Content uploaded by Jason Bried
Author content
All content in this area was uploaded by Jason Bried on Nov 03, 2020
Content may be subject to copyright.
BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit
publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to
critical research.
Influence of Long-term Greentree Reservoir Impoundment
on Stand Structure, Species Composition, and Hydrophytic
Indicators
Author(s): Gary N. Ervin, Lucas C. Majure, Jason T. Bried
Source: The Journal of the Torrey Botanical Society, 133(3):468-481. 2006.
Published By: Torrey Botanical Society
DOI: http://dx.doi.org/10.3159/1095-5674(2006)133[468:IOLGRI]2.0.CO;2
URL: http://www.bioone.org/doi/
full/10.3159/1095-5674%282006%29133%5B468%3AIOLGRI%5D2.0.CO
%3B2
BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the
biological, ecological, and environmental sciences. BioOne provides a sustainable online
platform for over 170 journals and books published by nonprofit societies, associations,
museums, institutions, and presses.
Your use of this PDF, the BioOne Web site, and all posted and associated content
indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/
terms_of_use.
Usage of BioOne content is strictly limited to personal, educational, and non-commercial
use. Commercial inquiries or rights and permissions requests should be directed to the
individual publisher as copyright holder.
Influence of long-term greentree reservoir impoundment on
stand structure, species composition, and hydrophytic indicators
1
Gary N. Ervin
2
, Lucas C. Majure, and Jason T. Bried
3
Department of Biological Sciences, PO Box GY, 130 Harned Biology,
Mississippi State University, MS 39762
E
RVIN
, G. N., L. C. M
AJURE
,
AND
J. T. B
RIED
(Department of Biological Sciences, PO Box GY, 130
Harned Biology, Mississippi State University, MS 39762). Influence of long-term greentree reservoir
impoundment on stand structure, species composition, and hydrophytic indicators. J. Torrey Bot. Soc. 133:
468–481. 2006.—Greentree reservoir management was initiated during a period when riparian forests of the
Mississippi Alluvial Valley were disappearing rapidly. Greentree reservoirs (GTRs) were intended to provide
a refuge for overwintering migratory waterfowl within a landscape of decreasing habitat availability.
However, GTRs frequently are flooded as much as 2.5 months longer than unimpounded bottomland
forests (UBF), and previous work has shown that such unnatural hydroperiods can have substantial negative
effects on such ecosystem attributes as tree assemblage composition and invertebrate production. We
conducted the present study to quantify (1) frequency of canopy gaps and (2) tree species composition in
GTRs, as compared with UBFs. In general, GTRs were quite similar to UBF stands, in terms of canopy
density, proportion of trees in the canopy versus mid-story, degree of stress exhibited by individual trees, gap
frequency, and species diversity. However, multivariate comparisons of GTRs versus unimpounded areas
indicated differences in species composition. Indicator Species Analyses and examination of the dominant
species showed clear differences between GTRs and UBFs, with GTR tree species being, on average, better
adapted to flooded conditions, based on wetland indicator status. Greentree reservoir canopies generally
were dominated by a Taxodium distichum –Acer rubrum –Quercus lyrata mix, whereas unimpounded forest
canopies were characterized as Liquidambar styraciflua – mixed Quercus stands. The midstories of the groups
were more similar to one another, with dominance by Acer–Carpinus caroliniana –Planera aquatica in
GTRs and Liquidambar –Carpinus –Acer in unimpounded stands. The GTRs included in this study had
been managed for 40 to 48 years, providing a long history of flooding which served to select for highly
adapted species assemblages.
Key words: bottomland forest, flood-adapted species, forested wetlands, GTR, multivariate analyses,
wetland management.
Bottomland forests (BF) in the southeastern
United States have faced marked reduction in
area and quality. Clearing of land for agricul-
tural purposes has reduced acreage of BF by
75%, from 8–10 million hectares to the 2–
3 million ha that remain in the lower Mis-
sissippi River Alluvial Valley alone, and
selective timber harvests have left many of
the remaining tracts of BF with poor quality
trees and severely degraded ecological function
(Reinecke et al. 1989, King and Allen 1996,
Kellison and Young 1997, Mitsch and Gosse-
link 2000). The hydrologic regimes in these
remaining BF vary widely, partly because of
the management of many tracts as greentree
reservoirs (GTRs) for migratory waterfowl
(Reinecke et al. 1989). Greentree reservoirs are
areas of BF that have been impounded by
levees for water retention. Water control
devices permit impoundment of water from
late autumn through early spring and sub-
sequent drawdown that are meant to parallel
natural BF hydrology. However, the continu-
ous, annual, extended periods of GTR flood-
ing are not representative of natural hydrolo-
gy, wherein a given ecosystem is flooded for
variable durations and at irregular intervals
within and among years. Furthermore, be-
cause of logistical limitations, many managers
are unable to remove water from GTRs at the
appropriate periods each year, resulting in up
1
This work was supported in part through grants
from the Mississippi State University Office of
Research and USGS grants 01HQGR0088 (WRRI)
and 04HQAG0135 (BRD) to GNE. The views and
conclusions contained in this document are those of
the authors and should not be interpreted as
necessarily representing the official policies, either
expressed or implied, of the U.S. Government.
2
Brook Herman, Cori Anderson, and Melissa
Smothers assisted with data collection for this
project. The following individuals provided helpful
commentary on earlier versions of this paper: A.
Ezell and R. M. Kaminski (MS State University), A.
Pierce (University of TN, Knoxville), B. Wehrle
(Noxubee NWR), and anonymous reviewers. E-mail:
gervin@biology.msstate.edu
3
Present address: Albany Pine Bush Preserve
Commission, The Nature Conservancy, 195 New
Karner Road, Albany, NY 12205.
Received for publication August 31, 2005, and in
revised form February 24, 2006.
Journal of the Torrey Botanical Society 133(3), 2006, pp. 468–481
468
to 2.5 months longer flooding in GTRs than
in adjacent unmanaged BF (e.g., Wehrle et al.
1995), thereby effectively reducing the period
of favorable growing conditions and creating
additional stress on the plant assemblage,
especially trees.
The pattern of canopy regeneration in
bottomland forests is characterized as gap-
phase replacement (King and Allen 1996) in
which disturbances that remove portions of
the canopy enable regeneration of woody and
herbaceous vegetation, including canopy trees.
Previous research (Young et al. 1990) has
indicated that canopy openness tends to be
greater in GTRs than in unimpounded BF
(UBF), with about 80%of the canopy in UBF
having a density of 90%or greater versus half
the canopy in GTRs having similar density.
One possible explanation for these differences
is the longer duration of flooding in GTRs,
which can increase tree mortality and the
tendency for trees to be windthrown (King
and Allen 1996). The mixture of late floodwa-
ter drawdown and recurring years of flooding
can increase the physiological stresses on GTR
trees and subsequently alter the resulting
forest structure and species composition.
Trends in forest tree species composition thus
generally progress towards a more water
tolerant assemblage in all strata (King 1995).
It was demonstrated previously in two of the
GTRs included in the present study that shifts
to more flood tolerant species increased the
relative abundance of trees with lower timber
and wildlife values (Karr et al. 1990, Young et
al. 1995).
Although a few studies have assessed the
impact of GTR management on overstory
composition and regeneration, most have been
limited to either relatively young GTRs (King
1995, King et al. 1998) or to small numbers of
sites, sometimes with no replication (Karr et
al. 1990, Young et al. 1990, King 1995, King et
al. 1998). The present project included four
GTRs, which had been managed as such for
40 to 48 years, and four UBF stands to
quantify the effects of greentree reservoir
management on: (1) frequency of canopy gaps
and (2) characteristics of both Indicator
Species (based statistically on combined site
affinity and abundance) and dominant species
in the upper and mid-story canopies. Because
of tree mortality and increased disturbance
produced by extended periods of flooding in
GTRs, the frequency and size of canopy gap
formations was expected to be greater there
than in naturally flooded UBF. Species di-
versity and richness also were expected to be
greater in naturally flooded UBF than in
GTRs because of the reportedly higher levels
of tree stress and mortality encountered in
forests managed as GTRs (e.g., King 1995,
King et al. 1998), and it was expected that tree
species in the GTRs would be more flood-
tolerant, on average, than species in UBF
stands.
Materials and Methods. S
TUDY
S
ITES
. This
study was carried out on the Noxubee
National Wildlife Refuge (NNWR) in east-
central Mississippi, USA (Figure 1). Eight
stands of bottomland forest were selected for
the study; four had been managed as greentree
reservoirs since the mid-1950s or 1960s (GTR
1, 1955; GTR 2, 1958; GTRs 3 and 4, 1963),
and four remained unimpounded at the time
this study was conducted. One of the GTRs
(GTR 3) is managed largely for moist-soil
habitat in its northernmost reaches, but it
nevertheless contains a considerable area of
bottomland forest in its southern half whose
hydrology is equivalent to that of the adjacent
GTR 4. The four unimpounded stands were
selected based on their proximity to and
spatial interspersion among the GTRs, to
minimize environmental variability between
stand types. However, we also attempted to
select UBF stands such that they would be
influenced as little as possible by hydrologic
alterations within the refuge landscape, such
as GTR and other reservoir levees. The result
was a set of eight highly interspersed stands of
bottomland forest.
The study area as a whole is representative
of low-gradient Upper Gulf Coastal Plain
floodplain forests, and exhibits an 8.2 m
decrease in elevation from west to east across
the eight study sites (approximately 12.2 km).
A network of ten small-order streams criss-
crossed this area of NNWR: Chinchahoma
Creek, Cypress Creek, Hollis Creek, Jones
Creek, Loakfoma Creek, Minnow Branch,
Oktoc Creek, Shaw Creek, Talking Warrior
Creek, and the Noxubee River. All of the
smaller streams ultimately flow into the
Noxubee River at the southeastern perimeter
of the refuge (33.26uN, 88.72uW). Many
bottomland and wetland areas on the refuge
are managed for waterfowl, including ducks,
geese, herons, egrets, and other wading birds.
2006]
ERVIN ET AL.: GREENTREE RESERVOIR MANAGEMENT
469
However, much of the remaining bottomland
forest has not been managed actively and,
thus, provides opportunity for comparison of
presumably natural, unimpounded BF with
the variously managed forest lands on the
refuge property.
The source, duration, and depths of flood-
waters have varied historically within and
among the GTRs. At one extreme, there has
been repeated annual continuous flooding in
some GTRs from mid-November through
mid-March from overbank flows and waters
released from adjacent reservoirs (e.g., depths
to 0.6 m in GTR 1). The other extreme
includes occasional inundation of three or
four 5–7 day events from December through
March, largely resulting from overbank flood-
ing (Gray and Kaminski 2005). Because of
highly variable microtopography in river
floodplain ecosystems (see Hodges 1997),
depth and duration of flooding also vary
within stands. This diversity was presumed to
have been more or less equivalent in the GTR
and UBF areas because of their position
within a relatively small portion of the
Noxubee floodplain (,60 km
2
) and because
of the dense network of smaller-order streams
within the refuge lands, all of which have
F
IG.
1. Locations of the eight bottomland forest stands surveyed in this study. This area is located at
approximately 33.25uN, 88.80uW. The approximate location of each of the 160 plots was mapped based on
several GPS point data, collected at a subset of study plots, and known directions and distances among plots.
Differentially shaded polygons within the Noxubee NWR are soil mapping units from the STATSGO
database (USDA NRCS 1994). Note that GTRs 1, 3, and 4 and UBFs A, C, and D all lie within the same
soil mapping unit. An earthen levee and water control structures separate GTR 3 from GTR 4.
470
JOURNAL OF THE TORREY BOTANICAL SOCIETY [V
OL
.133
created a mosaic of ridge, flat, and terrace
topography.
To quantify potential within- and among-
site topographic and soil variability, we
obtained data on plot elevation and soils using
Geographic Information Systems (GIS) meth-
ods. A map of the NNWR was downloaded
from the Mississippi Automated Resource
Information System (MARIS), which provides
geospatial information for areas throughout
the state of Mississippi (<http://www.maris.
state.ms.us>). Using ArcMap within ArcGIS
versions 8.3 and 9.0, the NWR shapefile was
transformed from a raster file to a vector file
in order to clip its properties with those of the
digital elevation models (DEMs) of the
counties across which the study sites were
located. The DEMs were based on the USGS
National Elevation Dataset, which is a com-
monly used 30-m resolution digital elevation
map of the conterminous United States
generated from 1:24,000 scale USGS quad
data, with an estimated vertical accuracy of 6
3 m (Steve Walker, GIS Operations Manager,
MARIS, personal communication). Digital
elevation models of Noxubee, Oktibbeha,
and Winston Counties were obtained from
MARIS. The coordinate systems of all files
were modified to the North American Datum
1983 (the legal horizontal geodetic reference
datum used by the US federal government) in
order to properly overlay file properties and
accurately quantify plot elevations. Pointfiles
of the GPS locality data for each of the twenty
plots located within each of the eight sites were
created through Microsoft Excel and imported
into ArcMap to create shapefiles. Those data
then were converted to raster format and given
the same coordinate system as the DEM and
NNWR files. The NNWR raster file then was
merged with the county DEMs to delimit the
boundaries of NNWR within those counties.
This was accomplished using raster calculator
within spatial analyst in ArcMap. The raster
files of all the GPS points then were clipped
with the overall NNWR file containing the
elevation data for those plots. Elevation data
finally were extracted from the raster file for
use in between- and among-site comparisons
of topographical/elevation data.
Soils data were obtained and treated simi-
larly. Soils data were downloaded from
the USDA Natural Resources Conservation
Service National Cartography and Geospa-
tial Center (<http://www.ncng.nrcs.usda.gov>).
The data used were from the State Soil
Geographic Database (STATSGO; USDA
NRCS 1994), which represents a state-level
summary, or generalization, of the county-
level soil maps for the entire state. The soils
overlay used in ArcGIS was compared
visually with a published county-level soil
survey, which also was used to determine
characteristics of the soils in the study area.
T
REE
S
URVEYS
. Systematic surveys were
conducted during summer 2003 in the four
GTR stands and four UBF stands. In each
stand, 20 plots were spaced at 150-m intervals,
with the aid of a Ranging Rangefinder (Model
75, American Visionwear), beginning at an
arbitrarily selected point near one corner or
side of the stand. One exception was UBF C,
which was surveyed at 200-m intervals because
it was the largest area and was highly dissected
by non-wadeable channels of the Noxubee
river. Thus, our survey encompassed a total of
395 ha of bottomland forest on the NNWR
(45 ha 37 sites, 80 ha in UBF stand C).
The survey method was a modification of
the point-centered-quarter method, in which
we collected data on the nearest tree (stem $
10 cm DBH) in each quadrant, delimited by
the cardinal directions of a point-based plot.
Data collected for each of the four trees per
plot were: species, degree of apparent stress,
and position in the canopy (upper or mid-
story). The six stress classes used were defined
as: 0 to 5%canopy dieback, 6 to 20%,21to
50%,51to80%,81to95%,or96to100%
dieback, indicated as levels 1 through 6. This
was modified slightly from King (1995), who
used only four stress classes. Simpson’s di-
versity index was calculated for each site, from
the plot-wise species abundances:
Diversity ~X
n
i~1
1=p
i
ðÞ
2
,
where p5proportional abundance per species
per site, and n5number of species per site
(Barbour et al. 1999).
We measured canopy density with a spher-
ical mirror densiometer by taking the mean of
four canopy measurements, one facing each
cardinal direction. Prior to conducting densi-
ometer measurements, each plot was sub-
jectively rated as a gap or closed forest, based
on obvious canopy openings, presence of
successional vegetation (e.g., shrubs, vines),
intact stumps, downed boles, and similar gap
2006]
ERVIN ET AL.: GREENTREE RESERVOIR MANAGEMENT
471
characteristics. Areas that were determined to
be gaps, based on this subjective evaluation,
then were classified as natural or man-made
gaps. Classification as man-made gaps was
based on the presence of such indicators as
obvious cut stumps, presence of access roads,
and areas clearly maintained as blind locations
for refuge waterfowl hunts. In GTRs, numer-
ous such gaps are managed as moist-soil
herbaceous vegetation for waterfowl; other
gaps were formed during the construction of
service roads and hiking trails in both the
GTRs and UBF. Only naturally formed gaps
were considered in our evaluation of GTR
management on gap frequency and in analyses
comparing gaps with closed-canopy forest.
Finally, dominant species in each stand were
determined by the ‘‘50/20’’ rule (Tiner 1999).
Specifically, species whose abundance ac-
counted for the first 50%of all species
abundances, of the 80 trees included per stand,
were considered dominant, along with each
other individual species that alone comprised
at least 20%of total abundance within a stand.
S
TATISTICAL
A
NALYSES
. Statistical compar-
isons of greentree reservoirs versus unim-
pounded forests and of subjectively-defined
gaps versus non-gap plots were carried out
with Sigma Stat (v. 1.0, Jandel Scientific) and
SPSS (version 12.0, SPSS Inc.). Data were
tested for normality and homogeneity of
variance assumptions before running ANOVA
tests for differences between groups and
among sites. Where data were found to violate
these assumptions, Kruskal-Wallis ANOVA
on ranks was performed. In some instances,
additional parametric tests (two-way ANOVA
and ANOVA incorporating Stand as a block-
ing variable nested within Forest Stand Type
[DataDesk 6.0 for Windows, Data Descrip-
tion, Inc.]) were performed on data that did
not meet parametric assumptions; those anal-
yses were conducted only for descriptive
purposes, not for purposes of statistical in-
ference, and are described as such below.
Multivariate tests were performed on these
data using PC-Ord 4.27 for Windows (MjM
Software), in order to determine species
composition effects of GTR management
and to quantify indicator species for the two
forest stand types (GTR vs. UBF). Multi-
Response Permutation Procedure (MRPP;
analogous to a nonparametric, multivariate
ANOVA) was used to test for differences
between tree assemblages in GTRs and UBF.
Indicator Species Analysis (ISA) was used to
determine which species were most indicative
of each management regime. These tests were
performed on the data as a whole and on the
canopy trees and overtopped tree datasets
individually. Non-metric multi-dimensional
scaling (NMS) also was used to ordinate the
plot3species data in an effort to determine
underlying correlations between species and
the limited environmental and other covariate
data collected (elevation, wetland indicator
status, canopy density, etc.).
All tests were conducted using guidelines
suggested by McCune and Grace (2002). The
MRPP analyses were conducted using the
recommended weighting option (C
i
5n
i
/
Sn
i
), and the Euclidean (Pythagorean) dis-
tance measure, with groups defined by Forest
Type. Results of ISA were evaluated using
a Monte Carlo simulation of 1000 permuta-
tions on Indicator Values derived by the
method of Dufrene and Legendre (1997).
Site-wise MRPP was not performed, because
of the need to have conducted 28 pairwise
comparisons, but ISA was performed on site-
wise grouping of data, for descriptive pur-
poses. Although Indicator Species Analysis is
useful in determining relatively unique species
3site combinations (as a combination of high
site fidelity and relative abundance), ISA
provides little information on the overall
characteristics of vegetation. It was for this
reason that we also evaluated information on
the dominant species for each site, for
comparison between the two groups (GTR
vs. UBF).
Ordination via NMS was conducted on the
full dataset by conducting an initial explor-
atory run to determine the appropriate num-
ber of axes to capture optimal variance within
the data. Settings for this preliminary analysis
consisted of a step-down approach from 4 axes
to 1 axis, using 50 runs with the actual data to
evaluate stability (with an instability criterion
of 0.0005), and a minimum of 200 and
maximum of 500 iterations per run (McCune
and Grace 2002). Following stability analyses,
a Monte Carlo simulation was performed to
determine optimal dimensionality for the
ordination. This randomization analysis con-
sisted of 50 runs as well, and the combination
of stress analysis and Monte Carlo simulation
indicated that 3 axes provided the optimal
dimensionality for the ordination. This pre-
472
JOURNAL OF THE TORREY BOTANICAL SOCIETY [V
OL
.133
liminary analysis was followed by a final run
of the real data using the starting configura-
tion for the statistically best 3-dimensional
ordination, and using the same settings as the
preliminary analysis.
Data used as overlays to determine available
factors that appeared to correlate with vari-
ance in the species assemblage data included:
canopy density, plot elevation, stress rating,
species per plot, whether each plot was
determined to be a gap, number of trees
present as upper canopy trees, and mean
wetness coefficient per plot. Wetness coeffi-
cients were based on USFWS Region 2
Wetland Indicator Status for the species
identified (Table 1). Species with no Indicator
Status were given a rating of Upland, which
results in a coefficient of 25, as the most
conservative estimate of wetland affinity
(occurring only rarely in true wetlands).
Results. G
ENERAL
E
FFECTS OF
GTR
M
ANAGEMENT
. In general, forests managed as
GTRs were quite similar to unimpounded
forests (Figs. 2 and 3). Tree canopy density,
number of midstory trees per plot, and degree
of stress exhibited by canopy dieback all were
similar between the two stand types (P$0.37
for each; Fig. 2). Additionally, the number of
subjectively-defined natural canopy gaps was
similar in GTRs and UBF, as was Simpson’s
diversity index (P$0.60; Fig. 3). Because
the number of man-made gaps varied
among stands, gap frequency comparison
between stand types was based only on natu-
ral gaps (divided by total plots, minus arti-
ficial gaps). No difference was found between
GTRs (27 66%of plots were natural gaps)
and UBF (35 611%;P50.54), although
a difference of eight percent here might be
difficult to detect with a sample size of four
sites per stand type.
Mean species richness did differ between the
stand types. At both the plot and site scales,
more species were encountered in UBFs than
GTRs. This difference was small at the plot
scale with 2.7 60.1 species per plot in GTRs,
vs. 3.0 60.1 species per plot in UBF stands (P
50.012; Fig. 2). At the site scale, there were
Table 1. Wetland indicator status categories and
equivalent wetness coefficients (Reed 1988, Herman
et al. 1997). Note that signs of wetness coefficients
are reversed from Herman et al. (1997) to yield
a positive correlation between wetness coefficient
and indicators for sites with a predominance of
wetland species.
Indicator Status
Probability of occurrence
in Wetlands
Wetness
Coefficient
Obligate wetland
(OBL) .99%+5
FACW++4
Facultative
wetland
(FACW) 67–99%+3
FACW2+2
FAC++1
Facultative (FAC) 34–66%0
FAC221
FACU+22
Facultative upland
(FACU) 1–33%23
FACU224
Upland (UPL) ,1%25
NO (No formal Ind. Status – assigned
UPL for these analyses)
25
F
IG.
2. Stand characteristics, as related to stand
type and gaps versus closed canopy forest. Bars
represent means 61SE of all plots within the
indicated group: n580 GTR plots, 80 unim-
pounded forest plots, 44 natural gaps, and 103 plots
in closed forest. The comparisons indicated by
asterisks exhibited significant differences between
groups (P#0.03) in K-W ANOVA on ranks.
Species per plot did not differ significantly between
gaps and closed forest, (P50.06). Note condensed
scale in upper panel.
2006]
ERVIN ET AL.: GREENTREE RESERVOIR MANAGEMENT
473
almost 25%more species in UBFs (19.8 61.1)
than in GTRs (16.0 61.2; P50.06; Fig. 3,
Table 2). Twice as many species were exclusive
to unimpounded forest than to GTRs (twelve
vs. six species), and no species that were
exclusive to the mid-story in UBFs were
encountered in the GTRs, whereas 3 of 4
mid-story-only species from GTRs were found
in UBF plots (Table 2). Furthermore, species
exclusive to UBF had a significantly lower
F
IG.
3. Stand characteristics in GTRs versus unimpounded forest stands. Bars represent means 61SE of
all sites within the indicated group (four sites per group). Relative elevation is the per-stand mean of plot-
level elevation, relative to the mean elevation of points along the nearest downstream reach of streams
immediately to the north and south of each stand.
Table 2. Tree species recorded during this study. Wetland indicator status and wetness coefficients are
given in parentheses (see Table 1). Superscript letters indicate species found in only the canopy (C) or the
mid-story (M), and in only GTRs (G) or unimpounded forest (U). Species are grouped into unimpounded
forest or GTR Indicators based on maximum indicator values from Indicator Species Analysis, not
significant P-values. Names are those used by the PLANTS database (USDA NRCS 2005).
UBF Indicator Species GTR Indicator Species
Asimina triloba (L.) Dunal (FAC; 0){
U
Acer rubrum L. (FAC; 0)
Carpinus caroliniana Walt. (FAC; 0) Betula nigra L. (FACW; 3){
C,G
Carya cordiformis (Wangenh.) K. Koch (FAC; 0){
M,U
Carya glabra (P. Mill.) Sweet (FACU; -3)
C. ovalis (Wangenh.) Sarg. (FACU; -3)
U
C. glabra (P. Mill.) Sweet
C. ovata (P. Mill.) K. Koch (FACU; -3) var. hirsuta (Ashe) Ashe (NO; -5)
G
C. tomentosa (Lam. ex Poir.) Nutt. (NO; -5){C. pallida (Ashe) Engl. & Graebn. (NO; -5){
Crataegus L. spp. (Unknown; -5){
M,U
Diospyros virginiana L. (FAC; 0){
M
Fagus grandifolia Ehrh. (FACU; -3)
U
Fraxinus pennsylvanica Marsh. (FACW; 3)
Ilex opaca Ait. (FAC2;-1){
M,U
Ilex decidua Walt. (FACW2;2){
M
Liquidambar styraciflua L. (FAC+;1) Morus rubra L. (FAC; 0){
M
Liriodendron tulipifera L. (FAC; 0){
M,U
Nyssa biflora Walt. (OBL; 5){
M,G
Pinus taeda L. (FAC; 0) N. sylvatica Marsh. (OBL; 5)
Platanus occidentalis L. (FACW2;2){
M,U
Pinus echinata P. Mill. (NO; -5){
C,G
Quercus alba L. (FACU; -3) Planera aquatica J.F. Gmel. (OBL; 5)
G
Q. falcata Michx. (FACU2; -4){
M,U
Quercus laurifolia Michx. (FACW; 3){
Q. michauxii Nutt. (FACW2;2) Q. lyrata Walt. (OBL; 5)
Q. nigra L. (FAC; 0){Q. pagoda Raf. (FAC+;1)
Q. phellos L. (FACW2;2) Salix nigra Marsh. (OBL; 5){
C,G
Q. shumardii Buckl. (FACW2;2)
U
Taxodium distichum (L.) L.C. Rich. (OBL; 5)
Q. stellata Wangenh. (FACU; -3){
M,U
Ulmus americana L. (FACW; 3)
Symplocos tinctoria (L.) L’He´r. (NO; -5){
M,U
Ulmus alata Michx. (FACU+;-2)
U. rubra Muhl. (FAC; 0)
{Species present as minor components of canopy or mid-story; absent from Tables 3 through 6.
474
JOURNAL OF THE TORREY BOTANICAL SOCIETY [V
OL
.133
wetness coefficient than those found only in
GTRs (20.9 60.6 for 11 of the 12 UBF
species, vs.3.061.6 for 5 of the 6 GTR
species (mean 6SE)—other species did not
have Wetland Indicator Status or were not
identified to species; U
M-W
58.5; P50.02),
indicating a higher degree of adaptation to
flooding in GTR assemblages. Species encoun-
tered only in UBF stands were Quercus
falcata, Q. shumardii,andQ. stellata, along
with Asimina triloba, Carya cordiformis, Carya
ovalis, Fagus grandifolia, Liriodendron tulipi-
fera, Platanus occidentalis,andSymplocos
tinctoria. Betula nigra, Carya glabra var.
hirsuta, Nyssa biflora, Pinus echinata, Planera
aquatica,andSalix nigra were exclusive to
GTRs.
E
FFECTS OF
GTR M
ANAGEMENT ON
S
PECIES
C
OMPOSITION
. Comparisons of GTRs versus
UBF areas, using MRPP, indicated that these
groups were different from one another, based
on analyses of all trees together (A50.015; P
#0.00001), on canopy trees alone (A50.010;
P50.0001), and on midstory tree abundances
(A50.008; P50.002). Indicator Species
Analyses were used to provide lists of species
indicative of each stand type (GTR or UBF)
and each individual site (species with both
high fidelity to and high frequency within
a given type or site; Tables 3 and 4). Clear
differences were present between GTRs and
UBF areas, with GTR species having, on
average, higher wetness coefficients (indicating
wetter conditions) than UBF assemblages,
regardless of the groups used to derive ISA
results (group-wise or site-wise indicators).
These analyses also resulted in groups of
mid-story species whose average wetness
coefficients (WC) generally were lower than
the mean for canopy species. For example, the
mean WC for the GTR canopy species in
Table 3. Results of Indicator Species Analyses, by forest stand type.
Group
Indicator Species
a
Overall Canopy Mid-story
GTR Acer rubrum ** Taxodium distichum ** Acer rubrum **
Planera aquatica ** Planera aquatica **
Taxodium distichum ** Quercus lyrata *
Taxodium distichum *
UBF Carpinus caroliniana *Quercus michauxii *Carpinus caroliniana *
Carya ovata ** Carya ovata **
Liquidambar styraciflua ** Liquidambar styraciflua *
Pinus taeda *Ulmus alata *
Ulmus alata *
a
Indicator species were determined as those with a P-value #0.10 (*) or P#0.05 (**) following Monte
Carlo resampling of the data set.
Table 4. Results of Indicator Species Analyses, by individual stand.
Stand
Indicator Species
a
Overall Canopy Mid-story
GTR 1 Quercus pagoda ** Quercus pagoda ** Carya glabra **
GTR 2 Carya glabra var. hirsuta ** Nyssa sylvatica **
GTR 3 Acer rubrum ** Acer rubrum **
Taxodium distichum ** Taxodium distichum **
GTR 4 Quercus lyrata ** Fraxinus pennsylvanica *Quercus lyrata **
Quercus lyrata **
UBF A None
UBF B Liquidambar styraciflua ** Pinus taeda ** Liquidambar styraciflua **
Pinus taeda ** Quercus alba **
Quercus alba **
UBF C Quercus phellos ** Quercus phellos **
UBF D Quercus michauxii ** Quercus michauxii ** Carpinus caroliniana *
Fagus grandifolia **
a
Indicator species were determined as those with a P-value #0.10 (*) or P#0.05 (**) following Monte
Carlo resampling of the data set.
2006]
ERVIN ET AL.: GREENTREE RESERVOIR MANAGEMENT
475
Table 3 is 5, vs. a mean of 3.8 for mid-story
species, and in UBF the canopy species had
a mean WC of 2, vs. a mean of 21.0 for mid-
story trees. This pattern also was found for the
dominant tree species in Table 5.
Ordination analyses revealed an interesting
pattern to the data (Fig. 4). Of the three axes
resulting from this NMS optimization, axes
one and three correlated most closely with
mean WC per plot, of the covariate data
available for comparison. These two axes
combined accounted for a total of 42%of
the variation in WC among plots, with axis
one having a tight linear relationship to WC in
both stand types (r520.60). The most
interesting aspect of this analysis was that
the GTR plots all grouped within a much
smaller area in species-defined space than did
UBF sites, a pattern apparently influenced in
large part by wetness tolerance of the species
present.
R
ELATIONSHIP OF
E
LEVATION TO
S
PECIES
C
OMPOSITION
. Plot elevation did not differ
significantly between GTR and UBF stands
(F
1,6
50.95 and P50.37 using four stand
means per stand type; F
1,158
53.45 and P5
0.065 using 160 plot-level elevations, which
could be interpreted as pseudoreplication;
Figure 3). However, because the stands used
in these surveys were spaced across a landscape
gradient of about 8.2 m of elevation, there
were significant differences among stands
(F
7,152
54.03; P,0.001). Specifically, UBF
stand A plots were at a significantly higher
elevation than plots in GTRs 1 and 2, and
plots in UBF stand D and GTR 4 were of
higher elevation than GTR 2 (Bonferroni
Table 5. Dominant species in each of the study sites, as determined by the ‘‘50/20 Rule’’ (Tiner 1999).
Species are listed in order of relative abundance within stands, from top to bottom.
Stand Canopy Mid-story
GTR 1 Quercus pagoda Acer rubrum
Liquidambar styraciflua Liquidambar styraciflua
Carpinus caroliniana
Carya glabra
Planera aquatica
Quercus michauxii
Ulmus americana
GTR 2 Taxodium distichum Carpinus caroliniana
Quercus pagoda Acer rubrum
Q. laurifolia Liquidambar styraciflua
Q. lyrata
Q. michauxii
GTR 3 Taxodium distichum Acer rubrum
Acer rubrum Planera aquatica
GTR 4 Quercus lyrata Acer rubrum
Acer rubrum Carpinus caroliniana
Taxodium distichum Quercus lyrata
Fraxinus pennsylvanica
UBF A Acer rubrum Carpinus caroliniana
Fraxinus pennsylvanica Acer rubrum
Quercus lyrata Liquidambar styraciflua
Liquidambar styraciflua Ulmus rubra
Quercus michauxii Ulmus alata
Quercus phellos Quercus michauxii
UBF B Pinus taeda Liquidambar styraciflua
Liquidambar styraciflua Carpinus caroliniana
UBF C Quercus lyrata Liquidambar styraciflua
Liquidambar styraciflua Carpinus caroliniana
Quercus phellos Carya ovata
Acer rubrum
Carya ovalis
UBF D Quercus michauxii Carpinus caroliniana
Liquidambar styraciflua Acer rubrum
476
JOURNAL OF THE TORREY BOTANICAL SOCIETY [V
OL
.133
post-hoc comparison, overall a50.05). Plots
in UBF A were, on average, about 2.2 m
higher in elevation than GTRs 1 and 2, and
UBF D and GTR 4 plots were about 1.9 m
higher than plots in GTR 2. Despite the
relatively large 8 m elevation gradient from
the highest UBF plot to the lowest GTR plot,
the spatial interspersion of stand types resulted
in the detection of no other differences in
elevations among sites. Within-stand relief
(highest to lowest plot within each stand)
averaged 6.0 60.1 m across the eight stands
(with plots spread across as much as 1200 m of
forest; Fig. 1). Thus, the gradient in elevation
within stands was almost as great as the
elevation gradient from the highest stand in
one type to the lowest stand in the other type.
Elevation also was found to be correlated
poorly with the three major axes resulting
from NMS ordination of plots (a total of 4%
of the variation in species among plots along
those three axes was explained by elevation).
Another potential topographic factor that
could affect tree species composition, through
indirect effects on hydrology, is variation in
elevation within stands (among plots). This
was assessed by comparing the coefficient of
variation (CV) in plot-level elevation in UBF
versus GTR stands. Within-stand variability
in plot elevation did not differ between stand
types (CV
GTR
57.6%60.2%,CV
UBF
5
7.6%60.2%;F
1,6
50.02; P50.89). Finally,
plot elevation relative to elevation of the two
streams nearest each stand (one to the north
and one to the south) was compared among
stands and stand types. As with plot-level
elevations, no differences were detected be-
tween stand types with regard to elevation
relative to the nearest streams or with respect
to variability in relative elevation within
stands (Figure 3).
Soils varied little across the study area
(Figure 1, Table 6). Six of the eight stands
were located on the same soil type (Mathiston
silt loam), and the other two stands over-
lapped this and only two other soil types. Each
of the three soils was classified as poorly to
somewhat poorly drained, because of low
permeability or the presence of a fragipan
(Stough fine sandy loam). Each soil also was
characterized by slow runoff potential and low
topographic relief.
Discussion. We encountered a total of 41
tree species during this project, 26 of which
were present in both upper canopy and
midstory. Three species were exclusive to the
upper canopy, and 12 were found only in the
midstory. A recent survey of BF in the
Mississippi River floodplain found a similar
number of species (33) in a survey of 21.5 ha
F
IG.
4. Ordination (NMS) of plots, based on
abundance of species at each, relative to the two axes
most strongly correlated with mean wetness co-
efficient of the four trees present at each plot.
Correlation coefficients (Pearson’s r)were0.57for
Axis 1 and 0.24 for Axis 2. Axes 2 and 3 in this
ordination were almost equally correlated with mean
per plot Wetness Coefficient; axis 2 had a slightly
higher value for Kendall’s tau (rank correlation) and
thus was selected for graphical display. The final
stress rating was 19.4 for the optimal, 3-dimensional
solution (P50.03), which had high stability
(instability value of 0.00048) after 168 iterations.
Symbol size is weighted by mean wetness coefficient
for each plot.
2006]
ERVIN ET AL.: GREENTREE RESERVOIR MANAGEMENT
477
of deltaic forest in southeastern Louisiana
(White and Skojac 2002). Of those trees, 27
were found in the overstory and 24 in the
understory (6 in understory only). Nine of the
upper canopy species and 8 midstory species
recorded in our study also were found in the
same stratum by White and Skojac (2002). Of
the species recorded by White and Skojac
(2002), Ulmus spp., Quercus nigra, Acer
rubrum,andLiquidambar styraciflua were
among the six most important upper canopy
species, and A. rubrum, Ulmus spp., Crataegus
spp., L. styraciflua,andQ. nigra had the five
highest Importance Values in the understory.
The species assemblages in both GTRs and
UBF areas used in our work are similar to
other published studies on BF and GTRs in
the southeastern US (King 1995, King and
Allen 1996, King and Antrobus 2001).
G
ENERAL
E
FFECTS OF
GTR M
ANAGEMENT
.
Based on examination of dominant tree
species, the GTRs at NNWR typified the
poorly drained BF described by Hodges (1997,
e.g., his Figure 6). The UBF stands exhibited
characteristics of both better drained and
poorly drained forests (Hodges 1997). The
latter results from the diverse topography and
hydrology found within riparian forests that
extend from river edge through flats, sloughs,
and often onto or across terraces. Indicator
species of GTRs suggest over-extended peri-
ods of flooding (highly flood-tolerant Taxo-
dium, Planera,andQuercus lyrata), but UBF
indicators suggest another negative aspect of
historic management activities (Table 3 and
5). Carpinus and Liquidambar, indicators of
past highly selective timber harvests (Kellison
and Young 1997), were two key dominant
species in the canopy or midstory of all the
UBF stands surveyed in this study, with
Liquidambar present as a dominant in the
upper canopy and midstory of all but one site
(Table 5).
Eight of the twelve tree species found only
in unimpounded forest (Table 2) were mid-
story species, suggesting very different succes-
sional dynamics in the two forest types. This,
combined with the overall difference in wet-
ness adaptation among the eleven UBF-
exclusive species (wetness coefficient of 20.9
60.6) versus the six species found only in
GTRs (3.0 61.6), suggests the species
occurring in the GTRs were better adapted
to conditions of continuously flooded or
saturated soils, and thus could tolerate the
combined stresses of soil anoxia and overstory
shading. Finally, Quercus falcata, Q. shumar-
dii,andQ. stellata were among those species
encountered only in UBF stands, as well as
other species potentially useful for wildlife,
such as Asimina triloba, Carya cordiformis,
Fagus grandifolia,andLiriodendron tulipifera.
G
APS
V
ERSUS
C
LOSED
C
ANOPY
F
OREST
. The
finding that natural canopy gaps occurred
with essentially equal frequency in UBFs (6.0
61.6 natural gaps per 20 plots) as in GTRs
(5.0 61.2 gaps per 20 plots; P50.64) was
somewhat surprising (Fig. 3). This was espe-
cially so, given the recent storm damage that
had occurred in GTRs 1 and 2 (1998 tornado
and severe storms of Autumn 2002). This
result contradicts expectations because the
management activities in GTRs and associated
extended periods of flooding should have
resulted in shallowly rooted trees, subjecting
them to greater likelihood of windthrow.
The frequency of gaps overall appears
similar to that found in other naturally
flooded BF of the southeastern USA. For
example, King and Antrobus (2001), in a more
thorough gap survey in unimpounded bot-
tomland forest, found 29 new gaps formed per
Table 6. Soils on which the study plots were located. Soils information from USDA SCS (1973),
locations of plots derived from data displayed in Figure 1 (USDA NRCS 1994).
Soils Characteristics Stands (#of plots)
Mathiston silt loam Loamy alluvial soil; moderate permeability but with slow
runoff; acidic soils susceptible to ponding; intermixed
with some poorly drained soils as well as sandier
deposits along major stream channels; 0 to 2%slopes
GTR 1, 3, 4 (all)
GTR 2 (1 plot)
UBF A, C, D (all)
Stough fine sandy loam Loamy soil on terraces and uplands; moderate permeability
in upper strata but slower in fragipan; acidic soils with
slow runoff; 0 to 2%slopes UBF B (15 plots)
Urbo silty clay loam Clayey alluvium with low permeability and runoff; acidic
soils susceptible to ponding; 0 to 2%slopes
GTR 2 (19 plots)
UBF B (5 plots)
478
JOURNAL OF THE TORREY BOTANICAL SOCIETY [V
OL
.133
year in the Cache River floodplain of north-
eastern Arkansas, which equated to 0.1–0.4%
of the total study area (103 ha) each year. In
our study, the 44 natural canopy gaps we
encountered at our systematically placed plots,
had they been of similar size as those of the
Cache River BF (,100 m
2
/ gap) would have
amounted to about 0.11%of the 395 ha
surveyed.
Three of the species frequently encountered
as dominants in our study sites (Acer rubrum,
Quercus lyrata,andFraxinus pennsylvanica)
were found to be key gapmaker species in the
Cache River BF (King and Antrobus 2001).
King and Antrobus (2001) found that Frax-
inus usually formed gaps as a result of bole
snapping during dry periods, whereas Q.
lyrata frequently was subject to windthrow;
Acer rubrum was equally susceptible to snap-
ping and uprooting. Results were attributed to
the morphology of each species, as well as the
wetness of the site on which gap-making trees
had grown. Drier sites and drier years usually
were correlated with higher incidence of
snapped trees, whereas the opposite conditions
correlated with windthrow. Thus, contrary to
commonly accepted dogma, longer periods of
inundation do not necessarily result in greater
frequency of forest gaps; gap dynamics are
a complex phenomenon, dependent upon the
combination of site characteristics and traits
of the species present.
Previous studies of GTR managed forests
indicated that under increased duration and
frequency of flooding, levels of stress exhibited
by individual trees was much higher than in
naturally flooded BF. However, most of those
studies (e.g., King 1995, King et al. 1998)
investigated GTRs during earlier stages of
management. King (1995) studied two GTRs
in Texas during years 2 through 4 after initial
flooding, and King et al. (1998) reported
effects in GTRs during years 2 through 10 of
management. The GTRs included in the
present study, on the other hand, had been
exposed to management for 40 to 48 years;
thus, the longer period of management likely
has led to stronger selection for GTR tree
species tolerant of the markedly altered
hydroperiod and has resulted in lower rates
of stress-related gap formation.
E
FFECTS OF
GTR M
ANAGEMENT ON
S
PECIES
C
OMPOSITION
. As mentioned above,
dominant tree species in NNWR GTRs
typified those of poorly drained BF, and the
UBF areas possessed characteristics of both
better drained and poorly drained forests
(Hodges 1997). The GTR upper canopies
generally were dominated by a Taxodium –
Acer –Quercus lyrata mix, whereas UBF
upper canopies could be characterized as
Liquidambar – mixed Quercus stands. The
midstories of the groups were more similar,
with dominance by Acer –Carpinus –Planera
in GTRs and Liquidambar –Carpinus –Acer
in UBF stands. Results from Indicator Species
Analyses yield similar descriptions of the tree
assemblages in these sites.
Although the differences we observed, or
conditions leading to them, could have existed
prior to establishment of the GTRs we studied
because of the tendency for GTRs to be
constructed on lower elevation sites, we found
no evidence to support such topographic
differences. The spatial interspersion of the
stands we studied helped to ensure that plot-
scale elevation and variability in elevation did
not differ between GTR and UBF stands, and
species associations (as revealed through
ordination analyses) were not strongly corre-
lated with elevation. Differences in elevation
that existed among individual sites within the
GTR and UBF groups were virtually equiva-
lent to elevation differences between the two
groups, and those differences appeared to
result largely from the interdispersion of sites
along the west-to-east topographic gradient.
Furthermore, within stand topographic relief
(6 60.1 m) was almost three times greater
than the statistically significant differences
found between individual UBF and GTR
stands (1.9 m to 2.2 m). Although the eleva-
tion data used were of a coarse precision (6
3 m vertical accuracy), that imprecision should
be more or less homogeneous across this low-
gradient study area. Such Digital Elevation
Models and the National Elevation Dataset
are widely used in lieu of highly laborious on-
the-ground surveying that would be required
to obtain elevation data across such broad
spatial expanses.
Similar conclusions regarding historical
conditions could be reached for the potential
effects of soil drainage on tree assemblages.
Six of the eight stands we surveyed were
located on the same soil type, a poorly drained
loamy alluvium (Mathiston silt loam). Al-
though fifteen plots of one UBF site were
located on probably the best drained of the
2006]
ERVIN ET AL.: GREENTREE RESERVOIR MANAGEMENT
479
three soil types, five of the plots in that stand
were located on the more poorly drained Urbo
silty clay loam. Thus, the most likely explana-
tion for the differences observed between the
unimpounded forest and the GTR tree assem-
blages was the fact that the GTRs had been
managed under an altered flood regime for
four to five decades.
Conclusions. Greentree reservoir manage-
ment initially was intended to provide surro-
gate BF habitats in which overwintering
migratory waterfowl could find refuge within
a landscape mosaic of decreasing BF avail-
ability. However, periods of flooding in GTRs
frequently are longer than in UBF (Reinecke
et al. 1989, Wehrle et al. 1995), and previous
work has shown that there can be substantial
drawbacks to the creation of such unnatural
hydroperiods, in terms of effects on the plant
assemblage (Karr et al. 1990, Young et al.
1990, King 1995, King et al. 1998, Gray and
Kaminski 2005), and effects on other taxo-
nomic groups, such as invertebrates (Wehrle et
al. 1995). The mixture of late floodwater
drawdown and recurring years of flooding
can increase the physiological stresses on BF
tree species in young GTRs and subsequently
alter the resulting forest structure and species
composition, resulting in a more flood-toler-
ant assemblage, in all strata of a developing
forest (King 1995, Gray and Kaminski 2005).
It is clear that simulating natural hydrologic
regimes remains a key impediment in success-
ful management of greentree reservoirs. As
indicated by King and Allen (1996), creation
and management of GTRs could be an
effective means by which to achieve the
multiple goals associated with wetlands man-
agement (aesthetics, hunting, timber manage-
ment, water quality improvement, flood water
retention, etc.), ‘‘provided water-level control
and maintenance are substantially improved’’
(emphasis theirs). Results of the present study
suggest that current and historical hydroper-
iods in the GTRs we investigated have been
sufficient to produce a forest tree assemblage
that is highly adapted to extended inundation
and capable of persistence under such man-
agement. However, most of the important tree
species used by overwintering waterfowl
(mixed oak species, as are present in the
UBF) were absent or of less importance in
both the upper canopy and midstory in the
GTRs, where instead, species less desirable to
wildlife, such as Taxodium distichum, Acer
rubrum,andQuercus lyrata were present in
high frequencies, in direct conflict with the
primary objective of GTR management.
Literature Cited
B
ARBOUR
, M. J., J. H. B
URK
,W.D.P
ITTS
,F.S.
G
ILLIAM
,
AND
M. W. S
CHWARTZ
. 1999. Terrestrial
Plant Ecology. Benjamin Cummings, Menlo
Park, CA.
D
UFRENE
,M.
AND
P. L
EGENDRE
. 1997. Species
assemblages and indicator species: the need for
a flexible asymmetrical approach. Ecol. Monogr.
67: 345–366.
G
RAY
,M.J.
AND
R. M. K
AMINSKI
. 2005. Effect of
continuous versus periodic winter flooding on
survival of oak seedlings in Mississippi greentree
reservoirs, p. 487–493. In L. H. Fredrickson, S.
L. King, and R. M. Kaminski [eds.], Ecology and
management of bottomland hardwood systems:
the state of our understanding. University of
Missouri-Columbia, Gaylord Memorial Labora-
tory Special Publication No. 10, Puxico, MO.
H
ERMAN
, K. D., L. A. M
ASTERS
,M.R.P
ENSKAR
,A.
A. R
EZNICEK
,G.S.W
ILHELM
,
AND
W. W.
B
RODOWICZ
. 1997. Floristic quality assessment:
Development and application in the state of
Michigan (USA). Nat. Areas J. 17: 265–279.
H
ODGES
, J. D. 1997. Development and ecology of
bottomland hardwood sites. For. Ecol. Manage.
90: 117–125.
K
ARR
, B. L., G. L. Y
OUNG
,J.D.H
ODGES
,B.D.
L
EOPOLD
,
AND
R. M. K
AMINSKI
. 1990. Effect of
flooding on greentree reservoirs. Technical Com-
pletion Report to the USGS Water Resources
Research Institute for Project G1571-03.
K
ELLISON
,R.C.
AND
M. J. Y
OUNG
. 1997. The
bottomland forest of the southern United States.
For. Ecol. Manage. 90: 101–115.
K
ING
, S. L. 1995. Effects of flooding regimes on two
impounded bottomland hardwood stands. Wet-
lands 15: 272–284.
K
ING
,S.L.
AND
J. A. A
LLEN
. 1996. Plant succession
and greentree reservoir management: Implica-
tions for management and restoration of bot-
tomland hardwood wetlands. Wetlands 16:
503–511.
K
ING
,S.L.
AND
T. J. A
NTROBUS
. 2001. Canopy
disturbance patterns in a bottomland hardwood
forest in northeastern Arkansas, USA. Wetlands
21: 543–553.
K
ING
, S. L., J. A. A
LLEN
,
AND
J. W. M
C
C
OY
. 1998.
Long-term effects of a lock and dam and
greentree reservoir management on a bottomland
hardwood forest. For. Ecol. Manage. 112:
213–226.
M
C
C
UNE
,B.
AND
J. B. G
RACE
. 2002. Analysis of
Ecological Communities. MjM Software Design,
Gleneden Beach, OR.
M
ITSCH
,W.J.
AND
J. G. G
OSSELINK
. 2000. Wetlands,
3
rd
edition. John Wiley & Sons, Inc., New York,
NY.
R
EED
,J
R
., P. B. 1988. National List of Plant Species
That Occur in Wetlands: 1988 National Summa-
ry. Retrieved 08 July 2005 from U.S. Fish &
480
JOURNAL OF THE TORREY BOTANICAL SOCIETY [V
OL
.133
Wildlife Service. <http://www.nwi.fws.gov/plants.
htm/>
R
EINECKE
, K. J., R. M. K
AMINSKI
,D.J.M
OORHEAD
,
J. D. H
ODGES
,
AND
J. R. N
ASSAR
, J. R. 1989.
Mississippi Alluvial Valley, p. 203–247. In L. M.
Smith, R. L. Pederson, and R. M. Kaminski
[eds.], Habitat management for migrating and
wintering waterfowl in North America. Texas
Tech University Press, Lubbock, TX.
T
INER
, R. W. 1999. Wetland Indicators: A guide to
wetland identification, delineation, classification,
and mapping. Lewis Publishers, Boca Raton, FL.
U
NITED
S
TATES
D
EPARTMENT OF
A
GRICULTURE
,S
OIL
C
ONSERVATION
S
ERVICE
. 1973. Soil survey of
Oktibbeha County, Mississippi. US Government
Printing Office, Washington, DC.
U
NITED
S
TATES
D
EPARTMENT OF
A
GRICULTURE
,N
AT-
URAL
R
ESOURCES
C
ONSERVATION
S
ERVICE
. 1994.
State Soil Geographic (STATSGO) Database
for MS. Accessed January 2006. <http://www.
ncgc.nrcs.usda.gov/products/datasets/statsgo/index.
html>
U
NITED
S
TATES
D
EPARTMENT OF
A
GRICULTURE
,N
AT-
URAL
R
ESOURCES
C
ONSERVATION
S
ERVICE
. 2005,
The PLANTS Database, Version 3.5 (http://
plants.usda.gov). Data compiled from various
sources by Mark W. Skinner. National Plant
Data Center, Baton Rouge, LA.
W
EHRLE
, B. W., R. M. K
AMINSKI
,B.D.L
EOPOLD
,
AND
W. P. S
MITH
. 1995. Aquatic invertebrate
resources in Mississippi forested wetlands during
winter. Wildl. Soc. Bull. 23: 774–783.
W
HITE
,D.A.
AND
S. A. S
KOJAC
. 2002. Remnant
bottomland forests near the terminus of the
Mississippi River in southeastern Louisiana.
Castanea 67: 134–145.
Y
OUNG
, G. L., B. L. K
ARR
,J.D.H
ODGES
,B.D.
L
EOPOLD
,
AND
R. M. K
AMINSKI
. 1990. Influence
of greentree reservoir management on overstory
composition, canopy density, and hardwood
regeneration on Noxubee National Wildlife
Refuge. Proc. of the 20th Mississippi Water
Resources Conference, B. Jean Daniel, ed.
Mississippi Water Resources Research Institute,
Mississippi State University, pp. 105–115.
Y
OUNG
, G. L., B. L. K
ARR
,B.D.L
EOPOLD
,
AND
J. D. H
ODGES
. 1995. Effect of greentree reser-
voir management on Mississippi bottom-
land hardwoods. Wildl. Soc. Bull. 23: 525–
531.
2006]
ERVIN ET AL.: GREENTREE RESERVOIR MANAGEMENT
481
