nena masthead
SENA Home Staff & Editors For Readers For Authors

Sources of Variation in the Abundance and Detection of the Endangered Florida Grasshopper Sparrow
Michael F. Delany, Richard A. Kiltie, Stephen L. Glass, and Christina L. Hannon

Southeastern Naturalist, Volume 12, Issue 3 (2013): 638–654

Full-text pdf (Accessible only to subscribers.To subscribe click here.)

 



Access Journal Content

Open access browsing of table of contents and abstract pages. Full text pdfs available for download for subscribers.

Issue-in-Progress: Vol. 23 (1) ... early view

Current Issue: Vol. 22 (3)
SENA 22(3)

Check out SENA's latest Special Issue:

Special Issue 12
SENA 22(special issue 12)

All Regular Issues

Monographs

Special Issues

 

submit

 

subscribe

 

JSTOR logoClarivate logoWeb of science logoBioOne logo EbscoHOST logoProQuest logo


M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 638 2013 SOUTHEASTERN NATURALIST 12(3):638–654 Sources of Variation in the Abundance and Detection of the Endangered Florida Grasshopper Sparrow Michael F. Delany1,*, Richard A. Kiltie1, Stephen L. Glass2, and Christina L. Hannon2 Abstract - Information on factors affecting the abundance and detection of the endangered Ammodramus savannarum floridanus (Florida Grasshopper Sparrow) was required to determine appropriate management strategies and evaluate monitoring efforts. We examined annual point-count data, records of prescribed fire (2003–2008), observer variability, and landscape features for Three Lakes Wildlife Management Area to identify sources of variation in abundance and detection. The population of male Florida Grasshopper Sparrows was estimated to be 498 (95% CL = 354–641), which corresponded to 23.9 (95% CL = 17.0–30.7) males per km2. Over most or all observed covariate ranges, abundance estimates increased with mean elevation above sea level, distance from outer edge of optimal habitat, and with growing-season burns (P ≤ 0.025). Abundance declined with time since last burn (P ≤ 0.006). Estimates of probability of detection ranged from 0.041 to 0.101, depending on observer and prior detections. Probability of detection declined with time of day, day of year, and days since last burn (P ≤ 0.058). The current prairie burn regime of 2- to 3-year intervals should be maintained with preference for increased growing-season burns. Annual point-count surveys should be continued. Surveys should be conducted within 2 hours of sunrise between mid-April and mid-May. Future monitoring should incorporate covariates of abundance and detection during the collection of data and in their analysis for population estimates used for recovery criteria. Introduction Ammodramus savannarum floridanus Mearns (Florida Grasshopper Sparrow) (AOU 1957) is an endangered subspecies endemic to the south-central prairie region of Florida (USFWS 1999). Native prairie in Florida has been greatly reduced by agriculture (Shriver and Vickery 1999), and this probably caused the extirpation of the sparrow from some former breeding locations (Delany and Linda 1994). Habitat loss and population declines have continued (Delany et al. 2007a), and population-viability analysis found the subspecies vulnerable to extinction within the next 50 years (Perkins et al. 2008). The recovery objective is to down-list the sparrow to threatened status when ≥10 protected locations contain stable, self-sustaining populations of ≥50 breeding pairs (USFWS 1999). The metapopulation (Tucker et al. 2010) comprises seven partially isolated breeding populations, and previous estimates indicated fewer than 1000 individuals may exist (Delany et al. 2007a). Accurate spatial and temporal information on the status of grassland sparrows is needed to predict their ability to persist and to determine appropriate 1Florida Fish and Wildlife Conservation Commission, 1105 SW Williston Road, Gainesville, FL 32601. 2Florida Fish and Wildlife Conservation Commission, 1231 Prairie Lakes Road, Kenansville, FL 34738. *Corresponding author - mike.delany@myFWC.com. 639 M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 management strategies (Curnutt et al. 1996). Density and reproductive success of the Florida Grasshopper have been associated with the frequency and seasonality of prescribed fire (Delany et al. 1985, 2002; Perkins et al. 2009; Shriver and Vickery 2001; Shriver et al. 1996; Walsh et al. 1995), grassland patch size and edge effects (Perkins et al. 2003), and hydrology (Perkins and Vickery 2005). Monitoring programs on public lands provide an opportunity to examine local environmental factors associated with abundance and evaluate management actions. Populations of the Florida Grasshopper Sparrow are monitored on public lands using standard point-count surveys (Ralph et al. 1993) to determine relative abundance. Adjusting for factors influencing detectability is important in obtaining reliable population estimates (Kéry and Schmid 2004, Thompson 2002), especially for this rare and secretive species. However, previous estimates of Florida Grasshopper Sparrow abundance based on point-count surveys have not accounted for imperfect detection (Delany et al. 2007a, Tucker et al. 2010). Our objectives were to examine point-count data and environmental variables at Three Lakes Wildlife Management Area (WMA) for factors affecting the abundance of Florida Grasshopper Sparrows and to identify sources of variation in detection. Study Site and Population Florida’s dry prairie is a distinct floristic region characterized as flat, open expanses dominated by fire-dependent grasses, Serenoa repens (Bartram) Small (Saw Palmetto), low shrubs, and abundant forbs (Orzell and Bridges 2006). This complex grassland ecosystem is determined by topography, fire frequency, hydrology, and management history (Platt et al. 2006, Stephenson 2011). Dry prairie occupied by Florida Grasshopper Sparrows is treeless and ranges from thick (34% shrub cover), low (≤57 cm) Saw Palmetto scrub to grass pastures with a sparse (<10% shrub cover) or patchy cover of shrubs and Saw Palmetto maintained by recurrent fire at 2- to 3- year intervals (Delany et al. 1985). Florida Grasshopper Sparrows at Three Lakes WMA occupy a relict patch of native prairie (Bridges 2006). The southern portion of the 25,692-ha property comprises about 3000 ha of dry prairie (Fig. 1). Prescribed burns are conducted on the WMA during dormant (October–March) and growing (April–September) seasons at 2- to 3-year intervals. From 2003 to 2008, an average of 1253 ha was burned annually: 536 ha during the dormant season and 717 ha during the growing season. The area has not been grazed by cattle since 1987. The breeding aggregation of Florida Grasshopper Sparrows on Three Lakes WMA was described by Howell (1932) and Nicholson (1936) and is the northernmost extant known population of the subspecies. This was the most stable population on public land in terms of population trends (Tucker et al. 2010), with 61–149 males counted during annual point-count surveys conducted during the April to June breeding season from 1991 to 2011 (Florida Fish and Wildlife Conservation Commission [FWC], Kenansville, FL, unpubl. data). M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 640 Figure 1. Florida Grasshopper Sparrow point-count survey locations (•) on Three Lakes Wildlife Management Area, 2003–2008. 641 M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 Methods Population monitoring and available data The Florida Grasshopper Sparrow population on Three Lakes WMA was monitored annually during 2003 to 2008 using point-count surveys (Ralph et al. 1993) modified after Walsh et al. (1995). A grid system of 190 points 400 m apart was established (Fig. 1). The array of points covered most of the area occupied by Florida Grasshopper Sparrows at this location and remained fixed from year to year. Some points (n = 24) were not included in the analysis because of their placement in unsuitable vegetation (a dense successional stage); hence, 166 points were used. For most points, a survey was conducted on 3 separate days each year, but 34 of the points were surveyed only twice a year, and 21 were surveyed only once a year. Environmental conditions specific to each point and not changing among visits (e.g., distance to edge and elevation) were recorded and are described below as point covariates. The most elemental observation consisted of the number of male Florida Grasshopper Sparrows detected by sight or sound at a survey point during a 5-minute observation period during the April–June breeding season. Maximum detection distance at a point was not formally limited but was estimated to be 200 m. Surveys were made between sunrise and about 1000 hrs in the absence of rain and at wind velocities <10 km/hr. Environmental conditions specific to each visit to a count point were recorded and are described below as observation covariates. Abundance and detection modeling We modeled abundance and detectability of Florida Grasshopper Sparrows at Three Lakes WMA with software Package Unmarked ver. 0.9-0 (Fiske and Chandler 2011) for the R statistical environment (ver. 2.12.0; R Development Core Team 2010). Package Unmarked implements the N-mixture models of Royle (2004) and Dail and Madsen (2011) for estimating detectability and abundance from spatially and temporally replicated point-count surveys. Covariate preparation. Observation covariates were modeled as predictors of detectability and included the following: (1) number of days since last prescribed fire (weighted by burn unit areas contained within the 200-m-radius count-point areas), (2) an index summarizing whether the most recent fire occurred during the dormant season (0 = October–March) or the growing season (1 = April–August) (weighted by burn unit area contained within the survey-point area), (3) starting time (decimal hr, AM) for a point count, (4) ordinal date of a point count within a year, (5) average rainfall (inches) in the month preceding the survey month (data were the same for all surveys in a given month), and (6) within-year experience at a survey point (i.e., a binary variable indicating whether the observer had counted any sparrows previously at a point in a given year; Riddle et a l. 2010). Seven observers took part in the study, and observer identity was modeled as a seventh covariate (Table 1). Preliminary modeling indicated that two observers (AM and TH) tended to produce lower detectability estimates than the others; the observer covariate was therefore reclassified as a binary variable with level B indicating counts by AM and TH and level A indicating counts by any of the other observers. M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 642 Survey-point covariates (those not changing among visits) were modeled as predictors of abundance and included the following: (1) UTM northing (103 m), (2) UTM easting (103 m), (3) mean elevation (ft) above sea level of the area around the survey point (average of elevation GIS grid cells within that area), and (4) distance (m) from a point to the nearest edge of optimal habitat for Florida Grasshopper Sparrows (i.e., toward the perimeter of the study site). In addition, averages across years at points were determined for the observation covariates number of days since last prescribed fire, prior burning-season index, and average rainfall in the month preceding point counts so that they could be considered as potential influences on abundance. Although rainfall and elevation covariates were modeled in English units, they were converted to metric units for plots illustrating their modeled effects. Variance-inflation factors and tolerance were checked among the observation covariates and among the survey-point covariates with Fox and Weisberg’s (2011) vif function from the Car Package for R. The only pair of variables with problematical tolerance (<0.4) was UTM easting and mean elevation above sea level. These two variables were highly correlated (Pearson r = 0.75). UTM easting was omitted from the count mixture models presented here in favor of the more ecologically meaningful elevation variable (which also produced models with lower Akaike information criterion [AIC; Burnham and Anderson 2002] than when UTM easting was used instead). Alternative exploratory models indicated that effects of UTM easting were very similar to those of minimum elevation. Frequency histograms of the continuous covariate measurements were inspected for any marked skewness. All plots were reasonably symmetrically distributed except for distance to edge of optimal habitat, which was highly rightskewed; a square-root transformation was therefore applied to that variable. All covariates except for observer ID and experience were standardized to mean = 0 and SD = 1 before they were used in a model to help stabilize the numerical optimization algorithm (Fiske and Chandler 2012). Preliminary model development. A number of decisions were made in modeling the counts and presenting results. We used the Poisson rather than negative binomial for the count side of the mixture model because models using the negative binomial yielded unrealistically right-skewed abundance estimates, a problem to which that distribution is prone (Joseph et al. 2009, Kéry et al. 2005). The Poisson distribution’s assumption that the variance must equal the mean was made Table 1. Observer code, numer of point counts conducted, and percent of all counts conducted by 7 observers at Three Lakes Wildlife Management Area, 2003–2008. Observer Number of counts Percent of all counts AB 166 5.72 AP 275 9.48 AM 104 3.58 EB 26 0.90 GL 1594 54.93 HA 90 3.10 TH 647 22.30 643 M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 somewhat more robust by applying Package Unmarked’s nonparametric bootstrap function nonparboot (Fiske and Chandler 2011) with 1000 replicates and bias-correction (Efron and Tibshirani 1993) to estimate the model parameters. In any case, the directions and shapes of covariate effects, which are the focus of this report, appeared to be similar regardless of the mixture distribution assumed. Another necessary choice was whether to treat the population at each point as closed (constant) or open (potentially varying) (Dail and Madsen 2011). With the closed-population model, movement of individuals is allowed as long as the potentially countable population at each point is stable. We used a closed model because preliminary analyses suggested that variation in covariates affecting detection largely accounted for changes in counts at points over the course of the study. Our assumption of net population closure is consistent with the conclusion of Delany et al. (2007b) that there was negligible change in Florida Grasshopper Sparrow counts at Three Lakes WMA over a series of years preceding and partially overlapping those upon which results of the present study are based. Territoriality both sets an upper limit (≈8 birds) on the number of male Florida Grasshopper Sparrows in the area around the Three Lakes WMA points (Delany et al. 2007b) and likely accounts in part for effective population closure. Fitting N-mixture models by maximum likelihood methods requires setting a limit (K) over which integration of the likelihood model is performed (Dail and Madsen 2011, Royle 2004). Preliminary modeling indicated that K = 25 was sufficient to obtain stable parameter estimates when a Poisson distribution was assumed. All survey-point and observation covariates were initially entered in the Nmixture count model. Quadratic parameters also were initially included for the continuous covariates. These were subsequently reduced in a stepwise fashion by removing effects in any given model with the lowest z value until doing so caused AIC to rise by >2 units. This elimination process was first applied to quadratic effects and then to main effects. All modeled covariate effects were additive. The potential pitfalls of stepwise methods (e.g., Anderson 2008) apply here, so the reduced model should not necessarily be interpreted as describing a universal set of best predictors, but rather a set that economically accounts for the abundance and detectability implied by available count data. Absolute goodness of model fit as reflected in sum of squared errors (SSE) was assessed with the parametric bootstrap function parboot (1000 replicates) from Package Unmarked (Fiske and Chandler 2011). Abundance and detectability estimates were back-transformed from the scales (log and logistic, respectively) of the N-mixture model (Fiske and Chandler 2012, Royle 2004). Because all covariates of abundance were standardized to mean = 0, the estimated mean abundance was simply the exponentiated intercept. Mean detectability was estimated as the reverse logistic transformation at specified values of the binary variables for observer and experience, as all other detectability covariates were standardized to mean = 0. Back-transformed SEs were estimated by the delta method using either Package Unmarked’s backtransform function or the deltamethod function from the msm R package (Jackson 2011). M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 644 Covariate effects were plotted by estimating abundance or detectability (with 95% confidence limits) over the range of each value of the covariate while setting the other continuous covariates to 0. For detectability, effects of the continuous covariates are plotted only for observer = A, experience = 0 (no prior detection at a count point in a given year). Results After stepwise reduction, the covariates on abundance retained in the bootstrapped mixture model included mean rainfall in preceding month, mean index of most recent burn season and quadratic effects of minimum elevation, of distance to edge, and of mean number of days since last burn and mean index of most recent burn season (Table 2). Covariates retained for detection probability included index of prior experience, observer identification index, survey start time, mean number of days since last burn, and quadratic effects Table 2. Bootstrapped (n = 1000) parameterA estimates for the Poisson model after bias correction for Florida Grasshopper Sparrow abundance and detection at Three Lakes Wildlife Management Area, 2003–2008. 95% CL Parameter Estimate SE Lower Upper Abundance: Intercept 1.099 0.147 0.811 1.387 Minimum elevation (ft) above sea level 13.145 5.366 2.628 23.663 Minimum elevation (ft) above sea level2 –12.691 5.243 –22.966 –2.415 Edge distance 1.408 0.473 0.481 2.335 Edge distance2 –0.795 0.406 –1.590 0.000 Mean number of days since last burn –0.781 0.613 –1.992 0.420 Mean number of days since last burn2 0.686 0.727 –0.740 2.111 Mean index of most recent burn season 0.159 0.101 –0.038 0.356 Mean rainfall (in) in month preceding survey 0.172 0.068 0.039 0.304 Detection: Intercept –2.690 0.152 –2.987 –2.393 Index of prior experience 0.509 0.134 0.246 0.771 Observer identification index –0.471 0.133 –0.732 –0.211 Index of most recent burn season 0.260 0.241 –0.213 0.732 Index of most recent burn season2 –0.373 0.236 –0.836 0.089 Survey start time (decimal hr a.m.) –0.437 0.065 –0.566 –0.309 Julian date of survey 0.822 0.783 –0.714 2.375 Julian date of survey2 –1.050 0.793 –2.604 0.505 Mean number of days since last burn –0.395 0.064 –0.520 –0.270 AAbundance (survey point) and detection (survey point visit) covariates: minimum elevation was for the 200-m-radius survey area of a point; edge distance was the minimum distance (m) from the point to the edge of suitable habitat; mean number of days since last burn was for the point survey area; mean index of most recent burn season was for the point survey area (0 = dormant season, 1 = growing season); mean rainfall was for the month preceding the survey month at the Sunset Ranch; index of prior experience within a year at a site was 0 = no prior sparrows counted, 1 = ≥1 sparrow previously counted within a year; observer identification index: 0 = any observers except AM or TH (for whom observer index = 1). 645 M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 of most recent burn season index and of survey date (Table 2). The mean parametric bootstrap on SSE gave P = 0.064, so fit of the model qualified as adequate, albeit marginally. The parameter estimates indicated that Florida Grasshopper Sparrow abundance increased with mean elevation above sea level for all but the highest elevation values, with distance from the edge of optimal habitat, with mean burning season index (growing season), and with mean rainfall during the month preceding the survey month (Fig. 2). Estimated abundance tended to decrease at a decelerating rate with increasing mean number of days since last burn. Parameter estimates for detection covariates implied the following effects. Observers labeled group A were more likely to detect Grasshopper Sparrows when present than were those in group B (Fig. 3). Sparrows were more likely to be detected at a survey point after a prior detection in a given year than before. Detection probability evinced a decelerating trend as survey start times increased and as days since last burn at a survey point increased. Detection probability was fairly constant during the first 30 days of the survey season, then declined in the next 40 days. Detection probability showed a convex relationship to burn-season index, declining slightly toward index = 1 (growing-season burns) than index = 0 (dormant-season burns). Other combinations of observer and experience produced continuous effect plots with forms similar to those for observer = A and with experience = 0; the plots differed only in “elevation” as inferable from the top plot of observer and experience effects in Figure 3. Average Florida Grasshopper Sparrow abundance per point implied by the bootstrapped model was 3.001 males (SE = 0.441, 95% CL = 2.137–3.864). Total abundance for the sampled area was estimated as 498 males (95% CL = 354–641) and density as 23.9 males per km2 (95% CL = 17.0–30.7). Detection probability estimates (p) were as follows: for observer group = A and experience = 0, p = 0.064 (SE = 0.009, 95% CL = 0.046–0.081); for observer group = A and experience = 1, p = 0.101 (SE = 0.016, 95% CL = 0.070–0.133); for observer group = B and experience = 0, p = 0. 0.041 (SE = 0.007, 95% CL = 0.026–0.055); and for observer group = B and experience = 1, p = 0.066 (SE = 0.012, 95% CL = 0.042–0.090). Discussion Abundance With a proportion of the population undetected, previous estimates of Florida Grasshopper Sparrow abundance based on point-count surveys were biased low. However, by modeling abundance and detectability at the level of the survey point and then aggregating this local information by modeling the variation in abundance and detection among all points at Three Lakes WMA, we were able to estimate detection probability and abundance. Our estimate of total abundance (498 males, 95% CL = 354–641) was adjusted for imperfect detection and therefore was much larger than the 98–142 males counted each year during the 2003– 2008 point-count surveys (FWC, unpubl. data). However, confidence limits were M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 646 Figure 2. Modeled effects of the covariates of abundance of Florida Grasshopper Sparrows at Three Lakes Wildlife Management Area during point-count surveys, 2003–2008: elevation (A), distance to edge (B), days since last burn (C), season of burn index of the most recent fire (0 = dormant season, October–March; 1 = growing season, April–August) (D), and rain during month preceding month of survey (E). 647 M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 Figure 3. Modeled effects of covariates of detection probability of Florida Grasshopper Sparrow at Three Lakes Wildlife Management Area during point-count surveys, 2003– 2008: observer groups A and B, and experience (0 = no prior detection at a count point in a given year, 1 = prior detection at a count point in a given year) (A), season of burn index of the most recent fire (0 = dormant season, October–March; 1 = growing season, April–August) (B), survey start time (C), date of survey (D), and days since last burn (E). M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 648 wide and our estimate of abundance may be an effective maximum estimate for a population occupying all available habitat. Florida Grasshopper Sparrow density at Three Lakes WMA (0.24 territory/ha) was at the low end of the range (0.21–1.30 territories/ha, in grasslands) reported for other subspecies (reviewed in Vickery 1996). Our density estimate was close to the range (0.26–0.37 territory/ha) reported by Perkins et al. (2003), who used more labor-intensive spot-mapping methods (International Bird Census Committee 1970) at Three Lakes WMA. The Florida Grasshopper Sparrow is at the edge of the species’ range at Three Lakes WMA and therefore may be more variable in density and abundance than are other subspecies (Curnutt et al. 1996). Populations of grassland birds also vary over space and time in response to the instability of grasslands (Cody 1985, Wiens 1973). However, previous analysis found that the Florida Grasshopper Sparrow population on Three Lakes WMA was relatively stable compared with monitored populations on other public lands (Tucker et al. 2010). Variation in covariates associated with point-count surveys seemed to explain changes in abundance on our study area. Patterns of abundance may be more reliably detected in this stable population at Three Lakes WMA than in other populations declining for unknown reasons (see Delany 2007a, Tucker et al. 2010). Correlations between abundance and time post-burn suggested that Florida Grasshopper Sparrows at Three Lakes WMA responded positively to the effects of the 2- to 3-year fire regime. Previous studies quantifying habitat selection in relation to time post burn found that densities ranged from 0.05 to 0.75 territory/ ha during the first year following fire, but decreased to 0.01–0.18 territory/ha during the subsequent 1.5 years (Shriver and Vickery 2001, Walsh et al. 1995; but see Delany et al. 2002). Increased growth of shrub and Saw Palmetto with the exclusion of fire may eventually allow vegetation to reach a dense successional stage unusable by this ground-dwelling sparrow (Delany et al. 1985). Historically, Florida’s prairie ecosystem has probably been maintained by a frequent fire cycle caused by lightning during summer (May–August) thunderstorms followed by temporary flooding (Platt et al. 2006). Florida Grasshopper Sparrows are likely adapted to summer fires and may exhibit increased density and a prolonged breeding season following such burns (Perkins et al. 2009, Shriver and Vickery 1999, Shriver et al. 1996). The subspecies tends to use areas with sparse woody cover (Delany et al. 1985). Growing-season burns can reduce woody vegetation compared with repeated fires in the dormant season (Drewa et al. 2002, USFWS 1999) and may improve grassland conditions for the sparrow. Our results accord with previous work indicating a greater abundance of Florida Grasshopper Sparrows in areas burned during the growing season (Perkins et al. 2009, Shriver and Vickery 1999, Shriver et al. 1996). Grasshopper Sparrows are considered area-sensitive because their occurrence and density are positively associated with grassland patch size (Ribic et al. 2009, Vickery 1996). Proximity to edges and the landscape characteristics of adjacent non-grassland edges (especially woody vegetation) have also been associated 649 M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 with probability of occurrence and density of Grasshopper Sparrows (Delisle and Savidge 1996, Helzer and Jelinski 1999, Renfrew and Ribic 2002, Ribic and Sample 2001, Vickery 1996). Nests of grassland birds near edges may be more vulnerable to predation and parasitism than nests located in the interior of grassland patches (Johnson and Temple 1990, Winter et al. 2000). Interior grassland areas ≥400 m from non-prairie edge serve as reproductive source areas and contained higher territory densities for Florida Grasshopper Sparrows, whereas areas closer to edges were deemed population sinks (sensu Pulliam 1988) and had lower territory densities of sparrows (Perkins et al. 2003). Similarly, we found that edge effect strongly influenced the abundance of Florida Grasshopper Sparrows, with abundance peaking at 600 m from non-prairie edge. In the “low-relief landscape” of Florida’s dry prairie, small differences in elevation can result in distinct vegetation associations (Orzell and Bridges 2006). Florida Grasshopper Sparrows were more abundant and occurred in higher densities at 18.5–19.0 m above sea level than at other elevations, and birds may be responding to associated subtle differences in vegetation composition and structure in the selection of breeding territories. Higher-elevation prairie may burn more intensely than wetter low-lying areas and may provide more suitable breeding habitat for the sparrow. Florida Grasshopper Sparrow nests at higher elevations also may be less susceptible to flooding (Perkins and Vickery 2005). Similarly, territory density of Grasshopper Sparrows (A. s. pratensis) in Wisconsin was higher in dry upland pastures than at lower elevations (Renfrew and Ribic 2002, Ribic and Sample 2001) . Abundance of grassland birds and timing of their reproduction has been associated with rainfall. The influence of rainfall may be correlated with vegetation growth and changes in food availability (Ahlering et al. 2009, Cody 1985, Pulliam and Parker 1979). No information is available on the limitations of food resources for Florida Grasshopper Sparrows. However, high water levels can contribute to decreased productivity (Perkins and Vickery 2005), and flooding may be an important source of nest loss (Pranty 2000). The increased abundance of Florida Grasshopper Sparrows with increased rainfall during the month preceding a survey may be due to changes in food availability, or nest loss from flooding and re-nesting efforts. However, rainfall at Three Lakes WMA can be patchy, and the rain gauge located 3 km from survey points may not reflect precipitation over the entire study area. Habitat choices are assumed to be adaptive, with individuals occupying the best available locations (Fretwell and Lucus 1970, Lanyon and Thompson 1986). Many factors can influence avian habitat use (e.g., conspecific attraction, climate, food availability) and make interpretations of the selection process difficult (Cody 1985, Stamps 1988). Moreover, density is not necessarily correlated with habitat quality and nesting success in grassland birds (Vickery et al. 1992, Winter and Faaborg 1999). Nevertheless, Florida Grasshopper Sparrow selection of physical and vegetative components of the study area may have adaptive value manifested in increased fitness in nest success in recently burned areas (Delany M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 650 et al. 2002, Shriver and Vickery 2001) and at locations ≥400 m from non-prairie edge (Perkins et al. 2003). Detection Because Florida Grasshopper Sparrows are sedentary during the breeding season (Delany et al. 1995), all individuals at Three Lakes WMA were available for detection. However, the intermittent availability of Grasshopper Sparrows due to sporadic singing behavior and the amount of time spent on the ground makes them less likely to be detected during short (5-minute) count periods (Diefenbach et al. 2007), so our estimates of abundance and detection may still be biased low. Estimates also may be negatively biased because observers tend to undercount birds during unlimited-radius counts (Simons et al. 2007) and at distances ≥75 m in grasslands (Rotella et al. 1999). We found a strong inter-observer effect on the detection of Florida Grasshopper Sparrows that varied with experience (previous detections). Mean estimates of detection probabilities of Florida Grasshopper Sparrows among observers ranged from 0.04 to 0.101, with a greater detection probability among observers with within-year experience at a survey point. Variation in observer detection probability may also reflect the assignment of observers to survey points. Diefenbach et al. (2003) found that Grasshopper Sparrow detection probabilities among observers ranged from 0.44 to 0.66 in Pennsylvania. Grasshopper Sparrow detection ranged from 0.82 at 50 m to 0.64 at 75 m during point-count surveys in North Dakota (Rotella et al. 1999). Although our mean estimates of detection seem low compared with those in other studies, under some covariate conditions (e.g., early in the day and season), they were considerably higher (Fig. 3). Time of day and time of season were important sources of variation affecting the propensity of Florida Grasshopper Sparrows to sing and the probability of their detection. Detection was greatest at sunrise and declined sharply with time. Hochachka et al. (2009) found similar daily variation in Florida Grasshopper Sparrow singing probability at Kissimmee Prairie Preserve State Park. Changes in singing behavior over the course of the breeding season also may influence detection (Wilson and Bart 1985). We found that surveys conducted within the first 30 days of the breeding season (to mid-May) were more likely to detect sparrows than surveys conducted later in the breeding season. In contrast, Hochachka et al. (2009) did not find a strong correlation in day-to-day variation in singing probability of Florida Grasshopper Sparrows, but sampled only to 10 May. Diefenbach et al. (2007) also found no temporal changes in the availability of Grasshopper Sparrows (A. s. pratensis) within the breeding season during point-count surveys in Pennsylvania. Based on our results, the use of point-count survey data from later than 2 hours after sunrise or after mid-May to estimate population size or evaluate responses to management actions may provide misleading information by underestimating abundance. Conclusions Our estimate of abundance was made when Florida Grasshopper Sparrows seemed to be thriving. There was a significant reduction in Florida Grasshopper 651 M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 Sparrow occurrence and abundance over the WMA and at other breeding aggregations from 2008 to 2013, and the subspecies may be in jeopardy of extinction (M.F. Delany, pers. observ.). Evaluating the status of Florida Grasshopper Sparrow breeding aggregations will be essential in determining appropriate management actions, and annual point-count surveys at Three Lakes WMA should be continued to guide recovery efforts. To increase detection, point-count surveys should be conducted for 2 hours beginning at official sunrise and conducted early in the breeding season (before mid-May). Landscape features, time of day and day-of-year effects, and observer identity should be incorporated into analyses that account for an imperfect detection of this cryptic sparrow. Compared to spotmapping, the use of analytical methods described here may provide more efficient and reliable estimates of abundance for recovery criteria. However, other methods should be used (e.g., distance estimates) to verify our results. Multiple factors affected the abundance and detection of Florida Grasshopper Sparrows. Prescribed fire is the most important management tool for maintaining suitable dry prairie for Florida Grasshopper Sparrows, and the fire regime at Three Lakes WMA (2–3 year burn interval, and growing-season burns) should continue. A better understanding of the influence of the patchiness and the intensity of prescribed fire on sparrow distribution and abundance is needed. The large grassland at Three Lakes WMA should be maintained, and encroaching woody vegetation should be removed to increase the amount of contiguous habitat for Florida Grasshopper Sparrows. Acknowledgments We thank D. Dail for kindly providing R code for comparing open-and closed-population models. R. Butryn provided Figure 1. B. Ames, A. Blackford, H. Harter, A. Prince, and E. Rushton assisted with point-count surveys. B. Crowder, K. Miller, J. Meyers, T. O’Meara, D. Perkins, J. Rodgers, Jr., and an anonymous reviewer provided helpful comments on previous drafts of this paper. Literature Cited Ahlering, M.A., D.H. Johnson, and J. Faaborg. 2009. Factors associated with arrival densities of Grasshopper Sparrow (Ammodramus savannarum) and Baird’s Sparrow (A. bairdii) in the upper Great Plains. Auk 126:799–808. Anderson, D.R. 2008. Model Based Inference in the Life Sciences. Springer, New York, NY. 184 pp. American Ornithologists’ Union (AOU). 1957. Check-list of North American Birds. Fifth Edition. Lord Baltimore Press, Inc., Baltimore, MD. 691 pp. Bridges, E. 2006. Landscape ecology of Florida dry prairie in the Kissimmee River Region. Pp. 14–42, In R. Noss (Ed.). Land of Fire and Water: The Florida Dry Prairie Ecosystem. Proceedings of the Florida Dry Prairie Conference. E.O. Painter Printing Company, DeLeon Springs, FL. 241 pp. Burnham, K.P., and D.R. Anderson. 2002. Model Selection and Multimodel Inference. Second Edition. Springer, New York, NY. 488 pp. Cody, M.L. 1985. Habitat selection in grassland and open-country birds. Pp. 191–251, In M.L. Cody (Ed.). Habitat Selection in Birds. Academic Press, Orlando, FL. 558 pp. M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 652 Curnutt, J.L., S.L. Pimm, and B.A. Maurer. 1996. Population variability of sparrows in space and time. Oikos 76:131–144. Dail, D., and L. Madsen. 2011. Models for estimating abundance from repeated counts of an open metapopoulation. Biometrics 67:577–587. Delany, M.F., and S.B. Linda. 1994. Characteristics of occupied and abandoned Florida Grasshopper Sparrow territories. Florida Field Naturalist 22:106–109. Delany, M.F., H.M. Stevenson, and R. McCracken. 1985. Distribution, abundance, and habitat of the Florida Grasshopper Sparrow. Journal of Wildlife Management 49:626–631. Delany, M.F., C.T. Moore, and D.R. Progulske. 1995. Territory size and movements of Florida Grasshopper Sparrows. Journal of Field Ornithology 66:305–309. Delany, M.F., S.B. Linda, B. Pranty, and D.W. Perkins. 2002. Density and reproductive success of Florida Grasshopper Sparrows following fire. Journal of Range Management 55:336–340. Delany, M.F., M.B. Shumar, M.E. McDermott, P.S. Kubilis, J.L. Hatchitt, and R.G. Rivero. 2007a. Florida Grasshopper Sparrow distribution, abundance, and habitat availability. Southeastern Naturalist 6:15–26. Delany, M.F., P.S. Kubilis, R.G. Rivero, and K.R. Rogers. 2007b. Assessment of hurricane effects on Florida Grasshopper Sparrow populations and habitat. Final report to the US Fish and Wildlife Service, South Florida Ecological Services Office, Vero Beach, FL. 74 pp. Delisle, J.M., and J.A. Savidge. 1996. Reproductive success of Grasshopper Sparrows in relation to edge. Prairie Naturalist 28:107–113. Diefenbach, D.R., D.W. Brauning, and J.A. Mattice. 2003. Variability in grassland bird counts related to observer differences and species detection rates. Auk 120:1168–1179. Diefenbach, D.R., M.R. Marshall, J.A. Mattice, and D.W. Brauning. 2007. Incorporating availability for detection in estimates of bird abundance. Auk 124:96–106. Drewa, P.B., W.J. Platt, and E.B. Moser. 2002. Fire effects on resprouting of shrubs in headwaters of southeastern Longleaf Pine savannas. Ecology 83:7 55–767. Efron, B., and R.J. Tibshirani. 1993. An Introduction to the Bootstrap. Chapman and Hall/CRC, Boca Raton, FL. 436 pp. Fiske, I., and R.B. Chandler. 2011. Unmarked: An R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43:1–23. Fiske, I., and R.B. Chandler. 2012. Overview of unmarked: An R package for analysis of unmarked animals. Available online at http://cran.r-project.org/web/packages/unmarked/ vignettes/unmarked.pdf. Accessed 15 January 2013. Fox, J., and S. Weisberg. 2011. An R Companion to Applied Regression, Second Edition. SAGE Publications Inc., Thousand Oaks, CA. 512 pp. Fretwell, S.D., and H.L. Lucas. 1970. On territorial behavior and other factors influencing habitat distribution in birds. Acta Biotheoretica 19:16–36. Helzer, C.J., and D.E. Jelinski. 1999. The relative importance of patch area and perimeter- area ratio to grassland breeding birds. Ecological Applications 9:1448–1458. Hochachka, W.M., M. Winter, and R.A. Charif. 2009. Sources of variation in singing probability of Florida Grasshopper Sparrows, and implications for design and analysis of auditory surveys. Condor 111:349–360. Howell, A.H. 1932. Florida Bird Life. Coward-McCann, New York, NY. 579 pp. International Bird Census Committee. 1970. An international standard for mapping methods in bird census work recommended by the International Bird Census Committee. Audubon Field Notes 24:722–726. Jackson, C.H. 2011. Multi-state models for panel data: The msm package for R. Journal of Statistical Software 38:1–29. 653 M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 Johnson, R.G., and S.A. Temple. 1990. Nest predation and brood parasitism in tallgrass prairie birds. Journal of Wildlife Management 54:106–111. Joseph, L.N., C. Elkin, T.G. Martin, and H.P. Possingham. 2009. Modeling abundance using N-mixture models: The importance of considering ecological mechanisms. Ecological Applications 19:631–642. Kéry, M., and H. Schmid. 2004. Monitoring programs need to take into account imperfect species detectability. Basic and Applied Ecology 5:65–73. Kéry, M., J.A. Royle, and H. Schmid. 2005. Modeling avian abundance from replicated counts using binomial mixture models. Ecological Applications 15:1450–1461. Lanyon, S.M., and C.F. Thompson. 1986. Site fidelity and habitat quality as determinants of settlement pattern in male Painted Buntings. Condor 88:206–210. Nicholson, W.H. 1936. Notes on the habits of the Florida Grasshopper Sparrow. Auk 53:318–319. Orzell, S.L., and E.L. Bridges. 2006. Floristic composition of the south-central Florida dry prairie landscape. Pp. 64–99, In R. Noss (Ed.). Land of Fire and Water: The Florida Dry Prairie Ecosystem. Proceedings of the Florida Dry Prairie Conference. E.O. Painter Printing Company, DeLeon Springs, FL. 241 pp. Perkins, D.W., and P.D. Vickery. 2005. Effects of altered hydrology on the breeding ecology of the Florida Grasshopper Sparrow and Bachman’s Sparrow. Florida Field Naturalist 33:29–40. Perkins, D.W., P.D. Vickery, and W.G. Shriver. 2003. Spatial dynamics of source-sink habitats: Effects on rare grassland birds. Journal of Wildlife Management 67:588–599. Perkins, D.W., P.D. Vickery, and W.G. Shriver. 2008. Population viability analysis of the Florida Grasshopper Sparrow (Ammodramus savannarum floridanus): Testing recovery goals and management options. Auk 125:167–177. Perkins, D.W., W.G. Shriver, and P.D. Vickery. 2009. The effects of fire on the breeding ecology of Florida Grasshopper Sparrows (Ammodramus savannarum floridanus) and Bachman’s Sparrows (Aimophila aestivalis). Florida Field Naturalist 37:121–145. Platt, W.J., J.M. Huffman, and M.G. Slocum. 2006. Fire regimes and trees in Florida dry prairie landscapes. Pp. 3–13, In R. Noss (Ed.). Land of Fire and Water: The Florida Dry Prairie Ecosystem. Proceedings of the Florida Dry Prairie Conference. E.O. Painter Printing Company, DeLeon Springs, FL. 241 pp. Pranty, B. 2000. Three sources of Florida Grasshopper Sparrow mortality. Florida Field Naturalist 28:27–29 Pulliam, H.R. 1988. Sources, sinks, and population regulation. American Naturalist 132:652–661. Pulliam, H.R., and T.A. Parker III. 1979. Population regulation of sparrows. Fortschritte Der Zoologie 25:137–147. R Development Core Team. 2010. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. Available online at http://www.R-project.org. Accessed 24 January 2011. Ralph, C.J., G.R. Geupel, P. Pyle, T.E. Martin, and D.F. DeSante. 1993. Handbook of field methods for monitoring landbirds. General Technical Report PSW-GTR-144. Pacific Southwest Research Station, Forest Service, US Department of Agriculture, Albany, CA. 41 pp. Renfrew, R.B., and C.A. Ribic. 2002. Influence of topography on density of grassland passerines in pastures. American Midland Naturalist 147:315–325. Ribic, C.A., and D.W. Sample. 2001. Associations of grassland birds with landscape factors in southern Wisconsin. American Midland Naturalist 146:105–121. M.F. Delany, R.A. Kiltie, S.L. Glass, and C.L. Hannon 2013 Southeastern Naturalist Vol. 12, No. 3 654 Ribic, C.A., R.R. Koford, J.R. Herkert, D.H. Johnson, N.D. Niemuth, D.E. Naugle, K.K. Bakker, D.W. Sample, and R.B. Renfrew. 2009. Area sensitivity in North American grassland birds: Patterns and processes. Auk 126:233–244. Riddle, J.D., R.S. Mordecai, K.H. Pollock, and T.R. Simons. 2010. Effects of prior detections on estimates of detection probability, abundance, and occupancy. Auk 127:94–99. Rotella, J.J., E.M. Madeen, and A.J. Hansen. 1999. Sampling considerations for estimating density of passerines in grasslands. Studies in Avian Biology 19:237–243. Royle, J.A. 2004. N-mixture models for estimating population size from spatially replicated counts. Biometrics 60:108–115. Shriver, W.G., and P.D. Vickery. 1999. Aerial assessment of potential Florida Grasshopper Sparrow habitat: Conservation in a fragmented landscape. Florida Field Naturalist 27:1–36. Shriver, W.G., and P.D. Vickery. 2001. Response of breeding Florida Grasshopper and Bachman’s Sparrows to winter prescribed burning. Journal of Wildlife Management 65:470–475. Shriver, W.G., P.D. Vickery, and S.A. Hedges. 1996. Effects of summer burns on Florida Grasshopper Sparrows. Florida Field Naturalist 24:68–73. Simons, T.R., M.W. Alldredge, K.H. Pollock, and J.M. Wettroth. 2007. Experimental analysis of the auditory detection process on avian point counts. Auk 124:986–999. Stamps, J.A. 1988. Conspecific attraction and aggregation in territorial species. American Naturalist 131:329–347. Stephenson, K.E. 2011. Distribution of grasslands in 19th-century Florida. American Midland Naturalist 165:50–59. Thompson, W.L. 2002. Towards reliable bird surveys: Accounting for individuals present but not detected. Auk 119:18–25. Tucker, J.W., G.R. Schrott, M.F. Delany, S.L. Glass, C.L. Hannon, P. Miller, and R. Bowman. 2010. Metapopulation structure, population trends, and status of Florida Grasshopper Sparrows. Journal of Field Ornithology 81:267–277. US Fish and Wildlife Service (USFWS). 1999. Recovery for the Florida Grasshopper Sparrow. Pp. 4-387–4-391, In South Florida multi-species recovery plan. Vero Beach, FL. Available online at http://www.fws.gov/verobeach/MSRPPDFs/FloridaGrasshopperSparrow. pdf. Accessed 14 November 2011. Vickery, P.D. 1996. Grasshopper Sparrow (Ammodramus savannarum). No. 239, In A. Poole and F. Gill (Eds.). The Birds of North America. Academy of the Natural Sciences, Philadelphia, and American Ornithologists’ Union, Washington, DC. 24 pp. Vickery, P.D., M.L. Hunter, Jr., and J.V. Wells. 1992. Is density an indicator of breeding success? Auk 109:706–710. Walsh, P.B., D.A. Darrow, and J.G. Dyess. 1995. Habitat selection by Florida Grasshopper Sparrows in response to fire. Proceedings of the Annual Conference of the Southeastern Association of Fish and Wildlife Agencies 49: 340–347. Wiens, J.A. 1973. Pattern and process in grassland bird communities. Ecological Monographs 43:237–270. Wilson, D.M., and J. Bart. 1985. Reliability of singing bird surveys: Effects of song phenology during the breeding season. Condor 87:69–73. Winter, M., and J. Faaborg. 1999. Patterns of area sensitivity in grassland-nesting birds. Conservation Biology 13:1424–1436. Winter, M., D.H. Johnson, and J. Faaborg. 2000. Evidence for edge effects on multiple levels in tallgrass prairie. Condor 102:256–266.