Access to this full-text is provided by Wiley.
Content available from Ecosphere
This content is subject to copyright. Terms and conditions apply.
The role of understory phenology and productivity in the carbon
dynamics of longleaf pine savannas
SUSANNE WIESNER,
1
CHRISTINA L. STAUDHAMMER,
1
CHLOE L. JAVAHERI,
1
J. KEVIN HIERS,
2
LINDSAY R. BORING,
3
ROBERT J. MITCHELL,
3
AND GREGORY STARR
1,
1
Department of Biological Sciences, University of Alabama, Tuscaloosa, Alabama 35487 USA
2
Tall Timbers Research Station, 13093 Henry Beadel Dr., Tallahassee, Florida 32312 USA
3
Joseph W. Jones Ecological Research Center, Newton, Georgia 39870 USA
Citation: Wiesner, S., C. L. Staudhammer, C. L. Javaheri, J. K. Hiers, L. R. Boring, R. J. Mitchell, and G. Starr. 2019. The
role of understory phenology and productivity in the carbon dynamics of longleaf pine savannas. Ecosphere 10(4):
e02675. 10.1002/ecs2.2675
Abstract. Savanna ecosystems contribute ~30% of global net primary production (NPP), but they vary
substantially in composition and function, specifically in the understory, which can result in complex
responses to environmental fluctuations. We tested how understory phenology and its contribution to
ecosystem productivity within a longleaf pine ecosystem varied at two ends of a soil moisture gradient
(mesic and xeric). We used the Normalized Difference Vegetation Index (NDVI) of the understory and
ecosystem productivity estimates from eddy covariance systems to understand how variation in the under-
story affected overall ecosystem recovery from disturbances (drought and fire). We found that the mesic
site recovered more rapidly from the disturbance of fire, compared to the xeric site, indicated by a faster
increase in NDVI. During drought, understory NDVI at the xeric site decreased less compared to the mesic
site, suggesting adaptation to lower soil moisture conditions. Our results also show large variation within
savanna ecosystems in the contribution of the understory to ecosystem productivity and recovery, high-
lighting the critical need to further subcategorize global savanna ecosystems by their structural features, to
accurately predict their contribution to global estimates of NPP.
Key words: drought; fire; Normalized Difference Vegetation Index; net ecosystem production; prescribed fire; savanna
ecosystems.
Received 4 September 2018; revised 20 February 2019; accepted 26 February 2019. Corresponding Editor: Dawn M.
Browning.
Copyright: ©2019 The Authors. This is an open access article under the terms of the Creative Commons Attribution
License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
E-mail: gstarr@ua.edu
INTRODUCTION
Understory vegetation is thought to play a key
role in carbon dynamics of ecosystems across the
globe. Uncertainties exist, however, regarding
the total carbon found in this vegetation strata
(Johnson et al. 2017, Refsland and Fraterrigo
2017), and these uncertainties are expected to
increase as climatic extremes become more com-
mon and severe (IPCC 2014). These shifts in cli-
mate, particularly drought, will have a direct
effect on ecosystem structure (both understory
and overstory), function (phenology and physiol-
ogy), and ultimately their management strategies
(Rich et al. 2008, Becknell et al. 2015). Quantify-
ing phenological changes and understory carbon
dynamics is particularly difficult using coarse
remote sensing techniques due to the canopy
variation in forests and/or uneven sensor
obstruction (Picotte and Robertson 2011). Fur-
thermore, critical variation of the understory
occurs at fine (1–10 m
2
) spatial scales (Hiers et al.
2009, Starr et al. 2015). This is especially the case
for savanna and woodland ecosystems where
❖www.esajournals.org 1April 2019 ❖Volume 10(4) ❖Article e02675
canopies can be relatively open and less uniform,
and therefore harder to classify using coarse spa-
tial scales of satellite-derived remote sensing
products (Patenaude et al. 2004, Simard et al.
2011, Smith et al. 2014).
Understanding savanna structure and function
is critically important, as savannas cover ~15% of
the global land surface (Ma et al. 2013) and con-
tribute ~30% to global net primary production
(NPP, Field et al. 1998, Fei et al. 2017), thus repre-
senting a significant portion of the global carbon
cycle (Bombelli et al. 2009). Nevertheless,
savanna ecosystems vary across the globe based
on their underlying geology and regional climates
(Bond et al. 2003, Vaughn et al. 2015). Differences
among and within savannas often result in com-
plex responses to environmental fluctuations
(Boke-Ol
en et al. 2016, Starr et al. 2016). For
example, savanna ecosystems were shown to
exhibit different responses to drought, even
within the same region (Starr et al. 2016). This
finding has been linked to variations in soil water-
holding capacity, which in turn has led to varia-
tions in plant functional types among sites
(Vaughn et al. 2015). Differences in plant func-
tional types can drive changes in phenological
patterns, as well as differential response to distur-
bances (Rich et al. 2008, Bernhardt-R€
omermann
et al. 2011, Wu et al. 2016). Thus, inclusion of site-
specific phenology is extremely important in
making accurate predictions regarding physio-
logical patterns and ecosystem carbon dynamics.
These patterns, however, are poorly represented
in global dynamic vegetation models and are a
source of systematic errors in regional to global
scale model predictions (Boke-Ol
en et al. 2016).
Complicating our understanding of ecosystem
carbon and energy dynamics is the fact that dis-
turbance events result in differing responses
between the understory and the overstory (Rich
et al. 2008). The understory in forests and wood-
lands can contribute roughly 30–50% to the
annual carbon fluxes of an ecosystem (Powell
et al. 2008, Bracho et al. 2012). However, this
response varies across the landscape and may
change with increasing frequencies of the distur-
bance (Mitchell et al. 2014). A study on North
American savannas revealed that the understory
was more capable of rapidly utilizing newly
available resources following a prolonged
drought compared to the overstory (Rich et al.
2008). This rapid shift to resource availability off-
set the whole ecosystem decline in function and
accelerated recovery from the disturbance (Rich
et al. 2008). In contrast, Ma et al. (2013) showed
that increasing temporal variability in phenology
coincided with decreasing tree–grass ratios,
which suggest that higher woody fractions in the
understory promote higher stability and resili-
ence to future climate change. Collectively, these
studies highlight the influence of site-specificabi-
otic and biotic factors that drive savanna over-
story and understory phenological patterns and
ultimately the ecosystem’s carbon dynamics.
Savanna understory vegetation also plays a
key role in enabling the efficient spread of sur-
face fires through these ecosystems, which main-
tains the ecosystem’s structure and function
(Barlow et al. 2015). With a bare or patchy
understory, fire would not effectively spread
across the floor and therefore enable the expan-
sion of woody shrubs (Mitchell et al. 2006), lead-
ing to a shift in the ecosystem structure (Peterson
and Reich 2007), and hence a shift in ecosystem
carbon dynamics. Changes in structure can also
precipitate declines in biodiversity (Walker and
Peet 1984) which further alter the carbon cycling
in savanna ecosystems (Mitchell et al. 1999, Kirk-
man et al. 2016a). However, the degree to which
fires interact with understory regrowth, espe-
cially during drought, has yet to be fully under-
stood. How fire alters the overall carbon fluxes of
the ecosystem, as a function of understory
regrowth, varies substantially across different
ecosystems. A study conducted in a longleaf pine
ecosystem found that physiological activity and
hence productivity returned to pre-fire levels
only 30–60 d post-fire (Starr et al. 2015). A sepa-
rate study in a mixed longleaf slash pine forest in
north central Florida (USA) suggested that 70%
of understory N and C pools are consumed by
managed burns, but recover within three years
post-fire (Lavoie et al. 2010). These studies, how-
ever, were short in nature—following a single
fire cycle, showing the need for longer term
observations to cover greater environmental con-
ditions and multiple fire cycles.
Longleaf pine savannas of the southeastern
United States are ideal for these studies since
their canopies are open and dominated by lon-
gleaf pine, Pinus palustris (Mill.), while their
understories host a remarkably high quantity of
❖www.esajournals.org 2April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
species when compared to other temperate
ecosystems (Walker and Peet 1984, Kirkman
et al. 2004). Up to 50 species per m
2
can be
found within the understory, mostly on poorly
drained soil (Walker and Peet 1984, Kirkman
et al. 2001), whereas drier and sandier soils often
have lower biodiversity in the understory
(Goebel et al. 2001, Kirkman et al. 2001). The rel-
ative species diversity in longleaf pine savannas
is often attributed to differences in nutrient
availabilities in the soil, as more mesic sites usu-
ally have higher proportions of clay in the soil,
which has been shown to drastically lower nutri-
ent availability (Metz 1952, Christensen 1977,
Pellegrini 2016).
Quantifying understory phenology, as a func-
tion of differences in understory composition,
and its contribution to regional and global car-
bon budgets can be tedious and time-consuming
and may even be problematic (Johnson et al.
2017). However, the Normalized Difference
Vegetation Index (NDVI; Boke-Ol
en et al. 2016)
has been demonstrated to be advantageous in
determining the magnitude of ecosystem carbon
and energy fluxes (Potter 1999), as well as in
quantifying understory phenology (Glenn et al.
2008). For example, NDVI has been used to
extract phenophases (greenup, maturity, senes-
cence, and dormancy; Zhang et al. 2003) and pre-
dict leaf phenological patterns (Xu et al. 2014) for
temperate deciduous trees. NDVI was also suc-
cessfully used to quantify phenology patterns at
FLUXNET sites; however, the accuracy
depended on the plant functional types present
at each site (Balzarolo et al. 2016). Other studies
have used NDVI estimates to quantify food
availability and nutritional quality for avian her-
bivores, showing that using handheld instru-
ments to measure NDVI had a higher accuracy
compared to satellite-derived measures (Hogrefe
et al. 2017).
The spatial resolution of NDVI measures is
critical for interpretation of ecological relevance.
Studies have found that utilizing temporal and/or
spatial resolutions that are too coarse lowers pre-
diction accuracy when studying interannual phe-
nological patterns or biomass productivity (Gatis
et al. 2017). The course nature of satellite-derived
NDVI measures in evergreen forests has been
shown to be a better measure of ecosystem struc-
ture and function, rather than gross ecosystem
productivity (GEP; Restrepo-Coupe et al. 2016).
These studies highlighted the usefulness of
NDVI measures and its limitations with respect
to coarse spatial scales. This underlines the need
for increased use of ground-based remote sens-
ing techniques, especially when site-specific char-
acteristics play an important role in the
productivity of the ecosystem of interest
(Hogrefe et al. 2017).
Using sub-canopy remote sensed products, the
overall objective of this study was to determine
the contribution of understory phenology to
ecosystem CO
2
fluxes at the ends of a soil mois-
ture gradient with respect to prescribed fire and
seasonal drought. We used monthly understory
NDVI measurements to quantify productivity at
two longleaf pine sites representing the ends of
this gradient (mesic and xeric). We hypothesized
that (1) the xeric site would recover more rapidly
from drought as it evolved under lower soil
moisture conditions; (2) fire would lower under-
story NDVI at both sites, with the mesic site
understory recovering more rapidly post-fire due
to higher soil water availability; and (3) under-
story NDVI would be associated with increased
ecosystem productivity at both sites, with the
xeric site exhibiting greater influence of the
understory on overall carbon fluxes due to its
more open canopy.
MATERIALS AND METHODS
Study site
The study was conducted from August 2011
through to March 2016 in southwest Georgia,
USA (31.2201°N, 84.4792°W), within the South-
eastern Coastal Plain, at the Joseph W. Jones Eco-
logical Research Center (JWJERC; Fig. 1). The
area receives on average 1310 mm of rain annu-
ally and is climatically categorized as humid sub-
tropical (Christensen 1981, Goebel et al. 2001).
The long-term average monthly air temperatures
range from 9.6°C to 27.6°C in January and July,
respectively.
Locations for this study were established in
two sites that represent the xeric and mesic ends
of an edaphic gradient in longleaf pine ecosys-
tems at the JWJERC and are located ~5.0 km
from one another (Fig. 1; Kirkman et al. 2001).
The xeric site was characterized by deep sandy
soils classified as typic quartzipsamments
❖www.esajournals.org 3April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
without an argillic horizon in the upper 300 cm
of soil, with a water-holding capacity of approxi-
mately 18 cm of water per meter of soil, because
of the high porosity and underground drainage
in the soil (Goebel et al. 2001, Kirkman et al.
2001). The mesic site had a soil composition that
consisted of sandy loam that rested on top of
either clay or sandy clay loam soils. The mesic
Fig. 1. Land cover map of the Joseph W. Jones Ecological Research Center in southeast Georgia, USA, showing
the (A) mesic and (B) xeric sites. Blue circles show eddy covariance tower locations, and orange circles are the
tower footprints at the three sites.
❖www.esajournals.org 4April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
site soils were not well drained and characterized
as arenic paleudults. The argillic horizon of these
soils was within ~95 cm, and the water-holding
capacity was 40 cm per m of soil in the upper
300 cm (Goebel et al. 2001, Kirkman et al. 2001).
The relatively small discrepancies between the
site characteristics contributed to a large distinc-
tion between the sites’species composition and
vegetation structure (Kirkman et al. 2001).
Both sites were comprised of an open longleaf
pine canopy in the overstory and wiregrass (Aris-
tida beyrichana Trin.) in the understory, while the
composition of secondary species varied by site
(Kirkman et al. 2001). Both ecosystems were
woodland systems with basal areas of 11.1 and
18.4 m
2
/ha at the xeric and mesic sites, respec-
tively, of which ~18% were oak trees at the xeric
site versus 6% at the mesic site. Due to differ-
ences in soil drainage, understory species rich-
ness was significantly higher at the mesic site
compared to the xeric site (Kirkman et al. 2016a).
The understory of the xeric site, however, had a
higher density of native hardwoods relative to
the mesic site, with the scrub oak species Quercus
laevis Walt. and Quercus margaretta Ashe found
most prevalently. The mesic site had a lower con-
centration of oaks present, and instead, a higher
population of common persimmon Diospyros vir-
giniana L. found most frequently in the under-
story (Goebel et al. 2001). The overstory leaf area
index (LAI) for the mesic site ranged from 0.65 to
1.1 m
2
/m
2
, while at the xeric site, it ranged from
0.22 to 0.65 m
2
/m
2
(Addington et al. 2006,
Wright et al. 2013). Both the mesic and xeric sites
have been thoroughly described in previously
published works by Mitchell et al. (1999), Kirk-
man et al. (2001), Ford et al. (2008), and Starr
et al. (2015).
Four productivity plots, 50 950 m in size,
were established prior to this study at each site.
These plots were located within the flux footprint
of each site’s eddy covariance tower (~500 m
radius from the tower). Understory composition
at both sites was estimated annually during each
fall within the four plots, with the exception of
2014 and 2016, due to a change in the sampling
protocol from annual measurements to 2-yr mea-
surement cycles in 2013. The understory biomass
was estimated using 0.75-m
2
clip plots. A plot
frame was randomly tossed from each of seven
litter trap positions, which had been previously
established within the plots at the two sites,
resulting in a total of 28 replicates at each site.
Where the frames touched the forest floor, all live
and dead vegetation smaller than 1 meter in
height was clipped and brought to the laboratory
for analyses. The vegetation was then classified
by growth form into forbs, ferns, legumes, other
grasses, and woody plants. The biomass from
the plots was then dried to a constant weight.
Drought indices
The Palmer Drought Severity Index (PDSI) for
the study period was obtained from the National
Oceanic Atmospheric Administration (NOAA)
and the National Centers for Environmental
Information (NCEI), Ashville, North Carolina,
USA (National Oceanic and Atmospheric
Administration, Department of Commerce n.d.).
The index describes severity of a drought on a
scale from 5to+5, where more negative num-
bers indicate more severe drought. The index is
computed using air temperature, rainfall, and
soil moisture measurements at climate stations
around the globe (Palmer 1965, Dai et al. 2004).
Prescribed burns
The two sites were burned frequently prior to
the study, averaging a two-year fire return inter-
val since 1994 with the last burn prior to the
study occurring in 2009. During our study, the
sites experienced 3 burns, during springs of 2011,
2013, and 2015. Strip head fires were set every
30–50 m (depending on the local fire conditions
at the time of the burn) starting downwind of the
management unit and moving upwind (Hiers
et al. 2009). No damage to overstory trees was
recorded following the prescribed fires, and as
fires typical to this surface fire regime are low
intensity (O’Brien et al. 2016), flame heights are
typically below 2 m, thus unable to reach over-
story crowns which are on average 23 m above
the forest floor (Robertson and Ostertag 2007,
Varner et al. 2007, Whelan et al. 2013). While the
two study sites had prescribed fires on a two-
year cycle, there was no control site that was
absent of prescribed fires, due to rapid ecosystem
movement toward an alternative stable state
when deprived of prescribed fires (O’Brien et al.
2010). If prescribed fire is withheld for as little as
4 yr, the ecosystem structure, composition, and
function change (Kirkman et al. 2013).
❖www.esajournals.org 5April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
In addition to data collected describing under-
story vegetation composition, we sampled
aboveground litter and understory biomass
before and after the fires, using methods
described in Ottmar et al. (2007). The number of
clip plots sampled at the two sites varied from 10
to 20, such that a standard error of the mean of
~15% was achieved (Whelan et al. 2013). In 2011,
2013, and 2015, we collected aboveground bio-
mass within two weeks pre- and post-fire using
paired 0.75-m
2
clip plots at the mesic and xeric
sites (Table 1). Clip plots at both sites were
located along a transect and sampled every 25 m
starting at the base of each flux tower and
extending within the flux footprint (~500 m) in
the direction of the prevailing wind (Whelan
et al. 2013). Harvested clip plot litter and bio-
mass were dried to a constant weight and mass,
and fuel consumption was calculated as the dif-
ference of pre-burn and post-burn dry weight
(Whelan et al. 2013).
NDVI measurements
Understory NDVI was measured monthly
using a Tetracam ADC camera attached to a tri-
pod, which was placed consistently at 1.60 m
above each subplot for each measurement.
Because the xeric site was spatially more com-
plex than the mesic site, four plots were estab-
lished within the footprint at the xeric site,
whereas the mesic site was sufficiently repre-
sented by only three plots. In both sites, four 1-
m
2
subplots were located 10 m from the center of
each plot, one in each of the four cardinal direc-
tions (north, east, south, and west). Because the
camera field of view could not span the entire
1m
2
, four pictures were taken in each of the four
subplot corners, which were processed electroni-
cally and averaged during the post-processing
procedure. Before each subplot measurement,
the camera was calibrated using a white Teflon
calibration chip, to account for changes in reflec-
tive values due to weather conditions. Post-pro-
cessing was performed using the Pixelwrench II
software (Tetracam, Chatsworth, California,
USA), which calculates understory NDVI from
the red (Red) and near-infrared (NIR) reflectance
values of each image as
NDVI ¼ðNIR RedÞ=ðNIR þRedÞ:
Ecosystem flux measurements
Eddy covariance towers, positioned at each
site, measured net ecosystem exchange of CO
2
(NEE), as well as meteorological variables such
as rainfall, air temperature (T
air
), and photosyn-
thetic active radiation (PAR) above the canopy
(Whelan et al. 2013). In addition, soil moisture
(SWC) and soil temperature, as well as ground
heat flux, were measured in a soil array adjacent
to the tower. NEE was measured via an
open-path eddy covariance method through a
simplification of the continuity equation and
converted from lmolm
2
s
1
to g C/m
2
per
30 min. Missing half-hourly data were gap-filled
using separate functions for NEE during day-
time and nighttime. When photosynthetically
active radiation (PAR) was ≥10 lmolm
2
s
1
,
daytime NEE data were gap-filled using a
Michaelis–Menten approach, and when PAR was
<10 lmolm
2
s
1
, nighttime NEE data were
gap-filled using a modification of the Lloyd and
Taylor (1994) approach, both on a monthly basis.
Where too few observations were available to
produce stable and biologically reasonable
parameter estimates, annual equations were used
to gap-fill data by site. Gross ecosystem exchange
(GEE) was calculated from the difference of
NEE and ecosystem respiration (R
eco
) as GEE =
NEE +R
eco
. NEE, GEE, and R
eco
were then
summed up to monthly estimates. Daytime R
eco
was estimated using nighttime gap-filling equa-
tion to partition NEE. A detailed explanation of
the equipment and flux processing method can be
found in Starr et al. (2016) and Whelan et al.
(2013). Finally, we converted NEE and GEE to net
ecosystem productivity (NEP) and gross ecosys-
tem productivity (GEP), where positive values
Table 1. Understory biomass and fuel consumption
pre- and post-prescribed burns in 2011, 2013, and
2015 at the mesic and xeric sites.
Year Site
Dry weight
pre-burn
(g/m
2
)
Dry weight
post-burn
(g/m
2
)
Fuel
consumption
(g/m
2
)
2011 Mesic 1052.2 334.9 717.3
Xeric 780.5 425.7 354.8
2013 Mesic 957.7 523.5 434.2
Xeric 937.0 917.6 19.4
2015 Mesic 1643.1 440.8 1202.3
Xeric 1808.1 668.0 1140.1
❖www.esajournals.org 6April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
indicate C uptake by the ecosystem and negative
values C release to the atmosphere.
Statistical analysis
Prior to the analysis of NDVI and ecosystem
fluxes, we formulated simple linear mixed models
to quantify differences in SWC, T
air
,andPAR
between the two sites, using site, year, fire-cycle
time (FCT), and their interactions as independent
variables. The factor FCT was included to describe
the amount of time that had passed since pre-
scribed burning. Following Whelan et al. (2013),
FCT1,2,and3correspondedtothefirst, second,
and third 28 d following fire, respectively. These
categories comprised spring and the beginning of
summer. FCT 4 and 5 represented the next 140
and further 225 d following FCT 3, encompassing
late summer and fall. Pre-fire (pf) indicated the
time before the next fire (~10 months). We used
the lme function from the package nlme in R ver-
sion 3.4.3 (R Core Team 2013), including a ran-
dom effect to account for monthly repeated
measurements and the subplot design, as well as
an autocorrelations structure to account for the
correlation between adjacent time points.
We then estimated linear mixed models for
each of the response variables: monthly mean
NDVI and monthly sums of NEP, GEP, and R
eco
.
For the NDVI model, we included fixed effects
for site, PDSI, FCT, mean monthly PAR, and
mean SWC, as well as the interactions of site
with PDSI and FCT, FCT and PAR, and FCT.
The NEP, GEP, and R
eco
models included fixed
effects for year, month, FCT, site, PDSI, and the
interactions of site with FCT, PDSI, and NDVI,
as well as the interaction between FCT and
NDVI. Mean PAR and SWC were also included
as fixed effects in these models. We also tested
the models using T
air
instead of PAR, which
resulted in similar trends of the response vari-
ables, as well as similar AIC (Akaike informa-
tion criterion) values. We could not include both
variables as their high correlation (>0.8) would
have led to multicollinearity in the independent
variables. We chose PAR for our final models to
quantify changes in NDVI and ecosystem fluxes
in response to changes in incoming energy. For
each response variable, we used a modified
backward selection criterion to choose the best,
most parsimonious model. Non-significant
(P>0.05) variables were eliminated, as long as
it resulted in a lower (better) AIC value. Where
interactions were significant, the underlying
simple effects were retained in the model
regardless of significance. We examined the nat-
ure of significant differences of all interactions
using the lsmeans function in R (Lenth 2016),
whereby marginal means are computed holding
all other independent variables at their average
values.
RESULTS
Annual rainfall was ~800 mm during 2011 and
2012 (Fig. 2A), lowering the PDSI (Fig. 2A) to
<4. Annual rainfall sums increased to
~1500 mm during 2013, which lead to an
increase in PDSI to >0, followed by another
decrease during 2015 (Fig. 2A). Soil moisture
was significantly lower at the xeric site compared
to the mesic site for all years and levels of FCT.
Additionally, SWC was significantly lower at
both sites during the drought (Fig. 2B), while
SWC returned to pre-drought levels at the end of
2012 as rainfall recovered (Fig. 2A). Mean
monthly PAR decreased during the wetter years,
but this difference was not significant (Fig. 2C).
Average PAR was significantly higher at the xeric
site (by ~30–60 lmolm
2
s
1
) compared to the
mesic site for all years except 2015, when
PAR was significantly higher at the mesic site (by
15 lmolm
2
s
1
). Furthermore, PAR was signifi-
cantly higher at the xeric site compared to the
mesic site for all levels of FCT, except for pre-fire.
Values of T
air
were higher (>30°C) during 2011
and then decreased below 30°C during summer
of the following years (Fig. 2D). Understory
NDVI during the growing season in 2011 was
lower at both sites with gradual increases in sub-
sequent years (Fig. 3A) and further increases in
the summer months of 2014 and 2015. During
winter months, NEP and GEP were lower at the
xeric site compared to the mesic site (Fig. 2B, C).
R
eco
decreased slightly at both sites during 2015,
but GEP was only slightly lower during the win-
ter months of that year (Fig. 3D).
Understory NDVI phenology under drought and
fire
Forbs, grasses, and wiregrass comprised a
greater proportion of the understory at the mesic
site, specifically outside of drought years,
❖www.esajournals.org 7April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
whereas woody understory species were more
abundant at the xeric site, especially during the
drought (Fig. 4).
Model results indicated that NDVI varied signifi-
cantly by site, but its effect depended on SWC, PAR,
PDSI, and time since fire (FCT). Pre-fire (pf) under-
story NDVI was slightly higher at the mesic site
(~0.48), compared to the xeric site (~0.43, Table 2,
Fig. 5). Following the prescribed fire, NDVI at the
xeric site decreased during FCT 1 (~0.26). How-
ever, no significant change was observed from
FCT 1 through FCT 3 (28–54 d post-fire, Fig. 5A).
NDVI at the mesic site also decreased to ~0.41–
0.43 during FCT 1 and FCT 2, followed by an
increase in FCT 3 to ~0.47 and a further increase
during FCT 4 to 0.66, which was significantly dif-
ferent from NDVI during FCT 2 (Fig. 5). During
FCT 5, NDVI at the mesic site significantly
decreased to 0.53. NDVI at the xeric site signifi-
cantly increased during FCT 5 (~0.66), followed
by a significant decrease in FCT 5 to ~0.42.
NDVI significantly decreased with an increase
in PDSI at both sites, with less of a decrease at the
xeric site (Fig. 5B). In addition, the xeric site had
significantly lower NDVI for all levels of PDSI.
NDVI significantly decreased for the levels of
FCT 5 and pre-fire but showed no significant
change for FCT 1–3, when average PAR increased
to 500 lmolm
2
s
1
. For the interaction of FCT 4
and PAR, NDVI significantly decreased by ~0.5
when PAR increased, but exceeded 1, when PAR
was below 100 lmolm
2
s
1
(Fig. 5). NDVI
decreased by 0.3 with increases in SWC (from
17% to 24%), but these effects were not signifi-
cantly different by site.
Understory NDVI effects on NEP
Net ecosystem productivity was significantly
different by site and was driven by NDVI, FCT,
PAR, time since fire, and PDSI (Table 3). The
mesic site NEP significantly decreased with
increases in NDVI compared to the xeric site
Fig. 2. Monthly (A) summed rainfall (mm) and Palmer Drought Severity Index (PDSI), that is, gray line with its
corresponding y-axis on the right-hand side of panel (A), mean (B) soil water content (SWC; %), (C) photosynthetic
active radiation (PAR; lmols
1
m
2
), and (D) air temperature (T
air
;°C) from August 2011 through March 2016.
❖www.esajournals.org 8April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
Fig. 3. Monthly (A) mean values of the Normalized Difference Vegetation Index (NDVI) and summed (B) net
ecosystem productivity (NEP), (C) gross ecosystem productivity (GEP), and (D) ecosystem respiration (R
eco
)ing
C/m
2
from August 2011 through March 2016. Arrows indicate the time of prescribed fire at the mesic and xeric
sites.
Fig 4. Annual understory live biomass by plant type at the mesic and xeric sites from 2011 to 2015.
❖www.esajournals.org 9April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
(Fig. 6A, Table 3). Monthly NEP was signifi-
cantly lower at the xeric compared to the mesic
site pre-fire but increased (greater carbon uptake)
following the fire (Fig. 6B). NEP at the mesic site
significantly increased following the fires during
FCT 1 and 2 (up to 40 g C/m
2
per month). Fol-
lowing FCT 2, NEP decreased at the mesic site to
values similar to those of the xeric site (20–10 g
C/m
2
per month). Both sites were weaker carbon
sinks when drought increased (Fig. 6C). NEP sig-
nificantly increased when PDSI moved toward
zero, with the mesic site being a greater carbon
sink than the xeric site.
As NDVI increased, so did NEP during FCT 2
and FCT 3, which was statistically significant.
During FCT 4, the opposite pattern was
observed, where higher NDVI resulted in lower
NEP (Fig. 6D). Photosynthetically active radia-
tion significantly increased the carbon sink
strength at both sites, with no significant differ-
ence between the mesic and xeric sites (Fig. 7A).
Understory NDVI effects on GEP
Similar to that of NEP, there was a significant
difference in GEP by site which was correlated
with NDVI, FCT, and PDSI (Table 3). There was
also a significant effect of NDVI with FCT and a
simple effect of PAR, which was not significantly
different by site (Table 3). When NDVI was
below 0.5, monthly GEP at the mesic site was sig-
nificantly higher (160 g C/m
2
per month) than
the xeric site (145 g C/m
2
per month). When
NDVI (>0.75) increased, GEP decreased slightly
at the mesic site, but increased at the xeric site
such that it had higher productivity than the
mesic site (Fig. 8A).
Preceding the prescribed burn GEP was not
significantly different at the xeric site, and this
persisted through FCT 2, although NDVI was
lower (Fig. 8B). During FCT 3, GEP increased
from 140 to 160 g C/m
2
per month at the xeric
site, which then remained at that rate for FCT 4
and FCT 5. The mesic site experienced a slight
decrease in GEP during FCT 1, followed by an
increase to approximately 170 g C/m
2
per
month and a steep decrease in GEP during
FCT 3 and FCT 4 (140 g C/m
2
per month).
Then, in FCT 5, the mesic site increased its pro-
ductivity to roughly 155 g C/m
2
per month.
Higher NDVI significantly increased GEP
Table 2. Type 3 tests of fixed effects for the model of
the Normalized Difference Vegetation Index
(NDVI).
Effect v
2
value df P-value
(Intercept) 43.59 1 <0.0001
PDSI 1.042 1 0.307
Site 47.17 1 <0.0001
FCT 18.58 5 0.002
SWC 68.57 1 <0.0001
PAR 19.84 1 <0.0001
Site:FCT 36.06 5 <0.0001
Site:SWC 115.98 1 <0.0001
PDSI:Site 5.42 1 0.0198
Notes: df, degrees of freedom (numerator). P-value
corresponds to chi-squared statistic with df.
Fig. 5. Least square mean values of Normalized Difference Vegetation Index (NDVI) by site and (A) fire-cycle
time (FCT) and (B) Palmer Drought Severity Index (PDSI; 5 to 5), and by (C) FCT and photosynthetic active
radiation (PAR). Error bars indicate 1 standard error (SEM). FCT 1, 2, and 3 correspond to the first, second, and
third 28 d following fire. FCT 4 and 5 represented the next 140 and further 225 d following FCT 3. Pre-fire (pf)
indicated the time before the next fire (~10 months).
❖www.esajournals.org 10 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
during pre-fire, and FCT 2 and 3, but decreased
its magnitude during FCT 4 across both sites
(Fig. 8D). Interestingly, GEP increased with
decreasing PAR and PDSI at both sites (Figs. 7B
and 8C).
Understory NDVI effects on R
eco
Model results indicated that R
eco
was signifi-
cantly different by site and also depended on
PDSI, FCT, and NDVI. Increases in understory
NDVI increased R
eco
at the mesic site, but only
slightly changed the magnitude of R
eco
at the
xeric site. These changes in R
eco
, however, were
not significant across NDVI levels (Fig. 9A). R
eco
was significantly lower at the mesic site for all
levels of NDVI, compared to the xeric site.
Fire affected R
eco
differently by site (Fig. 9B).
R
eco
significantly decreased at the mesic site,
following the fire during FCT 1 (122 g C/m
2
per
month), slightly increased during FCT 2 (127 g
C/m
2
per month) and decreased further during
FCT 3 and 4 (~120 g C/m
2
per month). R
eco
at the
xeric site decreased only slightly during FCT 1–2,
followed by an increase to ~145 g C/m
2
per
month during FCT 3 and 4. The mesic site
increased its R
eco
during FCT 5, back to ~135 g
C/m
2
per month, whereas the xeric site decreased
R
eco
to ~140 C/m
2
per month. Negative values of
PDSI increased R
eco
at both sites. However, this
was more pronounced at the xeric site (Fig. 9C).
For levels of PDSI >3, both sites had monthly
R
eco
rates of ~120 g C/m
2
per month, with signifi-
cantly higher rates at the xeric site throughout
this study. Increases in PAR significantly
decreased the magnitude of R
eco
, with no signifi-
cant difference by site (Fig. 7C).
DISCUSSION
Our study demonstrates that understory vege-
tation in longleaf pine ecosystems plays a com-
plex role in patterns of carbon uptake and
release. The understory phenology and composi-
tion contribute significantly to the overall pro-
ductivity of this ecosystem. More importantly,
our results reveal that site characteristics play a
critical role in how understories mediate ecosys-
tem recovery from drought and prescribed fire.
We show that changes in understory biomass
(higher NDVI) significantly affected GEP during
recovery from fire. We found increased photo-
synthetic capacity between two and three
months after the fires when understory biomass
was high. However, this relationship shifted after
four months, resulting in a decrease in photosyn-
thetic activity when NDVI was greater (Jurik
1986, Hikosaka 2005). This altered response high-
lights the need to incorporate understory phenol-
ogy when studying ecosystem recovery and
resilience in open forest stands. It further demon-
strates that the capacity of ecosystems to seques-
ter carbon depends on changes in understory
biomass over the course of the year (Chen et al.
2015).
For example, minor decreases in GEP at the
mesic site occurred with increasing NDVI indi-
cated that the understory contributed less to GEP
and that the overstory was the primary driver of
ecosystem productivity at the site (Yang et al.
Table 3. Type 3 tests of fixed effects for the models of
net ecosystem productivity (NEP), gross ecosystem
productivity (GEP), and ecosystem respiration
(R
eco
).
Model Effect v
2
value df P(>v
2
)
NEP NDVI >0.01 1 0.965
FCT 61.91 5 <0.0001
PDSI 2.28 1 0.131
PAR 97.80 1 <0.0001
Site 1229.33 1 <0.0001
NDVI:Site 102.33 1 <0.0001
PDSI:Site 22.98 1 <0.0001
NDVI:FCT 13.21 5 0.022
FCT:Site 311.50 5 <0.0001
GEP NDVI 2.69 1 0.101
FCT 57.93 5 <0.0001
PDSI 0.49 1 0.482
PAR 15.08 1 <0.0001
Site 705.64 1 <0.0001
NDVI:Site 53.22 1 <0.0001
PDSI:Site 3.87 1 0.049
NDVI:FCT 13.16 5 0.022
FCT:Site 552.07 5 <0.0001
R
eco
NDVI 9.37 1 0.002
FCT 13.50 5 0.019
PDSI 1.59 1 0.208
SWC 2.49 1 0.115
PAR 447.69 1 <0.0001
Site 70.56 1 <0.0001
NDVI:Site 9.80 1 0.002
PDSI:Site 122.72 1 <0.0001
FCT:Site 466.00 5 <0.0001
Notes: df, degrees of freedom (numerator). P-value
corresponds to chi-squared statistic with df.
❖www.esajournals.org 11 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
2017). This resulted in lower photosynthetic
capacity at the site when NDVI was high, com-
pared to the xeric site, as it is dominated by lon-
gleaf pine and grass species in the understory
(~30%; Wiesner et al. 2018). Furthermore,
biomass and leaf area were higher at the mesic
site, compared to the xeric site, suggesting less
solar radiation was reaching the understory,
which in turn could decrease productivity rela-
tive to the xeric site (Starr et al. 2016). Even
Fig. 6. Least square mean values of net ecosystem productivity (NEP) by site and (A) Normalized Difference
Vegetation Index (NDVI), (B) fire-cycle time (FCT), and (C) the Palmer Drought Severity Index (PDSI), and by
(D) FCT and NDVI. Error bars indicate 1 standard error (SEM). FCT 1, 2, and 3 correspond to the first, second,
and third 28 d following fire. FCT 4 and 5 represented the next 140 and a further 225 d following FCT 3. Pre-fire
(pf) indicated the time before the next fire (~10 months).
Fig. 7. Least square mean values of (A) net ecosystem productivity (NEP), (B) gross ecosystem productivity
(GEP), and (C) ecosystem respiration (R
eco
) across levels of photosynthetic active radiation (PAR). Error bars indi-
cate 1 standard error (SEM).
❖www.esajournals.org 12 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
though basal area, leaf area, and understory
NDVI were lower at the xeric site, R
eco
was
higher with lower NDVI. These findings were
likely a result of greater maintenance respiration
of the oak dominated overstory during dor-
mancy (Amthor 1984, Wu et al. 2014). In addi-
tion, the understory at the xeric site had a
proportionally higher contribution to the flux of
Fig. 8. Least square mean values of gross ecosystem productivity (GEP) by site and (A) the Normalized Vege-
tation Difference Index (NDVI), (B) fire-cycle time (FCT), and (C) the Palmer Drought Severity Index (PDSI), and
by (D) NDVI and FCT. Error bars indicate 1 standard error (SEM). FCT 1, 2, and 3 correspond to the first,
second, and third 28 d following fire. FCT 4 and 5 represented the next 140 and further 225 d following FCT 3.
Pre-fire (pf) indicated the time before the next fire (~10 months).
Fig. 9. Least square mean values of R
eco
by site and (A) the Normalized Difference Vegetation Index (NDVI),
(B) fire-cycle time (FCT), and (C) the Palmer Drought Severity Index (PDSI). Error bars indicate 1 standard error
(SEM).
❖www.esajournals.org 13 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
GEP, compared to the mesic site. This effect was
a result of higher abundance of woody species in
the understory and therefore higher photosyn-
thetic activity during peak NDVI, as broadleaved
shrubs and trees are known to have higher pho-
tosynthetic capacity compared to grasses and
coniferous trees (Weaver and Mogensen 1919,
Klein et al. 2013, Renninger et al. 2015).
We show that longleaf pine savannas are resili-
ent to the disturbance of fire and drought, as a
result of their ability to rapidly mobilize non-
structural carbohydrates when photosynthate
supply is low, in agreement with a previous
study conducted at our sites (Aubrey and Teskey
2018). However, we found that the magnitude
and timescale of recovery was site-dependent.
The mesic site recovered more rapidly from the
disturbance of fire, possibly as a result of a
higher species diversity in the understory (Kirk-
man et al. 2016b), which increases the potential
of certain understory species to rescue ecosystem
function (Elmqvist et al. 2003, Starr et al. 2015,
Wiesner et al. 2018). Understory NDVI and NEP
at the mesic site increased substantially more
during the first three months following the fire,
indicating rapid plant regrowth in the under-
story, potentially as a result of more readily avail-
able nutrients following the fire (Lavoie et al.
2010). This is supported by a previous study con-
ducted at these sites, which found increased soil
respiration primarily at the mesic site for two
months after prescribed fire. Our findings and
the findings from Wiesner et al. (2018) suggest
that more rapid understory plant biomass
regrowth at the mesic site increased respiration
rates, as a result of higher maintenance and
growth respiration. The grass dominated under-
story at the site also likely accelerated its recov-
ery time, as grasses make use of stored and
available resources more rapidly compared to
woody species (Brewer 2011). Furthermore,
higher soil moisture conditions at the site
enabled faster resource acquisition (White et al.
2016), which enhanced plant biomass regrowth
in contrast to the xeric site.
On the contrary, lower understory NDVI at the
xeric site following the fire was likely a function
of lower fuel consumption compared to the
mesic site. Due to higher heterogeneity of the
understory vegetation on xeric sites, fire did not
spread as uniformly (Duncan et al. 2015). This
would also contribute to smaller magnitude in
biomass regrowth following the fires (Pinno and
Errington 2016), as reflected in the ecosystem
fluxes compared to the mesic site. Even though
average N mineralization levels in the upper
soils were found to be higher at xeric sites (Bor-
ing et al. 2004), the heterogeneity in fire effects
may have resulted in lower nutrient availability
within the upper soil layers post-fire, compared
to the mesic site. Furthermore, similar N mineral-
ization rates at mesic and xeric sites below the
upper soil surface (<15 cm), as well as higher
SWC and higher root biomass at the mesic site
(Hendricks et al. 2006), would enable greater
nutrient availability and uptake (Lavoie et al.
2010). Moreover, sandy soils at the xeric site can
promote leaching of nutrients (Nguyen and
Marschner 2013), thereby decreasing nutrient
availability and thus plant regrowth.
However, the xeric site experienced a substan-
tial increase in NDVI, GEP, and R
eco
4–8 months
following the prescribed burns. High R
eco
and
GEP at the xeric site were indicative of the sea-
sonal increase in greenness and therefore meta-
bolic activity of deciduous oaks (Jurik 1986,
Hikosaka 2005, Powell et al. 2005), thereby
increasing growth respiration, as more recent
photosynthates were invested into bud and leaf
expansion (Keel and Sch€
adel 2010, Alla et al.
2013, Kuster et al. 2014, Herrmann et al. 2015).
In support of this notion, we found a decrease in
NEP for increasing NDVI during that time, sug-
gesting that more plant biomass in the under-
story increased respiration. In contrast, three
months following the burn, R
eco
was lower at the
mesic site, but GEP continued to decrease for
FCT 4, which also resulted in lower carbon sink
capacity at the site. A similar response for the
mesic and xeric sites was found in Whelan et al.
(2013), who showed that the base respiration rate
decreased at the mesic site but increased at the
xeric site following the fires. Another explanation
for lower metabolic activity at the mesic site dur-
ing FCT 4 and 5 was that average T
air
was higher
(>20°C) compared to T
air
for FCT 1–3(<20°C, evi-
dent from Fig. 2D), which would cause stomata
to close to conserve water (Gonzalez-Benecke
et al. 2011, Samuelson et al. 2012). For example,
longleaf pine trees were shown to decrease their
growth rates in response to warmer summer
temperatures (Foster and Brooks 2001), whereas
❖www.esajournals.org 14 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
oak species are more resistant to higher T
air
(i.e.,
anisohydric response; Roman et al. 2015).
Understory species at the xeric site were
shown to be less drought resilient compared to
those of its overstory. Even though the xeric site
is chronically exposed to lower soil moisture con-
ditions, we found that understory NDVI was sig-
nificantly lower for all levels of PDSI compared
to the mesic site, showing that understory spe-
cies were not able to thrive at the level of that of
the mesic site under drought conditions. How-
ever, the magnitude of NDVI change (i.e., slope
of the regression) was lower at the xeric site
when severity of drought increased, indicating
that plant functional types at the site could buffer
drought conditions more effectively (Coble et al.
2017, Granda et al. 2018), such that understory
NDVI decreased by half as much as at the mesic
site. In addition, overall ecosystem productivity
at xeric sites was shown to recover more rapidly
from drought (Starr et al. 2016), compared to
more mesic sites, which was the result of
drought-resistant plant species in the over- and
understory (Wright et al. 2013). Oak components
have been shown to increase climate resilience
on xeric sites by facilitating longleaf pine regen-
eration (Loudermilk et al. 2016). We also showed
that differences in site productivity during
drought were mainly governed by differences in
photosynthetic activity, such that the xeric site
had similar levels of R
eco
over the range of PDSI
but could not maintain similar levels of GEP,
compared to the mesic site.
In this study, the variation in phenology of the
understory vegetation played an important role
in the ecosystem’s response to disturbance. The
effects of prescribed fire on NEP and post-fire
recovery were mediated not only by site type but
also by interactions with variation in climate
(PDSI). The ability to sustain relatively high pho-
tosynthetic capacity (and respiration) during
drought at both sites has profound consequences
in relation to future changes in climate, if
droughts become more severe and frequent
(IPCC 2014). However, prolonged drought could
change the system from a carbon sink to a
source, as R
eco
would increase over GEP, which
may lead to carbon starvation or hydraulic fail-
ure, depending on the severity of the event
(McDowell 2011, Zeppel et al. 2011, Klein 2015).
Furthermore, severe drought could cause a shift
in ecosystem structure, as a consequence of the
inability to manage these systems with pre-
scribed fire, as drier conditions limit prescription
options for managers (Mitchell et al. 2014,
Chiodi et al. 2018).
Future climate scenarios for the Southeastern
Coastal Plain include more frequent extreme
weather events, such as hurricanes and severe
droughts, as well as differences in the amount
and timing of precipitation (IPCC 2014). Longleaf
pine ecosystems have been hypothesized to exhi-
bit higher resistance to these projected changes,
leading to renewed efforts for their restoration.
The success of these efforts will bring to the fore-
front the importance of accurate and unbiased
estimates of their NEP.
Our results demonstrate large differences
within savanna ecosystems in the contribution of
the understory to ecosystem productivity and
recovery, highlighting the critical need to evalu-
ate how variation in savanna structure affects
their contribution to global estimates of NPP.
Furthermore, the inclusion of site-specific phe-
nology would improve predictions regarding
physiological patterns and ecosystem carbon
dynamics, as represented in global dynamic veg-
etation models (Boke-Ol
en et al. 2016). Further
studies are needed to quantify these differences
to fully understand how multiple disturbances
interact over a range of structural variations in
savanna ecosystems across the globe.
CONCLUSIONS
Understory phenology and fire are critical
drivers of productivity in many ecosystems
globally. We demonstrated that understory con-
tribution to ecosystem fluxes is not uniform
across environmental gradients within a savanna
ecosystem, which affected ecosystem recovery
times and overall productivity at the sites. As
wildfire activities increase under changing
climate (Westerling et al. 2006) and ecosystem
management efforts increasingly focus on savanna
ecosystems (Brudvig 2011, Noss et al. 2015),
understanding climate and fire controls on under-
story carbon assimilation across edaphic variation
will be critical for understanding ecosystem
resilience.
For the longleaf pine savannas in this study,
xeric sites characterized by lower overstory
❖www.esajournals.org 15 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
density were shown to have higher contributions
of the understory to overall ecosystem fluxes,
which increased their carbon sink capacity when
understory NDVI was high. Native hardwood
components to the understory of xeric sites
exerted strong influence of phenological patterns
observed in this study. More mesic and more pro-
ductive sites experienced lower variability in
ecosystem fluxes, as a result of the overstory dom-
inating ecosystem fluxes, due to high basal and
leaf area.
Site recovery from fire disturbance was context
specific to interannual rainfall. For example,
understories at more mesic sites recovered more
rapidly from prescribed fires, which increased
their carbon sink potential. In contrast, xeric sites
showed little changes in NDVI and therefore
NEP, GEP, and R
eco
following the fires, which
resulted in little change in their sink capacity. In
contrast, xeric site understories were less affected
by drought, but did not improve overall ecosys-
tem recovery from the disturbance.
Savanna ecosystems do not exhibit uniform
ecosystem productivity and recovery from dis-
turbance as a result of variations in structure
from the underlying geology, which should be
incorporated in global classifications of savanna
ecosystems.Wehighlighttheimportanceofincor-
porating site variations within savanna ecosystems
to accurately predict ecosystem function and recov-
ery from disturbances. Our results demonstrate
the need for more fine-scale studies, especially in
ecosystems with large structural variations, to
accurately predict global carbon fluxes.
ACKNOWLEDGMENTS
The authors thank the Forest Ecology Laboratory’s
personnel, with special thanks to Tanner Warren,
Andres Baron-Lopez, and Scott Taylor, for data collec-
tion and provision during the study at the Joseph W.
Jones Ecological Research Center. CS and GS acknowl-
edge support from the U.S. National Science Founda-
tion (DEB EF-1241881 and EF-1702996).
LITERATURE CITED
Addington, R. N., L. A. Donovan, R. J. Mitchell, J. M.
Vose, S. D. Pecot, S. B. Jack, U. G. Hacke, J. S.
Sperry, and R. Oren. 2006. Adjustments in hydrau-
lic architecture of Pinus palustris maintain similar
stomatal conductance in xeric and mesic habitats.
Plant, Cell and Environment 29:535–545.
Alla, A. Q., J. J. Camarero, and G. Montserrat-Mart
ı.
2013. Seasonal and inter-annual variability of bud
development as related to climate in two coexisting
Mediterranean Quercus species. Annals of Botany
111:261–270.
Amthor, J. S. 1984. The role of maintenance respiration
in plant growth. Plant, Cell and Environment
7:561–569.
Aubrey, D. P., and R. O. Teskey. 2018. Stored root car-
bohydrates can maintain root respiration for
extended periods. New Phytologist 218:142–152.
Balzarolo, M., et al. 2016. Matching the phenology of
Net Ecosystem Exchange and vegetation indices
estimated with MODIS and FLUXNET in-situ
observations. Remote Sensing of Environment
174:290–300.
Barlow, R. J., J. S. Kush, J. C. Gilbert, and S. M. Her-
man. 2015. Prescribed fire effects in a longleaf pine
ecosystem-are winter fires working? Page 551 in H.
A. Gordon, K. F. Connor, and J. D. Haywood, edi-
tors. Southern Silvicultural Research Conference.
United States Department of Agriculture, Ashe-
ville, North Carolina, USA.
Becknell, J. M., et al. 2015. Assessing interactions
among changing climate, management, and distur-
bance in forests: a macrosystems approach. BioS-
cience 65:263–274.
Bernhardt-R€
omermann, M., et al. 2011. Functional
traits and local environment predict vegetation
responses to disturbance: a pan-European multi-
site experiment. Journal of Ecology 99:777–787.
Boke-Ol
en, N., V. Lehsten, J. Ard€
o, J. Beringer, L. Eklundh,
T. Holst, E. Veenendaal, and T. Tagesson. 2016. Esti-
mating and analyzing savannah phenology with a
lagged time series model. PLoS ONE 11:1–15.
Bombelli, A., et al. 2009. An outlook on the Sub-Saharan
Africa carbon balance. Biogeosciences 6:2193–2205.
Bond, W. J., G. F. Midgley, and F. I. Woodward. 2003.
The importance of low atmospheric CO2 and fire
in promoting the spread of grasslands and savan-
nas. Global Change Biology 9:973–982.
Boring, L. R., J. J. Hendricks, C. A. Wilson, and R. J.
Mitchell. 2004. Season of burn and nutrient losses
in a longleaf pine ecosystem. International Journal
of Wildland Fire 13:443–453.
Bracho, R., G. Starr, H. L. Gholz, T. A. Martin, W. P.
Cropper, and H. W. Loescher. 2012. Controls on
carbon dynamics by ecosystem structure and cli-
mate for Southeastern US slash pine plantations.
Ecological Monographs 82:101–128.
Brewer, J. S. 2011. Disturbance-mediated competition
between perennial plants along a resource supply
gradient. Journal of Ecology 99:1219–1228.
❖www.esajournals.org 16 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
Brudvig, L. A. 2011. The restoration of biodiversity:
Where has research been and where does it need to
go? American Journal of Botany 98:549–558.
Chen, Y., Z. Liu, X. Rao, X. Wang, C. Liang, Y. Lin, L.
Zhou, X.-A. Cai, and S. Fu. 2015. Carbon storage
and allocation pattern in plant biomass among dif-
ferent forest plantation stands in Guangdong,
China. Forests 6:794–808.
Chiodi, A. M., N. S. Larkin, and J. M. Varner. 2018. An
analysis of Southeastern US prescribed burn
weather windows: seasonal variability and El Ni~
no
associations. International Journal of Wildland Fire
27:176–189.
Christensen, N. L. 1977. Fire and soil-plant nutrient
relations in a pine-wiregrass savanna on the coastal
plain of North Carolina. Oecologia 31:27–44.
Christensen, N. L. 1981. Fire regimes in southeastern
ecosystems. Pages 112–136 in H. A. Mooney, T. M.
Bonnicksen, N. L. Christensen, F. B. Lotspeich, and
W. A. Reiners, editors. Tech. cords. Proc. Confer-
ence on fire regimes and ecosystem properties.
Technical report WO-36, US Forest Service, Wash-
ington, D.C., USA.
Coble, A. P., M. A. Vadeboncoeur, Z. C. Berry, K. A.
Jennings, C. D. McIntire, J. L. Campbell, L. E. Rus-
tad, P. H. Templer, and H. Asbjornsen. 2017. Are
Northeastern U.S. forests vulnerable to extreme
drought? Ecological Processes 6:34.
Dai, A., T. Qian, A. Dai, K. E. Trenberth, and T. Qian.
2004. A Global Dataset of Palmer Drought Severity
Index for 1870–2002: relationship with Soil Mois-
ture and Effects of Surface Warming. Journal of
Hydrometeorology 5:1117–1130.
Duncan, B. W., P. A. Schmalzer, D. R. Breininger, and
E. D. Stolen. 2015. Comparing fuels reduction and
patch mosaic fire regimes for reducing fire spread
potential: a spatial modeling approach. Ecological
Modelling 314:90–99.
Elmqvist, T., C. Folke, M. Nystr€
om, G. Peterson, J.
Bengtsson, B. Walker, and J. Norberg. 2003.
Response diversity, ecosystem change, and resili-
ence. Frontiers in Ecology and the Environment
1:488–494.
Fei, X., et al. 2017. Eddy covariance and biometric
measurements show that a savanna ecosystem in
Southwest China is a carbon sink. Scientific
Reports 7:41025.
Field, C., M. Behrenfeld, J. Randerson, and P. Falk-
owski. 1998. Primary production of the biosphere:
integrating terrestrial and oceanic components.
Science 281:237–240.
Ford, C. R., R. J. Mitchell, and R. O. Teskey. 2008. Water
table depth affects productivity, water use, and the
response to nitrogen addition in a savanna system.
Canadian Journal of Forest Research 38:2118–2127.
Foster, T. E., and J. R. Brooks. 2001. Long-term
trends in growth of Pinus palustris and Pinus elliottii
along a hydrological gradient in central Florida.
Canadian Journal of Forest Research 31:1661–
1670.
Gatis, N., K. Anderson, E. Grand-Clement, D. J. Lus-
combe, I. P. Hartley, D. Smith, and R. E. Brazier.
2017. Evaluating MODIS vegetation products
using digital images for quantifying local peatland
CO
2
gas fluxes. Remote Sensing in Ecology and
Conservation 3:217–231.
Glenn, E. P., A. R. Huete, P. L. Nagler, and S. G. Nel-
son. 2008. Relationship between remotely-sensed
vegetation indices, canopy attributes and plant
physiological processes: What Vegetation Indices
can and cannot tell us about the landscape. Sensors
8:2136–2160.
Goebel, P. C., B. J. Palik, L. K. Kirkman, M. B. Drew, L.
West, and D. C. Pederson. 2001. Forest ecosystems
of a Lower Gulf Coastal Plain landscape: multifac-
tor classification and analysis. Journal of the Torrey
Botanical Society 128:47.
Gonzalez-Benecke, C. A., T. A. Martin, and W. P. Crop-
per. 2011. Whole-tree water relations of co-occur-
ring mature Pinus palustris and Pinus elliottii var.
elliottii. Canadian Journal of Forest Research
41:509–523.
Granda, E., A. Q. Alla, N. A. Laskurain, J. Loidi, A.
S
anchez-Lorenzo, and J. J. Camarero. 2018. Coex-
isting oak species, including rear-edge populations,
buffer climate stress through xylem adjustments.
Tree Physiology 38:159–172.
Hendricks, J. J., R. L. Hendrick, C. A. Wilson, R. J.
Mitchell, S. D. Pecot, and D. Guo. 2006. Assessing
the patterns and controls of fine root dynamics: an
empirical test and methodological review. Journal
of Ecology 94:40–57.
Herrmann, S., S. Recht, M. Boenn, L. Feldhahn, O.
Angay, F. Fleischmann, M. T. Tarkka, T. E. E.
Grams, and F. Buscot. 2015. Endogenous rhythmic
growth in oak trees is regulated by internal clocks
rather than resource availability. Journal of Experi-
mental Botany 66:7113–7127.
Hiers, J. K., J. J. O’Brien, R. J. Mitchell, J. M. Grego, and
E. L. Loudermilk. 2009. The wildland fuel cell con-
cept: an approach to characterize fine-scale varia-
tion in fuels and fire in frequently burned longleaf
pine forests. International Journal of Wildland Fire
18:315.
Hikosaka, K. 2005. Leaf canopy as a dynamic system:
ecophysiology and optimality in leaf turnover.
Annals of Botany 95:521–533.
Hogrefe, K., V. Patil, D. Ruthrauff, B. Meixell, M.
Budde, J. Hupp, and D. Ward. 2017. Normalized
Difference Vegetation Index as an estimator for
❖www.esajournals.org 17 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
abundance and quality of avian herbivore forage in
Arctic Alaska. Remote Sensing 9:1234–21.
IPCC. 2014. Climate Change 2014: Synthesis Report.
Page 151. in Core Writing Team, R. K. Pachauri,
and L. A. Meyer, editors. Contribution of Working
Groups I, II and III to the Fifth Assessment Report
of the Intergovernmental Panel on Climate Change.
IPCC, Geneva, Switzerland.
Johnson, K. D., G. M. Domke, M. B. Russell, B. Walters,
J. Hom, A. Peduzzi, R. Birdsey, K. Dolan, and W.
Huang. 2017. Estimating aboveground live under-
story vegetation carbon in the United States. Envi-
ronmental Research Letters 12:125010.
Jurik, T. W. 1986. Seasonal patterns of leaf photosyn-
thetic capacity in successional Northern hardwood
tree species. American Journal of Botany 73:131–
138.
Keel, S. G., and C. Sch€
adel. 2010. Expanding leaves of
mature deciduous forest trees rapidly become
autotrophic. Tree Physiology 30:1253–1259.
Kirkman, L. K., A. Barnett, B. W. Williams, J. K. Hiers, S.
M. Pokswinski, and R. J. Mitchell. 2013. A dynamic
reference model: a framework for assessing biodi-
versity restoration goals in a fire-dependent
ecosystem. Ecological Applications 23:1574–1587.
Kirkman, L. K., K. L. Coffey, R. J. Mitchell, and E. B.
Moser. 2004. Ground cover recovery patterns and
life-history traits: implications for restoration
obstacles and opportunities in a species-rich
savanna. Journal of Ecology 92:409–421.
Kirkman, L. K., L. M. Giencke, R. S. Taylor, L. R. Bor-
ing, C. L. Staudhammer, and R. J. Mitchel. 2016a.
Productivity and species richness in longleaf pine
woodlands: Resource-disturbance influences across
an edaphic gradient. Ecology 97:2259–2271.
Kirkman, L. K., P. C. Goebel, B. J. Palik, and L. T. West.
2016b. Predicting plant species diversity in a lon-
gleaf pine landscape. Ecoscience 11:80–93.
Kirkman, L. K., R. J. Mitchell, R. C. Helton, and M. B.
Drew. 2001. Productivity and species richness
across an environmental gradient in a fire-depen-
dent ecosystem. American Journal of Botany
88:2119–2128.
Klein, T. 2015. Drought-induced tree mortality: from
discrete observations to comprehensive research.
Tree Physiology 35:225–228.
Klein, T., I. Shpringer, G. Ben Fikler, S. Cohen Elbaz,
and D. Yakir. 2013. Relationships between stomatal
regulation, water-use, and water-use efficiency of
two coexisting key Mediterranean tree species. For-
est Ecology and Management 302:34–42.
Kuster, T. M., M. Dobbertin, M. S. G€
unthardt-Goerg,
M. Schaub, and M. Arend. 2014. A phenological
timetable of oak growth under experimental
drought and air warming. PLoS ONE 9:e89724.
Lavoie, M., G. Starr, M. C. M. Mack, T. A. T. Martin,
and H. L. H. Gholz. 2010. Effects of a prescribed
fire on understory vegetation, carbon pools, and
soil nutrients in a longleaf pine-slash pine forest in
Florida. Natural Areas Journal 30:82–94.
Lenth, R. V.. 2016. Least-squares means: the R Package
lsmeans. Journal of Statistical Software 69:1–33.
Lloyd, J., and J. A. Taylor. 1994. On the temperature
dependence of soil respiration. Functional Ecology
8:315–323.
Loudermilk, E. L., J. K. Hiers, S. Pokswinski, J. J.
O’Brien, A. Barnett, and R. J. Mitchell. 2016. The
path back: Oaks (Quercus spp.) facilitate longleaf
pine (Pinus palustris) seedling establishment in
xeric sites. Ecosphere 7:e01361.
Ma, X., et al. 2013. Spatial patterns and temporal
dynamics in savanna vegetation phenology across
the North Australian tropical transect. Remote
Sensing of Environment 139:97–115.
McDowell, N. G. 2011. Mechanisms linking drought,
hydraulics, carbon metabolism, and vegetation
mortality. Plant Physiology 155:1051–1059.
Metz, L. J. 1952. Weight and nitrogen and calcium con-
tent of the annual litter fall of forests in the South
Carolina Piedmont 1. Soil Science Society of Amer-
ica Journal 16:38–41.
Mitchell, R. J., J. K. Hiers, S. B. Jack, and R. T.
Engstrom. 2006. Silviculture that sustains: the
nexus between silviculture, frequent prescribed
fire, and conservation of biodiversity in longleaf
pine forests of the southeastern United States.
Canadian Journal of Forest Research 36:2724–2736.
Mitchell, R. J., L. K. Kirkman, S. D. Pecot, C. A. Wilson,
B. J. Palik, and L. R. Boring. 1999. Patterns and con-
trols of ecosystem function in longleaf pine-wire-
grass savannas. I. Aboveground net primary
productivity. Canadian Journal of Forest Research
29:743–751.
Mitchell, R. J., Y. Liu, J. J. O’Brien, K. J. Elliott, G. Starr,
C. F. Miniat, and J. K. Hiers. 2014. Future climate
and fire interactions in the Southeastern region of
the United States. Forest Ecology and Management
327:316–326.
Nguyen, T.-T., and P. Marschner. 2013. Addition of a
fine-textured soil to compost to reduce nutrient
leaching in a sandy soil. Soil Research 51:232–239.
Noss, R. F., W. J. Platt, B. A. Sorrie, A. S. Weakley, D. B.
Means, J. Costanza, and R. K. Peet. 2015. How glo-
bal biodiversity hotspots may go unrecognized:
lessons from the North American Coastal Plain.
Diversity and Distributions 21:236–244.
O’Brien, J. J., J. K. Hiers, R. J. Mitchell, J. M. Varner III,
and K. Mordecai. 2010. Acute physiological stress
and mortality following fire in a long-unburned
longleaf pine ecosystem. Fire Ecology 6:1–12.
❖www.esajournals.org 18 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
O’Brien, J. J., E. L. Loudermilk, B. Hornsby, A. T.
Hudak, B. C. Bright, M. B. Dickinson, J. K. Hiers,
C. Teske, and R. D. Ottmar. 2016. High-resolution
infrared thermography for capturing wildland fire
behaviour: RxCADRE 2012. International Journal
of Wildland Fire 25:62–75.
Ottmar, R. D., D. V. Sandberg, C. L. Riccardi, and S. J.
Prichard. 2007. An overview of the fuel characteris-
tic classification system—quantifying, classifying,
and creating fuelbeds for resource planning. Cana-
dian Journal of Forest Research 37:2383–2393.
Palmer, W. C. 1965. Meteorological drought. Volume
30. US Department of Commerce, Weather Bureau,
Washington, D.C., USA.
Palmer Drought Severity Index. n.d. National Oceanic
and Atmospheric Administration. Department of
Commerce, Washington, D.C., USA.https://www.
ncdc.noaa.gov/temp-and-precip/drought/historical-
palmers/
Patenaude, G., R. A. Hill, R. Milne, D. L. A. Gaveau, B.
B. J. Briggs, and T. P. Dawson. 2004. Quantifying
forest above ground carbon content using LiDAR
remote sensing. Remote Sensing of Environment
93:368–380.
Pellegrini, A. F. A. 2016. Nutrient limitation in tropical
savannas across multiple scales and mechanisms.
Ecology 97:313–324.
Peterson, D. W., and P. B. Reich. 2007. Fire frequency
and tree canopy structure influence plant species
diversity in a forest-grassland ecotone. Plant Ecol-
ogy 194:5–16.
Picotte, J. J., and K. M. Robertson. 2011. Validation of
remote sensing of burn severity in south-eastern
US ecosystems. International Journal of Wildland
Fire 20:453–464.
Pinno, B., and R. Errington. 2016. Burn severity domi-
nates understory plant community response to fire
in xeric jack pine forests. Forests 7:83.
Potter, C. S. 1999. Terrestrial biomass and the effects of
deforestation on the global carbon cycle. BioScience
49:769–778.
Powell, T. L., H. L. Gholz, K. L. Clark, G. Starr, W. P.
Cropper, and T. A. Martin. 2008. Carbon exchange
of a mature, naturally regenerated pine forest in
north Florida. Global Change Biology 14:2523–
2538.
Powell, T. L., G. Starr, K. L. Clark, T. A. Martin, and H.
L. Gholz. 2005. Ecosystem and understory water
and energy exchange for a mature, naturally regen-
erated pine flatwoods forest in north Florida. Cana-
dian Journal of Forest Research 35:1568–1580.
R Core Team. 2013. R: a Language and Environment
for Statistical Computing. Third edition. R Founda-
tion for Statistical Computing, Vienna, Austria,
Page 299.
Refsland, T. K., and J. M. Fraterrigo. 2017. Both canopy
and understory traits act as response–effect traits
in fire-managed forests. Ecosphere 8:e02036.
Renninger, H. J., N. J. Carlo, K. L. Clark, and K. V. R.
Sch€
afer. 2015. Resource use and efficiency, and
stomatal responses to environmental drivers of oak
and pine species in an Atlantic Coastal Plain forest.
Frontiers in Plant Science 6:317.
Restrepo-Coupe, N., A. Huete, K. Davies, J. Cleverly, J.
Beringer, D. Eamus, E. van Gorsel, L. B. Hutley,
and W. S. Meyer. 2016. MODIS vegetation products
as proxies of photosynthetic potential along a gra-
dient of meteorologically and biologically driven
ecosystem productivity. Biogeosciences 13:5587–
5608.
Rich, P. M., D. D. Breshears, and A. B. White. 2008.
Phenology of mixed woody-herbaceous ecosys-
tems following extreme events: net and differential
responses. Ecology 89:342–352.
Robertson, K. M., and T. E. Ostertag. 2007. Effects of
land use on fuel characteristics and fire behavior in
pinelands of Southwest Georgia. Pages 181–191 in
R. E. Masters and K. E. M. Galley, editors. 23rd Tall
Timbers Fire Ecology Conference: Fire in Grass-
land and Shrubland Ecosystems. Tall Timbers
Research Station, Tallahassee Florida, USA.
Roman, D. T., K. A. Novick, E. R. Brzostek, D. Dragoni,
F. Rahman, and R. P. Phillips. 2015. The role of iso-
hydric and anisohydric species in determining
ecosystem-scale response to severe drought.
Oecologia 179:641–654.
Samuelson, L. J., T. A. Stokes, and K. H. Johnsen. 2012.
Ecophysiological comparison of 50-year-old lon-
gleaf pine, slash pine and loblolly pine. Forest Ecol-
ogy and Management 274:108–115.
Simard, M., N. Pinto, J. B. Fisher, and A. Baccini. 2011.
Mapping forest canopy height globally with space-
borne lidar. Journal of Geophysical Research: Bio-
geosciences 116:248.
Smith, A. M. S., et al. 2014. Remote sensing the
vulnerability of vegetation in natural terrestrial
ecosystems. Remote Sensing of Environment 154:
322–337.
Starr, G., C. L. Staudhammer, H. W. Loescher, R.
Mitchell, A. Whelan, J. K. Hiers, and J. J. O’Brien.
2015. Time series analysis of forest carbon dynam-
ics: recovery of Pinus palustris physiology follow-
ing a prescribed fire. New Forests 46:63–90.
Starr, G., C. L. Staudhammer, S. Wiesner, S. Kunwor,
H. W. Loescher, A. F. Baron, A. Whelan, R. J. Mitch-
ell, and L. Boring. 2016. Carbon dynamics of Pinus
palustris ecosystems following drought. Forests
7:98.
Varner, J. M., J. K. Hiers, R. D. Ottmar, D. R. Gordon,
F. E. Putz, and D. D. Wade. 2007. Overstory tree
❖www.esajournals.org 19 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
mortality resulting from reintroducing fire to long-
unburned longleaf pine forests: the importance of
duff moisture. Canadian Journal of Forest Research
37:1349–1358.
Vaughn, N. R., G. P. Asner, I. P. J. Smit, and E. S. Rid-
del. 2015. Multiple scales of control on the structure
and spatial distribution of woody vegetation in
African savanna watersheds. PLoS ONE 10:
e0145192.
Walker, J., and R. K. Peet. 1984. Composition and spe-
cies diversity of pine-wiregrass savannas of the
Green Swamp, North Carolina. Vegetatio 55:163–
179.
Weaver, J. E., and A. Mogensen. 1919. Relative transpi-
ration of coniferous and broad-leaved trees in
autumn and winter. Botanical Gazette 68:393.
Westerling, A. L., H. G. Hidalgo, D. R. Cayan, and T.
W. Swetnam. 2006. Warming and earlier spring
increase western US forest wildfire activity. Science
5789:940–943.
Whelan, A., R. Mitchell, C. Staudhammer, and G. Starr.
2013. Cyclic occurrence of fire and its role in carbon
dynamics along an edaphic moisture gradient in
longleaf pine ecosystems. PLoS ONE 8:e54045.
White, A. C., A. Rogers, M. Rees, and C. P. Osborne.
2016. How can we make plants grow faster? A
source–sink perspective on growth rate. Journal of
Experimental Botany 67:31–45.
Wiesner, S., C. L. Staudhammer, H. W. Loescher, A.
Baron-lopez, L. R. Boring, R. J. Mitchell, and G.
Starr. 2018. Interactions among abiotic drivers, dis-
turbance and gross ecosystem carbon exchange on
soil respiration from Subtropical pine savannas.
Ecosystems 21:1639–1658.
Wright, J. K., M. Williams, G. Starr, J. McGee, R.
Addington, and R. J. Mitchell. 2013. Measured and
modelled leaf and stand-scale productivity across a
soil moisture gradient and a severe drought. Plant,
Cell & Environment 36:467–483.
Wu, J., Z. Liu, G. Huang, D. Chen, W. Zhang, Y. Shao,
S. Wan, and S. Fu. 2014. Response of soil respira-
tion and ecosystem carbon budget to vegetation
removal in Eucalyptus plantations with contrasting
ages. Scientific Reports 4:6262.
Wu, J., S. Wurst, and X. Zhang. 2016. Plant functional
trait diversity regulates the nonlinear response of
productivity to regional climate change in Tibetan
alpine grasslands. Scientific Reports 6:35649.
Xu, H., T. Twine, and X. Yang. 2014. Evaluating remo-
tely sensed phenological metrics in a dynamic
ecosystem model. Remote Sensing 6:4660–4686.
Yang, H., X. Yang, M. Heskel, S. Sun, and J. Tang. 2017.
Seasonal variations of leaf and canopy properties
tracked by ground-based NDVI imagery in a tem-
perate forest. Scientific Reports 7:1–10.
Zeppel, M. J. B., H. D. Adams, and W. R. L. Anderegg.
2011. Mechanistic causes of tree drought mortality:
recent results, unresolved questions and future
research needs. New Phytologist 192:800–803.
Zhang, X., M. A. Friedl, C. B. Schaaf, A. H. Strahler, J.
C. F. Hodges, F. Gao, B. C. Reed, and A. Huete.
2003. Monitoring vegetation phenology using
MODIS. Remote Sensing of Environment 84:471–
475.
❖www.esajournals.org 20 April 2019 ❖Volume 10(4) ❖Article e02675
WIESNER ET AL.
