Occupancy Modeling of Woodpeckers: Maximizing Detections for Multiple Species With Multiple Spatial Scales

The role of cavity excavators in forest landscapes has cascading effects involving numerous species of birds, mammals, insects, and fungi (Bull and Jackson 2011). As a result, woodpeckers are often considered keystone species (Martin and Eadie 1999; Aubry and Raley 2002). Many forest woodpeckers are associated with habitats that include large trees and dead wood for foraging and nesting (Mikusiński 2006; Drever et al. 2008), and their sensitivity to timber harvest is well recognized (Imbeau et al. 1999; Roberge and Angelstam 2006; Bull et al. 2007). Because many forest birds have similar habitat requirements, managing for woodpecker diversity should also benefit general forest bird diversity (Martin and Eadie 1999; Drever and Martin 2010). Indeed, Mikusiński et al. (2001) and Roberge and Angelstam (2006) have shown a correlation between woodpecker richness and other forest bird richness at the landscape scale.

The pileated woodpecker (Drycopus pileatus) is generally associated with mature or old-growth forest types (Bull and Jackson 2011) and excavates cavities that are much larger than most other woodpecker species and provide roosting, nesting, and food caching opportunities for various secondary cavity users such as flammulated owl (Otus flammeolus), American kestrel (Falco sparverius), common goldeneye (Bucephala clangula), American marten (Martes americana), fisher (Martes pennant), and numerous species of bats (Bonar 2000; Aubry and Raley 2002; Martin et al. 2004; Bull and Jackson 2011). Excavated cavities additionally facilitate ecological processes by encouraging decomposition directly as well as indirectly by exposing wood for insect and fungal attack (Aubry and Raley 2002). Besides being an ecological engineer, pileated woodpeckers may depress insect outbreaks that negatively affect the commercial value of forest stands (Aubry and Raley 2002; Edworthy et al. 2011).

The American three-toed woodpecker (Picoides dorsalis) is also generally associated with mature or old-growth forest types (Imbeau et al. 1999; Leonard 2001; Hoyt and Hannon 2002). American three-toed woodpeckers prefer large snags in moderately burned stands, which may restrict distributions in some areas to recently burned forests (Hutto 1995; Kotliar et al. 2008). Because of their association with natural disturbances, American three-toed woodpeckers are considered susceptible to habitat loss due to fire suppression and salvage logging practices (Imbeau et al. 1999; Leonard 2001; Hoyt and Hannon 2002). In Idaho, it is considered a sensitive species for which population viability is a concern due to predicted downward trends in habitat suitability that would reduce the existing distribution (IDFG 2005). Monitoring of American three-toed woodpeckers is difficult because although they are generally sedentary, they can have irruptive movements that track with insect outbreaks (Yunick 1985). Similar to the American three-toed woodpecker, the Eurasian three-toed woodpecker (P. tridactylus) is considered to be a valuable indicator of species richness in European coniferous forests (Roberge and Anglestam 2006).

Monitoring populations of forest woodpeckers can be useful for informing management decisions regarding these and other forest species (Aubry and Raley 2002; Drever and Martin 2010). Usefulness of monitoring efforts, however, relies on the metrics estimated and methods used. Quantitative estimates of abundance, survival, and fecundity are generally considered ideal metrics for monitoring wildlife populations (Anderson and Gutzwiller 2005; Lancia et al. 2005). However, it can be difficult to obtain estimates of abundance or demographic rates for many populations and the cost of such studies cannot be justified in many cases, particularly over large spatial scales and for multiple species. Estimating the proportion of area occupied by the species is an attractive alternative that has been utilized for monitoring numerous species, including birds (Collier et al. 2010; Bruggeman et al. 2011; Hansen et al. 2011), terrestrial mammals (Moritz et al. 2008; Ahumada et al. 2011), primates (Karanth et al. 2010), bats (Weller and Baldwin 2012), amphibians (Jackson et al. 2006; Gould et al. 2012), and reptiles (Zylstra et al. 2010; Sewell et al. 2012). This method is based on detection– nondetection data and can be used over relatively large spatial scales to monitor trends in occupancy simultaneously for multiple species (Schultz et al. 2012). Additionally, with the use of multiple observation occasions, it is possible to estimate the probability of detecting a species, which can greatly improve accuracy of occupancy estimates (Pollock et al. 2002; MacKenzie et al. 2003, 2006).

Forest birds are commonly surveyed using the point-count method where an observer remains stationary and records all birds seen or heard over a defined period of time within a defined distance of the observer (Hutto et al. 1986; Lancia et al. 2005). Woodpeckers are generally thought to be conspicuous, owing to their distinctive calls, drumming patterns, and bold colors (Blackburn et al. 1998). Numerous studies have used point-count methods for surveying woodpeckers, particularly during concurrent surveys for other bird species (Hutto 1995; Imbeau et al. 1999; Kotliar et al. 2008; Krementz et al. 2012). However, woodpeckers typically have larger territories and vocalize less frequently than most song birds (Blackburn et al. 1998; Farnsworth et al. 2002), suggesting that a substantial proportion of individuals may not be detected using standard point-count methods. Johnson et al. (1981) suggested broadcasting recorded calls to survey avian species with these characteristics more efficiently. Shackelford and Conner (1997) noted that vocally mimicking a barred owl (Strix varia) often induced woodpeckers to respond by vocalizing or moving closer to the source of the sound; the authors reported a 71% increase in woodpeckers detected after vocally mimicking a barred owl call compared with using silent point counts in Texas. Similarly, Kumar and Singh (2010) detected more than twice as many individuals and a greater number of woodpecker species using playback of recorded calls in tropical forests.

In this study, our primary goal was to assess the distribution and area of occupancy of pileated and American three-toed woodpeckers within the Selway–Middle Fork Clearwater Collaborative Forest Landscape Restoration Program (CFLRP) project area in the Nez Perce–Clearwater National Forest using a single sampling scheme with a rigorous ability to collect data from multiple species with disparate spatial scales. The CFLRP is a federally sponsored program with the purpose of encouraging collaborative, science-based ecosystem restoration of priority forest landscapes. In this CFLRP landscape, the pileated and American three-toed woodpeckers are considered a management indicator species (species whose populations are thought to reflect the effects of management activities on various habitats) and a “species of greatest conservation concern” (IDFG 2005), respectively. Secondarily, we were interested in comparing results from an occupancy analysis using silent point counts and playback surveys that incorporated estimates of detection probability for these two woodpecker species.

We conducted our study in the Clearwater Mountains of north-central Idaho (46.097°N, −115.690°W), on the Nez Perce–Clearwater National Forest. The topography is mountainous with areas of steep, rugged terrain and few open valleys and meadows. Elevation ranges from 440 to 2,075 m, and annual precipitation ranges from 106 to 174 cm (Natural Resource Conservation Service 2010). The climate is Pacific maritime with cold, snowy winters and short, warm summers. The habitat is primarily mixed coniferous forest on the mountain slopes with narrow or no riparian areas along streams. At low to mid-elevations, the forest is comprised primarily of Douglas fir (Pseutotsuga menziesii), western larch (Larix occidentalis), grand fir (Abies grandis), and western red cedar (Thuja plicata); at higher elevations the forest transitions to subalpine fir (Abies lasiocarpa), Engelmann spruce (Picea engelmannii), mountain hemlock (Tsuga mertensiana), and lodgepole pine (Pinus contorta), with an increasing incidence of mixed shrub fields (Alnus spp., Salix spp, Ceanothus spp., Phyocarpus spp., Sorbus spp.) and mountain meadows. The National Forests have been managed under multiple-use and roadless/wilderness frameworks, which has resulted in a mixed pattern of stand structures and canopy covers, ranging from open clear-cuts, shrub fields, and regenerating forest to mature forest and old growth stands (Multiple-Use and Sustained Yield Act 1960).

During a typical point-count survey many woodpeckers that are present may remain undetected for numerous reasons (e.g., conspicuousness, study design, observer experience). Using presence–absence data as an estimate of occupancy is termed a “naïve estimate” and when detection probability (the probability of detecting a species when it is present, hereafter, p) is less than one, naïve estimates of occupancy are biased low (MacKenzie et al. 2006). We incorporated estimates of p to correct naïve estimates, resulting in unbiased estimates of occupancy.

We used a spatially balanced sampling design to select 44 1-km² sampling units from the western portion of the Middle Fork CFLRP Project (Stevens and Olsen 2004). We used a hierarchical multiscale sampling strategy where each sampling unit was composed of four survey stations to facilitate simultaneous sampling of other species at appropriate scales (Pavlacky et al. 2012). The hierarchical design permits simultaneous estimates of large-scale occupancy (ψ) at the sampling unit level and small-scale occupancy (θ) at the survey station level (Pavlacky et al. 2012). The latter can be interpreted as availability and is defined as the probability of the species occupying a survey station, given it is present within the sampling unit. Stations were positioned 250 m from the edge of the sampling unit such that there was 500 m between the four points, which is consistent with other woodpecker research (Raley and Aubry 1993; Hartwig et al. 2002; Wrightman and Saab 2005). We used the harvest history from the Nez Perce–Clearwater National Forest to stratify our sample grids between actively managed landscapes (i.e., those areas with some form of timber harvest) and unmanaged landscapes (i.e., those without a history of any timber harvest). We allocated our sampling effort to 70% actively managed landscapes and 30% unmanaged landscapes.

We used playback surveys to detect presence of both woodpecker species (Johnson et al. 1981). Surveys were conducted between 0600 and 1100 hours and all four stations within a sample unit were surveyed on the same morning. The survey protocol consisted of a 6-minute period of silent listening, a 6-minute playback survey for American three-toed woodpeckers, and a 6-minute playback survey for pileated woodpeckers, always in that order. If a dominance structure exists among woodpecker species, broadcasting calls from species of greater dominance may reduce detections of subordinate species. Though information on the dominance structure between these species is lacking, we choose to play the American three-toed woodpecker calls first due to its significantly smaller size, believing it would most likely be the subordinate species. Playback surveys consisted of alternating 30 seconds of calls and drumming from the species of interest, and 30 seconds of silent listening. If a species was detected during the silent listening phase, the phase was continued for potential detections of other species and the call playback phase was still conducted for that species. However, as a logistical time-saving measure, once we detected a species during its call playback phase (e.g., 2 minutes into the pileated call playback phase a pileated was detected) we discontinued the survey. We did not survey stations in proximity to running water such that audibly detecting woodpeckers was inhibited. We used a Foxpro NX3 digital game caller (FOXPRO Inc., Lewistown, PA) to broadcast calls and rotated direction of the caller 120° after each 1-minute call cycle, completing two rotations during each 6-minute playback survey. We used a volume level such that field technicians could not hear the recording at >250 m away; however, the ability of woodpeckers to hear the recording at greater distances was not known. We detected woodpeckers visually or by call, and recorded in which of the six 1-minute intervals of the survey the detection was made.

Before each survey began, we measured habitat variables within 50 m of the survey station. This allowed birds to settle after initial disturbance from entering the site before beginning each survey. The habitat variables included number of snags >23 cm in diameter at breast height (DBH) and >3 m high (Wrightman and Saab 2008), height of the base of the canopy measured with clinometers and a rangefinder, and percent ground covered with dead and downed trees with >23 cm diameter (course woody debris, hereafter CWD). All habitat variables were estimated visually from the sampling point to reduce movement that might affect woodpecker activity. We calculated naïve occupancy as the proportion of sampling units a species was detected separately for detections during the silent period and the playback period.

Studies on habitat use have indicated that pileated woodpeckers use old-growth forests with ≥60% canopy closure and use is related to density of snags and downed trees and absence of logging (Bull and Holthausen 1993). Three-toed woodpeckers appear to select habitat with mature and old-growth forests for foraging and roosting (Goggans et al. 1989) and forage in areas with trees of greater DBH compared with that available (Kotliar et al. 2008). We drew from these key findings and general landscape ecology concepts, and developed unique hypotheses to build a suite of a priori conceptual models. We identified landscape metrics that best captured the conceptual models and used program FRAGSTATS 3.3 (McGarigal et al. 2002) to calculate the metrics around each of our sample stations and sample units. We buffered sampling stations by a 250-m radius and the centroids of sample units by 1,250 m, resulting in an area roughly the size of a breeding pair of pileated woodpeckers' home range (490 ha; Mellen et al. 1992; Bull and Holthausen 1993).

For habitat classes, we used layers from the LANDFIRE data set (2006) including canopy cover and canopy height. We updated these layers with data from recent forest harvests using a tassel-cap soil transformation (Healey et al. 2005) of paired LANDSAT Thematic Mapper images in the DeltaCue add-on to ERDAS Imagine (Intergraph Inc., Norcross, GA). We used the Spatial Analyst extension in ArcGIS (ESRI Inc., Redlands, CA) to resample habitat layers and apply a minimum mapping unit of 1 ha. We collapsed the number of categories in the LANDFIRE data due to sparse data. The resulting categories were % landscape with 0–9.9%, 10–39.9%, 40–69.9%, and 70–100% canopy cover and % landscape with <5 m, 5–9.9 m, 10–24.9 m, and 25–50 m canopy height. We limited the potential large-scale occupancy covariates in our a priori models to 40–69.9% and 70–100% canopy cover, and 25–50 m canopy height as we felt these metrics would be most useful for describing mature and old-growth forest structure. Additionally, we included the station-scale covariates of number of snags, CWD, and canopy height as potential small-scale occupancy covariates.

We divided the 6-minute playback survey into equal periods to create occasions and used stations within a unit as our replicates (Pavlacky et al. 2012). One of the assumptions of occupancy estimation is that detections at a station are independent of each other; that is, detections of an individual species are not more or less likely, subsequent to first detection (MacKenzie et al. 2006). As this was not likely to be true given our method of playback surveys, we used a removal design and only considered detection histories up to first detection at each station for a given species (Farnsworth et al. 2002). This design is unable to estimate unique detection probabilities for each occasion and requires a constraint, such as constant p among occasions (MacKenzie et al. 2006). We examined our data by minute of survey for a constant decline in detections, as would be expected under the assumption of a constant p (Pavlackey et al. 2012). If this were true, we used the first 3 minutes for occasion 1 and the second 3 minutes for occasion 2 for each survey station. If equal p could not be assumed, we divided the 6-minute playback period into the fewest number of occasions of equal length such that the last two periods showed a steady decline in detections and a constant p could be assumed over these periods. Because of the limitations of the removal model and the limited number of observation occasions, we did not consider any covariates to describe p.

Models were fit and parameters estimated for pileated woodpecker and American three-toed woodpecker separately using program MARK (MARK Version 6.1, www.phidot.org, accessed 27 September 2011). We used Akaike's information criterion corrected for small sample sizes (AIC_c) to compare models and considered any models with ΔAIC_c < 2 of the best-fit model to be equally parsimonious (Burnham and Anderson 2002).

We surveyed 167 stations in 44 units for detection–nondetection of pileated and American three-toed woodpeckers from 12 April to 17 June 2012. We were unable to conduct counts at nine sample stations within sample units either due to time constraints or noise interference.

Pileated woodpeckers were detected at 22 stations in 17 units during the silent listening period, resulting in a naïve estimate of occupancy at the unit scale (ψ) of 0.39 (Table 1). During the playback surveys, we detected pileated woodpeckers at 44 stations in 26 units, increasing the naïve estimate of ψ to 0.59. Frequency of calls decreased from the first 3 minutes to the second 3 minutes, so we used a 2-sampling occasion model and assumed a constant p for the 6-minute playback survey. The top supported model describing occupancy for the pileated woodpecker was the null model (i.e., a single time- and habitat-invariant estimate for each parameter ψ, θ, and p; Table 2). The p for each 3-minute period of the playback survey was 0.31, resulting in 0.52 probability of detecting pileated woodpeckers during the 6-minute playback survey. Accordingly, accounting for imperfect detection, our corrected estimate of ψ was 0.70. Furthermore, given that pileated woodpeckers were present at the sample unit scale, the probability of occupancy for any single sampling station (i.e., availability, θ) was estimated to be 0.73.

The three models that included habitat covariates and their possible influence on ψ had ΔAIC_c < 2 (Table 2). Considering the greater number of parameters in these models and the only minor improvement in deviance estimates, there was very little support for any model with habitat covariates (Burnham and Anderson 2002; Arnold 2010).

American three-toed woodpeckers were detected at seven stations in six units during the silent listening period, for a naïve estimate of ψ  =  0.14 (Table 1). During the playback surveys, American three-toed woodpeckers were detected at 19 stations in 15 units, increasing the naïve estimate of ψ  =  0.34. Detections of individuals were low during the first 2 minutes of the playback survey, peaked during minute 3, and decreased over the remaining 3 minutes. Accordingly, we fitted models using three 2-minute occasions, allowing p in the first occasion to differ from a constant p in the remaining two occasions. Thus, from the null model, estimates of p were 0.13 during the first 2 minutes of the playback survey and 0.33 for minutes 3–4 and 5–6. The probability of detecting American three-toed woodpeckers during the entire 6-minute playback survey was 0.61. The unbiased estimate of ψ for American three-toed woodpeckers was 0.71. However, given that American three-toed woodpeckers were present at the sample unit scale, the estimated probability of occupying any sampling station (θ) was only 0.26.

There were three apparently equally parsimonious models describing occupancy for the American three-toed woodpecker, the null model and two models with covariates describing ψ (Table 3). Because there was little improvement in estimated deviance with additional covariates, there was little support for models more complex than the null model (Burnham and Anderson 2002; Arnold 2010).

We estimated occupancy of pileated and American three-toed woodpeckers at two spatial scales while accounting for the probability of detecting each species. Our method of dividing a single observation into multiple occasions and using a removal framework allowed estimation of p without the typical requirement of performing surveys during repeat visits to each station. By using the hierarchical multiscale framework in our analysis, we were able to tease apart small-scale availability from detection probability, resulting in a more informative analysis of occupancy for both species.

Our estimates of occupancy suggest that both pileated woodpeckers and American three-toed woodpeckers were widely distributed throughout the Selway–Middle Fork CFLRP area. For both species, the probability of occupation at any randomly selected 1-km² survey unit was about 70%. Detection probabilities over the 6-minute playback survey were similar; 0.52 and 0.61 for pileated and American three-toed woodpeckers, respectively. When corrected for detection probability, our estimate of occupancy increased from 0.59 to 0.70 for pileated woodpeckers and more than doubled from 0.34 to 0.71 for American three-toed woodpeckers. Failing to correct for imperfect detection would have resulted in significantly different conclusions regarding the distribution and area of occupancy of these species in our study area. However, simultaneously sampling for multiple species is not without some trade-offs. We standardized our surveys by always playing American three-toed woodpecker calls before those of pileated woodpeckers. How this might influence the probability of detection of a pileated woodpecker is unknown; if pileated woodpeckers are attracted to or avoid the calls of American three-toed woodpeckers, there maybe be some consistent bias in our detection probability estimate.

Whereas both woodpecker species showed similar patterns in large-scale occupancy (i.e., at the sample unit scale), estimates of small-scale occupancy were rather disparate between the two species. Within survey units where pileated woodpeckers were present, our models predicted that the species would occupy areas covering three of the four survey stations. American three-toed woodpeckers were estimated to occupy areas covering only one of the four survey stations within units occupied. These estimates are reflected in the species' respective home range estimates. Mellen et al. (1992) estimated the average summer home range for 11 individual pileated woodpeckers in coastal Oregon of 478 ha and noted that home ranges for pairs were even larger after chicks had fledged. Bull and Holthausen (1993) reported home ranges for seven breeding pairs from June to March between 321 and 630 ha with an average of 407 ha in northeastern Oregon. Territory size of American three-toed woodpeckers has not been widely documented; however, Goggans et al. (1989) estimated home ranges for three individuals after the breeding season at 53, 147, and 304 ha. In a study of Eurasian three-toed woodpeckers in Germany, average nesting-season home ranges for 10 pairs was estimated to be 86 ha (Pechacek 2004). Our results suggest that although the two species appear to occupy the same proportion of 1-km² units in our study area, American three-toed woodpeckers appear locally rare and are less likely to be detected because of their lower availability, indicating that available habitat is not saturated with birds.

One of the benefits of the occupancy framework that we applied is the ability to model occupancy as a function of environmental covariates (MacKenzie et al. 2006). As pileated woodpeckers are often considered a management indicator species of mature forest characteristics (Bull and Jackson 2011) and as American three-toed woodpeckers are generally associated with mature or old-growth forest types (Imbeau et al. 1999; Leonard 2001; Hoyt and Hannon 2002), we hypothesized that the percentage of a landscape composed of large trees or heavy canopy cover would influence the occupancy of pileated or American three-toed woodpeckers. However, we did not find strong evidence that any of our environmental covariates helped explain variation in occupancy at either scale for either species better than a simple “null” model. This result was unexpected and warrants further investigation. It is possible that our covariates are not representative of the pattern we were attempting to detect, imprecisely estimated, measured at an inappropriate scale, or that our sample size was insufficient. However, on the basis of our results, we suggest that the assumption of the general association of these woodpecker species with mature forests to be continually challenged with the best analytic methods such that the specifics of habitat requirements for each species become better understood. Such information would allow managers to decide the appropriateness of using pileated woodpeckers as a management indicator species for mature forest characteristics. Furthermore, we feel that if future work across Idaho on American three-toed woodpeckers shows corrected occupancy estimates consistent with ours, their designation as a “species of greatest conservation need” in the state may be unwarranted due to their wider-than-originally expected occurrence.

Our use of playbacks greatly increased the number of detections, resulting in approximately a twofold increase in naïve estimates of occupancy over silent surveys. This method, however, violates an assumption of independence in detections among the six 1-minute intervals and requires the use of a removal model for calculating unbiased estimates of occupancy. The removal model uses only first detections at a survey station for estimating p and generally results in reduced precision compared with a nonremoval model unless number of sampling occasions is increased (MacKenzie and Royle 2005; MacKenzie et al. 2006). Precision of our estimates were poor, particularly p for American three-toed woodpeckers. Poor precision in our estimates of p may also have been the result of variation in detection probability through the season due to breeding behavior. Birds are typically less vocal during incubation than during breeding and pileated woodpeckers response to playback call is known to vary with nesting chronology (Raley and Aubry 1993). The timing of our fieldwork (mid-April to mid-June) spanned three phases of breeding: courtship, incubation, and hatching (Leonard 2001; Bull and Jackson 2011), and thus our detection probability represents detectability across these phases. Recognizing factors such as these and incorporating them into the modeling framework generally improves parameter estimates. Our use of the removal design, coupled with few observation occasions, hindered our ability to incorporate these types of covariates into estimates of p. Repeat visits over time could improve nesting chronology-specific estimates of p and have the additional advantage of improved precision of occupancy estimates (MacKenzie et al. 2002). However, this would come at the cost of a relatively large increase in effort and expense. We feel that the removal design we used balanced the need to correct naïve occupancy estimates for detection probabilities with the practical logistical constraints of limited budgets and personnel.

Natural resource managers need to ensure that the metrics they collect regarding wildlife populations are accurate, yet often they have limited budgets to work with that preclude techniques that provide abundance or demographic rates. Estimating the proportion of area occupied by a species is an attractive alternative. But when detections are imperfect (<1.0), naïve occupancy estimates are biased low and using such data as the basis for management decisions would be imprudent. Furthermore, with ever-shrinking budgets, wildlife managers are increasingly interested in multiple-species sampling frameworks that are robust to disparate population scales. Our application of a hierarchical, multiscale occupation framework allowed us to use the same sampling stations for species with dissimilar territory sizes, yet tease apart availability from detection probability, resulting in greatly improved parameter estimates.

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 for the article.

Table S1. Data table containing detection histories of pileated woodpeckers (Drycopus pileatus) and American three-toed woodpeckers (Picoides dorsalis) in the Clearwater Mountains of north-central Idaho,2012. Unit_ID  =  unique unit number. Each unit contained four sampling points (A, B, C, D). 1–3 indicates the first 3-minute occasion that pileated woodpeckers were surveyed, 4–6 indicates the second 3-minute occasion. 1–2 indicates the first 2-minute occasion that American three-toed woodpeckers were surveyed, 3–4 indicates the second 2-minute occasion, and 5–6 indicates the third 2-minute occasion. The number 1 indicates that a woodpecker was detected during that occasion, a 0 indicates that the woodpecker was not detected, and a dot (.) indicates that the survey was not conducted for the occasion, either due to the species being detected in an earlier occasion or due to time constraints or noise interference. Management indicates whether the unit was in a managed stand (Ma) or unmanaged stand (Un). PLAND CanCov  =  percentage of the unit with 40–69.9% canopy cover, or 70–100% canopy cover. PLAND CanHeight  =  percentage of the unit with canopy height of 25–50 m. NSnags  =  number of snags >23 cm in diameter at breast height (DBH) and >3 m high within 50 m of each survey station (A–D). Height  =  height of the base of the canopy at each survey station (A–D). CWD  =  percent ground covered with dead and downed trees with >23 cm diameter within 50 m of each survey point (A–D).

Found at DOI: 10.3996/042013-JFWM-031.S1 (32 KB XLSX)

Reference S1. Aubry KB, Raley CM. 2002. The pileated woodpecker as a keystone habitat modifier in the Pacific Northwest. Pages 257–274 in Laudenslayer WF, Shea PJ, Valentine BE, Weatherspoon CP, Lisle TW, editors. Proceedings of the symposium on the ecology and management of dead wood in western forest. Berkeley, California: USDA Forest Service General Technical Report PSW-GTR-181.

Found at DOI: 10.3996/042013-JFWM-031.S2; also available at http://www.fs.fed.us/psw/publications/documents/gtr-181/001_TOC.pdf (261 KB PDF)

Reference S2. Goggans R, Dixon RD, Seminara LC. 1989. Habitat use by three-toed and black-backed woodpeckers, Deschutes National Forest. Oregon Technical Report. 87-3-02. Portland, Oregon: Oregon Depart of Fish and Wildlife and USDA Forest Service; 43 p.

Found at DOI: 10.3996/042013-JFWM-031.S3 (2.2 MB PDF)

Reference S3. Hartwig CL, Eastman DS, Harstead AS. 2002. Forest age and relative abundance of pileated woodpeckers in southeastern Vancouver Island. USDA Forest Service Gen. Tech. Rep. PSW-GTR-181.

Found at DOI: 10.3996/042013-JFWM-031.S4 (233 KB PDF)

Reference S4. Multiple-Use and Sustained Yield Act. 1960. Public Law 86-517. 86th Congress, June 12, 1960.

Found at DOI: 10.3996/042013-JFWM-031.S5; also available at http://www.fs.fed.us/emc/nfma/includes/musya60.pdf (58 KB PDF)

Reference S5. Raley CM, Aubry, KB. 1993. Protocol for pileated woodpecker call surveys and nest searches. Olympia, Washington: U.S. Forest Service, Pacific Northwest Research Station. Unpublished report.

Found at DOI: 10.3996/042013-JFWM-031.S6 (346 KB PDF)

Reference S6. Wrightman C, Saab V. 2008. Management indicator species surveys on the Payette National Forest 2008: Field testing of methods. Bozeman, Montana: USDA Forest Service, Rocky Mountain Research Station. Unpublished Report.

Found at DOI: 10.3996/042013-JFWM-031.S7; also available at http://www.fs.usda.gov/Internet/FSE_DOCUMENTS/stelprdb5292698.pdf (10.3 MB PDF)

We thank John McClung and Lance Swenson for field assistance and Dan Davis and Rema Sadak from the Nez Perce-Clearwater National Forest who provided substantial support and funding for the work.

Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.