What Azure blues occur in Canada? A re-assessment of Celastrina Tutt species (Lepidoptera, Lycaenidae)

Abstract The identity of Celastrina species in eastern Canada is reviewed based on larval host plants, phenology, adult phenotypes, mtDNA barcodes and re-assessment of published data. The status of the Cherry Gall Azure (Celastrina serotina Pavulaan & Wright) as a distinct species in Canada is not supported by any dataset, and is removed from the Canadian fauna. Previous records of this taxon are re-identified as Celastrina lucia (Kirby) and Celastrina neglecta (Edwards). Evidence is presented that both Celastrina lucia and Celastrina neglecta have a second, summer-flying generation in parts of Canada. The summer generation of Celastrina lucia has previously been misidentified as Celastrina neglecta, which differs in phenology, adult phenotype and larval hosts from summer Celastrina lucia. DNA barcodes are highly conserved among at least three North American Celastrina species, and provide no taxonomic information. Celastrina neglecta has a Canadian distribution restricted to southern Ontario, Manitoba, Saskatchewan and easternmost Alberta. The discovery of museum specimens of Celastrina ladon (Cramer) from southernmost Ontario represents a new species for the Canadian butterfly fauna, which is in need of conservation status assessment.


Introduction
Blues of the genus Celastrina Tutt, commonly known as azures, are perhaps the most familiar spring butterflies in Canada, occurring in all ecoregions except the high arctic. Despite their ubiquity, their identification and taxonomy is difficult, with species boundaries and nomenclature having a long history of controversy and confusion. Forty years ago, all North American Celastrina taxa were generally considered to represent variation within a single species described from Europe, C. argiolus (L.) (Langston 1975). This view remained essentially unchanged for another twenty years, with the exception of a second taxon, C. nigra (Forbes) recognized by Miller and Brown (1981) and Scott (1986). A global revision of Celastrina and related genera further entrenched the concept of only two North American species (Eliot and Kawazoé 1983). However, with a more detailed study of the genus in North America additional cryptic species were gradually recognized by some (Opler and Krizek 1984, Pratt et al. 1994, Scott and Wright 1998, Wright and Pavulaan 1999, Pavulaan and Wright 2005. Celastrina taxonomy is still unsettled, with recent comprehensive North American checklists varying between three (NABA 2001) and nine recognized species (Pelham 2008). A summary of some of the changing concepts, particularly in the historical literature, is given by Pratt et al. (1994) and Pavulaan (2014).
The conservative morphological variation between most Celastrina species, coupled with adult seasonal polyphenism, has been a major impediment to Celastrina taxonomy and dictated a gradual refinement of species concepts. Comparative data on molecular variation, physiology, development and ecology for sympatric or closely parapatric populations are therefore particularly important in evaluating species concepts, yet such data are largely lacking (but see Pavulaan 2014). To provide a taxonomic reference point for Canada's Celastrina populations and to stimulate further study, the identity of Ontario Celastrina populations is re-assessed based on published and novel data on phenology, larval host plant use and mtDNA variation. Ontario provides a unique geographic arena where biological and biogeographical attributes of putative species can be examined. Here, three species purportedly occur in sympatry: C. lucia (Kirby), C. serotina Pavulaan & Wright and C. neglecta (Edwards) (Layberry et al. 1998, Pavulaan andWright 2005). A fourth species, C. ladon (Cramer), has been reported from adjacent parts of Ohio and Michigan (Nielsen 1999). With potentially as many as four species present in Ontario, life history traits and diagnostic characters of Celastrina were studied and compared among two ecoregions, the Lake Erie region in southernmost Ontario and the Ottawa region in eastern Ontario. These regions were chosen as both have a long history of entomology with a comparatively large data pool on Lepidoptera, and represent separate ecoregions with all three (and potentially four) eastern Canadian Celastrina species present.

Current concepts of eastern Canadian Celastrina
Four Celastrina species are currently attributed to the Canadian fauna, three of them found in the East. The fourth species, C. echo (Edwards), is strictly western and al-  (Pelham 2011) The phenology and larval host plant are key diagnostic features used to distinguish C. serotina. Also, the ventral hindwing pattern is stated to be paler whitish grey on average than C. lucia, with heavily marked forms being rare. The taxonomy of Ontario C. serotina is particularly relevant since life histories and specimens of these populations formed part of the original species description (Wright and Pavulaan 2005). Celastrina serotina has also been reported from Québec, New Brunswick, Nova Scotia and Prince Edward Island (Layberry et al. 1998), and recently from Manitoba (based on larval collections, leg. T. Rapati; eButterfly 2015, record #EB-3473).
Celastrina ladon, the Spring Azure, has not been reported in Canada in the sense of the modern concept of the species, where the diagnostic male wing scale morphology ( Fig. 1) separates it from all other Celastrina (see also Omura 2015; Wright and Pavulaan 1999). Literature reports of C. ladon in Canada consist of previous concepts where C. ladon and C. lucia were considered to be conspecific (e.g. Wright and Pavulaan 1999). Celastrina ladon has subsequently not been included in the Canadian fauna (Hall et al. 2014). Older reports of C. ladon from southern Ontario may have included true C. ladon, but these records cannot be distinguished from C. lucia without voucher material. The Spring Azure is known from adjacent parts of southern Michigan (Nielsen 1999) and Ohio (Wright 1998).

Methods and materials
Specimens examined during this study included those deposited in the Canadian National Collection of Insects, Arachnids and Nematodes (CNC), in addition to Celastrina records with voucher photographs on eButterfly (2015). Forewing androconial scales of male Celastrina were examined using a Leica 205C dissecting scope. Vouchers of reared specimens are deposited in the CNC.

DNA barcodes
Molecular variation of Celastrina species was assessed using the COI barcode fragment, with DNA extraction, PCR amplification, and sequencing performed at the Canadian Centre for DNA Barcoding (CCDB), following standard protocols (CCDB 2013). Public barcode sequence records were available for three North American species (C. lucia, C. echo and C. neglecta), and the Eurasian C. argiolus and Asian C. morsheadi (Evans). Novel sequences were generated for 31 eastern Ontario specimens (Suppl. material 1), initially identified as C. neglecta (five wild-collected specimens and five reared from larvae collected in late July to August), C. serotina (10 specimens reared from larvae feeding on cherry galls in mid June), and C. lucia (four specimens collected in May).
DNA sequences were analyzed on the Barcode of Life Data Systems website (BOLD, www.boldsystems.org). The dataset was filtered to include only records with sequences greater than 600 base-pairs in length, and with voucher specimen photographs and collection data that made independent species identification possible. Sequence variation was analyzed using the Kimura-2-Parameter (K2P) distance model and the neighbor-joining (NJ) algorithm as implemented on BOLD. Voucher specimen data is given in Suppl. material 1.

Larval development and host plants
Larvae were collected from the wild to compare phenology and voltinism of C. lucia and C. serotina, and to obtain comparative study specimens unambiguously associated with the current concept of C. serotina. Celastrina serotina is univoltine with a peak flight after that of C. lucia (Pavulaan and Wright 2005), so larvae develop later in the season with the resulting pupae entering diapause until the following spring. Larval sampling was carried out in eastern Ontario (Table 2) by directed visual searches and the use of a beating sheet. Numerous species of flowering shrubs were sampled, with most effort directed to sampling Cornus, Viburnum and Prunus. Larvae were reared indoors under natural light:dark conditions and at a constant 20 °C, reflecting the June mean daily temperature of 20.4 °C for Ottawa (Environment and Climate Change Canada 2015).

Flight phenology
As a proxy for mean seasonal abundances of Celastrina taxa, observation records spanning from 1895-2014 were compiled from the Ontario Butterfly Atlas (Macnaughton et al. 2015). Each unique location-date record was treated as one observation event, regardless of Celastrina abundance during that event. Observation frequency (abundance) by date was assessed for two ecoregions, the Great Lakes-St. Lawrence Mixed Forest of easternmost Ontario and the Carolinian Forest of southernmost Ontario  (Scott 1995). The southern Ontario dataset included 1056 records from Brantford, Elgin, Essex, Kent, Lambton, Middlesex, Niagara and Norfolk counties; eastern Ontario data consisted of 2145 records from Ottawa-Carleton, Lanark, Russell, Prescott, Glengarry, Stormont and Dundas counties. Three Celastrina taxa were considered to occur in each region, but to avoid a priori assumptions about species identities, all Celastrina records were combined. Assessing flight peaks based on phenological data combined for multiple taxa could underestimate the number of taxa, if relative abundance discrepancies are large and flights overlap. Emergence patterns were therefore independently assessed through field surveys of eggs, larvae and adults 1-2 × per week in 2015. These data were supplemented with Celastrina records and accompanying voucher photographs available on eButterfly (2015).
To assess between-region differences in adult emergence times due to climatic differences, phenology data were examined using a simple degree-day model (e.g. Kelker et al. 1990; Dearborn and Westwood 2014) using the formula: where DD = degree-days, T max and T min = daily maximum and minimum temperatures, respectively, and LTT = the lower threshold temperature of insect development. LTT is the temperature at which physiological development is negligible, for the species and life stage under study. LTT values of 6°C to 10°C are generally implemented for insects, with values in the lower range corresponding to temperate-zone species (e.g. Kelker et al. 1990). As Celastrina are cold-adapted and some of the first lepidopterans to emerge from winter-diapausing pupae, LTT was set at 6 °C. A start date of April 1 st was chosen as DD accumulation values were zero prior to this date (for all values of LTT between 6 °C and 10 °C). Daily temperature data were obtained for 2009-2015 for two stations, Ottawa (city station) for the eastern Ontario region, and London for the southern Ontario region (Environment Canada 2015). A few instances of missing daily maximum/minimum temperature data were estimated by averaging the corresponding temperature from the preceding and following day. Daily maximum and minimum temperatures were calculated based on a six-year average from 2009-2015. London was chosen as representative of the Lake Erie region as it is inland from the Lake Erie shoreline and therefore less prone to cooling climatic effects of onshore winds along the immediate shoreline region.

Wing Scale structure
Examination of forewing scale structure in male Celastrina specimens from southern and eastern Ontario led to the discovery of four specimens of C Carmichael. Two female specimens are likely also C. ladon, one from Normandale, 28.May.1956, J.R. Lonsway, and one from St. Williams with the same date and collector as the male. All are from the Carolinian forest region of Lake Erie (Fig. 6). Celastrina ladon is therefore confirmed as part of the Canadian fauna for the first time. Although other literature and even photo records may exist, voucher specimens are needed to verify identification, at least until phenotypic variation and distribution of C. ladon in southern Ontario is better documented. Unvouchered previous records are of little value in ascertaining Celastrina identities in southern Ontario, underscoring the importance of voucher study specimens even when a species is thought to be common and well-known.

DNA barcodes
DNA barcode data were available for three North American Celastrina species (C. echo, C. neglecta, and C. lucia, based on independent identification), and representative Eurasian C. argiolus from seven countries (Fig. 2). Eastern Ontario specimens initially identified as C. lucia, C. neglecta and C. serotina are all considered to represent C. lucia based on the larval rearing, adult phenology and wing pattern, as discussed below.
Nearly all samples of C. lucia, C. neglecta and C. echo shared an identical DNA barcode. A single haplotype (h03, Fig. 2) was dominant across the continent, representing 76 of 79 individuals and occurring in all three species. Three additional haplotypes (h01, h02, h04; Fig. 2) differed by only a single base-pair, i.e. 0.15% divergence, and were represented by a single individual each (Fig. 2). Comparison of these four North American Celastrina haplotypes to others in the BOLD database using the sequence identification search engine showed that the extremely conserved genetic variation was not a sampling artefact, with virtually no variation in samples from across North America including Mexico, and including samples identified as all nine North American species in addition to the Mexican C. gozora (Bdvl.). This lack of mtDNA genetic differentiation between distinct species occurs also in other North American blues, as a Figure 2. Neighbour-joining tree of DNA barcode sequences for Celastrina, with specimen voucher number and country of origin at branch tips. North American samples include 79 samples represented by four haplotypes, with h03 shared among three species (n=76) and remaining three haplotypes with one sample each of C. lucia (h01) and C. neglecta (h02, h04). Voucher data is given in Suppl. material 1.
result of introgressive hybridization and possibly infection by the endoparasitic bacterium Wolbachia (Gompert et al. 2008). Further research with other molecular markers is needed in Celastrina. Although the DNA barcode sequence is not taxonomically informative for North American species, it does corroborate separate species status of C. argiolus, which differed by a minimum of 1.4% (mean 1.9%).

Larval development and host plants
Eggs and larvae of C. lucia were found on flower buds and inflorescences of nine species of shrubs in eastern Ontario (Table 3). Based on correlative adult abundance, plant community composition and frequency of larvae on these hosts, Prunus serotina, Cornus alternifolia L. C. rugosa Lam., Viburnum cassinoides L. and V. lentago L. are the most commonly used larval host plants of spring C. lucia in this region. Prunus pensylvanica L., P. nigra Aiton and Amelanchier species were also searched, but these shrubs bloom very early in the spring with flowers already senescing during peak Celastrina abundance, and no larvae were found (Table 3). Viburnum rafinesquianum Schult. is rarely used, possibly also due to the later flowering phenology; only one larva was found in searches of 20 shrubs at two different sites (#7, 6; Table 2). Three mature larvae were found feeding on flower buds of Celastrus scandens L. in open limestone alvar habitat (site #6). This is the first record of Celastrus as a host of Celastrina, and adds the family Celastraceae to the list of known host plants (Scott 1986).
Other deciduous shrubs flowering during and after the spring flight season of Celastrina were sampled opportunistically, but failed to yield larvae, even when larvae were common on other shrub species at the same sites. These included Ilex verticillata (L.) A. Gray, Ilex mucronata (L.) (both Aquifoliaceae), and Lonicera tatarica L. (an introduced invasive shrub), Cornus racemosa Lam., Vaccinium angustifolium Ait., and Gaylussacia baccata (Wangenh.) K. Koch. An extensive search of the introduced Viburnum lantana L. at one site (#2) yielded one half-grown larva, which died several days later in captivity feeding on this plant. Ilex is the sole host of Celastrina idella Wright and Pavulaan, but is thought to be toxic to C. lucia (Wright and Pavulaan 2005). Cornus sericea L. is a common host of boreal C. lucia populations, but searches for larvae in the study area (site #1) were unsuccessful, despite the patchy but common occurrence of this shrub. Virtually all of the host plants recorded above have completed flowering prior to the onset of summer C. lucia flights, which strongly favor Spiraea alba Du Roi as oviposition sites and larval hosts ( Table 3).
Most of the C. lucia host shrubs present in the eastern deciduous forest are absent in the boreal region further north. Within the host genera Cornus and Viburnum, C. sericea and V. edule (Michx.) Raf. occur widely in the boreal region, but only sporadically in certain plant communities. By contrast, species of Ericaceae are ubiquitous and constitute the main larval hosts in many parts of the ecoregion, particularly plant communities on acidic substrates such as granite barrens, sand plains and bogs. Host plants documented along the James Bay highway in northern Québec in June 2015 (BCS, unpubl. data) included Cornus sericea, Rhododendron groenlandicum (Oeder) Kron & Judd and Kalmia polifolia Wangenh. Searches on V. edule and Prunus pensylvanica failed to yield eggs or larvae.
Larvae found on different plant genera exhibited different colour morph frequencies. Larvae on Cornus alternifolia were mostly very pale, pastel-green with little patterning (Fig. 3, right), compared to those on Viburnum lentago, which were darker green and more patterned (Fig. 3, left). Three larvae from Celastrus flowers were very dark green with little patterning. Differences in larval colours and pattern may represent hostplant-induced variation, not previously documented in Celastrina but known to occur in other Lepidoptera (e.g. Sandre et al. 2013).
Of approximately 120 gall-feeding larvae found on 18 Prunus serotina trees heavily infested with eriophyid galls, 28 were retained for rearing. Based on size and duration to pupation, approximately 75% were penultimate or ultimate instar, but younger instars were present also. A similar age distribution was observed among Celastrina larvae on other hosts at the same time and location, based on collection of 30 larvae from Viburnum lentago and Cornus alternifolia. Twenty-two of 28 larvae from galls survived to pupation, with five adults emerging between June 29th and July 2 nd (summer phenotype) and three more emerging within 9 days at room temperature (spring phenotype) after a 95-day treatment of winter diapause conditions at 5°C in a conventional refrigerator. The remaining 14 pupae failed to merge and were dissected, revealing fully developed but desiccated adults, which could be assigned to either summer or spring phenotype by comparison to pinned specimens (Fig. 10). In total, 13/22 (59%) and 9/22 (41%) individuals displayed spring versus summer phenotype.
Similar results were obtained from rearing of cherry-gall feeding larvae collected in June of 2004 (RAL), where some pupae yielded summer-phenotype adults in the same year, and some entered diapause to emerge as spring-phenotype adults the following year (Suppl. material 1).
Phenology of cherry gall-feeding larvae was not notably different from that of larvae on other hosts, contrary to the prediction that larvae should appear later based on a later flight period in late May to late June, after that of C. lucia (Pavulaan and Wright 2005). Mature larvae found on 11 th June would have to be derived from adults flying at least three weeks earlier, assuming 5d for egg hatch and 16d for larval growth Cornus alternifolia (right column) at site #3 (Table 2). even under constant temperatures of 19 °C or more (Table 4). No small larvae were present after June 20 th . The gall-feeding larvae showed the same size/age distribution as C. lucia larvae collected from Viburnum and Cornus. Neither the larval phenology  Wagner and Mellichamp (1978) nor the summer-emerging adults resulting from gall-feeding spring larvae support that gall-feeding larvae represent a separate species, i.e. C. serotina. Furthermore, both May and August larvae, initially thought to represent C. lucia and C. neglecta (Suppl. material 1), can yield either summer adults from non-diapausing pupae or spring adults from diapausing pupae. The alternative taxonomic explanation is that gall-feeding larvae are C. lucia, utilizing an unusual plant resource that is, however, similar to a Prunus flower bud in size, shape, tissue consistency, and likely phytochemistry. With a relatively long spring flight period and short flowering phenology for a given host species, C. lucia must use a suite of hosts to match larval development to host phenology. Galls extend the temporal availability of Prunus as they are present longer than flower buds. The total flight season for C. lucia is over a month in a given year (Table 6), yet any particular hostplant provides optimal forage for a considerably shorter period. For example Prunus virginiana is one of the first hosts to have flower buds, but once flowering begins, females avoid them in favour of other host species.

Degree days
Comparing degree-day accumulation to flight abundances provides a standardized comparison of flight seasons between southern and eastern Ontario (as defined here), where different climatic conditions prevail. In other words, peak adult emergence is expected to have similar degree-day (DD 6 ) accumulation values (dictated by physiological developmental constraints) in regions with differing climates, even though flight times could have quite different calendar dates. Furthermore, DD 6 accumulation can be used to assess if climatic conditions are amenable to producing multiple yearly generations (multivoltinism).
Cumulative DD 6 during the spring and summer months was greater for southern compared to eastern Ontario (Table 5). In mid-to late April, DD 6 accumulation in eastern Ontario lags behind that of southern Ontario by 4-6 days. As the season progresses, the time lag between the two regions diminishes to 2-3 days, for May to the  end of July (Table 5). The faster DD 6 accumulation in southern Ontario is correlated with a slightly earlier spring Celastrina peak in that region, occurring on average three days earlier (Figs 4, 5). Large differences between abundance peaks (more than one week) observed between regions are therefore not likely attributable to regional variation in development times of the same species, assuming similar development rates and thresholds between regions. Is it possible that summer abundance peaks represent the offspring of spring Celastrina? Currently, spring and summer Celastrina are treated as separate species, and some have maintained that Celastrina flying subsequent to the spring flight appear too soon for this to be possible (e.g. Saunders 1875). In eastern Ontario, the median abundance dates occur on May 11 th and July 12 th (Fig. 4; 50% of observations for the period prior to June 12st or after June 20 th ). In southern Ontario, however, the situation is different, as there is an abundance peak with a mean date of June 15 th , after a spring peak on May 8 th . The time lag between the first two seasonal peaks is therefore between May 11 th -July 12 th in Eastern Ontario and May 8 th -June 15 th in southern Ontario, corresponding to an average degree-day (DD 6 ) of 750 and 381, respectively (data not shown; degree-day trends in Table 5). These DD 6 values likely represent a slight overestimate of actual degree-days available for completion of a generation, since the between-peak time lag does not account for the fact that most eggs are probably laid after the peak flight period. This is due to females emerging later than males and being less commonly observed, as is true for nearly all butterflies (Scott 1986).
Average life cycle duration of non-diapausing Celastrina in Ontario (35d total; egg = 5d, larva = 17d, pupa = 13d at 22 °C; Table 4), has an accumulated DD 6 value of approximately 560, considerably greater than the maximum estimated DD 6 of 381 available in southern Ontario, but less than the DD 6 of 750 in eastern Ontario. Degree-day modelling data therefore indicates that there are enough degree-days between the first and second abundance peaks to permit development of a second generation in eastern Ontario but not in southern Ontario, and the two peaks in the latter region cannot therefore represent the same species.

Adult phenology
Celastrina phenology in eastern Ontario exhibits a bimodal pattern, with a well-defined spring and summer peak. Median spring abundance (i.e., 50% of records) occurs on May 8 th and median summer abundance on July 12 th (spring and summer periods divided by the trough midpoint at June 21 st ). Celastrina abundance drops sharply between June 5 th and June 24 th ; in other words, azures of any kind are very rarely observed in eastern Ontario during this period (Fig. 4). This is opposite to the pattern seen in southern Ontario, where a June 15-19 th abundance peak occurs in addition to a May and July/August peak (Fig. 5).  Another notable difference in Celastrina phenology between eastern versus southern Ontario is the magnitude of spring (April-May) versus summer (July onwards) abundance peaks. In southern Ontario, there are considerably fewer spring than summer records, the converse of the pattern in eastern Ontario. Celastrina abundance also persists further into the summer in southern Ontario, not declining significantly until after Aug 21 st , compared to steady declines after mid-July in eastern Ontario (Fig. 5).
The bimodal abundance pattern in eastern Ontario reflects at minimum two entities, a spring-and a summer-flying Celastrina, previously considered to be C. lucia and C. neglecta, respectively. The time lag between spring and summer emergences, and the rearing results and phenotype comparisons discussed below, indicate that eastern Ontario spring and summer Celastrina represent two broods of the same species, C. lucia.
Although there is no evidence of a third peak (in eastern Ontario) intercalated between the first and second as would be expected for C. serotina, it is possible that such an abundance signature is hidden by virtue of C. serotina being much rarer than C. lucia and C. neglecta. However, the 2015 observations on larval and adult phenology do not support this (Table 6). No "flush" of freshly emerging adults appeared after the peak of C. lucia adults, and there was no detectable difference in age (size) of cherrygall feeding larvae compared to other larvae. What, then, is the true identity of Celastrina previously attributed to C. serotina? To address this, all eButterfly (2015) C. serotina records with voucher photographs were examined, consisting of 28 records with dates ranging from 14 May to 26 June. Both worn lucia-like individuals and freshly emerged neglecta-like individuals are identified as serotina, the primary means of identification apparently being date. Fourteen individuals were visually indistinguishable from either worn individuals of C. lucia or fresh, lightly marked (form "violacea") individuals thereof. Ten individuals were fresh with a chalky-white ventrum and small, sharp spots and little to no marginal markings, like those of the June-flying, southern Ontario entity here assigned to C. neglecta. Specimens identified as C. neglecta tended to occur further south than C. lucia (Fig. 8).
In southern Ontario, spring Celastrina are rare compared to the abundance of azures seen from June onwards (Layberry 1996). Saunders (1875) noted that Celastrina were absent prior to late May in the London area. This pattern is reflected by fewer spring vs. summer observations, and the presence of an additional June flight peak that is absent in eastern Ontario. Comparison of June Celastrina from southern Ontario to those from other areas reveals that these also differ phenotypically (Figs 9, 10). As discussed under C. neglecta in the Conclusions section, this taxon is here deemed to be C. neglecta.
The spring/summer abundance discrepancy in southern Ontario was also noted by Layberry (1996), who stated that spring Celastrina were rare and local and could not possibly produce the abundance of ubiquitous summer Celastrina. This discrepancy can be explained by the localized occurrence of C. lucia (near its southern range limit) and C. ladon (restricted to Carolinian woods) in spring, followed by the much more common C. neglecta in late May-June and again in Late July-August.  (Figure 4), C. neglecta data is based on assumption of two annual flights, the first peaking in mid-June and with a generation time similar to that of C. lucia (750 degree-days). Celastrina ladon data is based on assumption of a single, earlier flight and lower overall abundance compared to C. lucia, but with similar abundance changes and length of flight period. The sum of all predicted Celastrina abundances is compared to actual observation frequencies in Figure 5. The complex abundance peaks for southern Ontario are at least in part a result of combined data for multiple species. Degree-day modelling can however be used to approximate the apparent abundance peaks. Given a spring peak of C. lucia on May 8, and an average DD 6 accumulation of 750 to reach the second-brood peak (based on the eastern Ontario phenology), summer C. lucia would be expected to peak on July 11 th on average. A corresponding, although weak, peak occurs in southern Ontario between July 5 th and 14 th (Fig. 5). Assuming similar physiological development rates and parameters for C. neglecta, a summer peak of 750 DD 6 after the June 15th peak would be expected, corresponding to August 5 th . This correlates well with the observed peak between July 30 th and Aug 3 rd (Fig. 5).

Identification and distribution of Canadian Celastrina neglecta
To establish comparative phenotypes of C. neglecta and summer-brood C. lucia, southern Ontario specimens collected during the June flight peak ( Figure 5) were compared to July specimens from eastern Ontario (summer peak; Figure 4). This provided a conservative estimate of phenotypic variation in Celastrina neglecta, which differs in having darker, smaller and more sharply defined ventral spots, brighter white ventral ground colour, more reduced marginal markings, a solid white dorsal hindwing fringe, and more pronounced dark marginal shading of the forewing apex (compare Fig. 10g-k to 10l-o). To define the distribution of C. neglecta, a conservative approach was taken to avoid construing summer C. lucia with first or second generation C. neglecta. For Ontario and Québec, specimens were identified as C. neglecta only if they met two criteria, i.e. matching the C. neglecta phenotype as above, and a collection date between late May and late June, prior to the onset of the summer C. lucia flight. For the Prairies and Maritimes region where flight period is expected to be later compared to Ontario, all available specimens previously identified as C. neglecta were evaluated. Specimens previously identified as C. neglecta from all parts of the Canadian range revealed that true C. neglecta occurs from easternmost Alberta to southern Ontario. Specimens from eastern Ontario, Québec and the Atlantic region match the summer C. lucia phenotype, consistent with the notion that Maritimes Celastrina all represent a single, partially bivoltine species (Maritimes Butterfly Atlas 2015). Two Nova Scotia specimens reared from Aralia (CNC) previously identified as C. serotina (Pavulaan and Wright 2005) were also re-identified as summer brood C. lucia.
In Canada, C. neglecta is sympatric with C. lucia in nearly all parts of the neglecta range. Most summer records from the Prairie Provinces proved to be C. neglecta (Fig.  8), although summer brood C. lucia occur also in southern Manitoba (Fig. 10), and are expected in Saskatchewan based on a single recent record from as far west as Edmonton, Alberta. In Ontario, C. neglecta has a more restricted southern distribution compared to bivoltine C. lucia populations, so far documented to about 44°N (Fig. 8b). The maximum northeastern extent is currently at the eastern edge of the Oak Ridges Moraine (Rice Lake Plains) and the southern Napanee Limestone Plain (Fig. 8b). In southern Ontario it is the most common Celastrina, and both C. lucia (Fig. 9) and C. ladon (Fig. 7) have a more localized occurrence. Field work is needed to definitively establish the northern range limit, especially in the regions of Georgian Bay, Bruce Peninsula, and the Frontenac Arch. No Québec vouchers were located but the species could be expected in regions know for southern species, such as the southern Richelieu River valley and the Lake Champlain region.

Conclusions
The Canadian Celastrina fauna is revised to consist of four species: C. lucia (all provinces and territories), Celastrina neglecta (southern Ontario to eastern Alberta), C. ladon (Carolinian zone of southernmost Ontario), and C. echo (southern British Columbia and southwestern Alberta). From eastern Ontario eastward, what was previously treated as three Celastrina species is revised to a single, facultative bivoltine species, C. lucia.
Adults of C. lucia flying from early to mid-spring, in a relatively prolonged emergence, give rise to a second and possibly a partial third generation in July to September. Larval rearing, phenology, and seasonal emergence patterns show no evidence of C. serotina as a separate gall-feeding species distinct from C. lucia, and C. serotina is therefore removed from the Canadian fauna. Whether or not nominate C. serotina (described from Rhode Island) is a valid species, or simply represents late-emerging C. lucia that utilize cherry galls, needs to be re-evaluated. Molecular markers such as microsatellites could prove to be particularly valuable in advancing the taxonomy of Celastrina, given that the COI barcode marker is taxonomically uninformative here.

Celastrina lucia
Two additional possibilities in the identity of the species here assigned to C. lucia warrant comment. It is conceivable that C. neglecta is present as a univoltine, summer-flying entity that is phenotypically similar to and unrecognized within summer-brood C. lucia. This would require that the June-flying Celastrina in southern Ontario be C. serotina, and that Celastrina neglecta in eastern Ontario overwintering as pupae delay emergence until July. Both of these conditions are improbable; the identity of June Celastrina in southern Ontario is most likely C. neglecta as discussed below, and there are no known temperatezone Lycaenidae that overwinter as pupae and delay emergence until July. Eastern Ontario summer Celastrina also have the appearance of pale C. lucia (Figs 10, 11).
The second possibility is that the eastern Ontario taxon represents a species distinct from nominate C. lucia, that is C. lucia 'of authors' in the sense of Pratt et al. (1994), based on larger size, wing pattern differences, and differing host plant preferences.   This interpretation remains to be thoroughly evaluated, particularly by examining latitudinal gradients of the character traits in question. For now, we favour the simplest taxonomic hypothesis, where this taxon represents C. lucia with facultative bivoltine populations, clinally variable phenotypes and regional host plant preferences.
Although consistently stated to be univoltine in the literature, Celastrina lucia is here interpreted to be facultatively bivoltine (and possibly trivoltine) in southern Canada (Fig. 9), with northern, boreal populations being univoltine. In addition to climatic conditions, voltinism may be regulated by host plant availability (Shapiro 1975), explaining why more southerly populations of C. lucia could be strictly univoltine (Pavulaan 2014). Plasticity in voltinism is perhaps not surprising given that the Eurasian sister species C. argiolus, occupying very similar ecological niches, is also well known to be facultatively bivoltine (e.g. Ebert 1993). Celastrina echo is well-known to be bivoltine in western North America, and some western C. lucia populations can produce a second generation under laboratory conditions (James and Nunnallee 2011). Similar mechanisms of geoclimatically variable voltinism are common and taxonomically widespread in Lepidoptera, although perhaps less prevalent in temperate butterflies. As Celastrina is primarily a tropical group, multivoltinism is likely an ancestral evolutionary trait, with univoltinism a derived trait adaptive for climatic or host plant limitations.
Rearing data indicate that a proportion of spring individuals of Ontario C. lucia enter diapause the following spring (Eberlie 1997;Layberry 2004; this study). Summer observations are 45% fewer than in spring (Fig. 4), suggesting that roughly half of the individuals resulting from the spring brood enter diapause. 59% of pupae reared in 2015 similarly did so. Triggers for facultative bivoltinism are in part environmental, as flight phenology shifts later into the spring with latitudinal climatic amelioration. Warmer spring temperatures as a result of climate change are expected to favour northward expansion of bivoltinism in C. lucia. This was recently documented in Alberta with the first recorded summer brood C. lucia (G. Anweiler, pers. comm; photo examined), in an area with a century of butterfly surveying (Pohl et al. 2009). Celastrina lucia therefore provides an excellent opportunity to study the effects of climate change on developmental thresholds.
Larvae of C. lucia are polyphagous, but show preferences for several genera in different families (Table 3) and feed almost exclusively on flowers and fruits. Celastrina lucia uses a variety of host plants with differing flowering phenologies to span the duration of a relatively lengthy flight period. As part of this dietary strategy, C. lucia also feeds opportunistically on leaf galls of Prunus serotina and P. virginiana, which has been documented in Québec, Ontario and Manitoba, but is likely a geographically more widespread phenomenon.

Celastrina neglecta
In southern Ontario, a third Celastrina species appears in late spring after an initial May flight of both C. lucia and C. ladon. The appearance of this species is too soon after the first flight of Celastrina to represent a second annual generation. Adult wing phenotype is similar to the summer brood of C. lucia, but differs in having darker, smaller and more sharply defined ventral spots, more reduced marginal markings, a solid white dorsal hindwing fringe, and a less evenly checkered forewing fringe (Table 7). The differences between C. neglecta and summer C. lucia requires more study, and the diagnosis and accompanying figures given here should be treated as a guideline for further research rather than a definitive diagnostic tool.
In Ontario, this taxon was recognized as distinct from C. lucia 140 years ago by Saunders (1875), who considered it to be the most common Celastrina in the London area, appearing in late May to early June. Pavulaan and Wright (2005) assigned Saunders' records to C. serotina (although Saunders (1869) states that specimens were reared from larvae found on Cornus). The abundance of this species in the absence of Prunus serotina in southern Ontario (R. Cavasin, pers. comm.), and the larval host plant records discussed below, indicate that this species is not C. serotina. What name to apply to this taxon is however not straight-forward. The differential diagnosis of C. serotina and C. neglecta is based primarily on phenology, voltinism, and to some extent on host plant (Pavulaan and Wright 2005). Pavulaan and Wright (2005) state that neglecta has a single summer flight after that of C. serotina in Canada, but when C. neglecta has a spring flight, it is before that of C. serotina. The phenology of C. neglecta as proposed by Pavulaan and Wright (2005) seems counterintuitive as it states that C. neglecta has a summer flight in the north but then adds an earlier, spring flight southward. Other facultatively bivoltine Lepidoptera generally have additional flights later not earlier in the year. Celastrina neglecta is more intense blue with more white suffusion dorsally, and a weaker ventral maculation pattern compared to C. serotina (Pavulaan and Wright 2005). Of course the name of this species hinges on the identity of the lectotype specimen of C. neglecta, which surprisingly has not been considered in detail. Until this situation can be thoroughly reviewed, the identity of the June/August Celastrina of southern Ontario is most parsimonious with the current concept of C. neglecta. Many southern Ontario specimens are also very similar to the Manitoba taxon argentata (Fletcher), which is currently considered a synonym of C. neglecta (Pelham 2011). The distribution, similar phenotype and phenology of Great Lakes C. neglecta Great Plains argentata, together with Colorado C. humulus Scott and Wright 1999 certainly suggest that these taxa all represent the same species.
Canadian host plant records that are probably attributable to C. neglecta include Ceanothus americanus (based on late June larvae from Northumberland Co., Ontario; Eberlie 1997; 1998; ovipositing female, Northumberland Co., R. Cavasin, photo examined); Cornus amomum Mill. (late June oviposition and larvae at Point Pelee, J. Cossey, photo examined), and Cornus drummondii C.A. Mey (late June to early July larvae from Essex County, J. C. Lucier, Ontario Butterfly Atlas 2015). Host plants of populations in the prairies are completely unknown; both Cornus and Ceanothus are sparse or absent where these populations occur.

Research needs
Surprisingly, there are still many large gaps in our understanding of Celastrina taxonomy and biology. The most urgent need for Canadian Celastrina research is vouchered surveys for C. ladon in southern Ontario, so that potential conservation needs can be established. Regions where C. neglecta, C. lucia and/or C. ladon occur in sympatry provide an excellent opportunity for comparative study, where time series of vouchers are needed to establish diagnostic as well as habitat and host plant differences. Along similar lines, latitudinal transects of voucher series and host use are needed to examine the transition from southern to boreal C. lucia. Lastly, controlled-environment rearing studies of all taxa would establish plasticity in voltinism and developmental requirements and diapause triggers. The use of degree-day modeling could easily be fine-tuned as a useful comparative tool for Celastrina taxa and populations, and to model geographic variation of Celastrina emergence. Dearborn and Westwood (2014) used a similar approach to predict emergence of an endangered skipper.