Introduction

Of all the land plants the orchids (Orchidaceae) are among the most beautiful and charismatic. Found on all continents except Antarctica, the Orchidaceae is one of the most diverse families of flowering plants with approximately 20,000 species (Smith et al. Smith 2004). In Maryland, 21 genera and 51 species are known (Knapp and Naczi  unpublished data) occupying a diverse array of habitats from dry to wet substrates in forested to open-sunny conditions (Brown and Brown 1984). In the Catoctin Mountains of Frederick Co., Maryland, 27 species (native and non-native) have been informally reported (Wieg and unpublished data). Of these 27 species, 21 were readily occurring at the onset of this study. Four are listed as threatened or endangered (Maryland Natural Heritage Program 2010): longbract frog orchid (Coeloglossum viride var. virescens, yellow fringed orchid (Platanthera ciliaris), greater purple fringed orchid (Platanthera grandiflora), and yellow nodding ladie’s tresses (Spiranthes ochroleuca). Two are listed as rare (Maryland Natural Heritage Program 2010): brown widelip orchid (Liparis liliifolia), and palegreen orchid (Platanthera flava var. herbiola).

The understanding of how species respond to natural or management-induced habitat changes is a prerequisite for successful species conservation (Hegland et al. 2001). Species populations are known to be highly variable over a short time scale due to many environmental conditions (e.g., competition, climate). Therefore, long-term population data are needed to obtain reliable information on the life history and population dynamics of any species (Waite and Hutchings 1991; Fieberg and Ellner 2001). Many long-term studies of various terrestrial orchid groups show natural population fluctuations or effects of temporal environmental conditions (Tamm 1972; Hutchings 1987; Mood 1989; Willems and Meiser 1998; Gillman and Dodd 1998; Shefferson et al. 2003; Kery and Gregg 2004; Pfeifer et al. 2006).

The effects of white-tailed deer (Odocoileus virginianus Boddaert Boddaert) herbivory on vegetation and plant community structure is well known (Augustine and Frelich 1998; Gill and Beardall 2001; Rooney 2001; Russell et al. 2001; Horsley et al. 2003, Rooney and Waller 2003; Côté et al. 2004; Krueger and Peterson 2006; Mudrak et al. 2009; Freker et al. 2013). Overtime, elevated levels of herbivory can lead to density decline and extirpation of herbivory intolerant plants (Rooney and Dress 1997a; Fletcher et al. 2001). Regionally, a number of studies have shown impacts of herbivory on various species (Whigham 1990; Whigham and O’Neill 1991; Rooney and Dress 1997b; McGraw and Furedi 2005). Numerous studies have shown browsing by white-tailed deer at densities greater than 15–20 deer/mi2 can influence forest regeneration success (Hough 1965; Behrend et al. Behrend 1970; Marquis 1981; Tilghman 1989). Langdon (1985) noted that deer impacts on plant communities consists of three primary effects: (1) failure of plants to reproduce, (2) alteration of species composition which occurs when deer remove preferred browse species and indirectly create opportunities for less preferred or unpalatable species to proliferate, and (3) extirpation of highly palatable plants, especially those that were naturally uncommon or of local occurrence.

During the course of this study the deer population of the Catoctin Mountains became sizeable and an obvious ‘browse line’ developed (National Park Service 2008). Porter (1991) estimated the deer density at Catoctin Mountain Park, located within the study area, to exceed 40 deer/km2 (100 deer/mi2) and research has established that such high deer densities have negative impacts on plant and animal species (Anderson 1994; Alverson et al. 1998; Augustine and Frelich 1998; de Calesta 1994; McShea and Rappole 2000).

The initial intent of this long-term study was to document changes to orchid demographics in a large area over time. The unanticipated declines documented stimulated an investigation into possible causes. This post-facto effort links the declines of orchids to the deer population, using deer harvest data as a surrogate for population. The objectives of this study are to raise awareness of the decline in orchid species in the Catoctin Mountains and show the main cause of this decline is most likely herbivory by white-tailed deer.

Study area

The Catoctin Mountains study area is approximately 485 km2 (301 mi2). Located in Frederick Co., Maryland and comprises the easternmost ridge of the Blue Ridge Mountains (Fig. 1; Schmidt 1993). The Catoctin Mountains are oriented in a northeast/southwest direction and stretch approximately 80 km from their origin at South Mountain near Emmitsburg, MD, to the south just past Leesburg, VA, where they become undifferentiated into the Piedmont Physiographic Province (Schmidt 1993). The point of greatest elevation is just southwest of Cunningham Falls State Park at 538 m (1,765 ft). The Catoctin Mountains are located in the transition between two of Köppen’s climatic types; the Cfa, of Humid Subtropical climates, and Dfa, of Humid Continental climates (Markus et al. 2006). Average weather conditions at Catoctin Mountain Park, located near Thurmont, MD, indicate that the climate is cool-temperate with a mean annual temperature of 12.0 °C (53.6 °F) and a mean annual precipitation of 115.2 cm (45.3) in (NOAA 2013).

Fig. 1
figure 1

Map of the Catoctin Mountains study area with State of Maryland depicted to the left. Circles locations of a survey site

Floristically, the Catoctin Mountains are dominated by deciduous forests comprised primarily of oak (Quercus L. spp.). NatureServe (2011) documents six forested systems present within the study area: Appalachian Hemlock (Tsuga Canadensis (L.))-Northern Hardwood Forest, Central Appalachian dry oak-pine (Pinus L. spp.) forests, Central Appalachian pine-oak rocky woodlands, central Appalachian steam and riparian forests, North-Central Appalachian acidic cliff and talus, and northeastern interior dry-mesic oak forests.

The study area is located within the Catoctin-South Mountain Region of the Blue Ridge and is underlain by Catoctin metabasalt greenstone (Schmidt 1993; Reger and Cleaves 2008). The area contains a number of large protected lands including Catoctin Mountain Park, Cunningham Falls State Park, Frederick Municipal Forest, Gambrill State Park, and Camp David, the Presidential retreat. Culling of the deer herd was not allowed on Catoctin Mountain Park until 2010 (Loncosky personal communication).

Materials and methods

Twenty-one species of orchids were inventoried annually at 167 sites, during various times of year beginning in 1968 and ending in 2008 using traditional Natural Heritage Program inventory methods, namely the Observational Data Standard (Fig. 1; NatureServe 2006). The surveys were conducted by the second author or, occasionally, with assistance of other knowledgeable individuals. This reduced the likelihood that a site would not be thoroughly explored and helped limit issues with varying survey efforts among distinct surveyors over the years of the survey. Each site was thoroughly explored during surveys to ensure an accurate census. Individuals were counted using a hand held tally counter with the results of each site census recorded in a field notebook. The detailed locations of these sites are mapped and available upon request. They are not included here because a number of these species are considered by the Maryland Natural Heritage Program to be vulnerable to collecting. The study sites are located throughout the Catoctin Mountains and stretch nearly 50 km (31 mi) north to south and 16 km (10 mi) east to west (Fig. 1). The majority of these sites (142) are located in the northern portion of the Catoctin Mountains, where the Mountains become wider and occupy more landmass. Numerous sites have more than one species of orchid that are not easily detected at the same time of year due to distinct flowering and fruiting periods between species. This required several site visits throughout the year to accurately census the orchids at a given site. The total number of years that each individual species was censused varied (Table 1) as species were encountered at different times during the study and not all species were sampled each year.

Table 1 Orchid summary statistics

Because species were not sampled each year, missing data were estimated using the regression substitution method (Kauffman et al. 2003; Little and Rubin 1987). This method calculates a value for the missing variable by calculating a trend from each species at an individual site. To prevent simply reinforcing the trend line from which the missing variable is calculated some error is added (Little and Rubin 1987; SPSS 2004).

After a complete dataset was constructed, data for each species were summarized by years of inventory, total number of years, number of sites, highest census number with year, final census number, actual percent decline (calculated using highest census versus final census), and percent of data missing per species. These types of data (year and census) lend themselves to trend analysis using ordinary least squared analysis (Gotelli and Ellison 2004). These analyses were conducted using Systat version 11 (SPSS 2004). Each species was graphed showing total census on the Y-axis and year on the X-axis. The corresponding best fit line, R2 value and p-value were calculated.

No white-tailed deer population estimates are available for Frederick County or the Catoctin Mountains. White-tailed deer harvest data is available for Frederick County. These data were acquired from Brian Eyler (Wildlife and Heritage Service Deer Project Leader—Maryland Department of Natural Resources) and were used to provide an index of deer population size (Roseberry and Woolf 1991).

An inverse correlation analysis comparing the overall orchid census from 1987 to 2008 to the annual Frederick County white-tailed deer harvest during the same time period was completed. The year 1987 was selected for this analysis because this is the first year a complete dataset is available for all 21 species of orchids surveyed during the study.

Results

Nineteen species had significant declines, three species disappeared, one species was stable across the study and one expanded. Data is presented in three arbitrarily assigned categories for ease of presentation: species that disappeared, species with >90 % decline, and species with <90 % decline. Seven species showed a total decline of over 90 % (Table 1; Fig. 2), and nine showed declines from 51 to 87 % (Table 1; Fig. 3). Platanthera flava var. herbiola, did not decline, and P. ciliaris experienced significant growth (Table 1; Fig. 3). The R2 values are presented on each species census graphs (Figs. 2, 3). All regressions had calculated p-values of <0.005.

Fig. 2
figure 2

Species with a >90 % total decline, including the ‘species that disappeared’. Census (Y-axis), year (X-axis) with name for each species abbreviated along the Y-axis. Top row: A. hyemale, C. maculata var. maculata, C. odontorhiza var. odontorhiza, C. parviflorum var. pubescens. Middle row: E. helleborine, L. liliifolia, P. orbiculata, S. lacera var. gracilis. Bottom row: S. ochroleuca, T. discolor

Fig. 3
figure 3

Species with a <90 % total decline. Census (Y-axis), year (X-axis) with name for each species abbreviated along the Y-axis. Top row: C. viride var. virescens, C. acaule, G. spectabilis, G. pubescens. Middle row: I. verticillata, P. ciliaris, P. clavellata, P. flava var. herbiola. Bottom row: P. grandiflora, P. lacera, S. cernua

Species that disappeared

Three species disappeared during the study period with census totals becoming zero (Table 1; Fig. 2). These were the greater yellow lady’s slipper (Cypripedium parviflorum var. pubescens), lesser round-leaved orchid (Platanthera orbiculata) and Spiranthes ochroleuca. The number of sites for P. orbiculata and S. ochroleuca (9 and 4, respectively), years of survey (26 and 24), and initial number of individuals (59 and 41) are very similar. The loss of C. parviflorum var. pubescens is more striking as over 28 years there were more sites (17) and a larger number of individuals (127).

Species with >90 % decline

Seven species showed a total decline of >90 % (Table 1; Fig. 2). Among these species is the only non-native species of orchid known in the Catoctin Mountains, broadleaf helleborine (Epipactis helleborine). The six other species are Adam and Eve orchid (Aplectrum hyemale), summer coralroot (Corallorhiza maculata var. maculata), autumn coralroot (Corallorhiza odontorhiza var. odontorhiza), Liparis liliifolia, northern slender lady’s tresses (Spiranthes lacera var. gracilis), and the crippled crainfly (Tipularia discolor). Liparis liliifolia showed an increase in 2008 (Fig. 2). After averaging only 4 plants/year census from 2002 to 2007, 27 plants were found in 2008. Of these species the decline of C. odontorhiza is the most striking with a census high of 977 individuals in 1986 declining to just 70 individuals in 2008. The R2 values for these species are among the highest documented during the study, all of which range from 0.85 to 0.94 (Fig. 2).

Species with a <90 % decline

Nine species showed declines of <90 % (Table 1; Fig. 3). These species are Coeloglossum viride var. virescens, moccasin flower (Cypripedium acaule), showy orchid (Galearis spectabilis), downy rattlesnake plantain (Goodyera pubescens), large whorled pogonia (Isotria verticillata), small green wood orchid (Platanthera clavellata), Platanthera grandiflora, green fringed orchid (Platanthera lacera), and nodding lady’s tresses (Spiranthes cernua). Cypripedium acaule and G. spectabilis are arguably the most common terrestrial orchids in the Catoctin Mountains. These showed declines from 1,168 and 1,319 individuals to 128 and 66 individuals, respectively. Five of these species (C. viride var. virescens, I. verticillata, P. clavellata, P. grandiflora, and P. lacera) showed an obvious yet unexpected census increase in 2008 (Fig. 3). The R2 values for these species are more variable than the >90 % group. Goodyera pubescens shows the highest R2 value (0.97) of all species in this study. Only P. grandiflora (R2 = 0.53) and C. viride var. virescens (R2 = 0.75) have R2 values <0.85.

Species that did not decline

Two species did not show declines. These are Platanthera ciliaris and Platanthera flava var. herbiola. Platanthera flava var. herbiola shows a very slight decline (16 plants) but no highly correlated R2 values (Table 1; Fig. 3). Platanthera ciliaris shows an overall census increase, but 94.8 % of this growth is found at a single site called Whiskey Springs Pond. If this site is removed from the dataset, the species declines by over 94 % with R2 = 0.94 (Table 1; Fig. 3). From 1984 to 1988 an average of 14 plants were observed at Whiskey Springs Pond. After the implementation of a periodic mowing regime, beginning in 1989, the average annual census has increased to 227 plants.

Relationship between orchid census and deer harvests

Though deer harvest data is not a perfect replacement for deer population data, it does illustrate trends. In the 1900’s white-tailed deer were nearly extirpated from the State of Maryland (Maryland Department of Natural Resources 2013). In Frederick County, the number of individual deer harvested from 1960 to 1980 increased from 229 to 710, a nearly threefold increase. From 1980 to 2000, the harvest showed exponential growth going from 631 individuals to 7,843 individuals, a 12-fold increase (Fig. 4). From 2001 to 2008 the number of deer harvested became more erratic. The harvest peaks at 8,578 in 2002, decreases to 6,884 in 2006, then increases once again to 8,238 in 2008 (Fig. 4). The Inverse Correlation Analysis comparing the total deer harvest in Frederick County, to the overall orchid census from 1987 to 2008 yielded a R2 value of −0.93 (Fig. 4).

Fig. 4
figure 4

Inverse correlation of the deer harvest of Frederick County to overall orchid census. Squares no. of deer harvested, Circles individual orchids census

Discussion

Recent studies of long-term orchid population data documented annual fluctuations in orchid species (Alexandersson and Agren 1996, Gillman and Dodd 1998, Pfeifer et al. 2006, Rasmussen and Whigham 1998). The data collected in this study show no such annual fluctuations. This makes an explanation based on weather patterns or natural species fluctuations doubtful. Only after compiling these data did the severity and consistency of the trends become evident. Though there are many potential factors that may be contributing to these declines, including invasive species and non-target impacts to native pollinators from chemical spraying for non-native gypsy moth (Lymantria dispar L), insufficient data exist to conduct scientifically meaningful tests.

The impact of white-tailed deer herbivory was an obvious potential cause of this decline and an independent dataset existed to examine this factor. Studies on the impacts of herbivory to understory herbs are numerous and show herbivory represents a significant threat (Whigham 1990; Anderson 1994; Augustine and Frelich 1998; Ruhren and Handel 2000, 2003; Fletcher et al. 2001; Knight 2004). Regionally, deer herbivory is believed to be so intense it may cause the extinction of American ginseng (Panax quinquefolius L.), a now rare herbaceous plant (McGraw and Furedi 2005).

The deer harvest data for Frederick County, shows a significantly high inverse correlation (R2 = −0.93). The inverse correlation argues that the white-tailed deer population of Frederick County is the most likely cause of this decline. Given the level of urbanization and development in Frederick County, it is expected that the majority of the deer harvested from Frederick County came from within the study area. The public lands in the Catoctin Mountains account for 88 % of all the publicly held lands available for hunting in Frederick County (Maryland Department of Natural Resources 2013). Although deer population density data are not available within the study area, it is reasonable to assume that trends in the study area would mirror county-wide trends.

The increase in orchids in 2008 was unexpected and is likely a response to a decline in the deer population. The deer harvest dropped from nearly 9,000 individuals in 2001 to 7,000 in 2006. Liberalized bag limits are likely the result of the harvest increase in 2007 to 2008 (B. Eyler pers. comm). We expect as the white-tailed deer population continues to decline the response in orchid species will continue to be favorable. Seedlings of many terrestrial species are subterranean and seeds may still be present in the seed bank (Rasmussen and Whigham 1998). Future inventory should be conducted to determine the current orchid census at a subset of these sites given the recent implementation of deer control efforts at Catoctin Mountain Park. Deer exclosure studies should be conducted to further test the hypothesis that deer herbivory is causing this decline and to document overall herbaceous species response. It is likely that other plant groups have seen a very similar decline (i.e. Trillium, Lilium, Carex) but given no dataset exists it can only be inferred from a lack of diversity throughout the study area or a response to deer exclosures.

The lack of overall decline in Platanthera flava var. herbiola is caused by a count of 270 individuals in 2008, up from just 90 in 2007 (Fig. 3). The only species that showed an increase during this study period was P. ciliaris. The single site that explains this growth is owned and managed by the State of Maryland. Platanthera ciliaris is a pyrophytic species requiring open conditions such as open woods, roadsides, and seepage slopes (Sheviak 2002). To mimic the disturbance requirements of this rare species, the site has been mowed periodically beginning in 1989 (D. Rohrback pers. com.). Platanthera ciliaris has responded positively to the disturbance regime.

This study shows the value and utility of long-term datasets over a large area. This study also challenges the underlying idea that an area is protected just because it is publicly owned. Proper natural resource management is a prerequisite for species survival. In the case of this study, we were very fortunate to have a long-term dataset showing the declines that occurred. Most areas that support similarly-sized deer populations have no such datasets and damage to the flora will be witnessed but the actual declines of plant species will only be inferred due to a lack of botanical diversity. It is likely that adjacent states with similar deer populations, large parks with no easy access for hunters, and lands that do not allow hunting have seen or will see impacts to vegetation similar to these. Without long-term data sets as a point of reference, even catastrophic declines such as the ones published here, may go unnoticed.