class, summing these areas together, multiplying by the ratio of added to counted exotic spores, and then dividing by five, because five years of sample are contained in each sample. To provide relatively stable estimates of charcoal abundance, an area of 21.6 [mm.sup.2]-253.9 [mm.sup.2] (4.5-52%) of a microscope cover slip was scanned, and 10-64 marker spores were counted (MacDonald et al. 1991). Macroscopic charcoal could not be measured, because there was insufficient sediment remaining from many of the 5-yr samples.
Postfire vegetation dynamics
To test the hypothesis that there was a statistically significant, postfire sequence of peaks in pollen taxa in the forest around RLA, cross-correlation analysis (Gottman 1981) was performed on the raw (nonsmoothed) pollen accumulation rates (PARs), microscopic charcoal accumulation rates (CHARs), and sedimentation rates (cf. Green 1981, Clark et al. 1989). To reduce the possibility of chance correlations, the analyses were restricted to taxa present in a minimum of 10% of the samples. To meet the assumption of a stationary time series (Gottman 1981), cross-correlation analysis was performed on the residuals of each pollen record after detrending it with the highest order polynomial in which all parameters were significant at the P [less than] 0.05 level.
Cross-correlation analysis was conducted by lagging each pollen taxon against CHAR, Picea cf. glauca, and Gramineae. It is recognized that some peaks in CHAR may reflect large regional fires that did not reach RLA (Patterson et al. 1987, MacDonald et. al. 1991). Therefore, cross-correlation with pollen records was undertaken. Picea cf. glauca represents mature forest, whereas the family Gramineae is indicative of early postfire vegetation.
Postfire vegetation dynamics were also assessed by visually examining the 840-yr PAR records of all pollen taxa that exhibited significant cross-correlations with CHAR. This method involved first identifying local fires and then examining the pollen stratigraphy between consecutive fires. It was expected that large, local fires would be visible as a peak in CHAR, followed by a peak in sediment thickness due to increased erosion, and a disruption of the postfire pollen sequence suggested by cross-correlation analysis. It is not possible to specify which taxa will indicate a fire, because we are testing to determine if the postfire vegetation sequence is the same after all fires.
The 840-yr pollen records were smoothed with a three-term, centered, moving average to remove 5-15-yr fluctuations in the pollen records caused by factors such as interannual variations in pollen production, dispersal and deposition processes, and analytic errors in sediment and pollen sampling, preparation, and counting (Birks and Birks 1980), which might visually obscure the longer term postfire vegetation dynamics. Data smoothing has been used for similar reasons in other fossil pollen and charcoal studies (Ritchie 1985, Clark and Royall 1994).
The 840-yr record of sediment thickness exhibited a sigmoid form, with an approximately fivefold increase in sediment thickness toward the top of the core [ILLUSTRATION FOR FIGURE 4 OMITTED]. Sediment density exhibited an inverse trend (not shown), suggesting that the trend in sediment thickness reflected compaction of deeper sediments. The trend was removed with a second-order polynomial ([r.sup.2] = 0.82, n = 168, P [less than] 0.001). The residuals from the regression were presumed to reflect short-term variations in sedimentation due to postfire erosion. The residuals were smoothed with a three-term, centered, moving average.
Mean fire interval estimation
The mean fire interval was assessed using graphical methods and time series analyses. The graphical method involved inferring the occurrence of large, local fires in the charcoal, pollen, and sediment records using the method previously explained. The mean fire interval was estimated as the mean number of years between each of the inferred large local fires.
The time series method involved assessing the autocorrelation structure of the PAR records that have significant cross-correlations with CHAR, Picea cf. glauca, and Gramineae. Significant autocorrelations would indicate a periodic pattern of peaks and declines (Gottman 1981). It has been suggested that such periodicities in the pollen and charcoal records indicate the mean fire interval (Green 1981). To meet the assumption of a stationary timeseries, autocorrelation analysis was performed on the residuals of the 840-yr record of each pollen taxa, following detrending with a cubic polynomial.
The mean fire interval estimates provided by these two methods were compared with the fire cycle estimates in dendrochronological fire history studies of boreal white spruce forests. Note that the mean fire interval and the fire cycle are typically equivalent in boreal forests where burning appears to be largely independent of forest age (Johnson and Gutsell 1994).
Postfire vegetation dynamics
Twenty-one of the 34 identified terrestrial pollen and spore taxa were present in [greater than]10% of the samples [ILLUSTRATION FOR FIGURE 5 OMITTED], and were tested for significant cross-correlations against CHAR, Picea cf. glauca, and Gramineae. The CHAR, Picea cf. glauca, and Gramineae records exhibited significant cross-correlations with each other, 12 other pollen and spore taxa, and sediment thickness. Given the similarities, only the results obtained using CHAR are presented and discussed.
Significant positive correlations with Picea cf. glauca and Picea cf. mariana occur for several decades prior to the peak in CHAR at year zero [ILLUSTRATION FOR FIGURE 6 OMITTED]. That is, these taxa are greatest in abundance for several decades prior to the peak in CHAR. Significant positive correlations with the ground layer taxa of Equisetum, Gramineae, Galium, and Thalictrum occur for several decades after a peak in CHAR. This indicates increases in these taxa immediately following a peak in CHAR. Sediment thickness increases following a peak in CHAR, and peaks 35-yr after CHAR. These peaks are followed by sequential peaks in the shrubs Salix and Shepherdia canadensis, the deciduous trees Populus, Alnus, and Betula, and then the coniferous trees Larix, Pinus, Picea cf. mariana, Picea cf. glauca, and the moss Sphagnum. The cross-correlations are generally low (r [less than] 0.4), suggesting that the sequential peaks in pollen we have described are only weakly expressed in some postfire sequences.
The 840-yr PAR record of taxa with significant cross-correlations with CHAR exhibits components of the general pattern of pollen peaks previously outlined [ILLUSTRATION FOR FIGURE 7 OMITTED]. For example, marked declines in Picea cf. glauca, Picea cf. mariana, and Larix at AD 1185 are followed by sequential peaks in CHAR, Gramineae, Galium, detrended sediment thickness, Thalictrum, Salix, Shepherdia canadensis, Populus, Alnus, Betula, Larix, Pinus, Sphagnum, Picea cf. mariana, and Picea cf. glauca. In other cases, though, peaks in pollen do not conform to the complete sequence suggested by the cross-correlation analyses. For example, declines at AD148 in Picea cf. glauca, Picea cf. mariana, Alnus, and Pinus are followed by small peaks in CHAR and detrended sediment thickness, declines in Equisetum, Gramineae, Gallum, and Shepherdia canadensis, and increases in Larix, Salix, Alnus, Betula, Pinus, Sphagnum, Picea cf. mariana, and Picea cf. glauca.
Peaks in CHAR do not always coincide with pollen indicators of fire. For example, the peak in CHAR in AD 1525 has a concurrent decrease in Larix, but no other indicators of fire. In AD 1650, there is no peak in CHAR, but a fire is suggested by declines in all tree pollen, followed by sequential peaks in sediment thickness, Equisetum, Gramineae, and Thalictrum.
Mean fire interval
We infer a total of 12 large local fires from the 840 yr record of pollen, CHAR, and sediment thickness [ILLUSTRATION FOR FIGURE 7 OMITTED]. Eleven of the inferred fires are marked by declines in Picea cf. glauca, followed by a peaks in sediment thickness, whereas the 12th fire is marked by a decline in Picea cf. glauca and a peak in CHAR [ILLUSTRATION FOR FIGURE 7 OMITTED]. Ten fires coincide with declines in Picea cf. mariana, another 10 with declines in Pinus, six of the fires coincide with declines in Larix, and another six with declines in Alnus. Seven of the fires are followed by peaks in Equisetum and six are followed by peaks in Gramineae. The interval between the 12 fires [ILLUSTRATION FOR FIGURE 6 OMITTED] ranges from 30 to 130 yr, with a mean fire interval of 69 yr.
Significant positive autocorrelation coefficients were present in the PAR records of 11 pollen and spore taxa, and the records of detrended sediment thickness and CHAR [ILLUSTRATION FOR FIGURE 8 OMITTED]. Autocorrelations are significant for the first 30 yr for most records, because peaks in abundance for each record typically last for several decades [ILLUSTRATION FOR FIGURE 7 OMITTED]. Significant positive autocorrelations between 45 and 80 yr occur for Galium, Shepherdia canadensis, and Larix. Significant positive autocorrelations between 95 and 185 yr occur for all records [ILLUSTRATION FOR FIGURE 8 OMITTED]. However, the autocorrelations are generally quite low (r [less than] 0.3).
Long-term vegetation dynamics
Long-term changes in the abundance of a number of taxa are apparent the pollen records [ILLUSTRATION FOR FIGURE 7 OMITTED]. Initial high levels of Picea cf. glauca end with a fire in AD 1185 and are followed by a gradual 800 yr increase in abundance back to pre-AD 1185 levels. Populus exhibits high abundance between AD 1250 and 1550, and after 1930. Gramineae exhibits peaks in abundance between AD 1200 and 1250 and between AD 1450 and 1750. Equisetum, Salix, and Shepherdia canadensis exhibit peaks in abundance between AD 1350 and 1750.
Postfire vegetation dynamics
The cross-correlation analyses indicate that there is a predictable postfire sequence of peaks in the relative abundance of herbs, shrubs, deciduous trees, and then white spruce [ILLUSTRATION FOR FIGURE 6 OMITTED]. This sequence is the same as that found in the chronosequence study of an Alaskan white spruce forest (Foote 1983). Although the cross-correlations were significant, they were all less than 0.4, and therefore did not explain all of the short-term variations in the pollen records. The pollen stratigraphy shows that the postfire pollen sequence is not the same after all fires [ILLUSTRATION FOR FIGURE 7 OMITTED], confirming the incomplete explanation of postfire vegetation dynamics provided by the cross-correlation analyses.
Analytic noise associated with pollen analysis (cf. Birks and Birks 1980) may account for some of the variation in the postfire pollen patterns. Further variation is probably the result of many small and/or closely recurring local forest disturbances such as fire, windstorm, or insect damage. In addition, each of the disturbances can vary in terms of characteristics such as size, location, seasonality, intensity, and the availability of propagules following the disturbance. For example, a small 4-ha fire that burned the north-facing slope of RLA, dominated by white spruce and paper birch, would change the pollen signal differently than would a 4-ha fire that burned the east-facing slope, dominated by white spruce and jack pine. Our stratigraphic results, therefore, support the conclusion of other studies (e.g., Johnson et al. 1994, Fastie 1995) that, due to natural variability in disturbances, individual sites will not necessarily undergo the deterministic vegetation dynamic suggested by chronosequence.
Although the pollen stratigraphies show that all pollen taxa do not peak in a predictable postfire sequence, they do show that the peaks tend to follow the sequence of herbs, shrubs, deciduous trees, and then conifers [ILLUSTRATION FOR FIGURE 7 OMITTED]. This sequence largely reflects the growth rates and ages at sexual maturation of the different plant species, most of which became established in the site within a few years following the disturbance (Johnson 1981, Rowe 1983). For example, the delayed peak in Picea cf. glauca pollen probably reflects delayed sexual maturation (MacDonald et al. 1991), because P. glauca does not become sexually mature and produce pollen until it is 45-60 yr old (Nienstaedt and Zasada 1990). The greater delay in the peak in P. cf. glauca pollen following the fires at AD 1185 and 1345 may represent delayed establishment. The sequential declines in herb, shrub, and deciduous tree taxa probably results from shading by the increasingly dominant Picea glauca (Lieffers and Stadt 1994, Okland 1995).
Mean fire interval
The mean fire interval estimate of 69 yr suggested by the inference of large, local fires [ILLUSTRATION FOR FIGURE 7 OMITTED] is shorter than the 96- and 113-yr fire cycle estimates for white spruce forests in northern Alberta (Larsen 1997) and Alaska (Yarie 1981), respectively. Indeed, it is more similar to the 30-70 yr fire cycle (no confidence intervals given) expected in boreal jack pine and aspen forests (Yarie 1981, Carroll and Bliss 1982, Purchase and LaRoi 1983) However, the mean fire interval estimates of 95-185 yr suggested by the dominant periodicity in pollen records [ILLUSTRATION FOR FIGURE 8 OMITTED] broadly overlap the 71-142 yr confidence interval for the 96-yr fire cycle estimate for white spruce forests in northern Alberta (Larsen 1997).
We believe that the 12 fire events we infer were indeed large, local fires, because 11 of them were marked by increases in sediment thickness, presumably due to erosion. Delayed peaks in sediment thickness are evident following the fires at AD 1185 and 1865, but the occurrence of these fires is clearly indicated by pollen and, respectively, CHAR and dendrochronological records. Note, however, that only nine of the 12 fires exhibited declines in the Picea cf. glauca PARs of [greater than]40%, when calculated as the difference between the largest and smallest PAR in the two samples before and after the inferred fire. The three fires with smaller declines are AD 1620 (20%), 1700 (17%), and 1940 (37%). based on computer simulations of fires with different sizes and proximities to RLA, it is expected that a 40% decline could only result from a fire of [greater than or equal to]1 [km.sup.2] that burned to the lakeshore (Sugita et al. 1997). Indeed, the fire at AD 1940 burned at an area of several square kilometers outside the RLA watershed, but [less than] 1 ha within the watershed (Sugita et al. 1997).
If we assume that there were only nine large, local fires, then the mean fire interval becomes 85 yr and is within the 71-142 yr confidence interval for the fire cycle in northern Alberta white spruce forests (Larsen 1997). By restricting our inference of fires to those events that were most likely to be large and local, we have the best chance of obtaining an accurate estimate of the mean fire interval. Because 99.81% of the boreal forest in northern Alberta is burned by fires that are [greater than or equal to]1 [km.sup.2] in size (Larsen and MacDonald 1995), the restriction of our observations to such fires should not seriously influence our mean fire interval estimate. The problem is in determining which fires are [greater than] 1 [km.sup.2]. It is possible that the fires in AD 1620 and 1700 were [greater than] 1 [km.sup.2] and exhibited low declines in Picea cf. glauca PARs, because the fires burned portions of the RLA watershed that were not dominated by white spruce. If a well-spaced network of pollen records within a 1-[km.sup.2] area was available (cf. Davis et al. 1994), it might be possible to determine which fires were large. These large fires could then be used to estimate the mean fire interval. Without such a network of sites, the estimation of the mean fire interval is problematic.
The low autocorrelation values are not surprising, as it is unrealistic to assume that boreal forest fires occur at highly regular intervals (Johnson and Gutsell 1994), and because pollen taxa exhibit different-sized peaks following different fires [ILLUSTRATION FOR FIGURE 7 OMITTED].
Long-term vegetation dynamics
We suggest three possible reasons for the long-term changes in species' abundances observed in the PAR and pollen percentage records [ILLUSTRATION FOR FIGURE 7 OMITTED]: climatic change, fires with different characteristics, and spatial rearrangement of forest dominants.
It is possible that the Medieval Warm Period (about AD 900 to 1300; Hughes and Diaz 1994) created warm and dry conditions in northern Alberta. This climate may have resulted in relatively poor conditions for white spruce, thereby allowing aspen and other understory taxa to increase in abundance between 1100 and 1400 AD. Mixed-wood forests dominated by aspen typify the area [approximately]200 km south of RLA (Rowe 1972). A return to cooler conditions during the Little Ice Age (about AD 1550 to 1850; Hughes and Diaz 1994) may have allowed the return to dominance of white spruce, which, in turn, caused the decline of understory taxa due to increased shading (Lieffers and Stadt 1994). However, the long-term climate records that are required to test this idea are not available for northern Alberta.
The long-term vegetation dynamics might also be the result of the characteristics of individual fires, such as the season, spatial pattern, or the amount of soil exposed by the fire. For example, the marked change from Picea cf. glauca to Populus dominance following the fire at AD 1185 might reflect a large spring fire that exposed large areas of mineral soil. The spring burn would allow the aspen seeds dispersed in spring to become established before white spruce seeds, which are released in fall; the exposed mineral soil would allow aspen to become easily established; and the large fire size should result in more light aspen seeds than heavy spruce seeds reaching the site (cf. Zasada 1971). Other explanations involving characteristics of individual fires are possible.
The long-term changes in pollen records might also result from a local rearrangement of the forest patches around RLA that does not require a change in the actual absolute abundance of any of the taxa over a larger area. For example, it is possible that if the aspen-dominated patch just north of RLA and the white spruce-dominated patch in the RLA watershed [ILLUSTRATION FOR FIGURE 2 OMITTED] shifted places due to characteristics of a fire, the change in the pollen record observed between AD 1100 and 1550 would be observed. The ability to identify alternative spatial arrangements is confounded by the different source areas of the pollen taxa (cf. Prentice 1985, Sugita 1993). A network of pollen sites within a 1-[km.sup.2] area would have helped us to resolve whether or not the long-term changes were due to forest patch rearrangement.
In this study, we used fine-resolution fossil pollen and charcoal analysis to reconstruct an 840-yr record of site-specific fire occurrence, short-term postfire vegetation dynamics, and centennial-scale changes in forest composition in a boreal white spruce forest. We found some general and repeated patterns in postfire vegetation dynamics, but also much variability following individual burns. The 69-yr mean fire interval we obtained was lower than expected, probably because it is difficult to recognize the size of the fires evident in the pollen record. Long-term changes in forest composition were apparent and might have been caused by climate change, different fire characteristics, or rearrangement of forest patches. The usefulness of a single site in reconstructuring the mean fire interval, postfire vegetation dynamics, and long-term vegetation changes is limited. A network of closely spaced sites might alleviate some of the uncertainties described here.
We thank Ian Campbell, Terry Carleton, Jim Clark, Charles Cogbill, Darwyn Coxson, David Foster, Ed Johnson, Stef MacLachlan, Katrina Moser, and three anonymous reviewers for comments on earlier versions of this manuscript. This research was supported by a McMaster Graduate Scholarship and a Northern Scientific Training Grant to C. P.S. Larsen, and a NSERC Operating Grant and a Northern Supplement Grant to G. M. MacDonald.
Airphoto Analysis Associates. 1979. Biophysical resource inventory of Wood Buffalo National Park. Canadian Parks Service, Prairie Region, Winnipeg, Manitoba, Canada.
Bergeron, Y., and M. Dubuc. 1989. Succession in the southern part of the Canadian boreal forest. Vegetatio 79:51-63.
Birks, H. J. B., and H. H. Birks. 1980. Quaternary palaeoecology. Edward Arnold, London, UK.
Budd, A. C. 1979. Budd's flora of the Canadian prairie provinces. Agriculture Canada, Publication 1662, Ottawa, Ontario, Canada.
Campbell, I. D., and J. H. McAndrews. 1993. Forest disequilibrium caused by Little Ice Age cooling. Nature 366: 336-338.
Carroll, S. B., and L. C. Bliss. 1982. Jack pine-lichen woodland on sandy soils in northern Saskatchewan and northeastern Alberta. Canadian Journal of Botany 60:2270-2282.
Clark, J. S., J. Merkt, and H. Muller. 1989. Postglacial fire, vegetation, and human history on 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.
Cogbill, C. V. 1985. Dynamics of the boreal forests of the Laurentian Highlands, Canada. Canadian Journal of Forest Research 15:252-261.
Davis, M. B., S. Sugita, R. R. Calcote, and L. E. Frelich. 1994. Historical development of alternate communities in a hemlock-hardwood forest in northern Michigan, USA. Pages 19-39 in R. May and N. R. Webb, editors. Large-scale ecology and conservation Biology. Blackwell Scientific Publications, Oxford, UK.
Dean, W. E. 1974. Determination of carbonate and organic matter in calcareous sediments and sedimentary rocks by loss on ignition: a comparison with other methods. Journal of Sedimentary Petrology 44:249-253.
DeGrandpre, L., D. Gagnon, and Y. Bergeron. 1993. Changes in the understory of Canadian southern boreal forest after fire. Journal of Vegetation Science 4:803-810.
Dix, R. L., and J. M. A. Swan. 1971. The roles of disturbance and succession in upland forest at Candle Lake, Saskatchewan. Canadian Journal of Botany 49:657-676.
Environment Canada. 1982. Canadian climate normals: 1951-1980. Temperature and precipitation. The North. Atmospheric Environment Service, Toronto, Ontario, Canada.
Faegri, K., P. E. Kaland, and K. Kryzywinski. 1989. Textbook of pollen analysis by Knut Faegri and Jobs. Iversen. Fourth edition. John Wiley, Chichester, UK.
Fastie, C. L. 1995. Causes and ecosystem consequences of multiple pathways of primary succession at Glacier Bay, Alaska. Ecology 76:1899-1916.
Foote, M. J. 1983. Classification, description, and dynamics of plant communities after fire in the taiga of interior Alaska. Research Paper PNW-307. U.S. Forest Service, Portland, Oregon, USA.
Gottman, J. M. 1981. Time series analysis. Cambridge University Press, Cambridge, UK.
Green, D. G. 1981. Time series and postglacial forest ecology. Quaternary Research 15:265-277.
Hansen, B.C. S., and D. R. Engstrom. 1985. A comparison of numerical and qualitative methods of separating pollen of black and white spruce. Canadian Journal of Botany 63: 2159-2163.
Hughes, M. K., and H. F. Diaz. 1994. Was there a "Medieval Warm Period," and if so, where and when? Climatic Change 26:109-142.
Johnson, E. A. 1981. Vegetation organization and dynamics of lichen woodland communities in the Northwest Territories, Canada. Ecology 62:200-215.
-----. 1992. Fire and vegetation dynamics: studies from the North American boreal forest. Cambridge University Press, Cambridge, UK.
Johnson, E. A., and S. L. Gutsell. 1994. Fire frequency models, methods, and interpretations. Advances in Ecological Research 25:239-287.
Johnson, E. A., K. Miyanishi, and H. Kleb. 1994. The hazards of interpretation of static age structures, as shown by stand reconstructions in a Pinus contorta-Picea engelmannii forest. Journal of Ecology 82:923-931.
Larsen, C. P.S. 1997. Spatial and temporal variations in boreal forest fire frequency in northern Alberta. Journal of Biogeography 24, in press.
Larsen, C. P.S., and G. M. MacDonald. 1993. Lake morphometry, sediment mixing, and the selection of sites for fine-resolution palaeoecological studies. Quaternary Science Reviews 12:781-792.
Larsen, C. P.S., and G. M. MacDonald. 1995. Relations between tree ring widths, climate, and annual area burned in the boreal forest of Alberta. Canadian Journal of Forest Research 25:1746-1755.
Lieffers, V. J., and K. J. Stadt. 1994. Growth of understory Picea glauca, Calamagrostis canadensis, and Epilobium angustifolium in relation to overstory light transmission. Canadian Journal of Forest Research 24:1193-1198.
MacDonald, G. M., C. P.S. Larsen, J. M. Szeicz, and K. A. Moser. 1991. The reconstruction of boreal forest fire history from lake sediments: a comparison of charcoal, pollen, sedimentological, and geochemical indices. Quaternary Science Reviews 10:53-71.
McAndrews, J. H., A. A. Berti, and G. Norris. 1973. Key to the Quaternary pollen and spores of the Great Lakes Region. Life Science Miscellaneous Publications, Royal Ontario Museum, Toronto, Ontario, Canada.
Nienstaedt, H., and Zasada, J. C. 1990. White spruce. Pages 204-226 in R. M. Burns and B. H. Honkala, editors. Silivics of North America: 1, Conifers. Agriculture Handbook 654. United States Department of Agriculture. Washington, D.C., USA.
Okland, R. H. 1995. Persistence of vascular plants in a Norwegian boreal coniferous forest. Ecography 18:3-14.
O'Sullivan, P. E. 1983. Annually laminated lake sediments and the study of Quaternary environmental changes: a review. Quaternary Science Reviews 1:245-313.
Patterson, W. A., III, K. J. Edwards, and D. J. Maguire. 1987. Microscopic charcoal as a fossil indicator of fire. Quaternary Science Reviews 6:3-23.
Pickett, S. T. A. 1989. Space-for-time substitution as an alternative to long-term studies. Pages 110-135 in G. E. Likens, editor. Long-term studies in ecology. Springer-Verlag, New York, New York, USA.
Prentice, I. C. 1985. Pollen representation, source area, and basin size: towards a unified theory of pollen analysis. Quaternary Research 23:76-86.
Purchase, J. E., and G. H. LaRoi. 1983. Pinus banksiana forests of the Fort Vermillion area, northern Alberta. Canadian Journal of Botany 61:804-824.
Rhodes, T. E., and R. B. Davis. 1995. Effects of late Holocene forest disturbance and vegetation change on acidic Mud Pond, Maine, USA. Ecology 76:734-746.
Ritchie, J. C. 1985. Late-Quaternary climatic and vegetational change in the Lower Mackenzie Basin, northwest Canada. Ecology 66:612-621.
Rowe, J. S. 1972. Forest regions of Canada. Canadian Forestry Service. Ottawa, Canada.
-----. 1983. Concepts of fire effects on plant individuals and species. Pages 135-154 in R. W. Wein and D. A. MacLean, editors. The role of fire in northern circumpolar ecosystems. John Wiley, New York, New York, USA.
Shugart, H. H., R. Leemans, and G. B. Bonan, editors. 1992. A systems analysis of the global boreal forest. Cambridge University Press, New York, New York, USA.
Stockmarr, J. 1971. Tablets with spores used in absolute pollen analysis. Pollen et Spores 13:615-621.
Stuiver, M., and P. J. Reimer. 1993. Extended 14C database and revised Calib 3.0 14C age calibration program. Radiocarbon 35:215-230.
Sugita, S. 1993. A model of pollen source area for an entire lake surface. Quaternary Research 39:239-244.
Sugita, S., G. M. MacDonald, and C. P.S. Larsen. 1997. Reconstruction of fire disturbance and forest succession from fossil pollen in lake sediments: potential and limitations. Pages 387-412 in J. S. Clark, H. Cachier, J. G. Goldammer, and B. Stocks, editors. Sediment records of biomass burning and global change. NATO ASI Series, Volume I 51. Springer-Verlag, Berlin Germany.
Swain, A.M. 1978. Environmental changes during the past 2000 years in north central Wisconsin: analysis of pollen, charcoal, and seeds from varved lake sediments. Quaternary Research 10:55-68.
Wright, H. E., Jr., D. H. Mann, and P. H. Glaser. 1984. Piston corers for peat and lake sediments. Ecology 65:657-659.
Yarie, J. 1981. Forest fire cycles and life tables: a case study from interior Alaska. Canadian Journal of Forest Research 11:554-562.
Zasada, J. 1971. Natural regeneration of interior Alaska forests. Pages 231-246 in C. W. Slaughter, R. J. Barney, and G. M. Hansen, editors. Fire in the northern environment: a symposium. Pacific Northwest Forest, U.S. Forest Service, Portland, Oregon, USA.
|Printer friendly Cite/link Email Feedback|
|Author:||Larsen, C.P.S.; MacDonald, G.M.|
|Date:||Jan 1, 1998|
|Previous Article:||Plant species richness in riparian wetlands - a test of biodiversity theory.|
|Next Article:||Soil chemistry and plants in Fennoscandian boreal forest as exemplified by a local gradient.|