Introduction

In temperate areas of North America and Europe, bog (peatland) vegetation is naturally rare and isolated, forming a low proportion of the natural landscape (Spitzer et al. 1999; Spitzer and Danks 2006; Whitehouse 2006; Whitehouse et al. 2008). In Wisconsin, peatlands occur primarily in central and northern areas (Curtis 1959). Prior to European settlement, peatlands occupied <1 % of the Wisconsin landscape (Hoffman 2002). Most of that vegetation is still extant, with only 9 % loss, and much of what is left, especially in northern Wisconsin, is relatively undegraded. Primary human impacts are roads, ditches, and logging along the margins; adjacent lands are more affected by timber harvesting, agriculture, off-road vehicles, and urbanization (pers. obs.; Wright et al. 1992; Epstein et al. 1997). Conversion to cranberry agriculture and peat harvesting have occurred more in central Wisconsin bogs (Curtis 1959). By contrast, in Europe bog vegetation is much destroyed and degraded by human activities, and its associated butterfly species are of high conservation concern (Spencer and Collins 2008; Turlure et al. 2009). The four bog-related vegetation types ranked highest in proportion of threatened butterfly species of their typical faunas (van Swaay et al. 2006).

Wisconsin bogs present the opportunity to study how localized butterfly populations function in relatively undisturbed sites in a naturally fragmented landscape. Prior studies (Nekola 1998; Nekola and Kraft 2002; Swengel and Swengel 2010, 2011; Johnson 2011) described patterns of bog butterfly population incidence and abundance. In this paper, we extend that work by examining change in bog butterfly abundance over time in our Wisconsin bog butterfly surveys during 2002–2013. We analyze time-series of population abundance in annually surveyed sites to describe the range of variation among years (patterns of fluctuation), relationships to climatic factors, and population trend (correlations with year) by bog type and subregion. These results should be useful for application to conservation of bog butterflies where they are vulnerable, and for understanding vulnerable butterflies in other fragmented vegetations.

Methods

Study region

We surveyed 75 bogs in the study region (Fig. 1). They were scattered across an area 367 km east–west by 169 km north–south (45.33–46.86°N, 88.21–92.56°W) spanning the entire breadth of northern Wisconsin (Swengel and Swengel 2010, 2011). At 20 of these sites, we also surveyed the lowland (wetland) roadside ditch through or adjacent to the bog. All sites could not be visited each year but most were visited more than once both within and among years. We consistently surveyed 29 bogs and 5 roadsides each year from the starting year (1999–2005) through 2013. We classified our study sites (Swengel and Swengel 2013) according to Nekola’s (1998) categories: (1) muskeg (black spruce Picea mariana-cottongrass Eriophorum spissum-wiregrass Carex oligosperma-Sphagnum savanna similar in elevation to surrounding uplands), (2) kettlehole (Sphagnum-leatherleaf Chamaedaphne calyculata mats, often floating on lakes or sunk in depressions much lower than the surrounding landscape), and (3) coastal peatland (tamarack Larix laricina-sedge, especially Carex lasiocarpa, mats with ridges of muskeg-like vegetation in estuaries along the Lake Superior coast).

Fig. 1
figure 1

Map (right) showing subregions where 29 long-term study bogs were located in northern Wisconsin (90 km between nearest northeast and north central muskegs; 117 km between nearest north central and northwest muskegs). Forty-six other bogs surveyed less often were north of the dashed line. Also shown are barrens surveyed in central Wisconsin as described in Swengel and Swengel (2010), with results presented in Online Resource 11

Surveys

We conducted most formal transect surveys in bogs during 2002–2013, with a few conducted in years before that (Swengel and Swengel 2010, 2011). Beginning in 2002, we surveyed in a rotation through the western, central, and eastern sections of the bog study region (Fig. 1), trying to cover one section per weekend, or more if a section was missed the previous weekend and/or if time allowed. We missed an occasional weekend per year due to weather or another commitment. Surveys occurred between 25 March and 12 September, usually from April to early May through mid-August.

We walked transect surveys along similar routes per visit to a site at a slow pace (about 2 km/h) on parallel routes 5–10 m apart (as described in Swengel and Swengel 2010, 2011), similar to Pollard (1977). We counted all adult butterflies observed ahead and to the sides, to the limit an individual could be identified, possibly with binoculars after detection, and tracked. Surveys occurred during a wide range of times of day and weather, occasionally in light drizzle so long as butterfly activity was apparent, but not in continuous rain.

Analyses

We analyzed 10 study species (Table 1), classified into two categories that correspond approximately to those (in parentheses) described by Spitzer and Danks (2006): (1) bog specialist (tyrphobiont)—restricted or nearly so to peatlands (all eight species known in Wisconsin), and (2) bog affiliate (tyrphophile)—breeding in bogs as well as other vegetations; limited to species of north temperate or boreal affinity (two species). These are Nekola’s (1998) ten peatland-obligate study species that complete their entire life cycle in bogs. As in Nekola and Kraft (2002), we identified the flight period per species per year by the first and last date we detected it across the study region.

Table 1 Total individuals in time series from 29 bogs and 5 bog roadsides used in long-term analyses

Our population abundance index is the peak survey count per site per year, standardized to survey time to create an observation rate (relative abundance) per hour per site. We surveyed sites multiple times per year both to verify the timing of the main flight period and to survey different target species. However, a “collated” index (e.g. sum of weekly counts throughout a species’ flight period in a year) was not possible because the number of visits per flight period varied both among sites and among years. We avoided pseudoreplication (counting the same individual on more than one survey) because our population indices all contain only a single sampling of individuals per generation per species at each site. One survey during the main flight period has been adequate for producing representative indices for comparisons of relative abundance within and among sites (Thomas 1983; Swengel and Swengel 2005; Schlicht et al. 2009). We assembled time series for sites surveyed consistently for a species annually from 2002 to 2005 through 2013.

We obtained climate data for winter (December–February), spring (March–May), summer (June to August), the growing season (April–September), and year from the Wisconsin State Climatology Office (2013). Our northeast, north central, and northwest (including muskegs, kettleholes, and coastal peatlands) subregions correspond to the same subregions in this climate dataset. We analyzed seasonal temperature, precipitation, and season-long snowfall total (from the prior year’s fall to the current year’s spring), and the Palmer Drought Severity Index, which becomes more positive in floods and more negative in droughts. This index uses temperature and precipitation data to evaluate potential evapotranspiration to measure environmental dryness [cumulative departure in surface water balance (Dai et al. 2014)]. The climate data were available up to 2011, except 2010 for snowfall. We matched climate data to butterfly abundance by subregion and year, and correlated butterfly relative abundance at the scale of the site time series to climate for up to one year after the timing of the climate variable.

We identified our first observed date (FOD) each year that we found each species at any bog site anywhere in the bog study region (Fig. 1). It was not possible to subdivide these dates by subregion because the interval between visits within a subregion was usually more than a week. We calculated the mean FOD for each species during 2002–2013, and correlated those means to the standard deviation (SD) for each mean FOD and to the difference in days between the earliest and latest FOD for each species.

Analysis was done with ABstat 7.20 software (1994 Anderson-Bell Corp., Parker, Colorado). Statistical significance was set at two-tailed P < 0.05. Since significant results occurred much more frequently than expected due to Type I statistical errors, we did not lower the critical P value further, as far more Type II errors (biologically meaningful patterns lacking statistical significance) would be created than Type I errors eliminated. We chose to use non-parametric tests because they do not require data to be distributed normally. All correlations were done with the Spearman rank correlation. We used the Mann–Whitney U test to test for significant differences in relative abundance of Jutta arctic between even and odd years. As reviewed in Swengel and Swengel (2013), Jutta arctic varies among parts of its range as to whether adults are found primarily or only in even or odd years, or readily in both even and odd years.

We performed analyses at the scale of the individual species at the individual site per year. We did not average or otherwise collate the survey results for a species in a year across sites prior to analysis for two reasons. First, the start year for the time series varied by site. Second, we wanted to retain local variation in the statistical tests. Because butterfly survey counts can lack a normal distribution, the mean may be skewed toward the high abundance of a few sites when most other sites had few or no individuals found. By analyzing at the scale of the site, rather than the region, each site was represented equally in statistical tests.

Results

First observed date (FOD)

FOD was more variable for spring than summer butterflies (Table 2). The mean FOD for each species during 2002–2013 correlated highly negatively with the SD of that date (r = −0.770, N = 10 species, P < 0.01) and span of that date (difference in days between earliest and latest FOD for a species over 12 years; r = −0.766, N = 10 species, P < 0.01). All correlations of FOD with spring temperature were negative (Table 2), significantly so for four species. One spring species, red-disked alpine, had a significant negative correlation of abundance with FOD (Table 3). Two spring and one summer species had significant positive correlations between FOD and the following year’s abundance (brown elfin, Frigga fritillary, common ringlet) (Table 3).

Table 2 Descriptive statistics on first observed date (FOD) as Julian date in northern Wisconsin during 2002–2013, with Spearman rank correlation coefficients (r) of FOD with spring temperature (N = 10 years, 2002–2011) averaged across the three northern subregions
Table 3 Spearman rank correlation coefficients (r) of first observed date with this year’s and next year’s abundance in northern Wisconsin

Annual variation and population trends

Brown elfin exhibited remarkably regular, possibly cyclic variation with low abundance synchronized over the entire study region (Fig. 2, Online Resource 1). Abundance patterns in each subregion for the other study species are provided in Online Resources 2–10. Jutta arctic showed dramatic consistent variation in abundance between even and odd years in Northeast (Online Resource 6) with odd years significantly higher than even (Fig. 3). This even–odd variation was not obvious in the other subregions (Online Resource 6) but was nonetheless significantly different in North Central in the opposite direction (Fig. 3). The other study species varied greatly in abundance among years but not in as distinctive patterns (Online Resources 1–5, 7–10).

Fig. 2
figure 2

Brown elfin abundance from all 27 long-term sites for which the species was analyzed

Fig. 3
figure 3

Mean Jutta arctic abundance at study sites by even and odd years, by subregion; significantly different in northeast and north central (two-tailed P < 0.05 in Mann–Whitney U tests). Number of sites per subregion is in parentheses

Significant trends of abundance over time (Table 4) were positive for spring butterflies (brown elfin, Freija fritillary, red-disked alpine) and negative for summer butterflies (common ringlet, bog copper, purplish fritillary) with the exception of the Dorcas copper, which was positive. However, common ringlet has been broadly increasing in abundance farther south in central Wisconsin barrens (Online Resource 11). Within species, the trend in roadsides was usually less favorable than in bogs in the same region (common ringlet, Dorcas copper, bog copper in northwest but not north central, purplish fritillary) (Table 4).

Table 4 Spearman rank correlation coefficients (r) of trend (butterfly abundance versus year) during 2002–2013 in northern Wisconsin, by southern extent of range

Climate

All ten study species had significant correlations with climate variables (Table 5), more so with precipitation than temperature. All correlations to winter climate variables indicated increasing abundance with coolness (4 species) and/or dryness (6 species). In correlations to current-year spring climate variables, four species favored dryness and/or warmth. However, brown elfin (the earliest spring-flying species) favored coolness, and the two latest summer-flying species favored wetness. Correlations of summer-flying species to current-year summer variables all related to moisture. The species that had higher current-year abundance with earlier first observed dates (Table 3) was a spring flier that favored a dry winter and spring (Table 5).

Table 5 Results of significant (P < 0.05) Spearman rank correlation of abundance each year of each study species in the study sites during 2002–2013 with climate factors, by species in order of flight period, grouped by tendency to prefer dry or wet climatic conditions. Results represent the conditions related to higher abundance

For some species, lag effects of climate variables on next year’s butterfly abundance were similar to those matching current-year climate and abundance (Table 5). Examples include Jutta arctic (warmth and dryness), and bog copper and purplish fritillary (wetness). Other species showed seasonal variation in temperature or precipitation preferences. For example, brown elfin favored cool and dry in the current winter and spring but warm and dry in the previous summer. Freija fritillary favored wetness in the previous summer but otherwise dryness in both the current and previous summer. Frigga fritillary favored warmth in the current spring but coolness in the previous year. Bog fritillary favored summer wetness for the current year’s adult abundance but dryness for next year’s abundance.

Three species (bog fritillary, bog copper, purplish fritillary) consistently showed an affinity for moistness (Table 5), although bog fritillary favored dryness in the previous year. Within a site, bog fritillary abundance covaried strongly with bog copper abundance later in the same year, and bog copper abundance covaried strongly with the following year’s bog fritillary abundance (Table 6). Bog fritillary and bog copper were the two study species more abundant in kettleholes and coastal peatlands than muskegs (Table 1). All bog copper’s consistent preferences for moistness occurred in muskegs (Table 5). Bog fritillary had no significant relationships to climate variables in muskegs. Bog copper had no significant climate correlations in kettleholes and coastal peatlands nor did bog fritillary in coastal peatlands, even though they dramatically varied in abundance there. Six of the other seven species consistently favored dryness and drought (Table 5), although varying in whether they also favored coolness or warmth. Results for the seventh species (Dorcas copper) were equivocal, since it favored cool and dry in the current-year winter, but warm and wet in the prior year’s growing season.

Table 6 Spearman rank correlation coefficients (r) of bog fritillary abundance to this year’s bog copper abundance and last year’s bog copper abundance within the same site, for all five subregions combined, during 2004–2013 at sites surveyed each of those years

Discussion

First observed date (FOD)

Although FOD can have some biases due to variation in sites and sampling intensity among years (van Strien et al. 2008), our study is less prone to these issues because sampling effort and sites were relatively similar among years, and observers were always the same. Furthermore, the consistent negative correlation of FOD and spring temperature (Table 2) was expected from the literature relating variation in butterfly flight periods to variation among years in seasonal development (examples in Sparks and Carey 1995; Sparks and Yates 1997; Forister and Shapiro 2003; Polgar et al. 2013), and provides a validation of the FODs. But the correlations of FOD to some species’ abundance (Table 3) also support that FOD relates both to phenology and butterfly abundance (van Strien et al. 2008).

Spring butterflies consistently varied more in FOD than summer butterflies (Table 2), as also noted in other butterfly faunas (Forister and Shapiro 2003; Polgar et al. 2013). The effect on adult butterflies could result from butterflies being easier to detect, but not more abundant, in warmer weather, which might explain why this effect was more evident for spring than summer butterflies. Adult butterflies might be more apparent in warm springs because of the greater amount of time spent flying (Cormont et al. 2011). The lag effect of FOD on next year’s adult abundance (Table 3) might result primarily from effects on immatures. Climatic influences on abundance may conflict between the current year and the following year, i.e., on different life stages. For example, in Europe the negative effect of winter warmth on larval survival of bog fritillary far outweighed the positive effects of spring and summer warmth on egg, pre-diapause larval, and pupal survival, and number of eggs laid (Schtickzelle and Baguette 2004; Radchuk et al. 2013).

Annual variation

Dramatic variation in abundance usually appeared to correspond to the frequently reported fluctuations in butterfly abundance related to climatic variation (examples in Dennis 1993; Pollard and Yates 1993; Roy et al. 2001; Warren et al. 2001). This acts separately on individual populations yet can produce remarkably synchronized variation in abundance among populations over much larger spatial scales than individuals in these populations typically disperse (Sutcliffe et al. 1996). However, even though brown elfin exhibited significant relationships to climate (Table 5), its time series also showed a consistent seemingly cyclic variation (Fig. 2, Online Resource 1), suggestive of density-dependent population regulation such as reported for holly blue Celastrina argiolus in Britain (Thomas et al. 2011: Fig. 4). The brown elfin’s larval host plants are heaths (Ericaceae) (Opler and Krizek 1984; Nielsen 1999). These plants appear abundant in muskegs, less so in kettleholes and coastal peatlands, although larval breeding requirements may be more limiting than host plant abundance as reported for other butterflies (Turlure et al. 2010; Thomas et al. 2011).

Jutta arctic exhibited dramatically consistent higher abundance in odd than even years in northeast (Fig. 3, Online Resource 6). In the adjacent north central subregion, Jutta arctic showed less dramatic but significantly higher abundance in even years (Fig. 3) but this did not appear as regular and predictable as the pattern in the Northeast. This shift in even/odd year abundance patterns among adjacent subregions is consistent with reports from elsewhere in the species’ range, as reviewed in Swengel and Swengel (2013). This species is reported to have a two-year life cycle (Scott 1986).

Population trends

Within species, the trend in roadsides was usually less favorable than in bogs in the same region (common ringlet, Dorcas copper, bog copper in northwest but not north central, purplish fritillary) (Table 4). This suggests the negative influence of human disturbance in the landscape, since bog sites (away from roadsides) had fewer human-caused degradations, such as the trenching, scarification, and re-grading we observed in roadsides.

The most northerly species (purplish fritillary) had a strongly negative population trend (Table 4), consistent with numerous others’ reports of northern species retreating uphill or northward (e.g., Parmesan 1996; Forister et al. 2010; Breed et al. 2013). Nekola and Kraft (2002) expressed concern about the long-term future in Wisconsin for the purplish fritillary. It is in a species complex, or conspecific, with arctic fritillary Boloria chariclea (Opler and Krizek 1984; Scott 1986; Johnson 2011), which in Europe is a butterfly likely to have the most severe range contractions with climatic warming (Settele et al. 2008; Heikkenen et al. 2010).

But southerliness of range otherwise did not correspond to trends (Table 4). Bog copper has the southernmost range of the tyrphobionts (Glassberg 1999) but had a negative trend. Freija fritillary and red-disked alpine are relatively northern species but had positive trends. While the significant trend for common ringlet in the bog study region was negative, this species was increasing in abundance 120–300 km farther south in central Wisconsin barrens (Online Resource 11). Over the last 55 years, this species has appeared to be colonizing the state from the northwest (Ebner 1970).

Thus, some of the trends in bogs (Table 4) may relate not to temperature (as assumed in a north–south analysis) but to precipitation, or changes in habitat characteristics (e.g. disturbances or drainage at bog edges). Furthermore, these species may tolerate climatic conditions outside of their observed geographic ranges. Warren et al. (2001) noted that the ranges of many British butterfly species appear limited by factors other than climate, since they are not occupying all areas that are climatically suitable. Unexpected outcomes such as downhill expansion of some high arctic butterflies in a Swedish arctic alpine national park may result from the complexity of seasonal variation in both temperature and precipitation (Franzén and Öckinger 2012). The large variation in annual abundance of bog butterflies in this study, and uncertainty about longer-term effects of climate, highlight the value of longer monitoring periods than in our study (Thomas et al. 2002).

Climate

Since this analysis tested for one kind of pattern (correlation) (Table 5), it would not detect many other kinds of climatic influences such as threshold or non-linear effects or multivariate interactions. Caution should be used in any application of these results to predict future bog butterfly responses to climatic variation.

This analysis did not produce consistent relationships such as southern species tolerating warmth more than northern species, or tyrphobionts preferring wetness and coolness (Table 5). The one season where all significant relationships indicated the same preference (cool, dry) was winter, which suggests that these conditions improved overwintering immature survival. Otherwise, in spring and summer, the species varied as to whether they favored cool or warm, moist or dry.

However, this analysis of butterfly abundance more often produced significant results related to moisture than to temperature (Table 5). Most of the study species clearly categorized into two guilds (regardless of range relative to Wisconsin), with a majority classifying as dry steppe species and the remaining minority (bog fritillary, bog copper, purplish fritillary) as wetland species. Only one species (Dorcas copper) was unclear how to classify in this regard because of conflicting climatic results between current-year and lag-year effects on abundance.

The relationships of the study species’ abundance to moisture (Table 5) consistently corresponded to their differences in abundances by bog type (Table 1, Online Resource 1; Nekola 1998; Swengel and Swengel 2011). Kettleholes (some of which float on a lake) and coastal peatlands are more consistently moist, while muskegs (or parts of muskegs) can be drier (Nekola 1998). Furthermore, differences in the species’ climatic preferences by bog type were also generally consistent with the differences in moistness among the bog types. Two of the three species in the wetland guild (bog fritillary, bog copper) were far more abundant in the wetter bog types (kettlehole and coastal peatlands). Both had more favorable trends in coastal peatlands (with a more maritime climate) than inland kettleholes (Table 4). All of bog copper’s significant preferences for moister weather occurred in the drier bog type (muskeg) (Table 5). Bog copper showed no significant climatic correlations in kettleholes and coastal peatlands, nor did bog fritillary in coastal peatlands, even though they varied dramatically in abundance there. However, bog fritillary did favor dryness in kettleholes as a lag-year effect. Climatic relationships might become apparent for these species in coastal peatlands with climate data restricted to weather stations along the Lake Superior coast, which is far cooler in the growing season than nearby inland locations (Curtis 1959; Zimmerman 1991).

The third species (purplish fritillary) in this wetland guild only occurred in muskegs in northwest. Nekola (1998) identified this species as one of five members of a high arctic dry sedge-heath guild. Purplish fritillary’s preference for moistness may serve as a temperature buffer, for this is the study species with the northernmost southern edge to its range (Opler and Krizek 1984; Glassberg 1999). The closely related B. titania in Switzerland likewise preferred a locally higher proportion of wetland, as well as more trees, which may also serve as a climatic buffer (Cozzi et al. 2008).

All species in the steppe guild occurred more abundantly or only in muskegs (Table 1). Three of the five taxa in Nekola’s (1998) high arctic guild (freija and frigga fritillaries, red-disked alpine) are in the steppe guild identified here. Dorcas copper is also in Nekola’s (1998) high arctic guild but is not clearly classified in our analysis of abundance related to climate data. Both tyrphophile study species are in the steppe guild and occurred in all three bog types. They also occur widely in non-wetland habitats: brown elfin in barrens with its ericaceous hosts and ringlets in grasslands, including barrens (Online Resource 11) (Opler and Krizek 1984; Glassberg 1999).

Five species showed conflicts among seasons in temperature or precipitation effects on abundance (Table 5), possibly related to which life stage was affected by the climatic variable (WallisDeVries and van Swaay 2006; Settele et al. 2008; Cormont et al. 2011; Radchuk et al. 2013). For example, bog fritillary favored summer wetness for the current year’s adult abundance but dryness for next year’s abundance, possibly due to how that moisture affects larvae. Schtickzelle and Baguette (2004) found complex variation in bog fritillary population response to climate depending on the season measured (including a preference for cool winters as found here), with an overall negative effect of warm temperatures.

Conservation implications

Tyrphobionts do not form a single guild and do not respond to vegetative factors similarly, as evidenced by significant differences in abundance by bog type (Nekola 1998; Swengel and Swengel 2011). Based on their abundance relationships to climatic variables, most tyrphobionts in this study allied as dry steppe species, while two tyrphobionts with relatively more southerly ranges allied as wetland species. Half of the steppe guild species had positive population trends, while two out of three wetland guild species and no steppe species significantly declined. The species did not respond consistently to temperature in relation to the southerliness of their range. Many tyrphobionts favored warmer growing season conditions and earlier phenologies. Thus, climate did not appear directly limiting on the ranges of these animal populations. They have experienced warmer climate since the last glaciation (e.g., the hypsithermal) (Pielou 1991; Ashworth 2001; Mauquoy and Yeloff 2008) as well as extreme variation in climate, as evidenced in the variation among years in first observed date. However, there may be a lag in butterfly population responses to climate change (Menéndez et al. 2006).

The overriding factors determining the future of this bog butterfly fauna appear to be habitat degradation and potential climate change beyond the range of variation typically experienced in this region between glaciations. The benefit of protecting bogs from degradation is evident in the poorer butterfly abundance trends in roadsides in this study. Elsewhere in the world, habitat destruction and alteration of both the bogs and the surrounding matrix threaten many bog butterfly populations (van Swaay et al. 2006; Spencer and Collins 2008; Turlure et al. 2009; Weking et al. 2013). Degraded or late-succession bogs have fewer hummocks, which are essential thermal refuges (Turlure et al. 2010). Thus, stopping and reversing habitat deterioration are climate change mitigations (Essl et al. 2012), as demonstrated more generally in Thomas et al. (2012). Maintaining a diversity of vegetative structures is necessary to allow larvae to thermoregulate successfully (Turlure et al. 2011). However, some climate change models indicate that the tipping point could be reached when peatland vegetation would no longer be able to persist in more southerly areas such as our study region (Mauquoy and Yeloff 2008; Essl et al. 2012). In that scenario, it would be unlikely that the conditions these species require would continue to exist.