Woodpecker Densities in the Big Woods of Arkansas Free

Bottomland hardwood forests of the Lower Mississippi Alluvial Valley (LMAV) have experienced habitat loss due to fragmentation and changes in hydrology (Twedt and Loesch 1999; King et al. 2005). However, the Big Woods (see study site description and Figure 1) of eastern Arkansas has remained relatively intact and is one of the largest stands of bottomland hardwoods in the Southeastern United States (The Nature Conservancy 2009). Therefore, this component of the LMAV provides a valuable model system for evaluating current densities and habitat use patterns of wildlife in this largely fragmented ecosystem. The large-scale loss of bottomland hardwood forests throughout the LMAV, in conjunction with the development of a recovery plan for the ivory-billed woodpecker Campephilus principalis, sighted in 2004 in the Big Woods (Fitzpatrick et al. 2005) by the U.S. Fish and Wildlife Service (USFWS 2006), has raised concerns by federal, state, and nongovernmental agencies on how best to manage bottomland hardwoods in the LMAV. In response to these concerns, the Lower Mississippi Valley Joint Venture (LMVJV) Forest Resource Conservation Working Group developed a set of habitat management guidelines and monitoring schemes for forest-interior birds, emphasizing woodpeckers and with a focus on the ivory-billed woodpecker (Wilson et al. 2007). One of their main objectives was to estimate densities of forest-interior birds, with an emphasis on woodpeckers.

In order to develop habitat guidelines for ivory-billed woodpeckers and woodpeckers in general, the LMVJV needs a better understanding of their habitat use. Unfortunately, habitat needs of ivory-billed woodpeckers are not well-known and current recommendations are primarily based on information collected approximately 65 y ago (Jackson 2002) by Tanner (1942). Under the ivory-billed woodpecker Recovery Plan (USFWS 2006), the USFWS requested assumption-driven research regarding ivory-billed woodpecker habitat use. Tanner's (1942) study, albeit old, provides the only baseline information from which to make assumptions directing research and management for this species. Tanner (1942) suggested that pileated woodpeckers Dryocopus pileatus and red-bellied woodpeckers Melanerpes carolinus could be used as surrogates to study ivory-billed woodpecker habitat use. Tanner believed that densities of 3.9 pileated or 19 red-bellied woodpeckers/km² would indicate suitable ivory-billed woodpecker habitat, based on densities of these species in the home range of 1 ivory-billed woodpecker. Tanner (1942) also indicated that in addition to abundances of pileated and red-bellied woodpeckers, abundances of other species of woodpeckers could be indicative of quality ivory-billed woodpecker habitat. Based on these hypotheses by Tanner (1942), if the Big Woods can support at least 1 ivory-billed woodpecker, then there should be at least 3.9 pileated or 19 red-bellied woodpeckers/km². Our objective here was to estimate densities of six woodpecker species in the Big Woods of Arkansas to evaluate the suitability of the region for ivory-billed woodpeckers.

Before we developed our sampling scheme, it was necessary to understand the habitat use of ivory-billed woodpeckers. Based on historical accounts (Jackson 2002), we know that ivory-billed woodpeckers had some habitat affinities including a selection for: 1) extensive continuous forest areas, 2) very large trees, 3) continuous supply of recently dead trees, 4) an open canopy, 5) certain tree species (e.g., sweetgum Liquidambar styraciflua, Nuttall oak Quercus texana, see forest type 3 below), and 6) avoidance of other tree species (e.g., overcup oak Q. lyrata, water hickory Carya aquatic, see forest type 2 below). These habitat affinities probably met both foraging needs, and nesting or roosting tree requirements. Whether any one, or some, combination of these variables met some limiting requirement is unknown, but Tanner (1942) suggested that forage availability was an important determinant for the presence of ivory-billed woodpeckers in a particular woodland tract.

The Big Woods of Arkansas encompasses about 222,500 ha of the Cache, Arkansas, and White River floodplains and Bayou DeView in the Arkansas Delta (Figure 1). Our study areas included Sheffield Nelson Dagmar, Trusten Holder, Rex Hancock Black Swamp, Bayou Meto, Mike Freeze Wattensaw, and Henry Gray Hurricane Lake Wildlife Management Areas of the Arkansas Game & Fish Commission, and Cache River (CRNWR) and White River (WRNWR) National Wildlife Refuges. There were approximately 28,000 ha in the Wildlife Management Areas, 25,000 ha in CRNWR, and 65,000 ha in WRNWR, hereafter “units.” This is the largest corridor of bottomland hardwood forest remaining in the LMAV north of Louisiana's Atchafalaya Basin. The bottomland hardwood forests that occur here include a wide range of species and community types that can tolerate inundation or soil saturation (Wharton et al. 1982), resulting in a complex mosaic of community types reflecting differences in the alluvial and hydrologic environment (Smith and Klimas 2002). We recognized five hydrologic forest types (Wilson et al. 2007): 1) swamp forests—bald cypress Taxodium distichum, bald cypress–water tupelo Nyssa aquatica; 2) wet bottomland forest—overcup oak–water hickory, black willow Salix nigra, laurel oak Q. laurifolia–red maple Acer rubrum; 3) moist bottomland forest—sugarberry Celtis laevigata, elm Ulmus spp., ash Fraxinus spp., oak–elm–ash, oak–sweetgum; 4) dry bottomland forest—cherrybark oak Q. pagoda, swamp chestnut oak Q. michauxii, post oak Q. stellata, blackgum N. sylvatica; and 5) levee forest—cottonwood Populus deltoides, sycamore Platanus occidentalis, pecan Carya illinoiensis, boxelder Acer negundo.

Forest management across the study area is difficult to characterize because ownership spans three federal agencies, two state agencies, and numerous private landowners. Over the past 10 y, about 50% of the Big Woods has been harvested using partial cuts with a target of retaining larger, older age-class trees of various sizes (J. Denman, USFWS, personal communication). On federal properties, this objective has been pursued for > 20 y (Denman and Karnuth 2005). Private landowners tend not to manage for older trees. The partial cuts, where portions of the overstory are removed, vary from 30 to 90% removal with an overall average of 50% overstory opening (Denman and Karnuth 2005). This management approach results in a multiple-canopied condition that is representative of the “Desired Forest Conditions” of the LMVJV Forest Resource Conservation Working Group (Wilson et al. 2007). About 10% of the Big Woods undergo forest management in any single year with private lands initiating more actions than on public lands.

The LMVJV developed an ivory-billed woodpecker habitat assessment scheme based on the above five forest types, with the objective of quantifying current habitat conditions in the Big Woods of Arkansas (Wilson et al. 2007, appendix 4). The sampling frame was centered on those areas with evidence of ivory-billed woodpecker existence with additional nearby areas added to increase the chance of observing ivory-billed woodpeckers. Individual units within the area of interest were broken down into stands of approximately 200 ha each. Within each stand (n  =  146), an average of four randomly allocated point-transects (n  =  587) were located (transect numbers ranged from two to five based on the size of the stand). Along each transect, five equidistant plots (80 m apart) were established. At WRNWR only, the four transects were connected and nonrandomly placed due to logistical constraints. At each of these plots (0.08 ha each, n  =  2,935), we selected our woodpecker survey plots across three strata pertaining to tree density (mean trees/ha, mean trees > 61 cm diameter at breast height/ha), tree health (mean snags/ha, mean epicormic branching score/transect), and predominant tree species per stand. Epicormic branching is disturbance-induced development of new branches on the bole of a mature tree (Barnes et al. 1998). The LMVJV calculated these habitat metrics at the transect or stand level (R. Wilson, USFWS, personal communication). Based on logistical constraints, we randomly chose one out of the four transects in each stand. When possible, we used the center plot of a transect line as the woodpecker survey location. Upon reviewing the habitat metrics along transects selected, we found that some forest types or habitat metrics that we were interested in surveying were poorly or not represented. We were particularly concerned that swamp forest stands were not well-represented in our initial sample because all sightings of ivory-billed woodpeckers in the Big Woods in 2004–2005 were in swamp forests (Fitzpatrick et al. 2005). To address this problem, we added four swamp forest stands to include this missing forest type.

We used a repeated-survey distance-sampling approach based on the recommendations of Wilson et al. (2007) to estimate densities of each woodpecker species. We attempted to survey each plot five times during three seasons: breeding season 2006 (2 May–17 June, n  =  82 plots), winter 2007 (16 January–25 February, n  =  63 plots), and breeding season 2007 (25 May–1 July, n  =  87 plots). Due to logistical constraints, we surveyed a different number of plots during each season. We did not survey many of our sites in WRNWR during summer 2006 due to lack of manpower and due to high flood levels during winter 2007. We began surveys in the morning when it was light enough to see 200 m and ended them no later than 1000 hours. Surveys were not conducted when wind speeds exceeded Beaufort scale No. 4 (i.e., 20–28 km/h) or during heavy rain. Any woodpecker seen or heard during the 10-min period was marked in one of four predetermined distance bands (0–25 m, 25–50 m, 50–100 m, and > 100 m). Because different technicians were used each survey season, we followed the advice of Robbins and Stallcup (1981) by: 1) training each set of observers in advance with appropriate books and sound recordings, 2) field training to familiarize observers with field conditions and local dialects, 3) examining field records of all observers for comparability during early phases of each survey season, and 4) rotating observers among plots so that surveys would be as comparable as possible and to reduce the possibility of overlooking or misidentifying species. Also, observers practiced estimating distances with laser rangefinders. As a final precaution before collecting field data, we conducted simultaneous surveys with all technicians independently field-recording woodpecker detections. Any problems with species identifications were discussed at the end of these initial surveys during the training period to correct future identification errors (Kepler and Scott 1981).

We used the Multiple Covariate Distance Sampling analysis engine in program DISTANCE 5.0 (Thomas et al. 2005; Marques et al. 2007) to model detectability and estimate densities of each woodpecker species (individuals/km²) during the three seasons. There were not enough detections to estimate stand- or unit-level detection probabilities for any species. We estimated a global detection function for each species per season from all data pooled together, allowing for more precise density estimates (Marques et al. 2007). We used these global detection functions to estimate densities first at the level of CRNWR, WRNWR, and Arkansas Game and Fish Commission Wildlife Management Areas, and then at the level of the Big Woods. We also modeled the effective detection radius to assess whether the detectability for a species was a function of habitat or area surveyed (Buckland et al. 2001). A larger value can be suggestive of a considerably larger area surveyed.

The Multiple Covariate Distance Sampling analysis engine only allows hazard-rate and half-normal key functions for modeling detectability. Our candidate set of models for explaining detectability for each species included these key functions with a cosine series expansion and observer effects. We used Akaike's Information Criterion corrected for small sample size (AIC_c; Burnham and Anderson 2002) to rank these models and we used density estimates from the top-ranked model. We did not model average parameter estimates because currently program DISTANCE 5.0 does not allow model-averaging within the Multiple Covariate Distance Sampling analysis engine.

To place our woodpecker density estimates into perspective, we reviewed the literature for woodpecker density estimates derived from the Southeastern United States, usually from bottomland hardwood habitats. However, comparing density estimates based on different studies should be done with caution because different methods for estimating densities were used (see Bull and Jackson 1995, p 15).

We detected 4,047 individual woodpeckers across 3,588 detections, which varied in number by species and survey season. Thus, most detections were of a single woodpecker. We had enough detections to estimate densities of downy woodpeckers Picoides pubescens, red-bellied, and pileated woodpeckers during all three survey seasons. Red-headed woodpeckers M. erythrocephalus, yellow-bellied sapsuckers Sphyrapicus varius, and northern flickers Colaptes auratus are primarily winter visitors to the Big Woods; thus, we were only able to estimate their densities during winter 2007. In no season were we able to estimate densities for hairy woodpeckers P. villosus (Supplemental Material, Table S1, http://dx.doi.org/10.3996/032010-JFWM-006.S1; Table S2, http://dx.doi.org/10.3996/032010-JFWM-006.S2; Table S3, http://dx.doi.org/10.3996/032010-JFWM-006.S3).

We found the effective detection radius for downy and red-bellied woodpeckers was constant across seasons and varied around 60 m (Figure 2) while the effective detection radius for pileated woodpecker was higher during the winter as compared to the breeding season. The pileated woodpecker detection radius was about 100 m. The best-fitting detectability models were hazard-rate cosine and hazard-rate cosine plus observer effects. The only data sets that showed weak effects from varying observers in model selection results were pileated woodpeckers during winter 2007, red-bellied woodpeckers during winter and breeding seasons 2007, and yellow-bellied sapsuckers during winter 2007. For all other species and season combinations, detectability models with effects from varying observers were at least 36 times more plausible than models without. Detectability estimates (standard errors in parentheses) from top models ranged from 0.06 (0.003) to 0.49 (0.06) across all species during all survey seasons (Figure 3). While most detectability estimates were around 0.10, the pileated woodpecker had consistently higher probability of detections compared to the other woodpeckers each season.

Although the 95% confidence intervals (CIs) overlapped, there was a general pattern for pileated and downy woodpecker densities to be lower during the winter season than during the breeding season (Figure 4). The opposite pattern was true for red-headed woodpeckers, yellow-bellied sapsuckers, and northern flickers when densities could only be estimated during the winter. Unlike the other woodpeckers, the red-bellied woodpecker density estimates increased with each successive survey season. Downy woodpecker densities were similar each season. Downy and red-bellied woodpecker densities were higher than pileated woodpecker densities each season. Pileated woodpecker densities were always lower than the other estimable woodpecker densities. We detected at least 4.5 pileated woodpeckers and at least 98.6 red-bellied woodpeckers/km² during the winter and at least 16.6 pileated woodpeckers and at least 72.9 red-bellied woodpeckers/km² during the breeding seasons (Table). We found that compared to all other estimates, our pileated and red-bellied woodpecker estimates were as high as or higher than most estimates from the literature. Also, for all other comparable woodpecker density estimates that we could locate, ours were as high as or higher than estimates from other studies (Table).

Density estimates of six woodpecker species in the Big Woods of eastern Arkansas during both the breeding (2006 and 2007) and winter (2007) seasons were higher than estimates from other parts of the Southeast. Our estimates were at least 28% and 280% higher for pileated and red-bellied woodpeckers, respectively, than the densities Tanner (1942) suggested coincided with suitable ivory-billed woodpecker habitat. Following Tanner's hypothesis that high pileated and red-bellied woodpecker densities are indicative of good ivory-billed woodpecker habitat, then it follows that the Big Woods of Arkansas should be a preferred location for ivory-billed woodpeckers.

Southern bottomland hardwood forests harbor a diverse and abundant woodpecker community. For example, of the eight woodpecker species detected by Shackleford and Conner (1997) in eastern Texas, the red-bellied woodpecker, northern flicker, downy woodpecker, red-headed woodpecker, and yellow-bellied sapsucker were significantly more abundant in bottomland hardwood forests than in either nearby longleaf or mixed pine–hardwood forests. However, these woodpecker densities are not static because overall woodpecker densities increased markedly during the autumn and winter (Bock and Lepthien 1975; Shackleford and Conner 1997). These elevated woodpecker populations during the nonbreeding season result, in part, from migrating and wintering northern flicker, red-headed woodpecker, yellow-bellied sapsucker, and possibly red-bellied woodpecker moving into these southern forests (Shackelford and Conner 1997; Shackelford et al. 2000; Leonard and Strout 2006). Additionally the high use of bottomland hardwoods during the nonbreeding season results from the abundance of soft mast, an important food source during this time of year (Leonard and Strout 2006).

Why might the Big Woods be so attractive to woodpeckers, especially during the nonbreeding season? Wakeley et al. (2007) compared bottomland hardwood bird communities across the Southeastern United States, including the CRNWR. They concluded that the bird community at CRNWR was organized along “the hydrologic gradient acting either directly through avoidance of flooded areas during the breeding season or indirectly through selection of habitats having greater structural development of understory and midstory vegetation” (Wakeley et al. 2007, p 435). They implied that the altered hydrologic cycle of the rivers in the Big Woods resulted in disturbance to the plant community at various spatial scales, which in turn was organizing the bird community. We suspect that the frequent and prolonged flooding in the Big Woods (Denman and Karnuth 2005) stressed trees, resulting in increased food availability for the woodpeckers. The great variety of foraging methods employed by woodpeckers (Conner et al. 1994) suggests that the altered hydrologic disturbance was not affecting a single aspect of the forest but was impacting the forest at a variety of levels (e.g., under-, mid-, and overstory), and the variable topographic relief in the Big Woods contributed to the patchy availability of different plant communities. One primary aspect of the prolonged inundation of water in the forests was increased tree mortality (Conner and Sharitz 2005; 2005a, 2005b), which is highly correlated with arthropod biomass, the primary forage item of woodpeckers (Conner et al. 1994). Also, the Big Woods is composed predominantly of mature hardwoods (Denman and Karnuth 2005). These older growth areas are preferred by most woodpeckers not only because of the available forage but also because some woodpeckers (e.g., pileated woodpecker) excavate their nest and roost cavities in large trees (Evans and Conner 1979; Renken and Wiggers 1993; Conner et al. 1994; Tanner and Hamel 2001).

P.B. Hamel (United States Forest Service, personal communication) investigated Tanner's hypothesis that high pileated and red-bellied woodpecker winter densities could be used to aid in searching for ivory-billed woodpecker as well as characterizing habitats used by ivory-billed woodpeckers. He estimated pileated and red-bellied woodpecker winter densities across the Southeastern United States in bottomland hardwoods, and usually documented densities equal to or higher than those estimated by Tanner (1942). However, smaller woodpeckers, especially the red-bellied woodpeckers, might not have the same minimum criteria for nesting and roosting tree circumferences as the ivory-billed woodpeckers, nor would the foraging needs be as high or specific as the ivory-billed woodpecker. Thus, Hamel suggested that pileated and red-bellied woodpecker densities could act as surrogates for ivory-billed woodpecker but that those densities may not be the most effective strategy for developing searches or managing habitats for ivory-billed woodpeckers.

We found the greatest seasonal variation in densities in yellow-bellied sapsuckers, red-headed woodpeckers, and northern flickers. Seasonal differences in densities of these three woodpecker species were primarily due to migratory patterns and varying life-history patterns. The yellow-bellied sapsucker was only present during the winter season because it is migratory and breeds far to the north (Walters et al. 2002). Red-headed woodpeckers occurred in relatively high densities during the winter because they are both migratory and nomadic in foraging (Smith et al. 2000). We detected red-headed woodpeckers during both breeding seasons, but not enough to reliably estimate densities. Lastly, the situation for the northern flicker is less clear to us. Although the northern flicker is described as “strongly migratory,” the southern populations are sedentary (Moore 1995). We originally anticipated detecting sufficient numbers of northern flickers during the breeding seasons to estimate densities, but we did not. We assume from our results that northern flickers moved away from the Big Woods into nearby areas during the breeding season.

We found that the bottomland hardwood forests of the Big Woods of eastern Arkansas harbor some of the highest densities of woodpeckers reported in the Southeastern United States. Our high woodpecker density estimates for the Big Woods are likely due to the abundance of older growth hardwoods. In addition, the disturbance resulting from flooding in the area provides ample opportunities for foraging and excavating nest cavities. If one accepts Tanner's (1942) hypothesis that high woodpecker densities coincide with suitable ivory-billed woodpecker habitat, then it follows that the Big Woods of eastern Arkansas provide suitable habitat for ivory-billed woodpeckers.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author.

REGION  =  labels for public land type (federal wildlife refuges or state management units); UNIT = labels for the specific names of management units in each property; POINT = alphanumeric codes denoting the names of stands and the points along transects surveyed; DISTANCE = distance to bird detections; NUMBER = numbers of birds detected together; OBS = observer code.

Table S1. Woodpecker species survey data, Spring 2006.

Found at DOI: 10.3996/032010-JFWM-006.S1 (68.5 KB XLS).

Table S2. Woodpecker species survey data, Winter 2007.

Found at DOI: 10.3996/032010-JFWM-006.S2 (160 KB XLS).

Table S3. Woodpecker species survey data, Spring 2007.

Found at DOI: 10.3996/032010-JFWM-006.S3 (214 KB XLS).

We thank A. Clifton, K. Ercit, A. Finfera, D. Hollis, C. Kovach, A. Miller, H. Pruett, P. Tidwell, and M. Strauser for help collecting data. We thank M. Blaney, R. Crossett, J. Denman, T. Foti, E. Johnson, A. Keister, P. Hamel, R. Hines, C. Hunter, M. Lammertink, A. Mueller, S. Reagan, K. Ribbeck, C. Rideout, B. Uhlein, and R. Wilson for logistical support. Thanks to P.F. Doherty, Jr, P. Hamel, M. Kissling, S.E. Lehnen, P. Lukacs, T. Nudds, and four anonymous reviewers for comments on earlier versions of this manuscript. We thank the Arkansas Game and Fish Commission and (USFWS) Cache and White River NWRs for providing study areas.

Funding was provided by the USFWS and the USGS Arkansas Cooperative Fish & Wildlife Research Unit. The use of trade, product, industry or firm names or products is for informative purposes only and does not constitute an endorsement by the U.S. Government or the USGS.