Study Site and Species We worked at Archbold Biological Station (ABS;
Swain 1998), a research facility in south-central Florida (27o10’ N,
81o21’W). ABS is one of the largest remaining protected tracts on the
Lake Wales Ridge (Weekley et al. 2008) and a hotspot for endemism
(Estill and Cruzan 2001; Turner et al. 2006). ABS includes over 2,100
ha of Florida scrub, southern ridge sandhill, flatwoods, and seasonal
pond habitats (Abrahamson et al. 1984). We located the experiment in
scrubby flatwoods (), a type of Florida scrub found on coarse sands
(Abrahamson et al. 1984; Menges 1999). Our study site was on Duette
sand (a hyperthermic, Grossarenic Entic Haplohumod), a moderately well
drained soil with low water holding capacity (Carter et al. 1989).
Scrubby flatwoods are dominated by resprouting shrubs (Menges and
Kohfeldt 1995; Maguire and Menges 2011; Schafer and Mack 2013) and
have fewer and smaller gaps than neighboring rosemary scrub (Dee and
Menges 2014). Scrubby flatwoods have low nutrient availability
(Schafer and Mack 2013) but show little evidence of nutrient
limitation during the first year after fire (Schafer and Mack 2018),
suggesting that post-fire increases in nutrient availability (Lavoie
et al. 2010, Schafer and Mack 2010, Ficken and Wright 2017) may be
important for supporting resprouting and subsequent shrub growth in
scrubby flatwoods. The fire regime in scrubby flatwoods is
characterized by high intensity crown fires moving through shrub
canopies. The presumed natural fire return interval is 8-16 years
(Menges 2007; Menges et al. 2017). Fires were historically ignited by
lightning occurring during the growing season, especially in late
spring (Platt et al. 2015, Noss 2019). At ABS, fires vary in severity
and patchiness due to vegetation type but not weather or season
(Menges et al. 2017). Fire temperatures are high, with maximum
temperatures reaching 488 C, one-minute mean temperatures ranging from
265 to 330 C, and mean residence times (minutes above 60C) ranging
from 5 to 14.7 minutes (Wally et al. 2006; Dean et al. 2015). We chose
a single burn unit (BU 30A) at ABS, which last burned in 1994 (11
years before the study initiation) with sufficient intensity to
top-kill all shrubs. We studied eight of the most common shrub species
found in scrubby flatwoods at ABS: three oak species (Quercus inopina
Ashe, Q. geminata Small, Q. chapmanii Sarg.), three ericaceous shrubs
(Lyonia fruticosa (Michx.) G.S. Torr., L. lucida (Lam.) K. Koch,
Vaccinium myrsinites Lam), and two palmettos (Sabal etonia Swingle ex
Nash, Serenoa repens (W. Bartram) Small) (nomenclature follows
Wunderlin et al. 2019). Other prominent species at this site (V.
darrowii Camp, Gaylussacia dumosa Andrews (A. Gray), Palafoxia feayi
A. Gray, Aristida stricta Michx., Bejaria racemosa Vent., Chapmannia
floridana Torr. & A. Gray, and Rhynchospora megalocarpa A. Gray)
did not have enough stems distributed throughout the site to use in
the experiment. Sand pine (Pinus clausa Chapm. ex Engelm.)Vasey ex
Sarg.) was an important tree in nearby areas but was sparse at our
study site.
Experimental Design To explore the effects of disturbance return
interval (DRI) and disturbance type (burning, mowing) on resprouting
and growth of eight shrub species, we designed our experiment to
separate effects of time-since-disturbance from DRI. Over a period of
six years (2005-2010), plots were burned or mowed once, twice, three
times, or six times. The burning and mowing treatments were arranged
so that every plot was burned or mowed in 2010, the sixth year of the
study (). This kept all plots at the same time-since-disturbance
following the application of the final treatment in 2010 and allowed
the analyses to focus on effects of DRI on resprouting responses.
Study Site and Plot Setup We had four replicates of each disturbance
type (fire, mowing) and DRI (1,2,3,6 years) combination, resulting in
32 total plots. The 32 plots were set up in two groups of 16. To avoid
potential effects of burns on mowed plots, the two types of plots were
separated by about 20 m. The 16 burned plots were in one N-S line and
the 16 mowed plots were in a second line upwind (to the east) of the
burned plots. Each plot was 20 m by 5 m, with a 15 m by 1 m belt
transect centered within the plot, which was used for vegetation
measurements. The long axis of each plot was oriented E-W (in the
direction of the most common wind) so that head fires would occur with
moderate intensity. All plots were > 20 m from the nearest sand
road. To restrict fire to appropriate areas, we used mowing to create
5 m firebreaks around each plot and > 10 m firebreaks around the
entire study area.
Burning and Mowing Treatments DRIs were assigned randomly and
independently for burned and mowed plots. Fires were set with drip
torches in individual plots according to the experimental design ().
We generally used easterly winds to burn the plots toward the road on
the western boundary of the burn unit, sending ash away from the mowed
plots. If there was insufficient fuel due to repeated burns, we added
fuel that had been mown from the adjacent firebreaks (especially
palmetto leaves and pine needles). Fires were lit repeatedly to burn
100% of all burn-assigned plots. Fires were allowed to burn out before
any water was used in mop-up. All fires were conducted in the spring
months (), were successful in burning the plots without escapes, and
top-killed nearly all shrub stems. We mowed plots according to the
same schedule as burning (), a few days after each burn was
accomplished. We used a Brown tree cutter (Brown Manufacturing
Corporation, Ozark, AL) and mowed to a height of about 10 cm. Mowed
vegetation was left in the plots.
Fire Measurements During each fire, we measured fire temperatures
using HOBO temperature dataloggers (Type K thermocouple logger part
H12002, Onset Computer Corporation, Bourne, MA). Prior to the fires,
dataloggers were buried 10 cm below the soil surface with double
stranded wire leads exposed at the soil surface (beneath any litter;
there was no duff in these plots). Dataloggers were installed at 5-m
intervals along the long axis of each plot within the belt transect.
Dataloggers were programmed to save temperatures every 2 seconds. From
the raw data, we derived maximum temperatures, maximum one-minute mean
temperatures, and residence times (minutes above 60 oC). We also
estimated flame lengths and recorded weather data (relative humidity,
temperature, wind speed, wind direction) during fires.
Vegetation Measurements and Analysis We collected vegetation data on
the eight most common resprouting shrubs from belt transects centered
within the plots. We counted the number of stems (ramets) of each
target species in the belt transect. In addition, at each 1-m
interval, we selected the nearest ramet of each target species (if
within 50 cm of the transect) and measured its height and maximum
crown diameter (to the nearest cm). If we did not find at least five
ramets of a species using this sampling method, we randomly measured
additional ramets within the plot to bring sample sizes to five. Ramet
location was determined by rooting position (ground level). We
estimated aboveground biomass from height and crown diameter using
regressions developed from harvests of these species at ABS (Menges
and Smith 2019). For the current study, we chose regression equations
for plants collected across a range of times-since-fire. We sampled
vegetation pre-treatment in spring 2005 and 6 months (October 2010), 1
year (April 2011), 2 years (April 2012), and 4 years (April 2014)
after treatments ended. We analyzed biomass, height, and number of
stems (all ln transformed) in relation to species, DRI, and
disturbance type (and their interactions) in repeated measures
analysis of variance (ANOVA). We chose the univariate procedure as it
is more powerful than the multivariate procedure (von Ende 1993;
Potvin et al. 1990). For univariate repeated measures ANOVA, we used
Mauchly’s W test for sphericity. Since the assumption of sphericity
was never met, degrees of freedom were adjusted using the
Greenhouse-Geisser estimated epsilon values (Potvin et al. 1990). We
also used repeated measures ANOVA to analyze treatment effects on
individual species’ biomass as well. For individual sampling times, we
used factorial ANOVAS to analyze the effects of DRI and disturbance
type on biomass. All analyses used SPSS version 22.
Non-structural carbohydrates (NSC) estimations and analysis We
collected belowground plant parts to analyze NSC in xylem within our
plots a few months after treatments (in 2010) and one year after
treatments (in 2011). At both times, we used hand tools to carefully
excavate storage organs (roots or rhizomes). However, we collected
different numbers of samples each time due to logistical constraints.
In 2010, we collected a single individual of each of six species from
within each plot. In 2011, we collected 5 samples from four of the
most abundant species in our plots (Lyonia fruticosa, L. lucida,
Quercus geminata, Q. inopina) from within each plot. We collected
samples outside the 1 m belt transect used for vegetation
measurements. Samples were immediately trimmed and the extracted xylem
oven-dried at 60 C for 72 h. Dry material was finely ground using a
Retsch mill (Retsch MM 400, Dsseldorf, Germany). The soluble and
non-soluble fractions of NSC were extracted using the perchloric acid
/ anthrone method (Morris 1948), on 20 mg of dried and powdered wood
(see Olano et al. 2006 for details on the technique). This procedure
distinguishes soluble and non-soluble fractions of NSC. Soluble sugars
(SS) include mono- and disaccharides such as glucose, fructose, and
sucrose. Non-soluble sugars (NSS) include polysaccharides such as
starch and fructans. SS and NSS were expressed as the percentage (w/w)
of dry mass. Total NSC was estimated as the sum of SS and NSS. Because
not all species were sampled in all years, we analyzed soluble and
insoluble fractions of NSC in 2010 and 2011 separately in relation to
species, disturbance type, and DRI using general linear models
(insoluble fraction was ln transformed). Effects of 2010 NSC levels on
biomass at six months and of 2011 NSC levels on biomass at one year
were analyzed with linear regressions.
Soil Nutrient Measurements and Analysis On 26 April 2011 and 16 April
2012, approximately one and two years after the final treatment
application, we collected and bulked three soil cores (5 cm diameter;
6 cm depth) from each plot to obtain one sample per plot. Immediately
after collection, soils were passed through a 2 mm sieve to remove
large roots and belowground stems. In 2011, sub-samples of soil were
weighed for measurement of gravimetric soil moisture, soil organic
matter, soil pH, and determination of extractable ammonium (NH4+),
nitrate (NO3-), and phosphate (PO43-). In 2012, sub-samples of soil
were weighed for measurement of gravimetric soil moisture and soil
organic matter. The inferences we can draw form soil data are
constrained by sampling limitations because we did not measure soil
properties pre-disturbance. However, we expect that soil properties
were similar among plots, at least within a disturbance type, since
they were all in the same burn unit and had experienced the same
management history. Gravimetric soil moisture was determined on soils
dried at 105C for 48 hours. Soil organic matter was determined using
the loss on ignition method (Nelson and Sommers 1996); soils were
incubated in a muffle furnace at 400C for 16 hours. Soil pH was
measured on a 1:1 ratio of air-dried soil and de-ionized water (Thomas
1996) with an electronic pH meter (Thermo Orion 320A, Orion Research,
Inc., Boston, MA). To measure inorganic N, 50 mL of 2 M potassium
chloride (KCl) was added to 10 g of field moist soil, shaken for 30
seconds, and allowed to stand overnight. We filtered solutions through
Fisherbrand Q2 filter paper pre-leached with 2 M KCl. Extracts were
stored in the refrigerator for one day then analyzed colorimetrically
on a spectrophotometer microplate reader (Quant Microplate
Spectrophotometer, Bio-Tek Instruments, Inc., Winooski, VT) at the
MacArthur AgroEcology Research Center (MAERC). Concentrations of NH4+
were determined using the trisodium citrate, salicylate-nitroprusside,
hypochlorite method (Sims et al. 1995, Mulvaney 1996) and
concentrations of NO3- were determined using the vanadium,
sulfanilamide, NEDD method (Miranda et al. 2001). To measure
extractable phosphorus, 30 mL of 0.05 M hydrochloric acid (HCl) and
0.0125 M hydrogen sulfate (H2SO4) were added to 15 g of field moist
soil, shaken for 5 min, and then immediately filtered through
Fisherbrand Q2 filter paper. Concentrations of phosphate (PO43-) were
determined immediately after filtration on a spectrophotometer
microplate reader at the MAERC using the malachite green method
(D’Angelo et al. 2001). Soil variables were analyzed using a full
factorial ANOVA with disturbance type (fire vs. mowing) and DRI
(annual, biennial, triennial, once in six years) as fixed factors. We
analyzed total inorganic N as the sum of ammonium and nitrate;
concentrations of NO3- were zero in all but one plot. One plot (burned
once in six years) had a NH4+ concentration over 3.5 times higher than
the plot with the next highest NH4+ concentration; we suspect NH4+ was
high for a reason other than the treatment, so data from this plot
were not included in analyses for 2011. We used multiple linear
regressions to analyze the effects of 2011 soil variables on biomass
in 2011 (one year post-treatment) and 2012 (two years post-treatment).
Most soil variables were ln transformed for analysis.
