Taxonomy and biogeography of the Nearctic Raphia Hübner (Lepidoptera, Noctuidae, Raphiinae)

Abstract The taxonomic status and biogeography of the North American Raphia species is reviewed using adult morphology, larval host plants, geographic phenotypic variation, and variation of mtDNA COI barcode sequences. Lack of diagnostic morphological differences, combined with relatively low mtDNA barcode divergences and clinal phenotypic variation in key geographic regions indicate that the six previously recognized species of North American Raphia are best interpreted as parapatric subspecies. Raphia frater abrupta Grote, stat. n., R. f. coloradensis Putnam-Cramer, stat. r., R. f. piazzi Hill, stat. n., and R. f. elbea Smith, stat. n., are accordingly revised to subspecies of R. frater Grote. Type locality restrictions are provided for Raphia abrupta and Raphia frater and a neotype is designated for Raphia frater var. coloradensis.


Introduction
Raphia Hübner is a small genus of the Holarctic region, with a single African species (Poole 1989) questionably congeneric. As the sole genus currently comprising the Raphiinae, the phylogenetic placement of Raphia has an interesting history. Most early In the central Great Plains, a large series of over 60 specimens from northern Nebraska (Cherry Co.) is so variable that scarcely two individuals are alike, varying from the granular, dark grey forewing and white hindwing of ssp. frater, to the even, light grey forewing and slightly fuscous hindwing of R. f. abrupta; an intermediate specimen is shown in Fig. 1i. Some individuals show the blotchy black and grey pattern (with a contrasting black medio-anal shade) characteristic of R. f. coloradensis. A shorter series from Kansas (Riley Co.) falls within the variation of the Nebraska population. Interestingly, Smith (1908) remarked that Denver, Colorado specimens varied more towards ssp. abrupta than ssp. coloradensis. Single specimens from southeastern Montana and southwestern North Dakota have relatively pale forewings, like ssp. coloradensis, but some have fuscous hindwing shading and a dark prothoracic collar like abrupta. The zone of clinal variation may therefore be more extensive than the specimens from the few available sites in the northern Great Plains indicate, so further surveying in the region from southeastern Montana and southwestern North Dakota southward to western Nebraska / Kansas and eastern Colorado (particularly along major river corridors) would be helpful. Such a large transition zone appears to be the result of continuous, flat topography with a single, widespread Populus L. species (P. deltoides Bartr.) that is utilized by both ssp. coloradensis (to the northwest) and ssp. abrupta (to the southeast). This transition zone corresponds closely with the suture zone first proposed by Remington (1968) and recently verified by others (Swenson 2010 and references therein).
The nature of the ssp. abruptafrater interface is somewhat different in the Northeast, and is seemingly more influenced by topography and host plant distribution (Fig. 2); at least three Populus species occur regionally among topography ranging from coastal floodplains to the Appalachian Mountains. Specimens from the Pocono Mtns. of Pennsylvania are R. f. frater, whereas nearby central Maryland (Ann Arundel Co.) specimens (Fig. 1h) show transitional features in having a forewing pattern much like ssp. frater, but with a fuscous hindwing and a darker prothoracic collar characteristic of ssp. abrupta. Coastal Maryland (Montgomery Co.) specimens are typical R. f. abrupta (Fig. 1e). As discussed below, the transition zone between ssp. frater and ssp. abrupta seems to be mediated by habitat and host plant differences, with frater largely associated with aspen (P. tremuloides Michx. and P. grandidentata Michx.; Fig. 2) and ssp. abrupta with cottonwood (P. deltoides; Fig. 2). Study of the populations on either side  of the Ohio River is needed because frater occurs throughout Ohio (Rings et al. 1992), whereas the few northeastern Kentucky specimens that were examined are mostly like ssp. abrupta, but show some ssp. frater traits, including a mostly white hindwing. The Ohio River valley is an important suture zone between other biota, but the relative limits of R. f. frater and ssp. abrupta from the northern Appalachians eastward appear to be further south than recognized suture zones (Swenson 2010).
In southern New Mexico where elbea and coloradensis meet, elbea is known from the Mimbres Mountains (Grant Co.) in the southwest, with the nearest documented coloradensis locations 180 km to the northeast in the Rio Grande valley, and 300 km to the east in Eddy County (Fig. 1). The phenotypic transition between the two taxa is more abrupt in southern New Mexico than it is in the Great Basin where elbea imperceptibly transitions to the very pale Great Basin forms of coloradensis. A series from Twin Falls, Idaho, and some individuals from Leavenworth, Washington (Fig. 1p), are indistinguishable from Arizona elbea (Fig. 1q). Raphia f. elbea therefore grades into R. f. coloradensis in low-elevation habitats of the northern Great Basin. Specimens from the Eddy Co., New Mexico population are most like coloradensis, but some individuals are again indistinguishable from elbea. The mtDNA haplotypes associate this population with coloradensis (Fig. 3), and the available larval hosts are Populus angustifolia James and P. deltoides (National Park Service 2014), but not P. fremonti Wats. with which elbea is most often associated. Analysis of Raphia populations from the lower Rio Grande valley of New Mexico is desirable given the geographically intermediate position between elbea and coloradensis, and the presence of Populus fremonti (Fig. 3).
Populations at the edges of the Great Basin can be extremely variable, much like the situation between abrupta and frater in the north-central Great Plains. Series of specimens from Waterton Lakes, Alberta; Okanagan Valley, British Columbia; Baker County, Oregon; and Leavenworth, Washington range from the typical dark grey frater to pale yellowish coloradensis (Fig. 1). The Leavenworth population is remarkable in that it exhibits phenotypes ranging from frater to coloradensis (Fig 1o) and elbea (Fig 1p).
The geographic structure of California populations is not well documented; typical cinderella occurs from the San Francisco Bay area southward through the Central Valley to Los Angeles Co., but R. frater is apparently absent from southeastern California and the southern Sierra Nevada. Northern California (including the Sierra Nevada) specimens are most like Great Basin coloradensis but the transition from cinderella to this paler form is subtle, with Siskiyou Mountains material appearing intermediate. The Siskiyou Mountains are part of a northern California -southern Oregon suture zone also identified for other flora and fauna (see Swenson 2010 and references therein). Areas in central Texas where abrupta and piazzi meet, and eastern New Mexico/west Texas where the ranges of abrupta, coloradensis and elbea converge, remain unstudied.
Extreme phenotypic variation is therefore the modal geographic pattern at suture zones. In all cases where we examined suture zones between putative taxa, phenotype variation was moderate to extreme, and specimens could not consistently be assigned to existing taxonomic categories. Similarly, mtDNA variation patterns show no evidence of sympatric, reproductively isolated taxa, as discussed below.
Host plants. 72% of the 132 larval collections of R. frater from across Canada summarized by Prentice (1962) came from trembling aspen, 17% from other Populus species (P. balsamifera L., P. trichocarpa Torr. & Gray, P. grandidentata, P. '× canadensis' (Alt.), P. nigra L. var. italica du Roi), and 3% from Salix spp. Three records from white birch and one from alder (both Betulaceae) reported by Prentice are exceptional and possibly accidental. Wagner et al. (2011) consider Populus to be the main hosts, and question the validity of records from birch and alder. The closely-related European species, R. hybris, is restricted to Populus. Raphia larvae possess an unusually large number of proleg crochets (Beck 1996), a trait also seen in the Populus-feeding genus Ufeus Grote (Noctuidae: Noctuinae); this trait is postulated to be an adaptation to maintaining a grip on the leaves of aspen species (Lafontaine and Walsh 2013), which tremble even in slight breezes. Based on these data, Populus, and to a lesser extent Salix L. (both Salicaceae), are certainly the primary and probably the only larval host plants. Salix may be used only incidentally or in certain regions/habitats; the parallel paucity of Raphia and Populus occurrence in the central Appalachian region is notable (Fig. 2). Eight species of Populus occur in North America, divided into four sections: Leucoides Spach (P. heterophylla L.), Aigeiros Duby (P. deltoides, P. fremonti), Tacamahaca Spach (P. balsamifera, P. trichocarpa, P. angustifolia) and Populus L. (P. grandidentata, P. tremuloides). One additional species, P. mexicana Wesmael of central Mexico, is the sole constituent of section Abaso Eckenwalder (Eckenwalder 1996).
Although it is reasonably certain that Raphia frater larvae are Salicaceae specialists, the geographic variation in host use and extent of specialization is not well understood. Nonetheless, it is possible to extrapolate broader host use patterns based on larval collections, host plant distributions and habitat associations. Below, we outline some potential scenarios of host use among R. frater subspecies.
Host plant records for R. f. frater (Prentice 1962) indicate that P. tremuloides is probably the dominant, and certainly the most geographically widespread host (with the caveat that the high proportion of trembling aspen collections may simply reflect sampling bias). In eastern North America, the southern range limit of R. f. frater corresponds closely with the combined southern limits of the two aspen species (section Populus: P. tremuloides and P. grandidentata; Fig. 2). Raphia frater populations in riparian habitats of southern Alberta, where P. tremuloides is scarce or absent, are associated with other Populus species that form a complex zone of hybridization and overlap among four species (P. deltoides, P. balsamifera, P. trichocarpa, and P. angustifolia) along the major river valleys (Brayshaw 1965, Floate 2004. In this region Raphia f. frater phenotypes transition to R. f. coloradensis (Fig. 3). The northernmost extent of P. deltoides and P. deltoides × balsamifera bybrids in the Red Deer River valley at about 52° latitude (Floate 2004) also coincides with the northernmost extent of R. f. coloradensis-like phenotypes; north of there where P. tremuloides is the dominant species of the Aspen Parkland ecoregion, only pure R. f. frater phenotypes occur. Similarly, transitional frater-coloradensis populations occur in southwestern Alberta, southern British Columbia and central Washington at the range edges of P. tremuloides, where P. trichocarpa becomes the dominant Populus (Fig. 3). In northern Labrador, R. f. frater is at its northeastern range limit (not shown), occurring beyond the range of Populus; Salix species are the presumed hosts.
Throughout most of the range of R. f. abrupta, P. deltoides is the only Populus species present. Swamp cottonwood (P. heterophylla) has a small eastern North American range, occurring primarily along the Mississippi and Ohio River valleys and along the Atlantic seaboard (see e.g., Sibley 2009, USGS 2013, so this may serve as a host in some areas. The hosts for the southwest Texas taxon R. f. piazzi are unknown, and may constitute willows rather than Populus, the latter being rare or absent where R. f. piazzi occurs (Fig. 3).
In the Pacific Northwest, R. f. frater is associated with P. tremuloides in northern Washington and British Columbia, with R. f. coloradensis of dry, low-elevation habitats associated with P. trichocarpa (L. Crabo, pers. comm.). Crumb (1956) documented a larval collection from the latter species in south-central Washington. Throughout most of the Pacific Northwest, the only Populus species are P. tremuloides at upper elevations, Figure 4. Neighbour-joining tree and associated sampling sites of mtDNA barcode haplotypes in R. frater. Haplogroup colour corresponds to that of sampling sites. Subspecies assignment based on morphology and sample size is indicated for each haplogroup. Width of triangles is proportional to number of haplotypes, height represents maximum divergence within haplogroup. and P. trichocarpa in low-elevation riparian habitats and drier soils in moist regions (Fig.  3). A third species, P. angustifolia, occurs locally in the eastern parts of the Pacific Northwest, with habitats similar to P. trichocarpa. There are no specific host records for the Californian R. f. cinderella, with both P. trichocarpa and P. fremonti being the most likely hosts. Along the east slopes of the Oregon Cascade Ranges to the north, R. f. coloradensis associates with P. trichocarpa (L. Crabo, pers. comm.) Dyar (1894) cites "poplar" as a foodplant for larvae from Yosemite. Subspecies coloradensis in the Sierra Nevada, to the east of the central Californian range of cinderella, is associated with P. tremuloides.
Arizona populations of R. f. elbea feed on P. fremonti (Crumb 1956;D. Wagner, pers. comm.), and again this is the only available Populus in much of the range of ssp. elbea, excepting the higher elevations in the mountain ranges of central Arizona where P. tremuloides and P. angustifolia occur (Fig. 3). A population of R. f. coloradensis in the Rio Grande valley of central New Mexico is also associated with riparian P. fremonti (Fig. 3). Aspen-associated, R. f. frater-like populations may occur at high elevations in Arizona, similar to the situation in Colorado, but this has not been documented. This raises the interesting possibility that frater occurs at higher elevations, with elbea occurring in the low-elevation floodplain. Surveying at the uppermost elevations of the Chiricahua and Santa Catalina Mountains of southeastern Arizona where aspen occurs have so far not yielded Raphia (BCS unpubl. data), but the more expansive range of P. tremuloides along, for example, the Mogollon Rim is poorly surveyed.
In summary, larval host plant associations of R. frater populations shows some broad congruencies between subspecies and Populus species distributions, but with limited evidence for high host fidelity: range edges of R. frater subspecies generally do not closely follow those of the various Populus hosts, suggesting that Populus availability rather than high host fidelity may be the limiting factor to Raphia distribution, and that climatic and topographic effects have a greater selective influence that does host plant specialization.
To what extent these congruencies reflect common co-evolutionary trajectories, and what factors drive intraspecific divergence, would be a fascinating and fruitful area of study.
Molecular variation. The deepest splits in mtDNA barcode variation (alluded to previously in Lafontaine and Schmidt 2010) segregate North American Raphia into four groups, but only one of these is private to a recognizable taxon (R. f. elbea). None of the remaining subspecies exhibited discrete haplotypic variation. Based on analysis of 192 specimens from localities across the range of R. frater representing all subspecies (Suppl. material 1), haplotypes segregated into five groups: 1) a large group from across most of the eastern, northern and central range portions that includes R. f. frater, R. f. coloradensis, R. f. abrupta and R. f. piazzi, varying by up to ~1.3% (Fig. 4); 2) a discrete group closest to haplogroup 1, consisting of two R. f. piazzi specimens (Fig. 4); 3) a divergent group of geographically disparate samples of Eastern R. f. frater (Ontario, Manitoba) and Californian R. f. cinderella specimens, differing by a minimum of ~2.2% from all other groups (Fig. 4); 4) a group private to R. f. elbea (Arizona, New Mexico, Utah) with a minimum ~1.0% divergence; 5) A group of Californian R. f. cinderella with a minimum divergence of ~1.8% from all other haplotypes.
The combination of California and eastern Canada samples in haplogroup 3 to the exclusion of all others was quite unexpected, given the geographic structuring of other haplogroups. Two haplogroup-3 populations (Bird Hill, Manitoba; Bruce Peninsula, Ontario) also exhibited group 1 haplotypes (Fig. 4), the only sampled populations to yield more than one haplogroup. Representative specimens from these sites were of the same phenotype and from the same sampling event. This haplogroup could therefore be a retained ancestral mtDNA polymorphism, or indicative of Wolbachia-induced mtDNA lineage sorting similar to that documented by Kodandaramaiah et al. (2013). Determining the underlying cause of this interesting variation will require study using nuclear gene markers and Wolbachia assays.
Considering the general lack of taxonomic resolution of North American Raphia in the barcode sequence, and comparing divergences among Palaearctic Raphia as a metric of mean species divergences within the genus, mtDNA variation is most parsimonious with a geographically structured, single-species interpretation. The contrast between often considerably different adult phenotypes and lack of significant mtDNA and morphological differentiation may reflect strong regional selection on bark-cryptic wing patterns, which in turn is dependent on dominant host trees that vary according to regional host preferences.  (1874).

Raphia frater Grote
Diagnosis. Despite variation in adult facies and lack of a particular diagnostic trait, R. frater is recognizable by the combination of a broad, rounded forewing, often conspicuous antemedial and postmedial band, obsolete medial band (rarely faint), an orbicular, reniform and usually also a claviform stigma that are clearly outlined, black shading in the anal angle of the hindwing, and the conspicuously bipectinate male antennae. Pseudopanthea palata (Grote) and Colocasia Ochsenheimer species share some superficial similarities with Raphia frater, but attention to the above-stated characters relative to those in Pseudopanthea McDunnough and Colocasia Ochsenheimer will provide an easy diagnosis of this unique species. Description. Head -Male antennae bipectinate, anterior rami 3× longer than segment length, posterior rami 3.3× longer; female antennae simple; eyes round, with short, sparse interfacetal setae, visible only at high magnification; labial palpus with second segment clothed in long strap-like scales ventrally; third segment 0.6 × length of second segment (when denuded) and smoothly scaled; occiput and frons with mix of grey and black scales, frons with transverse line of black scales; frons rounded and moderately protuberant when denuded. Thorax -vestiture dark grey to yellowish grey, thoracic collar sometimes contrastingly darker than dorsum; tarsi smoothly scaled, with transverse bands of black and light to dark grey; tibia with similar scaling but with faint or indistinct banding; femur with long, shaggy hair-like scales. Abdomen -lacking specialized secondary sexual structures such as coremata; vestiture of smooth, short grey scales; small, rounded dorsal tufts on segments A3, A4 and A5, consisting of densely set spatulate scales. Forewing -ground colour varies from a dark charcoal grey to pale yellowish ochre; antemedial band a parallel-sided, double black line, varying from slightly irregular and rounded to nearly linear, acute, and angled at the cubital vein; medial band obsolete, usually reduced to a black bar or two diffuse lines at costa adjacent to reniform stigma, but band sometimes visible as a faint, diffuse black line extending from bottom of reniform stigma to anal margin; postmedial band a single black line, sinuate and slightly sagittate at veins (often faint or absent in ssp. coloradensis and elbea), expanding to diffuse black patch at costa; orbicular stigma paler than ground colour, with black border and often with a diffuse dark pupil (orbicular often absent entirely in ssp. coloradensis and elbea); reniform stigma paler than ground colour, with a black border (border often lacking in coloradensis and elbea) and a diffuse black central crescent; subterminal band absent, faint, or diffusely sagittate with paler distal edging; terminal area often darker grey than subterminal area. Average size is greatest in subspecies frater, while abrupta and piazzi are smallest; forewing length varies from 16.2 mm and 18.5 mm in male and female R. f. frater to 13.7 mm and 15.2 mm in male and female R. f. abrupta, respectively. Hindwing -ground colour varying from white, white and dusted with fuscous grey (ssp. frater, coloradensis, elbea, cinderella, piazzi), or entirely pale fuscous grey (ssp. abrupta), females with more fuscous than males; crescentic discal spot diffuse or absent; postmedial band faint or absent, although nearly always with a contiguous diffuse black patch at anal angle. Male genitalia (Fig. 5) -uncus slightly compressed dorsoventrally, with slight medial bulge, apex blunt; valva tapering more or less evenly to a rounded point, sacculus poorly differentiated from remaining valva; ampulla long and flattened, 0.7 × length of valva width, projecting mesially; aedeagus stout and sausage shaped, 2.1 × longer than wide; vesica a simple kidney-shaped, unarmed chamber equal in length to aedeagus, tapering gradually into ductus. Female genitalia (Fig. 5) -bursa copulatrix membranous, lacking apparent differential sclerotization, including ostium, ante-and postvaginal plate; ductus bursae a simple rugose tube, 3.3 × longer than diameter, connecting subbasally to corpus bursae; corpus bursae a simple kidney-shaped chamber, slighter wider over apical two-thirds; ductus seminalis arising from basal end of corpus bursae, caudad of ductus bursae; papillae anales short, bluntly rounded, with a broadly joined base, with two types of unusual, modified setae: 1) a dense band of thin, evenly curved setae arising from base of papillae and curving up to caudal margin, outer surface of lobe virtually encircled by a dense setal crown; and 2) highly modified thick, spatulate setae densely set along caudal margin of lobe. Immature stages -larva described by Thaxter (1883) based on eastern specimens, and by Dyar (1894) from Yosemite, California (possibly referable to R. f. cinderella). Illustrations in McCabe (1991) (head capsule and mandible), Wagner (2005) and Wagner et al. (2011). Mature larva stout, tapered only slightly anteriorly, bluish green to apple green with a slightly translucent quality, pinacula yellow, a dorsal transverse yellow band on A1, A5 and A8 extends to just above spiracle; T2 with short horn-like process middorsally, reddish with yellow base; these bands with whitish anterior border, those on A5 and A8 partially bordered with reddish orange; prolegs green, anal prolegs with yellow and reddish orange; head whitish green, usually retracted into T1, ocelli black, labrum whitish; total length 40 to 30 mm. Thaxter (1883) states that male larvae are more slender and smaller. Cocoon tough and firm, incorporating debris; pupa cylindrical with a rounded abdomen, cremaster short but broad and thick, lacking hooks. Eggs laid in small clusters or overlapping in linear groups; early instars much more elongate 'semi-loopers' with A3 and A4 prolegs reduced (Wagner et al. 2011). Larvae rest along midrib of leaf underside. Comparison among larvae of Californian R. frater (Dyar, 1894), R. f. frater, R. f. abrupta and R. f. elbea indicate no discernible differences among these subspecies.
Biology and distribution. Raphia frater occurs in virtually all wooded or shrubby habitats of the boreal region since the larval hosts form a dominant part of most nonconiferous forest types. It can be one of the most common late spring noctuids in aspen-dominated boreal forests of central Canada. In the West it becomes increasingly more restricted to riparian areas, particularly major river systems in drier regions of the Pacific Northwest and the desert of the Southwest. Raphia frater has a nearly transcontinental distribution, absent only from the arctic and most of the subarctic. The records from northern subarctic Labrador are surprising, but are based on three CNC specimens from two localities, so the data appear to be authentic. Handfield (2011) cites records from the northeast shore of the St. Lawrence, but the species is not known from Newfoundland. The range is essentially continuous south to northern Mexico, although very spotty throughout the Atlantic states, and spotty or absent in the central to southern Appalachians. Nominal R. f. frater occurs across the boreal region south to the northeastern States, southern Great Lakes region, and northern Rockies / Pacific Northwest; R. f. abrupta occurs from the Great Plains southward to eastern Texas and eastward to the Atlantic seaboard; R. f. coloradensis occurs from western California to the eastern slope of the Rockies; R. f. elbea occurs from at least southeastern Utah through Arizona, southwestern New Mexico and into Mexico; R. f. cinderella is restricted to western and central California; and R. f. piazzi occurs from the Edwards Plateau into southern Texas. Raphia frater is univoltine across the boreal region and most of the west, with peak flight activity from late May to July. It is bivoltine in the eastern U.S., flying mostly in April to May, and July to August. In the Deep South, R. f. abrupta has three abundance peaks: March, May and a smaller flight (partial third brood?) in September (Brou 2014). Larvae are most common from late July to mid-August in Canada (Prentice 1962).  Grote (1864) simply stated the type locality as "Middle States," and no additional information is available on the holotype label data. We interpret this as referring to the region south of the New England States, and north of the southern States. Given the complex variation of North American R. frater, it is advisable to restrict the type locality. As Grote's material likely originated from the eastern United States, we restrict the Type locality to Mount Pocono, Monroe Co., Pennsylvania, from which we examined typical R. frater frater specimens. Raphia frater and R. abrupta are the oldest available names for this species, and were published simultaneously. As first revisers, we designate frater as the senior name (ICZN, Article 24.2.2). Syn. Saligena personata Walker, 1865 -[BMNH]. Type locality: United States.

Raphia frater frater Grote
Diagnosis and description. The nominal subspecies of R. frater typically has an even, powdery, dark grey forewing ground colour with all of the markings complete, consisting of the antemedial and postmedial band, and the orbicular, reniform and usually the claviform stigmas. Average forewing length is 16.3 mm (n = 9) in males, 18.6 mm in females (n = 9). The male hindwing is white with little or no dusting of black scales in the subterminal area, and with a pronounced, diffusely-edged black patch in the anal angle, this often with an adjacent black line formed by the terminus of the postmedial band; females usually have some fuscous scales on the hindwing, especially on a slight postmedial band. This subspecies generally lacks the form with contrastingly darker medio-anal and costal black patches that is prevalent in R. f. coloradensis, but it does occur rarely even in Atlantic Canada (Fig. 1g). The yellowish-ochre forewing scales typical of R. f. coloradensis are absent. Raphia f. abrupta differs in having a more angulate and linear antemedial band, a paler grey and less powdery-appearing forewing, duskier hindwing, and smaller size. As discussed in the section on Raphia frater, geographically intermediate populations are extremely variable with respect to these traits, and are considered to be transitional between subspecies frater and coloradensis/abrupta, the only two subspecies abutting the range of R. f. frater.
Biology and distribution. Raphia f. frater is primarily a boreal taxon, especially common in aspen (Populus tremuloides and P. grandidentata) dominated forests and the Aspen Parkland ecoregion of the Prairie Provinces. In the East, it extends south of the Great Lakes region into Pennsylvania, Ohio and Indiana, but apparently not southward into the southern Appalachians, which are essentially devoid of Raphia records. The transition zone between R. f. frater and R. f. abrupta extends from Maryland westward roughly along the Ohio River Valley to east-central Missouri, then northwestward through the northern Great Plains. The southeastern range edge of R. f. frater is virtually identical to that of both trembling and bigtooth aspens (Fig. 2). In the West, R. f. frater occurs south along mid-elevation mountain ranges of the Pacific Northwest into Washington, and southward along the Rocky Mountains. Specimens from high elevations in Colorado (Gilpin Co., 9500') and New Mexico (Sangre de Cristo Mtns., 7900') are of the typical frater phenotype, the coloradensis phenotypes occurring at lower elevations.

Raphia frater abrupta Grote, stat. n. Figs 1b-e, 2
Raphia abrupta Grote, 1864 Certila flexuosa Walker, 1865 Type material. Raphia abrupta -female holotype # 7675 [ANSP]. Type locality: not given; here restricted to Sycamore Landing, Seneca, Montgomery Co., Maryland. The female type bears no locality or collector label data, and since this is a widespread, geographically variable taxon, we restrict the type locality to Sycamore Landing, Seneca, Montgomery Co., Maryland; a series in USNM from this locality, collected by D. C. Ferguson, is phenotypically more similar to the female type than specimens from the Great Plains; it is also more likley that the holotype originated from the eastern US rather than the Great Plains, which were not well collected in the mid 1800's.
Certila flexuosa Walker -[BMNH; not examined]. Type locality: North America. Diagnosis and description. Raphia frater abrupta replaces R. f. frater from the central Great Plains eastward to the mid-Atlantic seaboard, and southward to eastern Texas and Florida. It is on average smaller with a more evenly-coloured forewing, a more linear, angulate antemedial band and a fuscous hindwing. Average forewing length is 13.7 mm (n = 9) in males, 15.2 mm in females (n = 9). The thoracic collar is often darker than the dorsal thorax, not concolorous as in R. f. frater. The wing facies of subspecies abrupta is in many ways intermediate between R. f. piazzi of central and southern Texas and R. f. frater to the north, but the exact nature of the interface between abrupta and piazzi in Texas remains unstudied.
Biology and distribution. Subspecies abrupta occurs south of the range of the aspen species favoured by R. f. frater larvae, and its riparian haunts suggest it feeds on eastern cottonwood (Populus deltoides), the only Populus species in much of its range. Swamp cottonwood (P. heterophylla) and willows (Salix spp.) may also be suitable hosts. This subspecies is apparently rare on the Atlantic seaboard and absent altogether in the Appalachians. We examined only a single historical specimen from New Jersey (Trenton), with records north of there assignable to R. f. frater. All Ohio records were attributed to R. f. frater by Rings et al. (1992), although specimens with a pale grey forewing and dusky hindwing, traits of the abrupta phenotype, rarely occur as far north as southernmost Ontario (Toronto) and southeastern Minnesota (Fillmore Co.). Diagnosis and description. Raphia frater piazzi is the least-known member of the group with a restricted distribution in central and southern Texas. Most similar in size and facies to R. f. abrupta, it is distinguished from that subspecies by the paler, more evenly grey forewing with sharper transverse lines than in R. f. abrupta. The biology and biogeographic relationship to R. f. abrupta, which occurs to the northeast of piazzi's range, is not known, and very few specimens of this taxon are present in collec-tions. An additional enigma is whether or not Rio Grande piazzi populations interact with the vastly-different looking Sonoran R. f. elbea.

Raphia frater piazzi Hill
Biology and distribution. Described from southernmost Texas, this subspecies is otherwise known only from the Edwards Plateau region; a single specimen from Sinton County to the southeast is phenotypically intermediate between abrupta and piazzi, but clearly more field work is needed to establish the limits of both subspecies. mtDNA barcode data of three piazzi specimens (Sinton Co. and Zavalla Co.) are very similar to the haplotypes of R. f. abrupta, R. f. frater, and R. f. coloradensis. Remarks. We were unable to obtain DNA sequence from topotypical specimens of piazzi from the lower Rio Grande near Brownsville, Texas. The unique haplotype of the Edwards Plateau specimens (Fig. 4) may represent nominal piazzi, but could equally represent a unique genetic lineage from the Edwards Plateau, with its unique fauna much of which is not shared with the Rio Grande fauna. Diagnosis and description. Within the range of coloradensis, specimens identical to the typical boreal R. f. frater are often present; in the most arid parts of the range of coloradensis in the southern Great Basin, coloradensis is more consistently pale ochre yellow with obsolete transverse lines and diffuse black costal/reniform blotches, overall very similar to elbea, but with less pronounced costal and reniform dark patches. Average forewing length is 14.9 mm (n = 9) in males, 16.8 mm in females (n = 6).

Raphia frater coloradensis
Biology and distribution. This subspecies occurs from southernmost British Columbia / Alberta to New Mexico, Utah, and California. It is most commonly associated with riparian, low-elevation habitats. Northern populations fly from late May to July in a single generation. Flight dates spanning from May into August in the Great Basin and Southern Rocky mountain region indicate a second or partial second generation.
Remarks. Raphia frater coloradensis is the most weakly-differentiated subspecies, and may simply be an ecologically induced phenotype of R. f. frater that occurs in the warmer, drier regions of the West. Several populations, spanning a large geographical area, have been identified that exhibit a large range of phenotypic variation, as discussed above in the 'Morphology' section. Specimens from Siskiyou Co., California and the east slope of the northern Sierra Nevada (Sierra Co.) are phenotypically very similar to Great Basin coloradensis, and we therefore treat pallula as a junior subjective synonym. DNA barcodes of two specimens from the northern Sierra Nevada (Sierra Co.) belonged to the frater-coloradensis-abrupta haplogroup (Fig. 4).
Diagnosis and description. Raphia frater cinderella is a Californian subspecies that is similar in size and colour to R. f. coloradensis, but with a more diffuse, poorly contrasting forewing pattern that usually lacks the pronouncedly darker reniform and costal dark patches. The forewing ground colour is also pale powdery grey, not pale ochre as it often is in coloradensis. The two taxa appear to intergrade in the Siskiyous and northern Sierra Nevada.
Biology and distribution. The range of this subspecies is restricted to central and southern California west of the Sierra Nevada. Fremont Cottonwood and willows are the most likely larval hosts, although records specific to this subspecies are lacking. Most collection dates are from June; Records from Stanislaus Co. for April -May and July may indicate a second generation.
Diagnosis and description. Raphia frater elbea is most similar to the pale yellowish-ochre forms of R. f. coloradensis, but differ from that subspecies in having both the costal and reniform dark patches more prominent; when present, the black medio-anal patch is also darker and more elongate; additionally, R. f. elbea appears to exhibit a unique, divergent mtDNA haplotype group.
Biology and distribution. This subspecies occurs from southeastern Utah and western New Mexico southward through Arizona into northern Mexico. In southeastern Arizona it occurs in riparian areas in association with the larval host, Populus fremonti. Flight records are from February to October, with most being from March to May and August to September, indicating at least two generations annually.

Conclusions
The North American Raphia populations exhibit considerable geographic variation in phenotype, previously segregated into six species. Despite these geographically structured phenotypic differences, diagnostic morphological differences in genitalia and larvae are not evident. Scrutiny of geographic contact zones between putative taxa revealed populations with extensive phenotypic and conservative molecular variation, rather than bimodal phenotypic variation coupled with deep molecular divergences that would be expected for sympatric, reproductively isolated taxa. Raphia frater larvae are not highly restricted to a host species or genus, but do specialize on Populus and Salix, with a pattern of regional host availability and possibly also preference. Differences in host plant suitability among the various species of Salicaceae remain unstudied. Assessment of morphology, mtDNA variation, and biogeography therefore leads us to conclude that the geographic segregates of North American Raphia are best treated as subspecies of a single species. The regional adaptation to habitats representing nearly all North American biomes, combined with relatively discrete geographic ranges of unique adult phenotypes, suggest a pattern of young or incipient species in the R. frater group.
The taxonomy and biogeography of the North American Raphia populations is a complex interplay between topography, host plant use, phenotypic variation and evolutionary history. This study is only the first attempt at a better understanding of this interesting group. Many questions remain unanswered: what are the exact geospatial and host plant patterns of the contact zone between R. f. abrupta and R. f. frater? Is there geographic overlap with altitudinal segregation in the West between aspen-feeding frater and cottonwood feeding elbea? Does the mtDNA haplogroup 3 represent Wolbachia infection? Do the lower Rio Grande / Edwards Plateau piazzi populations grade into abrupta? Raphia would provide a fertile area of study in understanding large-scale patterns of host plant use and biogeography of a widely distributed continental Lepidopteran. and/or data used in this study, and Jocelyn Gill for preparing the specimen images. Discussions with Lars Crabo and Don Lafontaine on Raphia biogeography and species limits provided insight and guidance during this project. Evgeny Zakharov, Paul Hebert and other members of the Barcode of Life Project at the University of Guelph, Ontario, Canada, provided DNA data. Molecular analyses were carried out through grants from the National Science and Engineering Research Council of Canada and Genome Canada through the Ontario Genomics Institute.