2012 NORTHEASTERN NATURALIST 19(4):673–684 Local Distribution Factors and Sampling Effort Guidelines for the Rare Frosted Elfin Butterfly Jason T. Bried1,2, Jenny E. Murtaugh3, and Amanda M. Dillon1,* Abstract - Callophrys irus (Frosted Elfin) is threatened under New York State conservation law and has fewer than 5 secure populations within the state. Published research on these populations is needed to support the development of a state recovery plan and monitoring program for the species. We assessed the relationship between adult occupancy (patch use) in the Albany Pine Bush Preserve and a suite of potential controlling factors. We then used the results in a simulation framework to quantitatively inform how many sites and surveys are needed for Frosted Elfin occupancy monitoring. Patch use was best explained by a model that assumed the same occupancy probability for each patch. The species was more likely to use patches with limited shrub cover and greater host plant density, yet showed a good chance (≥76%) of using even the smaller patches (<1 ha) with relatively sparse density (<1000 ramets ha-1). Detection probability depended primarily on observer and survey date, ranging from 0.34 to 0.94 among observers and from 0.35 to 0.96 across surveys. In the worst-case scenario (i.e., low detectability and low intrinsic occupancy rate), minimum effort for adult Frosted Elfin occupancy monitoring in habitat similar to the Albany Pine Bush may require at least 20 habitat patches surveyed 6 times each or at least 10 habitat patches surveyed 8 times each. Less effort (e.g., 10 sites × 4 surveys) will likely suffice if surveys are restricted to the period of peak abundance. Adult occupancy (or patch use) is probably the most efficient state variable for monitoring Frosted Elfin populations, and changes in detection-corrected occupancy rate or proportion of area occupied could be useful for conservation planning . Introduction Callophrys irus Godart (Frosted Elfin) is a small and inconspicuous brown lycaenid butterfly. It is univoltine and non-migratory and, although it has a broad geographic distribution, occurs in small, localized populations, many of which are declining (NatureServe 2009, Schweitzer et al. 2011). The Frosted Elfin is one of a suite of specialist disturbance-dependent lepidopteran species threatened by degradation of disclimax and early successional habitat in the northeastern United States (Wagner et al. 2003). Where their distributions overlap, it has similar habitat requirements to the federally endangered Lycaeides melissa samuelis Nabokov (Karner Blue) and the phenologically similar Erynnis persius persius Scudder (Persius Duskywing) (Schweitzer et al. 2011, Shapiro 1974, Wagner et al. 2003). Compared to these two species, the Frosted Elfin has a much broader geographic range, spanning nearly 15 degrees of latitude. 1Albany Pine Bush Preserve Commission, 195 New Karner Road, Albany, NY 12205. 2Current address - Department of Zoology, Oklahoma State University, 501 Life Sciences West, Stillwater, OK 74078. 3New York State Department of Environmental Conservation, Bureau of Wildlife, Wildlife Diversity Unit, 625 Broadway, Albany, NY 12233. *Corresponding author - adillon@albanypinebush.org. 674 Northeastern Naturalist Vol. 19, No. 4 Historically, the Frosted Elfin was distributed from southern Canada and the northeastern United States, south to Florida, and west to Texas and Wisconsin. Frosted Elfin may have been most widespread in the Great Lakes region and from southern New England down the coast and into the Carolinas, with scattered populations westward. It is now probably extirpated from Canada, Maine, and Illinois, and is listed as special concern, threatened, or endangered in 11 states (NatureServe 2009); see Brock and Kaufman (2003) for a contemporary range map. The same factors that have led to decline of the Karner Blue, including fire suppression and loss of historical habitat, are also having adverse effect on the Frosted Elfin (Pfitsch and Williams 2009, Smallidge and Leopold 1997, State of New York Endangered Species Working Group 1994). Frosted Elfin is threatened under New York State conservation law and has fewer than 5 major occurrences or viable populations within the state (NY Natural Heritage 2009). At least in New York, the species has generally different host plant requirements between its coastal and inland populations. The Lupinus perennis L. (Wild Lupine, hereafter “lupine”)-feeding variety is found in xeric and open, disturbance- dependent habitats in the upper Hudson Valley, with concentrations in the Albany Pine Bush and the Saratoga Sandplains. However, small populations also persist in Oneida and Genesee counties and on Long Island. Baptisia spp. (Wild Indigo) feeders are found primarily on Long Island, but also occur in the lower Hudson Valley (NY Natural Heritage 2009). The indigo feeder has suffered from the same factors (fire suppression, habitat loss) as the lupine feeder and was probably also greatly reduced by Lymantria dispar dispar (L.) (Gypsy Moth) spraying in the late 1950s (State of New York Endangered Species Working Group 1994). There has been some speculation that 2 species are involved, but some authorities (e.g., Schweitzer et al. 2011) continue to recognize a single species. There is scant published information on lupine-feeding Frosted Elfin (Swengel 1996), and the only substantial studies in the Northeast focused on the indigo feeder (Albanese et al. 2007, 2008). Research is needed on New York populations to support the development of a state recovery plan and monitoring program for the species (NY Natural Heritage 2009, State of New York Frosted Elfin Recovery Team 2011). There is a lack of fundamental information, including which factors may be influencing local population dynamics. In this study, we assess the relationship between adult occupancy (patch use) in an urban pine barrens preserve and a suite of potential controlling factors. We then use the results in a simulation framework to quantitatively inform how many sites and surveys are needed for Frosted Elfin occupancy monitoring. Occupancy is the most basic population state variable and is gaining popularity for conservation programs thanks in large part to advances in statistical methodology over the past decade (MacKenzie et al. 2002 and subsequent papers). Field Site The Albany Pine Bush Preserve (42°42'N, 73°52'W, elevation 79–110 m) is located in east-central New York State between the cities of Albany and Schenectady (Fig. 1). The area is characterized by a cold-temperate humid climate, and the 2012 J.T. Bried, J.E. Murtaugh, and A.M. Dillon 675 preserve sits upon the single largest parabolic dune field in the northeastern United States (Barnes 2003). Preserve land (≈1300 ha) is highly fragmented yet contains one of the world’s largest remaining inland areas of barrens, thickets, and forests dominated by Quercus ilicifolia Wang. (Scrub Oak), Q. prinoides Willd. (Dwarf Chestnut Oak), and Pinus rigida Mill (Pitch Pine). Other major community types include semi-natural grassland, Acer rubrum L. (Red Maple) swamp, Appalachian oak-pine forest, and successional hardwood forest (APBPC 2010, Barnes 2003). The landscape is intensively managed using controlled burns, seed collection and plantings, mechanical treatments, and herbicides (APBPC 2010, Bried and Dillon 2012, Bried and Hecht 2011, Malcolm et al. 2008), and protects several dozen species of rare and declining shrubland birds, herpetofauna, and Lepidoptera (Barnes 2003, Gifford et al. 2010, Hunsinger 1999). Methods Butterfly survey Twenty-eight sites were chosen (Fig. 1) from more than 55 lupine patches currently found on preserve land. To ensure spatial coverage of the landscape, approximately 75% of patches in each Karner Blue management unit were randomly selected. Karner Blue management units are scattered across the preserve and each includes a single lupine patch or group of patches in proximity to each other. No other criteria were used in selecting study sites. Figure 1. Location of study sites (lupine patches) in the Albany Pine Bush Preserve of east-central New York State. 676 Northeastern Naturalist Vol. 19, No. 4 Sites were divided into 3 groups of 8–10 sites each, with observers and starting sites randomly assigned to each group. With each survey, observers were rotated among the groups, and the site order rotated within each group. Three observers completed the first 2 surveys on 6 and 10 May 2011, and 3 new observers did 4 more surveys on 20, 25, 27, and 31 May 2011. The 10-day gap from the second to third surveys was due to an extended period of adverse weather. Surveys took place on partly sunny to clear days reaching at least 18 °C. Search routes followed a fixed zigzag pattern throughout each site, with observers walking at a slow and steady pace. The observer stopped once Frosted Elfin presence was confirmed or the site was fully traversed. Observers were trained in Frosted Elfin identification by experienced biologists, and used closeup photographs to help distinguish it from co-occurring and similar-looking elfins (Callophrys henrici (Grote & Robinson) [Henry’s Elfin], Callophrys niphon (Hübner) [Eastern Pine Elfin], Callophrys augustinus (Westwood) [Brown Elfin]). All wildflower and shrub species in bloom were noted at each site during each survey. Local distribution modeling We used the single-season occupancy modeling framework of MacKenzie et al. (2002, 2006) to compare local distribution factors while controlling for detectability. This approach uses a logit link function to flexibly incorporate variable effects (covariates) thought to influence occupancy and detection probabilities. Covariates for the occupancy parameter (ψ) included primary habitat type, management history (3 metrics), lupine density (2 metrics), nectar richness (2 metrics), shrub cover, and patch area (Table 1); the remaining factors in Table 1 were used as detectability (p) covariates. We also tested the combined effect of weather variables, p(temperature + wind speed + sky cover), the quadratic or peak effect of patch size, ψ(area + area2), and the interaction of time since planting and mowing, ψ(planting × mowing). All modeling was done using PRESENCE v3.1 (Patuxent Wildlife Research Center, United States Geological Survey, Laurel, MD). It was difficult to develop ecological hypotheses and associated models given the lack of information on Frosted Elfin occupancy dynamics. We therefore took an exploratory approach and first modeled the potential nuisance (detection) effects one at a time while holding occupancy constant, including the null model (no covariates, constant ψ and p) for reference (see also Bailey et al. 2004, Bried and Pellet 2012). We used Akaike’s information criterion adjusted for small samples (AICc) to rank the models (Burnham and Anderson 2002). Retaining the top detectability covariate and any competing covariates (those with ΔAICc < 2.0; Burnham and Anderson 2002), we then ran a second set of models to compare each occupancy effect. We found no evidence that the global models (those with the greatest number of parameters) were a poor fit to the data (following the method of MacKenzie and Bailey 2004), and therefore assumed that any reduced models were also valid (see p. 112 in MacKenzie et al. 2006). To control for wide-ranging data and facilitate numerical convergence, patch area and lupine density were standardized to unit mean and variance prior to an alysis. 2012 J.T. Bried, J.E. Murtaugh, and A.M. Dillon 677 Sampling design tradeoffs We evaluated competing designs for Frosted Elfin occupancy monitoring using the Single-season Occupancy Study Design Assistant (SODA v0, written by G. Guillera-Arroita). This software program uses simulations to help evaluate the tradeoff between site and survey replication (Guillera-Arroita et al. 2010). To find a worst-case number of sites (S) and surveys (K), we used the lowest detectability estimate from the best occupancy-detection model as input for SODA. Because the intrinsic occupancy rate in the Albany Pine Bush was high (i.e., Frosted Elfin was observed in 23 out of 28 study sites) and may not be representative of other recovery areas, we repeated the simulations across a gradient of occupancy rates (80%, 50%, 20%) to improve generality of the results. Depending on the project requirements, SODA allows the user to prioritize between maximizing estimator quality (minimize variance) or minimizing total effort (Guillera-Arroita et al. 2010). Here the simulation goal was to minimize total effort based on estimator quality (precision) for various combinations of S Table 1. Covariates used to model occupancy (patch use) and detection probabilities of Frosted Elfin imagos in the Albany Pine Bush Preserve, New York State. Factor Definition and measurement Date t, variation across surveys Habitat type 3 categories = Scrub Oak barrens, former (Black Locust) Robinia pseudoacacia L. clone, or other (old field, powerline corridor, mixed woods, restored parking lot) LupineA (1) ramet density, (2) greater than or less than 2.6 ramets m-2 Management Years since (1) Black Locust removal, (2) planting or interseeding, (3) last mowing NectarB (1) mean number of plant species in flower per survey, (2) cumulative number of plant species in flower Observer Six people rotating among three site groups Area Estimated area of the lupine patch Shrub cover Greater than or less than 16% total cover of woody species <2.5 m height (see Albanese et al. 2007) Sky cover Clear (<5% of survey time under cloud cover), mostly sunny (5–33%), partly sunny (33–66%), mostly cloudy (66–95%), overcast (>95%) Temperature Mean recorded for 3 min using a Kestrel® 2000 Pocket Weather Meter Time of day Start time (nearest minute) of a site survey Wind speed Mean recorded for 3 min using a Kestrel® 2000 Pocket Weather Meter AEstimated using complete census, restricted random sampling, complete random sampling, or adaptive cluster sampling. Complete census was used in the smallest patches and involved a direct count of ramets (defined at the soil surface) throughout the site. The 3 sampling designs included 30-m long transects with lupine counts taking place in a 2-m strip (for sites with sparse lupine) or in 0.5-m2 quadrats placed every meter (for sites with dense lupine). Restricted random transect placement was used at small (less than 0.5 ha) or narrow-elongate sites, and complete random placement at larger sites. Adaptive cluster sampling was used to estimate abundance of geographically rare and clustered lupine populations (Bried 2012). All lupine sampling was done in one growing season over 2007–2011, except for two sites sampled in 2005. The threshold (2.6) comes from Albanese et al. (2007), who found Wild Indigo density to explain the most variation in adult Frosted Elfin density. BRichness was counted in two ways: 1) any forb or shrub species, and 2) only the species believed by the New York recovery team to be most important for the Karner Blue. 678 Northeastern Naturalist Vol. 19, No. 4 and K. Assuming a maximum employable effort of 40 sites × 12 surveys during the Frosted Elfin flight period, we found root-mean square errors (RMSE) of the occupancy estimator for all combinations of S = {10, 20, 30, 40} and K = {4, 6, 8, 10, 12}, running 1000 simulations for each scenario. Lower RMSE indicates greater expected precision for the occupancy estimator. Results Detection probability was most strongly influenced by observer and survey date (Table 2), which together captured 94% of the model weight. The probability ranged from 0.34 to 0.94 among observers, and was high (>0.9) during the first two surveys before dropping significantly (<0.5) during subsequent surveys (Fig. 2). We therefore used the combined effect of observer and survey date to account for detectability variation in each model of Frosted Elfin patch use. The best model of Frosted Elfin patch use assumed the same occupancy probability for each patch and captured nearly 40% of the overall model weight (Table 2). The next best models included the effects of shrub cover and host-plant Table 2. Occupancy-detection model comparison for two sets of models, where AICc is the model Akaike Information Criterion for small samples, ΔAICc is the absolute difference in AICc with the best model, w is the model weight, and K is the number of parameters. Model setA AICc ΔAICc w K Detection factors (with occupancy fixed across patches) Observer 156.70 0.00 0.573 3 Date 157.57 0.87 0.371 7 Wind speed 162.11 5.41 0.038 3 WeatherB 163.56 6.86 0.019 5 Temperature 174.59 17.89 0.000 3 Sky cover 179.61 22.91 0.000 3 Constant (fixed across surveys) 182.40 25.70 0.000 2 Time of day 184.23 27.53 0.000 3 Habitat type 186.18 29.48 0.000 4 Occupancy factors (with best detection factors, observer and da te) Constant (fixed across patches) 159.17 0.00 0.391 8 Shrub cover 162.30 2.59 0.107 9 Lupine 1 (density) 162.66 2.95 0.089 9 Lupine 2 (threshold) 163.11 3.40 0.071 9 Management 2 × 3 163.47 3.76 0.060 9 Management 3 (mowing) 163.74 4.03 0.052 9 Nectar 1 (mean richness) 163.75 4.04 0.052 9 Management 2 (planting) 163.94 4.23 0.047 9 Management 1 (locust removal) 163.99 4.28 0.046 9 Area 164.08 4.37 0.044 9 Area + area2 164.56 4.85 0.035 10 Habitat type 168.10 8.39 0.006 10 Nectar 2 (cumulative richness) 191.27 31.56 0.000 9 ANumerals after lupine, nectar, and management correspond to the alternate measurements in Table 1. “Management 2 × 3” tests an interaction between time since planting and time since mowing. BIncludes the additive effect of temperature, sky cover, and wind speed. 2012 J.T. Bried, J.E. Murtaugh, and A.M. Dillon 679 abundance (ΔAICc < 4). As expected, the estimated occupancy probability was lower at sites above the shrub-cover threshold (0.73) than at sites below it (0.92). Also as expected, the probability was higher at sites above the lupine density threshold (0.91) than at sites below it (0.77), and climbed steadily with increasing density across sites (Fig. 3). The remaining occupancy covariates received minimal to no support (ΔAICc ≥ 4). We used the lowest estimated detection probability from the minimum adequate model, ψ(constant)p(observer + date), as input for SODA; the occupancy probability for this model was 0.822 (0.072 SE). Based on simulations, low Figure 2. Relationship of adult Frosted Elfin detectability to observers and surveys in the Albany Pine Bush Preserve. Figure 3. Relationship between adult Frosted Elfin patch use and host plant density in the Albany Pine Bush Preserve. 680 Northeastern Naturalist Vol. 19, No. 4 RMSE (e.g., <0.01) was achieved for many of the designs (Table 3). The greatest step gains in precision (i.e., largest decreases in RMSE between consecutive effort levels) for the occupancy estimator were achieved with increases from 10 to 20 sites and 4 to 6 surveys. Improvements were negligible (ΔRMSE < 0.004) for increases in effort beyond 30 sites × 8 surveys. With greater occupancy rate, the effort requirements reduce and tradeoffs become less important (Table 3). Discussion Adult Frosted Elfin patch use in the Albany Pine Bush was best explained by a model that assumed the same occupancy probability for each patch. This result suggests that the Frosted Elfin’s local distribution may be largely invariant of patch characteristics, consistent with expert opinions that the species has less stringent habitat demands than other rare and declining lepidoptera of early successional barrens and shrublands (State of New York Frosted Elfin Recovery Team 2011). Because the null model was the minimum adequate model, it is possible that none of the measured factors contributed anything significant to our understanding of variation in Frosted Elfin patch use. However, based on heuristic interpretation of the model weights (see p. 79 in MacKenzie et al. 2006), there was less than a 40% probability that this model was “best”. Adding the weight of evidence for shrub cover and lupine abundance, the probability jumps to 66%, which suggests these factors are worth considering. Larval food plant abundance is generally critical to butterfly population dynamics (Singer 1972). Prior to this study, the state recovery team speculated that lupine abundance is less important to the Frosted Elfin than it is to the Karner Blue. Lupine may indeed be a key factor driving Karner Blue patch-use dynamics in the Albany Pine Bush (Bried and Pellet 2012). The current study indicates that lupine abundance contributes to Frosted Elfin’s local distribution, with the Table 3. Root-mean square errors for the occupancy estimator under competing design parameters (K = number of surveys, S = number of sites), assuming a worst-case detection probability of 0.29 coupled with a gradient of intrinsic occupancy rates (ψ ). K Simulation input S 4 6 8 10 12 ψ = 0.80, pˆ = 0.29 10 0.0368 0.0270 0.0217 0.0188 0.0167 20 0.0272 0.0165 0.0108 0.0096 0.0090 30 0.0203 0.0113 0.0077 0.0061 0.0061 40 0.0160 0.0087 0.0060 0.0050 0.0044 ψ = 0.50, pˆ = 0.29 10 0.0922 0.0564 0.0323 0.0279 0.0286 20 0.0467 0.0202 0.0164 0.0138 0.0127 30 0.0302 0.0153 0.0100 0.0093 0.0089 40 0.0203 0.0102 0.0074 0.0068 0.0070 ψ = 0.20, pˆ = 0.29 10 0.2497 0.1498 0.0722 0.0476 0.0359 20 0.1641 0.0619 0.0228 0.0176 0.0116 30 0.0914 0.0265 0.0096 0.0080 0.0056 40 0.0528 0.0125 0.0064 0.0049 0.0037 2012 J.T. Bried, J.E. Murtaugh, and A.M. Dillon 681 estimated occupancy probability about 15% higher at patches containing greater than 2.6 ramets m-2. Notably, this threshold was derived from a study on density classes of coastal plain indigo feeders (see Albanese et al. 2007) rather than occupancy status of inland lupine feeders. Additionally, the occupancy probability was at least 76% across a broad range of lupine densities (Fig. 3) and patch sizes (mean ± SD = 1.85 ± 2.13 ha, range = 0.04–9.15 ha). This finding suggests the species has a good chance of using even the smaller patches (<1 ha) with relatively sparse host-plant abundance (<1000 ramets ha -1). Even in areas where host-plant density decreases, adult Frosted Elfin density may remain relatively stable if shrub cover is sparse and dominated by native species (Albanese et al. 2007). Like some other rare butterflies in the eastern United States (Hanula and Horn 2011), Frosted Elfin populations may be highly sensitive to the invasion and establishment of non-native plant species, in addition to the normal succession of open barrens habitat (Albanese et al. 2007, Bried and Gifford 2010). Patch-use probability was about 20% lower at sites exceeding the 16% shrub-cover threshold found by Albanese et al. (2007), but we were unable to readily distinguish the non-native and invasive shrub component during the course of this project. The fact that occupancy data supported a finding based on density classes suggests that increased woody structure has a general negative effect on Frosted Elfin population dynamics. Tree canopy is also a strong predictor of adult population density and late-instar larval distribution (Albanese et al. 2007, 2008), but tree cover was generally lacking from the sites in our study sample. Detection probability depended primarily on observer and survey date. This is not surprising given that observer differences are often strong in butterfly surveys and for animals in general (e.g., Bried et al. 2011, Kéry and Plattner 2007). Seasonal variation in detectability is also high in butterfly surveys (Pellet 2008), and detectability in our study dropped sharply after the first surveys on 6 and 10 May. This date effect was likely driven by an earlier than usual start to the flight season in 2011 along with an extended period of rainy weather between the second and third surveys. The brood likely peaked in abundance before or during the first couple surveys, leaving fewer butterflies and lower species’ detectability from the third survey on. The 3 observers with higher detection probabilities (D, N, J in Fig. 2) conducted the first two surveys, suggesting that observer variation was influenced by the unusual phenology and weather patterns and not just inherent differences in search image. Management implications This study suggests that land managers should focus on woody structure more than lupine abundance when it comes to the Frosted Elfin. Of course, where the Frosted Elfin and Karner Blue overlap, the federally listed species will remain the primary conservation target, and the focus on lupine will (and should) continue. Nevertheless, populations of both species can be severely limited by woody encroachment, such as tree canopy that exceeds 60% cover (Albanese et al. 2008, Grundel et al. 1998). Reducing invasions of trees (e.g., 682 Northeastern Naturalist Vol. 19, No. 4 Bried and Hecht 2011, Pfitsch and Williams 2009) along with dense shrub thickets (e.g., Bried and Gifford 2010, Hanula and Horn 2011) will simultaneously benefit the Frosted Elfin and Karner Blue. Partial canopy is needed, however, and the shade heterogeneity provided by scattered trees and low-density native shrubs can help sustain the Frosted Elfin life cycle (Albanese et al. 2007). Intensive management of woody encroachment and invasion is well underway in the Albany Pine Bush (APBPC 2010, Bried and Gifford 2010, Bried and Hecht 2011), and thinning of over-abundant Pinus strobus L. (White Pine) is recognized as a priority management need for the Saratoga Sandplains and Rome Sand Plains recovery areas (Pfitsch and Williams 2009). These activities targeted at rare butterflies may also have a beneficial effect on non-target animal communities such as native solitary bees (Bried and Dillon 2012) and shrubland birds (Gifford et al. 2010, Wood et al. 2011). Monitoring implications An important aspect of any species recovery program is population-monitoring protocol. In the worst-case scenario (i.e., low detectability and low intrinsic occupancy rate), minimum effort for adult Frosted Elfin occupancy monitoring in habitat similar to the Albany Pine Bush may require at least 20 lupine patches surveyed 6 times each or at least 10 lupine patches surveyed 8 times each. In the former scenario, it would be worthwhile to increase patch replication to ≥30 if possible. Increasing total effort beyond 30 sites × 8 surveys may provide only negligible improvements in estimator quality. However, in recovery areas that support fewer than 10 lupine patches, such as the contemporary Rome Sand Plains, more surveys may be needed to compensate for the limited spatial replication. In reality, the effort required under the worst-case scenario might exceed what is practical. If surveys coincide with peak abundance and the observers are welltrained and astute, then detectability should improve and the effort requirement would decrease. We reran the simulations using the peak survey probabilities (mean of 0.956 and 0.912) and found RMSE < 0.017 for all scenarios, suggesting that even the smallest effort (10 sites × 4 surveys, perhaps less) should suffice. It may be possible to predict optimal timing based on plant phenology, degreedays, or historical surveys. When in doubt, though, we recommend assuming the worst-case scenario or using Table 3 to decide on the most feasible option. Adult occupancy (or patch use) is probably the most logistically and statistically efficient state variable for monitoring Frosted Elfin populations, and declines in detectioncorrected occupancy rate or proportion of area occupied could be used to trigger concern and action. Acknowledgments Thanks to Grace Barber, Neil Gifford, Kyle Hodgson, and Emily Pipher for helping with the butterfly surveys, and to Neil Gifford for reviewing an earlier draft of the manuscript. Two anonymous reviewers provided comments that helped improve the manuscript substantially. This project was inspired by recent Frosted Elfin recovery planning led by the New York State Department of Environmental Conservation, Region 4. 2012 J.T. Bried, J.E. Murtaugh, and A.M. Dillon 683 Literature Cited Albanese, G., P.D. Vickery, and P.R. Sievert. 2007. Habitat characteristics of adult Frosted Elfins (Callophrys irus) in sandplain communities of southeastern Massachusetts. Biological Conservation 136:53–64. Albanese, G., P.D. Vickery, and P.R. Sievert. 2008. Microhabitat use by larvae and females of a rare barrens butterfly, Frosted Elfin (Callophrys irus). Journal of Insect Conservation 12:603–615. Albany Pine Bush Preserve Commission (APBPC). 2010. Management plan and final environmental impact statement for the Albany Pine Bush Preserve. Albany, New York. Available online at http://www.albanypinebush.org/commission/management_plans. htm. Accessed 22 March 2012. Bailey, L.L., T.R. Simons, and K.H. Pollock. 2004. Estimating site occupancy and species-detection probability parameters for terrestrial salamanders. Ecological Applications 14:692–702. Barnes, J.K. 2003. Natural History of the Albany Pine Bush: Albany and Schenectady counties, New York. Bulletin No. 502, New York State Museum, Albany, NY. Bried, J.T. 2012. Adaptive cluster sampling in the context of restoration. Restoration Ecology doi:10.1111/j.1526-100X.2012.00924.x. Bried, J.T., and A.M. Dillon. 2012. Bee diversity in scrub oak patches 2 years after mow and herbicide treatment. Insect Conservation and Diversity 5:23 7–243. Bried, J.T., and N.A. Gifford. 2010. Mowing and herbicide of scrub oaks in pine barrens: Baseline data (New York). Ecological Restoration 28:245–248. Bried, J.T., and J.A. Hecht. 2011. Comparison of herbicide treatments to control native aspen encroachment in pine barrens. Natural Areas Journal 31:288–293. Bried, J.T., and J. Pellet. 2012. Optimal design of butterfly occupancy surveys and testing if occupancy converts to abundance for sparse populations. Journal of Insect Conservation 16:489–499. Bried, J.T., K.E. Langwig, A.A. DeWan, and N.A. Gifford. 2011. Habitat associations and survey effort for shrubland birds in an urban pine barrens preserve. Landscape and Urban Planning 99:218–225. Brock, J.P., and K. Kaufman. 2003. Butterflies of North America. Kaufman Focus Field Guides, Houghton Mifflin Company, New York, NY. 284 pp. Burnham, K.P., and D.R. Anderson. 2002. Model Selection and Multi-model Inference: A Practical Information-theoretic Approach. Springer-Verlag, New York, NY. 496 pp. Gifford, N.A., J.M. Deppen, and J.T. Bried. 2010. Importance of an urban pine barrens for the conservation of early successional shrubland birds. Landscape and Urban Planning 94:54–62. Grundel, R., N.B. Pavlovic, and C.L. Sulzman. 1998. Habitat use by the endangered Karner Blue butterfly in oak woodlands: The influence of canopy cover. Biological Conservation 85:47–53. Guillera-Arroita, G., M.S. Ridout, and B.J.T. Morgan. 2010. Design of occupancy studies with imperfect detection. Methods in Ecology and Evolution 1:13 1–139. Hanula, J.L., and S. Horn. 2011. Removing an exotic shrub from riparian forests increases butterfly abundance and diversity. Forest Ecology and Management 262:674–680. Hunsinger, K.C. 1999. A survey of the amphibians and reptiles of the Albany Pine Bush. M.Sc. Thesis. State University of New York at Albany, Albany, NY. Kéry, M., and M. Plattner. 2007. Species richness estimation and determinants of species detectability in butterfly monitoring programmes. Ecological Ent omology 32:53–61. 684 Northeastern Naturalist Vol. 19, No. 4 MacKenzie, D.I., and L.L. Bailey. 2004. Assessing the fit of site-occupancy models. Journal of Agricultural, Biological, and Environmental Statistics 9:300–31 8. MacKenzie, D.I., J.D. Nichols, G.B. Lachman, S. Droege, J.A. Royle, and C.A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248–2255. MacKenzie, D.I., J.D. Nichols, J.A. Royle, J.A. Pollock, L.L. Bailey, and J.E. Hines. 2006. Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence. Academic Press, Elsevier, CA. 324 pp. Malcolm, G.M., D.S. Bush, and S.K. Rice. 2008. Soil nitrogen conditions approach preinvasion levels following restoration of nitrogen-fixing Black Locust (Robinia pseudoacacia) stands in a pine-oak ecosystem. Restoration Ecology 16:70–78. NatureServe. 2009. NatureServe Explorer: An online encyclopedia of life [web application]. Version 7.1. Available online at http://natureserve.org. Accessed 30 August 2010. New York Natural Heritage Program (NY Natural Heritage). 2009. Online conservation guide for Callophrys irus. Available online at http://www.acris.nynhp.org. Accessed 30 August 2010. Pellet, J. 2008. Seasonal variation in the detectability of butterflies surveyed with Pollard walks. Journal of Insect Conservation 12:155–162. Pfitsch, W.A., and E.H. Williams. 2009. Habitat restoration for lupine and specialist butterflies. Restoration Ecology 17:226–233. Schweitzer, D.F., M.C. Minno, and D.L. Wagner. 2011. Rare, declining, and poorly known butterflies and moths (Lepidoptera) of forests and woodlands in the eastern United States. US Forest Service, Forest Health Technology Enterprise Team, FHTET-2011-01. Morgantown, WV. 517 pp. Shapiro, A.M. 1974. Partitioning of resources among lupine-feeding Lepidoptera. American Midland Naturalist 91:243–248. Singer, M.C. 1972. Complex components of habitat suitability within a butterfly colony. Science 176:75–77. Smallidge, P.J., and D.J. Leopold. 1997. Vegetation management for the maintenance and conservation of butterfly habitats in temperate human-dominated landscapes. Landscape and Urban Planning 38:259–280. State of New York Endangered Species Working Group. 1994. Frosted Elfin species dossier (unpublished, draft completed July 1993, revised January 1994). New York State Department of Environmental Conservation, Albany, NY. State of New York Frosted Elfin Recovery Team. 2011. Frosted Elfin recovery plan (unpublished DRAFT). New York State Department of Environmental Conservation, Region 4, Schenectady, NY. Swengel, A.B. 1996. Observations of Incisalia irus (Lepidoptera: Lycaenidae) in central Wisconsin 1988–95. The Great Lakes Entomologist 29:41–62. Wagner, D.L., M.W. Nelson, and D. Schweitzer. 2003. Shrubland Lepidoptera of southern New England and southeastern New York: Ecology, conservation, and management. Forest Ecology and Management 185:95–112. Wood, E.M., A.M. Pidgeon, C. Gratton, and T.T. Wilder. 2011. Effects of oak barrens habitat management for Karner Blue Butterfly (Lycaeides samuelis) on the avian community. Biological Conservation 144:3117–3126.