66 
THE WILSON JOURNAL OF ORNITHOLOGY • Vol. 123, No. 1. March 2011 
Our focal species were three wintering grass¬ 
land birds that frequent pine savannas along the 
Gulf Coast: Bachman’s Sparrow, Henslow’s 
Sparrow, and Sedge Wren. Numerous studies 
over the last decade have examined wintering 
Henslow’s Sparrow ecology, but few have 
reported habitat preferences among a mosaic of 
localized habitat types. Bachman’s Sparrow and 
Sedge Wren rarely have been studied in winter, 
and most reports of Bachman’s Sparrows using 
RCW clusters are from the breeding season. We 
do not know of published studies on Sedge Wren 
habitat preferences in southern pine savannas 
(Herkert et al. 2001). Our objectives were to: (1) 
assess differences in wintering grassland bird 
communities and vegetation structure among 
upland, RCW, and bog stands, and (2) ascertain 
which vegetation structures and plant species 
predict occurrence within a stand by Henslow’s 
Sparrows and Sedge Wrens. 
METHODS 
Study Area. —De Soto National Forest (153,780 
ha) in southern Mississippi, USA is mostly 
managed for longleaf and slash pine ( Pinas 
elliottii). The majority of pine savanna habitats 
in DSNF are upland stands, and many of these 
areas have severe shrub {Ilex spp., Gaylussacia 
spp.) encroachment due to past fire suppression 
(Brooks 2010). The pine savanna ecotypes in 
DSNF are described as xeric sand barrens and 
uplands, subxeric sandy uplands, and seeps (Peet 
2006). 
We selected 27 study sites to sample grass¬ 
land birds. Ten sites were classified as upland 
stands, 10 as bog stands, six were managed 
RCW clusters, and one was classified as other. 
Our selection criteria were that each site needed 
>50% herbaceous cover and <50% shrub 
cover. Sites with >50% shrub cover were 
considered a priori to be unsuitable for grass¬ 
land birds. We established 100^00 m of 20-m 
wide transects in upland, bog, and other stands, 
depending on stand size (from <1 to >10 ha). 
We established tour-sided plots in RCW clusters. 
These plots were 0.2 to 1.0 ha encompassing the 
majority of the managed cluster. Canopy closure 
within and among stand types varied from 0 to 
>70%. 
Bird Surveys. —We sampled grassland birds 
using flush net surveys. Surveys were conducted 
on fixed-width transects between 28 November 
and 28 February in 2007-2008 and 2008-2009. 
We sampled 16 sites the first winter and 11 
different sites the second year. Each site was 
sampled three times per winter except for three 
that were burned before the third round of 
sampling. All sites sampled the same year were 
>500 m apart. Our flush netting protocol was 
modified from Shackleford et al. (2001) and 
Carrie et al. (2002) as described by Brooks 
(2010). When a bird was flushed, it was identified 
to species if possible. If identification was 
uncertain, we would attempt to capture the bird 
for identification following the capture technique 
described in Bechtoldt and Stouffer (2005). We 
used the same sampling method in RCW stands, 
but made multiple systematic, non-overlapping 
passes through plots until they were completely 
sampled. Grassland birds that could not be 
identified to species were excluded from spe¬ 
cies-specific analyses but were included in 
analysis of total grassland bird density. 
Vegetation Sampling. —We sampled vegetation 
structure and plant species composition at each 
site. Each 20-m wide transect in sites with 
transects was partitioned into 20-m intervals for 
vegetation sampling plots. We measured canopy 
closure in each plot using a spherical densitometer 
(Lemmon 1956), and basal area with a 10-factor 
prism (Avery 1967) and Biltmore stick (Jackson 
1911). We measured herbaceous and woody plant 
density using a randomly placed 3-cm diameter 
pole (Wiens 1974), and modal herbaceous and 
woody heights within 30-cm of the pole were 
visually estimated to the nearest decimeter. We 
used a randomly placed 1-m 2 frame to estimate 
percent cover of herbaceous and woody ground 
vegetation, and species composition. We estimat¬ 
ed the number of woody stems at ground level 
inside each 1-trr frame. We later grouped plant 
species into 15 guilds by combinations of lif e 
form (graminoid, forb, or woody) and U.S. Fish 
and Wildlife Service Wetland Indicator Status 
(W1S) (USDI 1988) to reduce the number of 
variables. We collected the same vegetation data 
for RCW clusters using 5—10 randomly placed 
sampling plots per cluster. Two 1-m 2 frame and 
four pole measurements were taken in each 
sampling plot lor all stand types. Many of our 
study sites had patchy distributions of herbaceous 
cover and shrubs. We used the coefficient of 
variation (CV) to measure patchiness for herba¬ 
ceous cover, woody cover, and herbaceous density 
among points in each transect (Wiens 1974). The 
CV was calculated using each individual mea- 
