Introduction
The Poweshiek skipperling, Oarisma poweshiek (Parker, 1870) (Lepidoptera: Hesperiidae), is an obligate tall grass prairie butterfly once commonly found in the north–central United States of America and southern Manitoba, Canada (McCabe and Post Reference McCabe and Post1977; Catling and Lafontaine Reference Catling and Lafontaine1986; Klassen et al. Reference Klassen, Westwood, Preston and McKillop1989; Committee on the Status of Endangered Wildlife in Canada 2014; Belitz et al. Reference Belitz, Hendrick, Monfils, Cuthrell, Marshall and Kawahara2018). The number of colonies across their range declined (Committee on the Status of Endangered Wildlife in Canada 2014; Smith et al. Reference Smith, Cuthrell, Runquist, Delphey, Dana and Nordmeyer2016), before the number of adults in remnant colonies also decreased since approximately the 1990s to potentially a few hundred individuals across its range (Swengel and Swengel Reference Swengel and Swengel1999; Committee on the Status of Endangered Wildlife in Canada 2014; Smith et al. Reference Smith, Cuthrell, Runquist, Delphey, Dana and Nordmeyer2016; Grantham et al. Reference Grantham, Hamel, Reeves and Westwood2020). The destruction of 99% of tall grass prairie (Samson and Knopf Reference Samson and Knopf1994) likely contributed to the initial decline, but causes of the decrease in remnant colonies over the last few decades are unknown (Committee on the Status of Endangered Wildlife in Canada 2014; Smith et al. Reference Smith, Cuthrell, Runquist, Delphey, Dana and Nordmeyer2016). Oarisma poweshiek is now listed as endangered in both Canada and the United States of America (Committee on the Status of Endangered Wildlife in Canada 2014; United States Fish and Wildlife Service 2015; Canada Gazette 2019) and critically endangered across the planet (Royer Reference Royer2020). A better understanding of O. poweshiek biology may facilitate successful recovery efforts.
Remaining O. poweshiek populations are found at the margins of the historical range in Manitoba, Canada (Committee on the Status of Endangered Wildlife in Canada 2014) and Michigan, United States of America (Belitz et al. Reference Belitz, Hendrick, Monfils, Cuthrell, Marshall and Kawahara2018). In Manitoba, the Poweshiek skipperling occurs in wet to mesic tall grass prairies composed of convoluted prairie, forest, and wetland plant communities (Catling and Lafontaine Reference Catling and Lafontaine1986; Committee on the Status of Endangered Wildlife in Canada 2014). Typical species supported in these plant communities include Sporobolus heterolepis A. Gray (Poaceae), Liatris ligulistylis (A. Nelson) K. Schumann (Asteraceae), Potentilla fruticosa Linnaeus (Rosaceae) in the prairie areas, Populus tremuloides Michaux (Salicaceae) in the forest areas, and Juncaceae Jussieu spp. in the wetland areas, both permanent and ephemeral (Catling and Lafontaine Reference Catling and Lafontaine1986; Committee on the Status of Endangered Wildlife in Canada 2014). Oarisma poweshiek occurs most often in tall grass prairie complexes approximately 10 km2 in area (Westwood et al. Reference Westwood, Westwood, Hooshmandi, Pearson, LaFrance and Murray2020) and at relatively higher abundances in sites that are comparatively less wet, as indicated by the presence of plants that prefer various soil moistures (Henault Reference Henault2017).
The Poweshiek skipperling in Manitoba flies from the last week of June through early August (Semmler Reference Semmler2010; Committee on the Status of Endangered Wildlife in Canada 2014). Females lay eggs on plants, and hatched larvae feed until they enter diapause for winter. Larvae are thought to overwinter near the base of plants (McAlpine Reference McAlpine1972; Borkin Reference Borkin1995; Committee on the Status of Endangered Wildlife in Canada 2014) and resume feeding in the spring, before pupating in mid- to late June (Layberry et al. Reference Layberry, Hall and Lafontaine1998; Committee on the Status of Endangered Wildlife in Canada 2014).
The substrates these butterflies use for egg laying in Canada are unclear, comprising several plant families. Dupont-Morozoff (Reference Dupont-Morozoff2013) observed ovipositions on Andropogon gerardi Vitman (Poaceae), Melilotus (Linnaeus) Miller spp. (Fabaceae), Solidago Linnaeus spp. (Asteraceae), and Quercus macrocarpa Michaux (Fagaceae). One of us (J.H., unpublished data) observed a female and male mating in the canopy of graminoids at a visually mesic location, after which the female immediately flew approximately 3 m to lay an egg on a S. heterolepis leaf (Supplementary material, Fig. S1; Henault Reference Henault2021). Several oviposition substrates have been reported in the United States of America (Table 1). However, reports of direct observations of larval feeding in natural habitats are scarce. Borkin (Reference Borkin1995) observed individual larval feeding bouts in natural prairie habitat (Wisconsin, United States of America), where larvae consumed S. heterolepis and Schizachyrium scoparium (Michaux) Nash (Poaceae). Larvae from ova of captured females were placed on small, potted clumps of prairie graminoids that had been dug from a road backslope (Minnesota, United States of America; Dana, personal communication); they consumed the species observed by Borkin (Reference Borkin1995) and also A. gerardi (Dana, personal communication).
Taxa used by Oarisma poweshiek for oviposition (all in field conditions) and larval feeding in Canada and the United States of America.
1 Y = direct observation, ? = feeding marks near eggs.
2 cr = captive-reared larva(e). Manitoba, Minnesota, North Dakota and Wisconsin: tall grass prairies; Michigan: tall grass prairie fens.
Butterflies lay eggs in microhabitats that contain host food plants and the microclimatic conditions (e.g., temperature or humidity) required for immature stages to develop (Ashton et al. Reference Ashton, Gutiérrez and Wilson2009; Krämer et al. Reference Krämer, Kämpf, Enderle, Poniatowski and Fartmann2012; Ewing et al. Reference Ewing, Menéndez, Schofield and Bradbury2020). If eggs are laid on the ground or an otherwise unpalatable substrate, larvae have been observed searching for host food plants once hatched (Kopper et al. Reference Kopper, Charlton and Margolies2000; Hellmann Reference Hellmann2002). In general, female butterflies use cues, including host plants and plant chemical composition, to identify suitable microhabitats (Wiklund Reference Wiklund1984; Lund et al. Reference Lund, Brainard and Szendrei2019).
The developmental rate of O. poweshiek in confirmed field microhabitats in Manitoba has not been reported. As a result, annual surveys of adults in Manitoba and the United States of America (Westwood et al. Reference Westwood, Dupont and Hamel2012; Grantham et al. Reference Grantham, Hamel, Reeves and Westwood2020) must be synchronised by using estimated larval development rates to predict adult eclosion (Dearborn and Westwood Reference Dearborn and Westwood2014; Dearborn et al. Reference Dearborn, Henault and Westwood2014–2022). A better understanding of immature development, followed by a comparison of field and weather station data, may increase the accuracy of predictions and subsequent timing of surveys.
Observations tracking the foraging behaviour and development of the same O. poweshiek individuals in natural habitats in North America have not been reported. The vegetation, physical structure, soil characteristics, and microclimate attributes of suitable microhabitats of immature skipperlings also have not been described. Identifying the larval host plants that Poweshiek skipperlings use in natural habitats is critical for prioritising future research directions and facilitating the recovery of this species. The objectives of this research were to: (1) identify larval host plants and foraging behaviour in natural tall grass prairie sites in Manitoba, (2) examine the vegetative, physical, and microclimatic characteristics of microhabitats where eggs are laid, and (3) document O. poweshiek developmental rates in natural habitats. We hypothesised that areas where females laid eggs may possess specific host plant, structural, or microclimatic attributes, which might be facilitated by soil characteristics, and that larvae may travel within microhabitats to feed.
Methods
Field research was carried out within the Manitoba Tall Grass Prairie Preserve in southeastern Manitoba, Canada (the Nature Conservancy of Canada interpretive centre, ∼49.153° N, 96.729° W). We selected a site where O. poweshiek was consistently present during collaborative surveys, using meandering transect walks (Royer et al. Reference Royer, Austin and Newton1998), from 2007 through 2017 (Westwood et al. Reference Westwood, Dupont and Hamel2012; Grantham et al. Reference Grantham, Hamel, Reeves and Westwood2020), where at least one of us had observed egg laying and where individuals had not been reintroduced (Assiniboine Park Zoo 2018). We studied O. poweshiek at three prairie patches of approximately 0.20-km2 area within a 2.6-km perimeter that were isolated from each other by wetlands and forests. This study area is stewarded with prescribed burns and mechanical and herbicide removal of encroaching vegetation by the Nature Conservancy of Canada (Grantham et al. Reference Grantham, Pelc, Neufeld, Greaves, Anderson and Hamel2021); the last site-scale disturbance was a wildfire in autumn 2011.
In 2018, from 25 June through 4 July and on 6 July, from approximately 10:00–17:00 local time, we searched for and followed adult Poweshiek skipperlings to observe female egg laying. We observed females from 27 June through 3 July 2018 and males and females from 26 June through 4 July 2018. The bases of shoots where females were observed laying eggs were marked with a metal stake and recorded with a geographic positioning system unit (Garmin Oregon 700; Garmin, Olathe, Kansas, United States of America).
Egg enclosures, July 2018 to July 2019
We centred plastic refuse pails that we previously had cut in half at each location where eggs were laid (n = 6), breaking the ground surface to a depth of 3 cm (base 49 cm wide, or 0.19 m2) and extending 28 cm above the ground surface (Fig. 1). Enclosures were removed after the first adults of the next generation were observed in the site (July 2019). Our primary goal was to mimic environmental conditions that are typically experienced by local larvae (i.e., with no mesh and with potentially cooler white walls) to facilitate unaltered behaviour while limiting larval escape. We clipped vegetation approximately weekly, 3–5 cm along the pails’ internal walls, and any plants that overlay the pails’ rims. On sampling dates, plants were clipped after variables were recorded to ensure a consistent sampling arena. We hypothesised that this surface area would accommodate all movements during the development of these small larvae, which are approximately 2–24 mm long (McAlpine Reference McAlpine1972). If we did not see a larva in an enclosure during two consecutive weeks, we ceased comprehensive observations. To avoid the chance that an egg or larva was present but not seen, we periodically checked the enclosures to confirm the absence of immatures until the end of the study.
Larval foraging behaviour
The distance from eggs to the tip of the leaf blade and the height above the ground were recorded. One to three times per week, J.H. located a larva to watch for at least one hour. Based on these direct observations, the host species that larvae consumed and the locations on plants that larvae occupied relative to the tip of the leaf (from apex of head capsule along the leaf) and to the ground (vertical distance), as well as the horizontal distance between sequential larval activity locations, were recorded. Larvae were oriented vertically, except on one occasion, when they were oriented horizontally; therefore, we measured the direction that the dorsal side of the larva’s body faced to determine a larva’s degree of exposure to the sun. By chance, J.H. observed larvae immediately after each moult, allowing us to accurately identify the instars. Photographing dislodged Immature 1 from the plant while hatching; after replacement on its host, Immature 1 rested near the duff layer. Thereafter, to reduce disturbance, including during windy conditions or if larvae flinched, we measured only the larval position relative to the leaf tip or ground, in addition to the host species, during some observation events. During the only major measurement disturbance, Immature 2 left its host before returning the next day to the original location. In this way, larvae appear not to have been substantially influenced by our observations. Larvae were measured (total length) with a ruler while on plants from 1 to 3 days after moulting and were not handled in order to minimise risks of stress, displacement, and damage.
The thermal development thresholds of O. poweshiek are unknown; Dearborn and Westwood (Reference Dearborn and Westwood2014) used 6 °C as the lower threshold, based on studies of other Lepidoptera that overwinter as larvae in this region. The diapause-inducing conditions likely comprise a combination of temperature and photoperiod, as has been observed for other butterflies (Kim et al. Reference Kim, Kim, Hong, Lee, Park, Je and Lee2014). Given this uncertainty, we assumed that larvae would be inactive approximately when the minimum daily temperature was below 0 °C and during a shorter photoperiod (late fall and early spring), with developmental conditions likely deteriorating from September to December. Field observations began 26 June 2018, before the predicted adult flight period, paused 14 September, with the first date with minimum daily temperature below 0 °C being 5 September, and resumed 8 May 2019, after the snow had melted but before minimum daily temperatures returned above 0 °C on 22 May.
Microclimate in microhabitats
We recorded temperature, relative humidity, and dew point at one-hour intervals (9 July 2018 to 18 July 2019, inclusive) using data loggers (HOBO® Pro V2 U23-001: Onset Computer Corporation, Bourne, Massachusetts, United States of America) in egg enclosures (n = 4) and their accompanying comparison plots (see below; n = 4). Data loggers were placed in open-ended plastic tubes at the boundary of the dead vegetation layer (Fig. 1) in shaded areas to equalise sun exposure (data loggers in egg enclosures: shade of walls; in comparison plots: shade of tall vegetation). Vegetation was flattened within approximately 5 cm of the tube openings to replicate air flow and the humidity of the clipped areas in the egg enclosures. We also acquired temperature data from the weather station used to synchronise surveys (Emerson, Manitoba; ∼ 35 km to the southwest, or 254° from the study site) from an online repository (Environment Canada 2020).
Relative positions of egg enclosures, target plots, and comparison plots along an Oarisma poweshiek adult flight track. A microclimate data logger is shown in the shade of the egg enclosure.
Attributes of egg enclosures and target plots, 2018
In the egg enclosures, we counted the number of shoots of all plant species and estimated the percent cover of graminoids, including Poaceae spp., Cyperaceae spp., Juncaceae spp., and Juncaginaceae spp.; forbs; shrubs, including trees; duff, including both dead pieces of vegetation and fine decomposing material; and bare ground. We measured the height of graminoids (of approximately 95% of the plants) and, in three random locations, the depth of duff. Soil moisture and electrical conductivity (i.e., concentration of molecular ions; Natural Resources Conservation Service 2014) were measured 10 cm below the surface to focus on roots instead of soil surface at two random locations using soil probes that were connected to an electronic display (Fieldscout TDR 150; Spectrum Technologies, Inc., Aurora, Illinois, United States of America). Measurements were conducted on 18 and 30 July 2018 after seven and eight days without substantial precipitation, respectively.
We also established “target plots” outside of egg enclosures (Fig. 1), which we sampled on 2 and 3 August 2018, using the same methods except that we took no soil measurements. By comparing prairie attributes at different radii from deposited eggs, we hoped to determine the size of prairie from which females receive oviposition stimuli. We identified plant species using dichotomous keys and visually using Looman and Best (Reference Looman and Best1987) and Leighton and Harms (Reference Leighton and Harms2014), before updating nomenclature to Tropicos.org (Missouri Botanical Garden 2021).
Attributes of egg enclosures and comparison plots, 2019
While following adults in 2018, we staked a metal rod at locations where females flew over immediately before the eventual oviposition location. Apparently, females were not stimulated above the oviposition threshold, as Singer (Reference Singer1971) had described for other Lepidopteran taxa. We established one circular “comparison plot” in 2019 at these locations to accompany each oviposition location while prioritising locations with similar sun exposures (Fig. 1).
We sampled egg enclosures and comparison plots using the same methods as in 2018, with the following exceptions. Only graminoids were enumerated, and these were split into two groups: (1) confirmed and possible larval and egg hosts, as observed in 2018, and our hypothesised additional hosts, Sorghastrum nutans (Linnaeus) Nash and Muhlenbergia Schreber spp., and (2) all other graminoids. On 9 May, 30 May, 11 June, and 5 July, we counted only the number of shoots of species in the confirmed and potential host group, and we estimated the proportion of graminoid shoots belonging to host or nonhost graminoid groups. The heights of all combined graminoids and of the two groups were independently measured. On 18 July 2019, the duff depth was subdivided into thatch (upper, dead pieces) and litter (lower, fine decomposing material). Soil variables were measured, as in 2018, on 9 and 10 May 2019. The light intensity was measured once per plot (the upper plane of the duff and below; in immediate succession; Triple Range Light Meter, Model 217, General Electric Company, Boston, Massachusetts, United States of America) in the absence of clouds on 23 July 2019, within a 3.5-hour period.
Lepidopteran nomenclature follows Pohl et al. (Reference Pohl, Landry, Schmidt, Lafontaine, Troubridge and Macaulay2018) primarily and GBIF.org (Global Biodiversity Information Facility 2021) secondarily.
Analysis
We report the relatively shorter distances to the tip, as measured with a ruler along the leaves, with greater precision than we report distances to the ground. When calculating the duration of subsequent feeding and resting activities, we used only events where larvae alternated between feeding and resting at least once. The mean angle, directionality (r), and significance of direction (Rayleigh test, which must include at least five samples) of larval movements and orientations while on plants were calculated as in Batschelet (Reference Batschelet1981).
Microclimate data recorded by data loggers were standardised by adding or subtracting recorded values (°C; %) relative to the bias of each data logger, which was determined by operating the loggers for one week in a building under equal conditions, before conducting analyses. Temperature data from the weather station (30 June–8 July 2018, inclusive) were used to supplement egg-period microclimate data from the data loggers. Degree-day accumulations were calculated following Dearborn and Westwood (Reference Dearborn and Westwood2014) to enable comparisons: from the first date of a period (biofix date; calibration), the mean daily temperature was calculated before subtracting the hypothesised lower development threshold (6 °C), using the standard model (standard), or accounting for degree-day accumulations between lower (6 °C) and upper thresholds (32 °C), using the double sine model (double sine).
We did not observe any larvae completing their entire life cycle. Because of this, we estimated date ranges of hypothetical developmental periods of O. poweshiek during any given year as follows: 30 June–27 September (inclusive) – first egg laid and active larvae; 28 September–21 May – diapause; 22 May–10 July – break of diapause through first adult eclosion; and 30 June 2018–10 July 2019 – first egg laid through first adult eclosion. We think that the dates when the minimum daily temperatures are consistently below 0 °C may provide a biologically relevant calibration date for future research targeting similar taxa. The postdiapause period encompasses all dates during which degree-days could be accumulated following winter. We also assessed the period used by Dearborn and Westwood (Reference Dearborn and Westwood2014) – 1 March through the first adult observation – to synchronise surveys with adult emergence. Because we did not operate field data loggers for the entire egg development period, we revised the following periods when analysing temperature, humidity, and dew point: active larvae – 11 July–27 September, and egg hatch (instead of egg laid) to adult eclosion – 11 July 2018–10 July 2019.
We standardised shoot counts in all plots to 0.19 m2, which was the area of egg enclosures, before analyses. Data were transformed so that accompanying residuals met the assumptions of normality before tested using Welch’s two-sample t-tests; if the assumptions continued to fail, we evaluated 95% confidence intervals instead. We subtracted the percent of S. nutans, Muhlenbergia spp., and Hypoxis hirsuta (Linnaeus) Coville (Hypoxidaceae), which were counted as shoots, from the estimated proportion of host plants before generating Figure 6. Plant composition and abundance were compared using permutational multivariate analysis of variance, with observed values compared to random expected values, and then were assessed by a permutation test (Anderson Reference Anderson2001; McCune and Mefford Reference McCune and Mefford2011). We used the Sorensen (Bray–Curtis) distance measure, with species weighed equally while integrating respective abundances, which is suitable for continuous and heterogeneous data sets (Bray and Curtis Reference Bray and Curtis1957; Magurran Reference Magurran1988; McCune and Mefford Reference McCune and Mefford2011). We calculated the means of the duff or thatch and litter depths from samples within each plot, as well as soil moisture and electrical conductivity from samples within each plot across sample days. Measured footcandles were converted to metric lux, and light intensity was derived using the following formula:
$${\rm{light}}\,{\rm{intensity}}\,\left( \% \right) = \left( {{\rm{lu}}{{\rm{x}}_{{\rm{below}}\,{\rm{plane}}}}/{\rm{lu}}{{\rm{x}}_{{\rm{above}}\,{\rm{plane}}}}} \right) \times 100$$
We used RStudio (RStudio Team 2021) to calculate summary statistics, conduct tests, and create Figure 6 (base R, version 4.1.2; R Development Core Team 2021; “ggplot2”, Wickham et al. Reference Wickham, Chang, Henry, Pedersen, Takahashi and Wilke2020; “svglite”, Wickham et al. Reference Wickham, Henry, Pedersen, Luciani, Decorde and Lise2021). We improved the resolution of figures in Adobe Illustrator (https://www.adobe.com/products/illustrator.html) while also creating illustrations. Permutational multivariate analysis of variance was conducted using PC-ORD (McCune and Mefford Reference McCune and Mefford2011). We used Adobe Premiere Pro (https://www.adobe.com/products/premiere.html) to stabilise the video.
Results
Oviposition
Poweshiek skippering females were observed laying six eggs during five separate oviposition episodes – that is, up to five different females. All egg laying occurred on separate plants: grasses A. gerardi – four eggs; S. heterolepis – one egg; and the forb H. hirsuta – one egg. Eggs were laid between 30 June and 3 July 2018.
Before laying an egg, females were observed flying close to vegetation, hovering above a small area, occasionally landing but not laying eggs, and finally landing and ovipositing before elevating and departing. Before oviposition, females were often observed landing on leaf blades and touching the leaf blade with the tip of their abdomen but not laying an egg. On one occasion, a female probed at three separate possible oviposition locations, laid an egg at a fourth location, and then immediately repeated this alternating sequence once before our observations of this excursion ended. This meant that the female searched eight locations and laid two eggs over the entire observation. One female was observed nectar feeding immediately before initiating flight patterns that are typical before oviposition.
In 2017, before formal observations, J.H. observed a female and male attempting to mate near the top of the vegetative canopy at a visually mesic location (Supplementary material, Video S1; Henault Reference Henault2021). After several mating attempts, during some of which the female might have blocked mating by using her wings, similarly to what Borkin (Reference Borkin1995) reported, the skipperlings mated perched on a graminoid for approximately 30–60 seconds before separating. Following this, the female immediately flew approximately 3 m away and laid an egg on a Cyperaceae sp. These behaviours were similarly observed during the mating observation in 2015.
Observations of immatures
The changing egg colour during maturation and the larval colour pattern and body shape were as described in McAlpine (Reference McAlpine1972). However, in that study’s descriptions, McAlpine (Reference McAlpine1972) did not report the bumps that we observed – each with a small black spot – regularly occurring across the cuticle of first-instar larvae. Not including our visits to monitor eggs, we observed larvae for 48.1 hours during 2018 (Table 2). The following sections describe our observations of immatures, as eggs and larvae, by type of foraging behaviour and developmental aspect.
The number of feeding, resting, and movement observations among all Oarisma poweshiek instars, and of days during which immatures were observed during the study period.
Oviposition location and hatching
The egg of Immature 1 was laid on the dorsal surface of a blade of H. hirsuta. After consuming the top of its eggshell to hatch, the larva also ate the sides but not the base of the eggshell attached to the leaf, behaviour that was also observed in Immatures 2 and 3. Immature 2’s egg was laid on a S. heterolepis leaf (dorsal surface; 7.0 cm from tip, 15.0 cm above ground), and the egg of Immature 3 was laid on A. gerardi (dorsal surface of the leaf; 2.5 cm from the leaf tip, 36.8 cm above the ground).
Immatures 4, 5, and 6 were laid as eggs on A. gerardi. Immature 5 was laid 4.9 cm from the leaf tip and 29.6 cm from the ground. The locations where the eggs of Immatures 4, 5, and 6 were laid were searched on 17, 16, and 17 subsequent visits, respectively, over the entire study period, but no larvae were found.
Feeding and resting patterns: locations and durations
Immature 1 was next observed one day later, on a shoot of A. gerardi that was 12.4 cm to the east (90°), to begin feeding on the youngest developing leaf of the plant. Immature 2 moved from its natal plant to Muhlenbergia richardsonis (Trinius) Rydberg (Poaceae), and Immature 3 moved to a S. heterolepis. During these initial stages in an immature’s development, feeding occurred at the tip of a leaf (Fig. 2).
Immature 3 (instar 2) feeding at the tip of a Sporobolus heterolepis leaf.
Whereas Immature 1 was observed only feeding at the tip, Immatures 2 and 3 ate at the tips of host plants during instar 1 and at the sides of blades in addition to the tip during instars 2–4. In addition to consuming the point of a leaf tip in a rotary movement around the tip, Immature 3 also ate the tip down to wider portions of the blade by chewing from one edge to the opposite edge in one feeding pass. The feeding patterns of Immature 2 on the sides of blades produced two staggered notches on alternate edges of a leaf (mean length of two adjacent notches = 0.7 cm, range = 0.2–1.1 cm). These notches that were made by Immature 3 sometimes occurred on the same side of the leaf blade. Immature 2 also ate longer stretches of one side of a leaf (mean = 0.9 cm, range = 0.5–1.3 cm). To eat the long sections, Immature 2 chewed from the edge of a blade to the midvein before restarting from the edge (one feeding pass = width of its head capsule). On one occasion, Immature 3 consumed a relatively longer section of leaf (approximately half the length of this larva) on both sides of a leaf to the midvein at the same location (Supplementary material, Video S2).
Immature 2’s dorsal surface while feeding appeared directional (r = 0.846), but a small sample size prevented testing, and its orientation while resting was not towards a specific direction (Supplementary material, Table S1). The dorsal surface of Immature 3 while feeding and resting was significantly directed northwest.
Immatures 1, 2, and 3 fed at or near the tip, crawled closer to the ground, and rested on a leaf blade, and then repeated the sequence by crawling to or near the tip to feed (Fig. 3; Supplementary material, Table S2). The durations of all immatures’ feeding bouts (Immature 1: n = 3, mean = 10 minutes, 95% confidence interval = 9–12 minutes; Immature 2: n = 4, mean = 21 minutes, 95% confidence interval = 15–27 minutes; Immature 3: n = 7, mean = 25 minutes, 95% confidence interval = 11–39 minutes) were significantly shorter than their resting bouts (Immature 1: n = 2, mean = 54 minutes, 95% confidence interval = 45–62 minutes; Immature 2: n = 11, mean = 95 minutes, 95% confidence interval = 54–136 minutes; Immature 3: n = 7, mean = 87 minutes, 95 % confidence interval = 50–124 minutes). In general, Immatures 1, 2, and 3 faced the tip of the leaf while feeding and the base while resting, except for one instance when Immature 2 rested facing the leaf tip.
Oarisma poweshiek larval host plant feeding and resting locations on shoots (bars on left, mean and range; bars on right, mean and 95 % confidence interval). We display larvae during each instar in which they were observed (Immature 1, purple; Immature 2, blue; Immature 3, yellow; approximately 7× size, to scale). Given the absence of measurements, we display Immature 1 feeding at the approximate distance it rested from the ground. Duff height is represented to scale.
Relocations between hosts
Immatures 1, 2, and 3 travelled both between shoots of the same species and between shoots of different species to feed, rest, and moult (Fig. 4). We observed Immature 2 using living and dead vegetation to travel, but we did not observe the substrates of Immatures 1 and 3. Immature 2 appeared to move more frequently during later instars (Table 3), and its movement was directional.
Oarisma poweshiek larval tracks in egg enclosures displayed as arrows (Immature 1, purple; Immature 2, blue; Immature 3, yellow). Immatures 1 and 2 are shown starting from same blade for illustrative purposes only.
The distance that Oarisma poweshiek larvae moved during each relocation (mean and range; total accumulated during larval stages), frequency at which larvae travelled to different host shoots between observed activity bouts, and direction during relocations (mean angle ± angular deviation) by immatures (number of movements).
Immature 1 consumed only A. gerardi, and Immature 3 ate only S. heterolepis during our observations. Immature 2 ate exclusively M. richardsonis during the first instar, S. heterolepis and M. richardsonis during the second instar, S. heterolepis and S. scoparium during the third instar, and only S. heterolepis during the fourth instar (Table 4).
Proportion of feeding observations on host plant species among all Oarisma poweshiek instars, percent (number of observations).
Immatures 1, 2, and 3 were always observed on host plants, whether feeding, resting, or moulting – including after accidental disturbance by observers. After all larvae finished feeding, they rested on the same shoot which they had previously fed on.
Growth
Immature 1 was 0.36 cm long, with measurements taken during instar 1, Immature 2 was 0.85 cm long when measured during instar 3, and Immature 3 was 0.61 cm and 0.62 cm six days after the first measurement, during instar 2. We did not observe the cause of death for any immatures. Immature 2 was last observed resting at the base of a host plant approximately 5.2 cm above the ground on 9 September 2018.
Other species
Other Lepidopterans (e.g., Geometridae) were also observed feeding, and their feeding marks were similar to those of Poweshiek skipperling larvae. Thus, we did not record marks that we were not certain were made by O. poweshiek. Ants and jumping spiders were observed in the enclosures; however, no direct predation was observed.
Larval development
Larvae were observed on shoots as their exoskeletons hardened and adjacent to a shrivelled exuvia after moulting. Immatures 2 and 3 completed each stage in the same number of degree-days and calendar days (Table 5). Using data recorded by data loggers in the egg enclosures, we calculated that larvae required a mean of 1700.0 degree-days (standard model; n = 4, 95% confidence interval = 1644.5–1755.5 degree-days) and 1723.9 degree-days (double sine; n = 4, 95% confidence interval = 1654.3–1793.5 degree-days) to develop from the first egg laid to the first adult that emerged (30 June 2018–10 July 2019, inclusive). During this same period, we calculated that data loggers in comparison plots accumulated a mean of 1693.4 degree-days (standard model; n = 4, 95% confidence interval = 1590.7–1796.0 degree-days) and 1722.0 degree-days (double sine; n = 4, 95% confidence interval = 1655.9–1788.0 degree-days; degree-day accumulated during constituent periods in Supplementary material, Table S3). The degree-days calculated for egg enclosure and comparison plot locations using each model did not significantly differ, based on overlapping confidence intervals. Calculation of degree-days using data from the nearest Environment Canada weather station located at Emerson, Manitoba (Environment Canada 2020) during the same period resulted in 1715.1 degree-days (standard model) and 1737.0 degree-days (double sine).
Degree-day accumulations in egg enclosures calculated using two models (mean ± standard deviation) and the number of calendar days required for individual Oarisma poweshieks to complete each development stage; only completed stages are displayed. Data from data loggers in all egg enclosures were used (n = 4).
During the period used in general to synchronise adult surveys (1 March–10 July, inclusive in the present study), we calculated that microhabitats in egg enclosures accumulated 635.7 degree-days (standard model; n = 4, 95% confidence interval = 609.8–661.5 degree-days) and 706.6 degree-days (double sine; n = 4, 95% confidence interval = 699.9–713.3 degree-days). A total of 667.2 degree-days (standard model) and 695.3 degree-days (double sine) were calculated using temperature data from the Emerson weather station during this period.
Microclimate in microhabitats
During the entire O. poweshiek life cycle, the relative humidity levels (mean and minimum) were recorded by most data loggers as lower in egg enclosures than in comparison plots (Table 6). The differences recorded by the data loggers occurred most frequently during the active larval period (Supplementary material, Table S4). The temperature and dew point recorded in the egg enclosures typically did not differ from that in the comparison plots during individual development periods or the entire life cycle.
The temperature (º C), relative humidity (%), and dew point (º C) during the entire O. poweshiek life cycle (mean (95% confidence interval)). Bolded values were significantly different from the values of at least three data loggers of the other plot type.
Attributes of egg enclosures and target plots, 2018
The relative abundance (Fig. 5) and counted number (Supplementary material, Table S5) of host plant shoots did not significantly differ between egg enclosures and target plots (Welch’s two-sample t-tests, P > 0.05; 95% confidence interval, confidence interval overlapped). All vegetation species enumerated are reported in Supplementary material, Table S6.
Relative abundance of host food species in 2018 and 2019. Species are scaled to their respective proportions.
Estimated proportion of shoots of consumed host species and all other graminoids (nonhost) in only egg enclosures during sampling. Data for August 2018 are calculated from shoot counts.
Graminoid height was shorter, duff percent cover was higher, and shrub percent cover was lower in the egg enclosures than in the target plots (Table 7). Duff depth and the percent cover of graminoids, forbs, and bare soil in the egg enclosures did not significantly differ from those in the target plots (duff depth, percent covers of graminoids and forbs, Welch’s two-sample t-tests; P > 0.05; percent cover of soil – 95% confidence interval; confidence interval overlapped).
Vegetative and physical characteristics (mean ± standard deviation) in Oarisma poweshiek egg enclosures (each plot type: n = 6), target plots, and comparison plots. The duff (cm) at 18 July 2019 represents the thatch component (see Methods). Measurements at each date followed by different letters are significantly different.
Attributes of egg enclosures and comparison plots, 2019
The estimated proportion of shoots of consumed host species in both the egg enclosures and the comparison plots between 30 May and 5 July 2019 was greater than on 9 May 2019 (Fig. 6). The relative abundance and counted number of host plant species, individually and as a group, also did not differ significantly between plot types on individual sampling dates (30 May, 11 June, and 5 July) nor as a mean from 30 May through 5 July (individually: Welch’s two-sample t-tests, P > 0.05; 95% confidence interval, confidence interval overlapped; as a group: permutational multivariate analysis of variance tests; P > 0.05; Fig. 5; Supplementary material, Table S5).
The graminoid height was significantly lower in the egg enclosures than in the comparison plots on 9 May 2019, whereas the mean depths of duff and litter were significantly shallower in the egg enclosures on all 2019 sampling dates (Table 7). The percent covers of duff, shrubs, living vegetation, and bare soil did not significantly differ between plot types (Welch’s two-sample t-tests; P > 0.05). The light intensity in the egg enclosures (n = 6, mean 71.0% ± standard deviation 10.9%) did not significantly differ from that in the comparison plots (mean 57.5% ± standard deviation 11.9%; t = 2.047, df = 9.910, P = 0.068). Soil moisture in egg enclosures was 28.5% ± standard deviation 9.0% volumetric water content, and electrical conductivity was 0.15 mS/cm ± standard deviation 0.07 mS/cm in 2018 and 41.9% ± 6.5% volumetric water content and 0.19 mS/cm ± standard deviation 0.05 mS/cm in 2019. Soil moisture and electrical conductivity between plot types (2019) did not significantly differ (soil moisture: t = –0.241, df = 9.939, P = 0.814; electrical conductivity: t = –0.103, df = 9.407, P = 0.920). The height of larval host plant species and the height of graminoid species we did not observe larvae consuming (i.e., nonhosts) in the egg enclosures did not differ from those in the comparison plots on 5 July 2019 (host plants: t = –1.351, df = 8.371, P = 0.212; nonhosts: t = –1.243, df = 9.548, P = 0.243).
Discussion
Egg laying
Females appeared to receive oviposition stimuli (Singer Reference Singer1971) on at least three levels. First, while in flight, the presence of suitable host food species for larvae and relatively shallower duff, as suggested by our findings from the egg enclosures versus comparison plots, might have attracted females to a general area. Within those initial general areas, smaller areas containing shorter graminoids, sparser shrub cover, and comprehensive duff cover may have then stimulated probing behaviour, as suggested by the egg enclosures versus target plots, followed finally by egg deposition or resumption of flight. In other Lepidopterans, sequential stimulations lead to oviposition (Wiklund Reference Wiklund1984; Rabasa et al. Reference Rabasa, Gutiérrez and Escudero2005), with egg deposition prompted by chemical stimuli (e.g., nitrogen concentrations) while the insect is in contact with the plant (Singer and McBride Reference Singer and McBride2010; Lund et al. Reference Lund, Brainard and Szendrei2019).
In the present study, eggs were laid on leaves that resembled grasses, at canopy positions that were relatively closer to leaf tips than to the bases. This was also observed in the female in 2017. As with Anthocharis cardamines (Linnaeus) (Lepidoptera: Pieridae) (Dempster Reference Dempster1997), O. poweshiek may have been attracted to potentially warmer temperatures in the vegetative canopy that might support egg development.
Attributes of microhabitats
In 2018 and 2019, the host plant diversity did not differ between oviposition locations compared to areas where females did not lay eggs (Fig. 5); however, several physical characteristics did differ (Tables 6–7). Oarisma poweshiek microhabitats appeared to be relatively more open and generally drier (based on shorter vegetation height and lower relative humidity). We cannot determine the degree to which the consequences of clipping vegetation, compared to flattening it, might have caused humidity differences. In grasslands in Europe, Hipparchia fagi (Scopoli) (Lepidoptera: Nymphalidae) larvae use host grasses isolated on the edges of gaps among vegetation that, overall, may provide drier microclimates for immature development (Möllenbeck et al. Reference Möllenbeck, Hermann and Fartmann2009). In the present study, the light intensity was higher in the egg enclosures (approximately 71%; test not statistically significant) than in the comparison plots (approximately 58%). However, the recorded temperatures did not differ, and the timing of plant growth was not accelerated. We recommend further research to determine whether the intensity of light may influence larval development.
Larval foraging behaviour
Our observations are the first to track the activities of individual O. poweshiek larvae in natural habitats. All larvae moved from their natal plant to feed on a different species. Immature 2 varied its graminoid forage (M. richardsonis, S. heterolepis, and S. scoparium) during different instars, but Immatures 1 and 3 each consumed only one species (A. gerardi and S. heterolepis, respectively) during our observations. Although captive-reared larvae have been observed consuming S. heterolepis and S. scoparium in the United States of America (Borkin Reference Borkin1995; Dana, personal communication), our direct observations of larvae in natural habitats eating A. gerardi and M. richardsonis are new across O. poweshiek’s range. Although captive-reared larvae fed only one species may still survive (Assiniboine Park Zoo 2018), polyphagy in natural habitats might vary the nutrition, defensive chemicals, and parasites ingested by larvae and thereby confer fitness benefits (Erebidae; Karban et al. Reference Karban, Karban, Huntzinger, Pearse and Crutsinger2010). These four grasses are consistently associated with O. poweshiek across its range (oviposition and larval feeding; Table 1), suggesting a relatively higher facilitation of survival, possibly with additional regionally preferred hosts (e.g., Nymphalidae; Meister et al. Reference Meister, Lindman and Tammaru2015).
Larvae also relocated over substantial distances to shoots of the same host plant species on which to live, suggesting that factors in addition to host species attract foraging larvae. Relocations appeared to occur more frequently at later instars, and accumulated movements were inconsistently directional (Table 4). Hellmann (Reference Hellmann2002) reported Euphydryas editha bayensis (Sternitzky) (Lepidoptera: Nymphalidae) larvae switching host plants following leaf desiccation. Feeding-induced plant defenses (Roslin et al. Reference Roslin, Syrjälä, Roland, Harrison, Fownes and Matter2008) may prompt O. poweshiek larvae to relocate once the risk of movement outweighs that of eating an activated host as larger, later-instar larvae (e.g., Sphingidae; van Dam et al. Reference van Dam, Hermenau and Baldwin2001).
As larvae matured, we observed feeding patterns changing to include notch feeding of various lengths and arrangements, in addition to tip feeding – a combination of patterns that is similar to that of wild larvae in Borkin (Reference Borkin1995) and captive-reared larvae in McAlpine (Reference McAlpine1972). The leaf blade instead of the tip might provide a less exposed and safer location for larger, more visible larvae to feed. We also observed other Lepidoptera (including Geometridae) creating visually similar feeding marks; therefore, we recommend acquiring more direct observations of larval O. poweshiek feeding in natural habitats to determine or confirm the species consumed by larvae in extant skipperling populations (including, e.g., M. richardsonis in Michigan; Pointon Reference Pointon2015; Belitz et al. Reference Belitz, Monfils, Cuthrell and Monfils2019).
All larvae fed at the tips of leaves near the top of the vegetative canopy (Fig. 3), possibly to take advantage of warmer temperatures at these positions (Pieridae; Jugovic et al. Reference Jugovic, Grando and Genov2017). After feeding, larvae rested closer to the ground, possibly to shelter from predators among vegetation (Erebidae: Stamp Reference Stamp1992). Turlure et al. (Reference Turlure, Choutt, Baguette and van Dyck2010) reported larvae of Boloria aquilonaris (Stichel) (Lepidoptera: Nymphalidae) occupying sun-exposed hosts to feed and travelling to positions within cool and humid moss hummocks to rest. In the present study, Immature 3, which consistently occupied the north and shaded side of leaves while feeding and resting (south-approaching sun), might have been adjusting its body temperature, as Capinera et al. (Reference Capinera, Wiener and Anamosa1980) had observed in a grass-feeding moth. The durations of sequential feeding and resting bouts among Immatures differed, with feeding periods being shorter than resting periods. This suggests that the degree of larval attraction to a certain location might change as they eat and rest, perhaps influenced by the amount of plant material ingested or, while resting, the degree of digestion.
Ovipositing Pyrgus malvae (Linnaeus) (Lepidoptera: Hesperiidae) are influenced by both the microclimates and host plant species that are required by larvae (Krämer et al. Reference Krämer, Kämpf, Enderle, Poniatowski and Fartmann2012). We think a similar mechanism in O. poweshiek, which may instead be influenced by plant architectures, microclimates, and the presence of larval host plants, might result in females laying eggs at locations suitable for the development of eggs but not of larvae. As a result, larvae may reposition to very small areas where suitable host species and microclimates for activities (e.g., feeding and resting) intersect. A “microhabitat” in this case would consist of a fuzzy radius from the location of oviposition, encompassing all suitable intersections that larvae could potentially navigate to. Following local measurements at observed locations of adult oviposition and of larvae, as well as hypothetical three-dimensional mapping and modelling of host plants and microclimates in microhabitats, such a suitable area for each respective life stage within a “microhabitat” may be termed a “nanohabitat”.
Given that we observed that host species emerge and potential larval development threshold resumes above 0 °C at the end of May, we predict that through the last two weeks of May to first week of June, larvae resumed feeding (if postdiapause larvae were to eat the same species). Poweshiek skipperling larvae likely consume species that are phenologically timed (e.g., Pieridae spp.; Posledovich et al. Reference Posledovich, Toftegaard, Wiklund, Ehrlén and Gotthard2015) to attain suitable sizes before pupation (e.g., Hesperiidae; Rosenwald et al. Reference Rosenwald, Lill, Lind and Weiss2017). As environmental conditions, including extreme temperatures and precipitation, change from spring through fall while larvae are active, certain plant species may facilitate larval survival better than others. For example, the nutrients or the physical toughness of their tissues may degrade less readily or the timing and degree of maturation, or cool- or warm-season physiology (Kindscher and Wells Reference Kindscher and Wells1995), may better support larvae at that time. In addition, we observed the relative shoot proportion of individual plant species changes throughout the year (Fig. 5). Hellmann (Reference Hellmann2002) reported that environmental changes led to degradation of plant tissue resulting in host switching by Euphydryas editha bayensis. Based on this, larval adaptations to navigate microhabitats to locate ephemerally suitable shoots and to digest multiple species would likely confer fitness advantages to O. poweshiek. This pattern may explain the polyphagy of graminoid-feeding skippers in natural habitats that have been observed globally (natural habitats: Wiklund Reference Wiklund1984; Dana Reference Dana1991; García-Barros and Fartmann Reference García-Barros, Fartmann, Settele, Shreeve, Konvicka and van Dyck2009; lab choice: Molleman et al. Reference Molleman, Halali and Kodandaramaiah2020; Nordmeyer et al. Reference Nordmeyer, Runquist and Stapleton2021). The Poweshiek skipperling appears to be attracted to any species that meet its survival needs, including for ovipositing and larval feeding, possibly influenced by availability and seasonal variation in plant quality. This skipperling may be defined most accurately as a varied specialist or limited generalist. Further research that focuses on the larvae’s survival on single host species versus a mixture of host species as environmental conditions change could help determine and explain foraging adaptations to prairie habitats.
Because moving larvae in the present study did not circle along the plastic walls of the enclosures and the heights of respective species of both confirmed and potential host and nonhost groups did not differ between insides and outsides of the enclosures, we interpret that the egg enclosures neither limited larval relocations nor detectably altered the growing conditions of plants. However, we suggest that future studies use control enclosures to more accurately determine the effects of microclimates. While we do not encourage disturbing the habitat more than necessary, we think that information that could be generated by installing additional enclosures is worth the risk of minor disturbance.
Larval development
All three immatures completed developmental stages in a similar number of degree-days and calendar days, appearing to substantially increase in amount from the second to third instars (Table 5). Their development, measured as body length, was consistent with McAlpine (Reference McAlpine1972). The combined record and estimate of the number of degree-days required by O. poweshiek to complete their life cycle in the present study of approximately 1700 degree-days (standard model) or 1725 degree-days (double sine) is reported here for the first time, based on estimated development periods and thermal thresholds and for only one generation.
Immatures required approximately 415 degree-days (standard model) and 385 (double sine) to complete instar 3. Once Immature 2 moulted to instar 4, only approximately 115 degree-days (standard model) and 145 (double sine) were available to be accumulated before winter. McAlpine (Reference McAlpine1972) reported that captive-reared larvae in Michigan, which is at a relatively lower latitude, diapause in the fifth instar. We think that O. poweshiek in Manitoba may have evolved fewer prediapause instars – that is, diapausing in instar 4 – or shorter instar durations – that is, moulting to instar 5 before diapause – to compensate for the shorter available season to develop while appropriately timing adult emergence (e.g., Lycaenidae: Fischer and Fiedler Reference Fischer and Fiedler2001, Reference Fischer and Fiedler2002). To accurately determine the developmental schedules, including factors that induce or release diapause, and both the lower and upper thresholds, including degree-day accumulations, of immatures, we recommend similar direct observations during spring and physiological studies of multiple generations of O. poweshiek.
Because the degree-day accumulations estimated to complete the O. poweshiek life cycle that were calculated using only the double sine model did not significantly differ between data loggers and the weather station, we recommend that conservation partners focus on the double sine model that uses data from confirmed microhabitats or the Emerson station to accurately synchronise adult surveys. Microhabitats of immatures during the survey-synchronisation period accumulated more degree-days than did microhabitats in Dearborn and Westwood (Reference Dearborn and Westwood2014; our study: standard model = 635.7 degree-days, double sine = 706.6; Dearborn and Westwood’s (Reference Dearborn and Westwood2014) study: standard model = 602.8 degree days, double sine = 653.8). This was likely caused by: data logger placements, which in our study were in verified microhabitats and in Dearborn and Westwood’s (Reference Dearborn and Westwood2014) study were in random locations in prairies where O. poweshiek adults had been observed; annual weather fluctuations, which in our study were based on one year and in the 2014 study were a five-year mean; or an unknown factor.
Broader conclusions
Sites in Manitoba that had been intentionally burned approximately three to six years earlier supported relatively higher abundances of Poweshiek skipperlings than did sites with more recent and older burns and sites where grazing took place (Dupont-Morozoff et al. Reference Dupont-Morozoff, Westwood and Henault2022). Disturbances such as a cutting action – haying – and prescribed burns may improve the suitability of habitats for butterflies (Swengel Reference Swengel1996; Thom and Daniels Reference Thom and Daniels2017) but ought to be compatible with the life history of these same butterflies to reduce mortality (Swengel and Swengel Reference Swengel and Swengel1999, Reference Swengel and Swengel2015). Our findings regarding seasonal larval positioning of Poweshiek skipperlings suggest that haying in late summer or fall, when larvae are relatively closer to the ground, instead of in mid-summer, when the eggs or larvae are located in the vegetative canopy, may positively influence the vegetative components of habitats without directly displacing or destroying larvae (e.g., grazing for Pieridae; Nippen et al. Reference Nippen, Dolek and Loos2021). To determine conservation disturbances that create or maintain the intersection of microhabitat attributes that are suggested by our study and are compatible with other species at risk, we encourage future research.
Our findings suggest foraging by O. poweshiek, and possibly other species, has multiple components that likely increase fitness. These components are likely: nutrition or diverse diet; sensing quality hosts if or when conditions change; and the ability to move and navigate over distance, at different frequencies, over substrates. It is possible that, over the last few decades, factors, such as habitat management, climate, or other wildlife, that facilitate these foraging traits have changed, resulting at least temporarily in reduced larval fitness and leading to population decline. Although the small number of individuals in the present study limited our capacity to develop statistical inferences, our findings could be used to inform habitat stewardship and reintroduction via captive-rearing and site-assessment efforts. With the findings from the present study providing an improved understanding of adult oviposition behaviours, larval foraging patterns, and microhabitat attributes, we can continue narrowing the list of possible causes of decline of O. poweshiek and generating potential solutions.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.4039/tce.2022.34.
Acknowledgements
The authors thank Laura Reeves, Matthew Russell, Mia Kirbyson, and Stefanie Sheard for their enthusiastic field assistance. The authors are grateful to have researched on Treaty 1 territory, ancestrally stewarded by diverse First Nations, Métis, and other Indigenous communities. They also thank Melissa Grantham, Cary Hamel, and the Manitoba Tall Grass Prairie Preserve Management Committee for facilitating field research; Andrew Park, Gard Otis, William Watkins, and Robert Dana for manuscript reviews; Pam Henault for figure reviews; members of the Poweshiek Skipperling International Partnership and peers for encouragement; Medea Curteanu for project advice; and Jeremy Hemberger for introductions to data visualisation techniques. Financial support was provided by the Canadian Wildlife Service (facilitated by M.C.), the Nature Conservancy of Canada, the University of Winnipeg, and Wildlife, Fisheries and Resource Enforcement Branch, Government of Manitoba (W.W.) and is much appreciated. They thank colleagues for critiques during conferences, two anonymous reviewers for focused suggestions, and The Canadian Entomologist team for facilitation.
Author contributions
J.H. and R.W. conceived the idea of the study as part of J.H.’s M.Sc. thesis; J.H. led fieldwork, data analysis, and drafting of the manuscript; J.H. and R.W. both contributed to the drafts and the final version of the manuscript.
Open access
