Winter Use of South Florida Dry Prairie by Two Declining Grassland Passerines

Abstract.

Grassland-obligate birds are undergoing some of the steepest population declines of all North American passerines. Few studies have addressed these species' patterns of habitat use in their winter ranges in the southeastern United States. The dry prairie of south-central Florida constitutes one of the largest areas of contiguous grassland remaining within this region and during winter supports a diverse group of migratory grassland-obligate birds. Modern land-use of this region has altered the ecosystem's natural fire regimes and shifted the vegetative community away from graminoids and forbs toward overabundance of woody-stemmed species such as the saw palmetto (Sereona repens). Simultaneously, the prairie's historic range has decreased with urban development and conversion to agriculture. In order to understand how these changes affect overwintering grassland birds, we documented the dry prairie's winter bird community and evaluated the effects of habitat characteristics and time since fire on the occurrence of the Grasshopper Sparrow (Ammodramus savannarum pratensis) and Sedge Wren (Cistothorus platensis), two winter residents. We surveyed birds via flush transects and used an information-theoretic approach to select models that best predicted the species' occurrence. Time since fire was the best predictor of the Grasshopper Sparrow both years of our study, and occupancy by the Grasshopper Sparrow was six times more likely if transects were burned within the previous year. The Sedge Wren favored longer intervals between fires, and its response to habitat covariates in the two years differed. These results highlight the need for dry prairie to be managed with natural (1–3 years) fire-return intervals to maintain wintering habitat for declining grassland birds.

Resumen.

Las aves que obligatoriamente viven en pastizales estân sufriendo algunas de las disminuciones poblacionales más pronunciadas de todos los paserinos de América del Norte. Pocos estudios han tratado los patrones de uso de habitat de estas especies mientras están invernando en el sudeste de Estados Unidos. La pradera seca del sur-centro de Florida constituye una de las mayores áreas remanentes de pastizal continuo dentro de esta región y durante el invierno mantiene un grupo diverse de aves migratorias que obligatoriamente viven en pastizales. El uso actual del suelo de esta región ha alterado el régimen natural del fuego del ecosistema y ha desplazado la comunidad vegetal de graminoideas y herbáceas de hoja ancha hacia una sobreabundancia de especies leñosas como Sereona repens. Simultáneamente, la distribución histórica del hábitat ha decrecido con el desarrollo urbano y la transformación para agricultura. Para entender cómo estos cambios afectan a las aves que invernan en los pastizales, documentamos la comunidad invernal de aves de las praderas secas y evaluamos los efectos de las características del hábitat y del tiempo transcurrido desde un incendio en la presencia de Ammodramus savannarum pratensis y Cistothorus platensis, dos residentes invernales. Muestreamos las aves a lo largo de transectas y empleamos un enfoque teórico de la información para seleccionar los modelos que mejor predijeron la presencia de las especies. El tiempo desde el incendio fue la variable que mejor predijo a A. s. pratensis en ambos años de nuestro estudio. La ocupación por parte de esta especie fue seis veces más probable si las transectas habían sido quemadas en el transcurso del año anterior. Cistothorus platensis se vio favorecida por intervalos más prolongados entre los incendies y su respuesta a las covariables de hábitat difirieron en los dos años. Estos resultados enfatizan la necesidad de que la pradera seca sea manejada con intervalos naturales (1–3 años) de retorno del fuego para mantener hábitat invernal para las aves de pastizal que están declinando.

Introduction

Grassland birds have undergone population declines that surpass those experienced by any other suite of avian species in North America (Peterjohn and Sauer 1999, Brawn et al. 2001, Brennan and Kuvlesky 2005). Many members of this group are short-distance migrants, and substantial research has been conducted during the breeding season to link environmental phenomenon with demographic trends. Despite evidence suggesting that some migratory passerines may be limited during the nonbreeding portion of the life cycle (Marra et al. 1998, Rappole et al. 2003), relatively few studies have sought to understand the role that degradation of overwintering areas may play in population declines (Herkert and Knopf 1998, Vickery et al. 1999, Vickery and Herkert 2001). For example, the effect that successional vegetation change has on patterns of wintertime use by this group remains uncertain, in spite of the suggestion by several authors (Hunter 1990, Lymn and Temple 1991, Brennan and Kuvlesky 2005) that afforestation and the lack of early successional vegetation may be causing widespread declines in winter habitat for grassland birds across the southern United States. Furthermore, although prescribed burning is the most common means of managing southeastern savannas, empirical evidence describing how winter migrants respond to various fire-return intervals is largely lacking.

The dry prairie of south-central Florida constitutes one of the largest remaining contiguous grasslands within the southeastern United States. This ecosystem can be described as a treeless, pyrogenic landscape, composed primarily of wiregrass (Aristida beyrichiand), various bluestems (Andropogon spp.), scattered saw palmetto (Seronoa repens), and low-growing runner oak (Quercus minima;USFWS 1999). Over the last half century, anthropogenic changes in the region have severely reduced its quantity and quality. The composition and structure of the plant community have shifted dramatically as a result of changes in the natural seasonality, intensity, and frequency of fire. Historical evidence suggests that the natural fire frequency was at least every 1–3 years (FNAI and DNR 1990, Huffman 2006), with high probability that most dry prairies burned practically every year during the growing season (Harper 1921, USFWS 1999). Current land-use practices incorporate prescribed burning to promote forage for livestock, but burning is often applied at less frequent intervals and primarily during the season of dormancy (Mealor and Prunty 1976). Saw palmetto and other woody-stemmed species such as gallberry (Ilex glabra) are favored over forbs and native grasses as a result of this divergence from natural fire patterns, and given time come to dominate most prairies (Richardson 1977, Olson and Platt 1995, Watts et al. 2006). Additionally, much of the remaining dry prairie is being continually lost to conversion to citrus groves, pastures of exotic grass, and residential development (USFWS 1999); estimates suggest that <2% of the dry prairie remains intact and unaltered (Noss et al. 1995).

The importance of dry prairie as winter habitat for migrant grassland birds, and the cumulative effects of ecosystem change on these species, are relatively unknown (USFWS 1999). If management goals for the remaining fragments of dry prairie are to consider the needs of wintering grassland obligates, the habitat preferences of these species during winter must be better understood. Therefore, we documented the winter community of grassland birds on a Florida dry prairie and investigated how two members of this group respond to habitat variation and prescribed burning. Specifically, we examined how the occurrence of the Grasshopper Sparrow (Ammodramus savannarum pratensis) and Sedge Wren (Cistothorus platensis), two migratory grassland obligates of high conservation concern, was related to variation in vegetative composition and time since fire. Research on their breeding grounds has demonstrated that Grasshopper Sparrows tend to prefer grasslands that are free of woody-stemmed plants, offer openings near the ground for foraging, and are frequently burned (Vickery 1996, Dechant et al. 2003a). During the breeding season, Sedge Wrens are typically associated with mesic prairies with tall, dense vegetation and may prefer longer fire-return intervals than most other grassland species (Dechant et al. 2003b, Herkert et al. 2001). The foods each species exploits also differ; during winter the Grasshopper Sparrow is principally granivorous (Vickery 1996), while the Sedge Wren relies primarily on insects (Herkert et al. 2001). Because of their dissimilar life histories and habitat preferences, we believed that these two species would adequately serve as indicators from which broader generalizations for the entire suite of overwintering grassland birds could be drawn.

Our study was based largely upon two hypotheses. First, we believed that coverage by saw palmetto and other woody shrubs would be the most influential variable affecting the occurrence of both birds. High levels of saw palmetto have been shown to affect some resident grassland birds negatively by altering habitat structure and reducing herb diversity (Fitzgerald and Tanner 1992, Tucker and Bowman 2006, Butler 2007). Secondly, because foraging conditions for ground-dwelling birds are improved following fire (Haggerty 2000, Tucker and Bowman 2006), we believed that the occurrence of each species would be strongly related to time since fire, although we expected the interval each species prefers to differ.

Methods

Study Site

We conducted our study within the historic range of the dry prairie in De Soto, Manatee, and Sarasota counties, Florida, during January and February of 2006–2007. We selected four study sites; three were privately owned ranches managed for livestock and the landowners' recreation, and the fourth was a state park managed to promote biodiversity and maintain natural ecosystems. Dry prairie was the major land cover at all study sites, but hardwood hammocks, pastures of exotic grass, and ephemeral wetlands were intermixed among the prairie fragments. At all four sites, management of dry prairie was similar, a combination of prescribed burning and roller-drum chopping. Among the sites, the only major difference in management was periodic or continuous grazing by cattle on the private ranches and total exclusion of cattle at the state park. During winter, the region's subtropical climate is mild and extremely dry. Rainfall for the months of December to February averages 16 cm, and mean daily temperature is approximately 17.7° C.

Avian Surveys

We identified patches of dry prairie within each study site in the field and through inspection of satellite imagery. On the private ranches we were able to survey all existing patches of dry prairie, but because of the size of the state park (16 194 ha) we randomly selected patches to survey. We surveyed birds by the flush-transect technique suggested by Fletcher et al. (2000). Using this method, two observers dragged a 25-m rope fitted with noise makers (2-L plastic bottles filled with rocks) across the top of the vegetation. A third observer was positioned near the rope's center, roughly halfway between the other two observers. We identified flushed birds in flight to species by distinctive flight patterns, call notes, or coloration. To aid in species identification, the direction of travel along transects was selected to prevent observers from facing directly into the sun. Transects measured 25 × 100 m, encompassed an area of approximately 0.25 ha, and were covered in a single pass. We covered each transect in 3.7 ± 1.9 min (mean ± SD). Our survey patches contained from 2 to 10 transects, according to patch size. Transects were spaced ≥200 m apart to minimize the probability of double counting individual birds. Each transect was surveyed twice each year, and if a transect was severely altered (e.g., burned) between two surveys, data from that transect were not used for analysis. Surveys began at sunrise and continued until 13:00 and were conducted only during favorable weather with no precipitation and winds <10 km hr-1.

The overlap in winter between the Eastern Grasshopper Sparrow (Ammodramus savannarum pratensis) and the endemic Florida Grasshopper Sparrow (A. s. floridanus; Delany 1985) could have biased our assessment of the migrant's habitat preferences if these two differed in patterns of occupancy. The subspecies are nearly indistinguishable in the field, making it doubtful that technicians could have differentiated between the two while conducting surveys. Despite intensive sampling during the breeding season (Delany et al. 2007, Butler 2007), however, the endangered Florida Grasshopper Sparrow has not been found near our field sites, so we believe that our results apply to migrants only.

Vegetation Surveys

Each year, after the second bird survey, we measured vegetation by a variation of the method suggested by Herrick et al. (2005). We assessed the percent of coverage of the vegetation's major functional groups along two 25-m lines perpendicular to each bird transect, the first at the starting point, and the second at approximately the 75-m mark. At half-meter intervals along each line, a metal pin-flag was dropped and all functional classes touching the pin were recorded. We also noted whether the pin-flag struck leaf litter or bare soil. All vegetation variables along both lines within a transect were averaged prior to data analysis.

Statistical Analyses

Detection rate. Before investigating the relationship between the two species and environmental variables, we assessed the efficacy of the flush-transect technique. If flushed birds landed on portions of the transect not yet covered, we recorded those individuals' reflushing. To quantify the probability of reflushing, we used the program MARK to develop closed-capture models (White and Burnham 1999). Reflushes were treated as recaptures in Huggins' closed-capture model. We believe that this recapture rate was analogous to detection probability under the following assumptions: (1) behavior by individual birds during initial and subsequent flushes is similar, and (2) the probability of detecting the flushing of a new individual and the probability of detecting the flushing of a previously observed individual are equal.

Often birds, particularly the weak-flying Sedge Wrens, were reflushed multiple times over the length of a transect. We believe that the assumption of similar behavior during the initial flush and subsequent reflushes was not reasonable as the bird became increasingly disturbed. Therefore, the detection probability was calculated with only the first reflush after each bird's initial encounter. We believe that assumption 2 was realistic because the three observers were spaced roughly 12 m apart along the rope and at least two were focused on the transect at all times, ensuring that if birds were flushed, they were not missed.

We estimated the models' parameters to include the effects of species, year, and all vegetation covariates that we suspected might influence flushing rates. Vegetation covariates included the coverage of saw palmetto, bunch grass, broadleafed woody vegetation, and the interactions between pairs of variables. We evaluated the relative plausibility of models with Akaike's information criterion adjusted for small sample size (AIC_c; Burnham and Anderson 2002). To facilitate comparisons further, we calculated a ΔAIC_c value for each model, then normalized these values to obtain an Akaike weight (w_i) for each model within the candidate set. We calculated a modelaveraged parameter estimate and standard error for detection probability across all models within the confidence set (Burnham and Anderson 2002).

Species occurrence. Our model-averaged parameter estimates for detection probability were high (see Results), so we concluded that robust methods to account for detectability in presence/absence surveys were not warranted. Instead, we assessed the environmental factors influencing occurrence of Grasshopper Sparrows and Sedge Wrens by creating logistic-regression models for each species in which the response variable was occupancy by an individual in either trial of the surveys. Our choice of the binomial distribution was a result of the natural dispersion of the two species; most occupied transects contained only single individuals, which prevented the use of other generalized linear models to fit the data.

Although we surveyed the same transects in both 2006 and 2007, land management often significantly altered the habitat in the intervening year; therefore, we analyzed years separately. Models used in the regression analysis included combinations of four vegetation variables (saw palmetto coverage, bunch-grass coverage, coverage by broad-leafed woody-stemmed plants, and forb coverage), coverage by dead vegetative litter, distance to the nearest edge, and, in Sedge Wren models, distance to the nearest wetland. We also included time since burning as a predictor, categorizing it as 1 year, 2 years, and ≥3 years. In 2007, we observed no Sedge Wrens on transects burned in the previous year, so for the Sedge Wren models in 2007, we created a binary variable in which burns 1 and 2 years old were combined and compared to burns ≥3 years old.

We calculated correlation coefficients between all possible pairs of data. If two variables were strongly correlated (r² ≥ 0.30) we chose only the variable that we believed to be most biologically meaningful for inclusion in regression models. For instance, measurements of vegetative density were correlated with coverage by saw palmetto as a result of the plant's growth structure. However, we believed that overwintering passerines would be more affected by the reduction in graminoids and forbs that occurs when coverage of palmetto is high (Olson and Platt 1995) than by overall vegetative density, so we chose saw palmetto coverage as the variable for inclusion in predictive models. We suspected a priori that birds' response to certain vegetation variables was possibly nonlinear, so we ran base logistic models with each vegetation variable alone and visually examined a plot of the residuals for signs of a quadratic response.

In our study, transects were nested within patches. Failing to acknowledge the nested nature of our data would have resulted in pseudoreplication (Hurlbert 1984). Therefore, we accounted for this potential lack of independence by creating mixed models, which allowed the intercepts and slopes of our logistic models to vary randomly by patch. The effect of each individual patch was a random effect, for which regular logistic regression would not have accounted. Fixed effects included the mean effect of transect-specific characteristics, such as vegetative coverage, across all patches. We fit all mixed models by using the NLMIXED procedure of SAS software (SAS Institute 2003) and then determined the relative plausibility of each model by using AIC_C and the respective AIC_c model weights (Burnham and Anderson 2002). We assessed the goodness of fit of the global model and each model in the candidate set by visually examining plots of the residuals for lack of fit and obvious outliers.

Though some inference can be gained from model selection, we were most interested in the effect individual covariates had on the occurrence of grassland passerines. Among candidate models, the size of the effect of covariates can differ, but with mixed models use of model averaging to incorporate this uncertainty into parameter estimates is not suitable (J. T. Peterson, University of Georgia, pers. comm.). Consequently, we report the parameter estimates from all models contained within the confidence set of models. We defined the confidence set of models as those models with an Akaike weight of ≥10% of the weight of the most parsimonious model (Burnham and Anderson 2002). It is unlikely that the effect of a one-unit rate of change, as represented by the models' parameter estimates, would be significant enough to alter avian response to many of the variables we measured. Therefore, to aid the interpretation of our results, we calculated scaled odds ratios for each parameter by exponentiating the product of the parameter estimate and a scaling value that we deemed biologically relevant. For example, we believed that a 10% increase or decrease in coverage by saw palmetto may be the minimum change needed to yield ecological consequences that may influence grassland birds, so for each model we multiplied the parameter estimate of saw palmetto by this value prior to calculating the odds ratio. As a result, the scaled odds ratios that we report can be interpreted as the number of times a bird is more or less likely to occur after a parameter has increased or decreased, respectively, equivalent to the value of the unit sealer.

Results

We analyzed data from 141 transects within 25 patches during 2006 and 127 transects within 24 patches during 2007. Mean patch size was 241.0 and 228.9 ha during 2006 and 2007, respectively. The results of both years pooled, we observed 859 individuals of 23 species, and nearly half were migrants that use Florida's dry prairie only during winter. Species considered dependent upon grassland, shrub-scrub, or early successional vegetation accounted for 13 of the 23 species (Table 1).

Detection Rate

The model-averaged parameter estimate for detection probability was high ( = 0.8855, SE = 0.0212; Table 2). The bestfitting model suggested that detection rates did not differ by species or by year and that probability of flushing increased slightly as coverage of broad-leafed woody vegetation increased (parameter estimate = 0.4959, SE = 0.2281). This model was 1.89 times more likely than the second best-fitting model, which contained no covariates and did not allow detection to vary by species or year. All models within the confidence set reported detection rates ≥87 % (Table 2).

Grasshopper Sparrow

We observed 148 Grasshopper Sparrows in 2006, 68 in 2007. The Grasshopper Sparrow was the species we encountered most frequently each year, on 50% and 34% of transects during 2006 and 2007, respectively (Table 1). Although the level of support varied, the model containing only the parameters for time since burning fit best in both years (2006: w_i = 0.5585; 2007: w_i = 0.2382; Table 3). In 2006, the model containing only time since burning was over twice as likely as the second best-fitting model and three times more likely than the third-ranked model (Table 3). In 2007, the model containing only the burn variable was supported less, and no single model garnered a majority of the overall Akaike weight.

Although the relative strength of support for the timesince-burn model in each year varied, the parameter estimates for both years were nearly identical (Table 4). Models containing only the parameter for burning showed a significant effect of 1-year-old burns. The biological effect of 1-year-old burns on Grasshopper Sparrow occupancy was dramatic; in both years, the probability of the Grasshopper Sparrow occurring in 1-year-old burns was double its probability of occurring in older burns (Figure 1). This result was further supported by scaled odds ratios that suggested that Grasshopper Sparrows were approximately six times more likely to occupy a transect that had been burned within the previous year than one that had not been burned in ≥3 years (Table 4).

No other parameters shared consistent effects in both years, suggesting that the habitat characteristics we measured were not as important to sparrow occupancy as time since a fire. In the set of models for 2006, occurrence of Grasshopper Sparrows tended to decrease with increases in saw palmetto, although this effect was not statistically significant and did not appear in any of the 2007 models (Table 4). In 2006, vegetative litter was a component of two models, in each of which it had a significantly negative effect on sparrow occurrence (Table 4). Conversely, in 2007 the model containing only litter was included in the confidence set but received little weight, and estimates for that variable suggested its effect on Grasshopper Sparrow occurrence was negligible (Tables 3 and 4). The model containing only forb coverage ranked the third highest in 2007 and received nearly 70% of the weight of the best-fitting model (Table 3). Parameter effects from that model suggested that forb coverage had a negative effect on sparrow occurrence (Table 4). This effect was not seen in the 2006 models.

Sedge Wren

Sedge Wrens were somewhat less abundant than Grasshopper Sparrows. We observed 25 individuals on 21 transects during 2006 and 45 individuals on 40 transects during 2007. None of the variables in the models affected the Sedge Wren consistently in both years of the study.

In 2006, the model containing only vegetative litter fit best (w_i = 0.3638) and was 1.72 times more likely than the secondranked model (bunch grass only, w_i = 0.2103) and >2 times more likely than the third-ranked model (saw palmetto + bunch grass + broad-leafed woody + forb + litter; w_i = 0.1586; Table 5). Percent coverage by litter affected Sedge Wren occupancy negatively in 2006 and was relatively strong in all models in which it was contained. For instance, the estimate from the best-fitting model suggests that occurrence of the Sedge Wren was 2.58 times less likely for every 25% increase in coverage by vegetative litter (Table 6). Bunch grass was positively related to occurrence of Sedge Wrens, though the effect was not as strong as that of litter; occurrence of the Sedge Wren was 1.53 times more likely for every 10% increase in coverage by bunch grasses (Table 6). In the 2006 set of models distance to the nearest wetland also played a role in occupancy by the Sedge Wren. It was 2.4 times less likely on a transect for every 100-m increase in distance to the nearest wetland (Table 6).

In 2007, the model containing only time since burning fit best (w_i = 0.4328) and was >2.5 times more likely than the second-ranked model, which contained only litter (w_i = 0.1620; Table 5). Sedge Wrens were 2.26 times less likely to occur on transects that had been burned in the previous 2 years than on transects that had not been burned in ≥3 years (Table 6). Neither the results of model selection nor their parameter estimates suggested that time since burning was important to Sedge Wren occurrence during the previous year. As in 2006, model-selection results for 2007 suggested that litter and bunch grass played a role in the wrens' occupancy (Table 5). The direction of the effect of both parameters, however, was the reverse of that in 2006. Additionally, in 2007 estimates of both these parameters were quite uncertain, and neither had statistically significant effects in any model in which it was included (Table 6).

Discussion

Results of this study demonstrate that the value of dry prairie as wintering habitat for some grassland birds declines relatively quickly (≥3 years) in the absence of fire. These findings support those of Hunter (1990), Lymn and Temple (1991), and others who have suggested that in the southeastern United States degradation of savannas has limited the extent of wintering habitat for migrant grassland birds. Over half the species we documented are tied to vegetative communities dependent on disturbance, and the needs of many of these are likely met best by periodic fires. Our results also illustrate, however, that a detailed understanding of each species' biology and response to management practices is critical to the development of conservation strategies for a dynamic system like the dry prairie, where small alterations in management (e.g., fire regime) will favor some species over others.

Variation in habitat and fire-return intervals affected the Grasshopper Sparrow and Sedge Wren differently, in the Sedge Wren, differently in the two years of our study. Our initial hypothesis of an inverse relationship between saw palmetto and grassland birds' occupancy was not strongly supported by our predictive models. In both species coverage by saw palmetto appeared in some models in the confidence set, but in all cases its effects were negligible. The effects of time since fire, however, remained relatively consistent and were particularly strong in predicting occupancy by the Grasshopper Sparrow, with its clear affinity for areas burned within the previous year. This association has also been documented for this species in the northern part of its breeding range (Herkert 1994, Madden et al. 1999) and for the Savannah Sparrow (Passerculus sandwichensis) in its winter range on coastal prairies in Louisiana (Baldwin et al. 2007). Granivorous species such as the Grasshopper Sparrow may select recently burned areas because richness and seed production of herbs are highest during the growing season immediately following a fire (Buckner and Landers 1979). Additionally, burning likely improves foraging conditions by removing dead vegetative thatch, allowing birds to move more freely near the ground. The relationship between fire and subsequent increases in seed production and availability has also been suggested to influence the abundance of other Ammodramus sparrows overwintering on southeastern savannas. For instance, in longleaf pine forests in the Southeast, numbers of Henslow's Sparrow (Ammodramus henslowii) are highest during the winter immediately following fires, likely as a result of increased food resources (Tucker and Robinson 2003, Bechtoldt and Stouffer 2005), and this association may have serious consequences for overwinter survival (Thatcher et al. 2006).

During the study's first year, our models suggested that fire affected the occurrence of the Sedge Wren only weakly, but in 2007 time since fire was the strongest parameter influencing Sedge Wren occurrence. That year, the probability of Sedge Wren occurrence increased with time since fire. The species preferred areas that had not been burned in >2 years and was absent from all transects that had been burned within the previous year. On coastal prairie in Louisiana, Baldwin et al. (2007) found that wintering Sedge Wrens were more common within areas that had not been burned in 2 or 3 years. Studies on the breeding grounds also suggest that Sedge Wrens favor areas that have not been burned for several years (Herkert 1991, Johnson 1997).

Although we found some evidence that fire affects Sedge Wrens negatively, it is clear that some level of natural disturbance through fire is required to maintain a prairie in the conditions to which Sedge Wrens are adapted (Dechant et al. 2003b). Our study, however, was unable to demonstrate the maximum fire-return interval needed to maintain habitat usable by Sedge Wrens. Yet our results do suggest that under certain conditions burning may have short-term detrimental effects on wintering Sedge Wrens. Unlike the Grasshopper Sparrow, the Sedge Wren is primarily insectivorous on south Florida's dry prairie (Herkert et al. 2001). Although the long-term effects of burning on arthropods can be positive (Anderson et al. 1989, Siemann et al. 1997), immediately after a fire the abundance and diversity of some insect species can decline temporarily (Swengel 2001). During the growing season of 2006 and into the first months of 2007, south Florida experienced severe drought (SFWMD 2007). These dry conditions may have inhibited insect production, exacerbating the negative effects of burning on insect populations, and perhaps explaining why Sedge Wrens avoided recently burned areas during the winter that followed the drought. Sedge Wrens responded similarly to burning followed by prolonged drought on tall-grass prairies in Kansas (Zimmerman 1993).

Our estimates of detection probability approached 90%, and we believe that this level of detectability would not have been attainable without use of the flush-transect technique. Other studies of wintering sparrows on the dry prairie have used more passive survey methods and consequently reported fewer observations of migrant grassland birds (Fitzgerald and Tanner 1992). The technique we employed would not be practical in savannas where trees or shrubs create obstacles, but in open prairies and grasslands, the technique of flushing may be one of the only means of surveying wintering species adequately. Identification of nondescript passerines can be difficult while birds are in flight, but we found that all observers were able to identify the common species of our study sites consistently after only a few days of training. We suggest that future studies of cryptic nonvocal passerines in open environments should use this survey method.

Many managed prairies in southern Florida are currently burned every 2 to 4 years. Our results suggest that the upper end of this interval may be too long to provide suitable wintering habitat for the Grasshopper Sparrow. Although a species' presence is not a valid surrogate for demographic parameters in an assessment of habitat quality (Van Horne 1983), other studies have shown that habitat preferences may be strongly tied to overwinter survival (Tucker and Robinson 2003, Thatcher et al. 2006). Therefore, we suggest that dry prairie should be burned at 2-year intervals in order to maintain habitat suitable for wintering passerines. This frequency is within the historical range suggested for dry prairie (FNAI and DNR 1990, USFWS 1999), and fires at this frequency will likely be unable to spread through certain areas because of discrepancies in fuel loads (Wade and Lunsford 1989). Areas left unburned should accumulate those resources important to species such as the Sedge Wren, while frequently burned portions will create quality habitat for the Grasshopper Sparrow and other closely related granivorous species.

Acknowledgments

We are indebted to the staff of Myakka River State Park as well as the Hall, Paul, and Carlton families for granting permission to conduct this study on their properties. The assistance of J. Good, J. Harn, M. Prasek, A. Sweet, and C. Yarborough was invaluable in the collection of data. We thank R. Cooper and J. Peterson for providing guidance and advice on data collection and analysis. The University of Florida Institute of Food and Agricultural Sciences provided both support and housing during our field work. Support was also given by the Florida Freshwater Fish and Wildlife Commission, and we especially thank T. Hines for his role in the commission's involvement. Funding for this project was provided by the U.S. Department of Agriculture—National Resource Conservation Service (NRCS) Bobwhite Restoration Project through the NRCS Agricultural Wildlife Conservation Center and Tall Timbers Research Station and Land Conservancy.

Literature Cited