Millennial-scale fire and vegetation h
hardwood forest of southeastern Wi
DON
Wis
, UT,
er 20
nt co
Kett
isten
tion
f hig
ale re
0 yea
crea
kettle lake in the northern unit of the Kettle Moraine State Forest
oist summers and cool,
y and July mean monthly
respectively, and mean
JOURNAL OF QUATERNARY SCIENCE (2011) 26(3) 318–325 ISSN 0267-8179. DOI: 10.1002/jqs.1456* Correspondence: C. J. Long, as above.
ommes et al., 2004), withbe characterized as having warm, m
dry winters (Keys et al., 1995). Januar
temperatures average –8.3 and 21.78C
annual precipitation is ca. 905mm (GE-mail: longco@uwosh.eduMarlon et al., 2009).
The primary focus of fire history reconstructions in the upper
Great Lakes region has been in pine-dominated boreal forest
of northern Minnesota and Wisconsin or the mixed hardwood
conifer forests of Michigan and northern Wisconsin (Swain,
1978; Gajewski et al., 1985; Cleland et al., 2004; Shulte and
Mladenoff, 2005; Hotchkiss et al., 2007; Tweiten et al., 2009).
Little information regarding long-term fire histories has
been gathered from the primarily deciduous forests south of
the ‘tension zone’ in south-eastern Wisconsin (Curtis, 1959).
This is a region where conifers are replaced by additional
deciduous taxa such as Quercus, Juglans and Fraxinus,
hereafter be referred to as mesic hardwood forests. The goals
of this study were to reconstruct long-term fire histories
from locations within the mesic hardwood forest vegetation
zone, and examine the relations between changes in fire
during the end of the last glacial maximum (Carlson et al.,
2004; Fig. 1). This site lies in sandy soils adjacent to an esker
and has a surface area of 3 ha, a maximum water depth of 4m,
simple bathymetry and perennial inflowing streams that drain
an area of 200 ha. Dominant forest taxa in the watershed today
are sugar maple, red and white oak, and basswood, with some
hickory and paper birch. Pinus resinosa (red pine) has been
planted in several small stands within the watershed as a timber
crop. Lake Seven (43.61478N, 88.14258 W, elevation 309m)
lies 5 km south of Butler Lake on the eastern flanks of the
Lake Michigan Lobe hummocky ridge area (Carlson et al.,
2004). It is a 10-ha kettle lake with a maximum water depth of
7m and no discernible inflowing or outflowing streams in the
80-ha watershed. The forest vegetation at Lake Seven is similar
to that at Butler Lake. The climate of both watersheds canlonger time scales and larger spatial scales (Power et al., 2008;particles can establish or extend fire histories and allow
vegetation–fire–climate relationships to be examined over
and lies in a low area surrounded by higher hummocky ridges
formed through glacial outwash deposition of the Green Bay
Lobe to the north, and the Lake Michigan Lobe to the westCOLIN J. LONG,1* MITCHELL J. POWER2 and BREANNE MC
1Department of Geography and Urban Planning, University of
2Department of Geography, University of Utah, Salt Lake City
Received 17 February 2010; Revised 21 September 2010; Accepted 27 Septemb
ABSTRACT: High-resolution charcoal analysis of lake sedime
mesic hardwood forest of south-easternWisconsin located in the
the sites, which lie within 5 km of each other, have had a cons
pollen record from one of the sites confirmed the regional vegeta
each site was temporally linked to regional drought. Periods o
drought 4200 years ago and, over the last 2000 years, shorter-sc
700 and 600 cal a BP. The fire histories indicate that the last 100
and suggest that the mesic hardwood forests may be resilient to in
# 2011 John Wiley & Sons, Ltd.
KEYWORDS: fire history; Holocene; hardwood forests; Wisconsin.
Introduction
Variations in climate have been a major cause of vegetation
change in the upper Great Lakes region since the retreat of
the Laurentide ice sheet 15 ka ago (Bartlein et al., 1984;
Wright, 1992) and the progression of tree species into the
region as climate conditions warmed over the Holocene is
well documented (Webb et al., 1983, Webb, 1987; Baker
et al., 1992). In addition, our understanding of late Holocene
disturbances, such as fire, in shaping regional vegetation
communities (Turner, 1989; Turner and Dale, 1998; Lynch
et al., 2004) suggests that current vegetation patterns reflect past
patterns of disturbance operating at a variety of temporal scales
(Heinselmann, 1973; Hotchkiss et al., 2007). Although strong
links have been documented between vegetation, fire and
climate over the last century (Westerling et al., 2006), our
understanding of these links on long time scales, particularly in
ecosystems with infrequent fire, requires millennial-scale fire
histories beyond the reach of most dendrochronological
studies (Whitlock et al., 2003). In areas where dendrochrono-
logical records are of short temporal length, or are sparse, the
examination of lake sediment for plant remains and charcoalCopyright
2011 John Wiley & Sons, Ltd.istory from a mesic
sconsin, USA
ALD1
consin Oshkosh, Oshkosh, WI 54901-8624, USA
USA
10
res was used to reconstruct the fire history from two sites in a
leMoraine State Forest. Pollen data from the region indicate that
t presence of mesic hardwood forest for the last 6500 years. A
history and the charcoal analysis indicated that fire frequency at
h fire occurrence occurred in connection with a region-wide
gional droughts were centred at 1800, 1650, 1100, 1000, 800,
rs have had lower fire frequencies than the previous 6500 years
ses in fire that may result from future climate change. Copyright
frequency and forest communities over the time period they
have persisted in south-eastern Wisconsin.
Study sites
Two sites, Butler Lake and Lake Seven, were selected for this
study as they are in areas that have been dominated by mesic
hardwood forest species for the last 7000 cal a BP (Webb
et al., 1983;Webb, 1987;Wright, 1992). Themajor species that
constitute mesic hardwood forests of the area are Acer
saccharum (sugar maple), Tilia americana (basswood) and
Quercus spp. (oak), with some Ulmus spp. (elm), Carya spp.
(hickory), Abies balsamea (balsam fir), Fraxinus spp. (ash),
Ostrya virginiana (ironwood) and Betula papyrifera (paper
birch), with a minor constituent of Fagus grandifolia (American
beech) (nomenclature follows Gleason and Cronquist, 1963).
Although the abundance of each species at each location may
have changed since 7000 cal a BP, the overall climate has been
within the range of tolerances for all these major tree taxa
(Bartlein et al., 1984; Webb et al., 1993).
Butler Lake (43.66258N, 88.13398 W, elevation 318m) is a

MILLENNIAL-SCALE FIRE AND VEGETATION HISTORY IN SE WISCONSIN, USA 319approximately 75% of that falling between April and October.
Fire season in this area occurs in spring, prior to summer rains,
as snowmelt and increasing temperatures lower fuel moisture
in grasses and forest litter (WDNR, 2005). Since the establish-
ment of Kettle Moraine State Forest in 1936, no wildfires larger
than 10 acres have been reported (J. Leiterman, Kettle Moraine
State Forest, personal communication). There is a lack of
L
a
k
e





M
i
c
h
i
g
a
n
Lake Superior
WI
MN
IA
IL
MI
43 N
o
46 N
o
oo92 W 88 W
WL
FL
HB
SW SR
CA
MI
Figure 1. Location of Radtke Lake (RL), Butler Lake and Lake Seven
(BL) and Gass Lake (GL) within the upper Great Lakes region. The
shaded area indicates the present-day extent of the mesic hardwood
forests in Wisconsin (Curtis, 1959). Also shown are the locations from
Booth et al. (2005), South Rhody Peatland (SR) and SylvaniaWilderness
(SW); Hotchkiss et al. (2007), Ferry Lake (FL); Tweiten et al. (2009),
Warner Lake (WL); and Booth et al. (2006), Hole Bog (HB)mentioned in
the text that show evidence of Holocene drought conditions.dendrochronological data for the reconstruction of historic fires
in mesic hardwood forests of Wisconsin, so most of the
information about fire regimes is inferred from other regions
such as central Michigan. The modern fire regime in the mesic
hardwood forests of south-eastern Wisconsin has been
characterized as of mixed severity with over 400 years between
fires (Dickman and Cleland, 2002).
Methods
The field and laboratory methods for each record were similar.
Sediment cores were collected from the deepest part of each
lake using a 5-cm-diameter modified piston sampler (Wright
et al., 1983). Cores were extruded in the field, wrapped in
cellophane and aluminium foil, and transported back to the
laboratory where they were refrigerated. In the laboratory, the
cores were sliced longitudinally, described and subsampled
for pollen and charcoal analysis. The chronology for each
core was based on 14C dates from terrestrial macrofossils and
sediment.
Pollen analysis
As the Holocene history of forest vegetation migrating into the
upper Great Lakes region has been well documented (Webb et
al., 1983, 1993; Davis, 1993) and into south-eastern Wisconsin
in particular (West, 1961; Winkler et al., 1986; Maher, 1982;
Webb, 1994), we examined pollen for vegetation reconstruc-
tion at Butler Lake to confirm the regional progression of tree
species into the study area. Samples of 1 cm3 were taken every
Copyright
2011 John Wiley & Sons, Ltd.40 cm (approximately every 600 years) for pollen extraction
following the procedures of Faegri et al. (1989). A known
amount of microspheres was added to each sample prior to
processing to calculate pollen accumulation rates. The pollen
samples were mounted in silicon oil and examined at
magnifications of 400–1000
. Pollen was identified to the
lowest taxonomic level possible based on modern pollen
atlases. A minimum of 300 terrestrial grains were identified.
Pollen grains that could not be identified were labelled
‘Unknown’. Terrestrial pollen percentages were calculated
using the sum of terrestrial pollen and spores. Percentages of
aquatic taxa were calculated based on total terrestrial and
aquatic pollen and spores. The pollen percentage diagram for
Butler Lake was compared with the pollen stratigraphy from
Radtke Lake and Gass Lake, which are located 24 km south and
52 km north-west of Butler Lake, respectively, and sampled at
approximately 200-year intervals (Webb, 1987) (Fig. 1; see also
Fig. 4 below).
Fire history reconstruction
Variations in the abundance of macroscopic charcoal found in
the lake sediment records were used to reconstruct the fire
history for each site (Whitlock and Larson, 2001). Sediment
sampling for charcoal followed Long et al. (1998). Subsamples
of 2–3 cm were taken at contiguous 1-cm intervals and soaked
in 5% solution of hydrogen peroxide for 24 h. The samples were
then gently washed through a series of nested screens with
mesh sizes of 250 and 125mm. The sieved samples were
examined at 50
magnification, and all charcoal particles
greater than 125mm were counted. Charcoal counts for each
sample were converted to concentration (particles cm1) and,
using the sediment deposition rate, to charcoal accumulation
rates (CHAR, particles cm1 a1) at constant time steps (18
years at Butler Lake, 14 years at Lake Seven) to minimize any
variations in the record that might arise because of changes in
the deposition rate. The CHAR record was then decomposed
into background and peak components (Higuera et al., 2008,
http://CharAnalysis.googlepages.com). Background charcoal is
the slowly varying trend in CHAR as a primary result of changes
in fuel abundance and composition (Marlon et al., 2006).
Peaks, which are positive deviations from the background
CHAR (BCHAR), represent input of charcoal as a result of a fire
episode (one or more closely timed fires; Long et al., 1998). The
BCHAR component was determined using a Lowess smoother
robust to outliers with a 500-year window width. The
background values for each time interval were then subtracted
from the total CHAR accumulation for each interval. The peaks
in the charcoal record (i.e. intervals with CHAR values above
background) were tested for significance using a Gaussian
distribution, where peak CHAR values that exceeded the 95th
percentile were considered significant (i.e. not the result of
natural signal noise or analytical error). This procedure was
done on every 500-year overlapping portion of the CHAR
record, producing a unique threshold for each sample. Once
identified, all peaks were screened to eliminate those that
resulted from statistically insignificant variations in CHAR
(Gavin et al., 2006). If the maximum count in a CHAR peak had
a >5% chance of coming from the same Poisson-distribution
population as the minimum charcoal count with the proceed-
ing 75 years, then the peak was rejected (Higuera et al., 2008).
We calculated three measures of fire activity from each lake
record. First, the mean fire-episode return interval (FRI: average
time between fire episodes), which was determined using the
estimated dates of the CHAR peaks. Second, the fire-episode
frequency (fires episodes ka1), which was calculated using a
tricube-weighted regression (Cleveland, 1979), to smooth theJ. Quaternary Sci., Vol. 26(3) 318–325 (2011)