Although we found 10% fewer species overall compared to over a century of historical records assembled by Donnelly (2004a,b,c), every species we did not encounter was already very rare or a suspected vagrant. This amounted to roughly equal proportions within each suborder and these apparent range retractions were species specific and not phylogenetically constrained. We discovered five new species and there is increasing evidence pointing to ongoing marked distributional flux of odonate populations owing to climate change (Hassall & Thompson, 2008; Corser, 2010; Ott, 2010), and it is clear that extant groups of these mobile insects have traversed huge spans of time and space across the earth over the past few 100 Myr (Coope, 2004; Grimaldi & Engel, 2005). Below, we explore the nature of this northeastern NA Odonata hotspot in the context of potential causes of high levels of insect diversity (Mayhew, 2007).
Tropical and temperate conservatism
We found evidence that older groups demonstrate temperate, as opposed to tropical conservatism as Weins et al. (2006) predicted, and furthermore regions that they have inhabited the longest – in this case the freshwaters of the Appalachian Mountain forests in Eastern/central NA – contain the highest richness levels while serving as refugia during episodes of deteriorating climate. As outlined by Carle and Cook (1984) and Carle (1995) in the case of the basal dragonfly families (Fig. 3), we found that temperate Eastern/central NA has been a major evolutionary repository of global Anisopteran diversity since at least Mesozoic times > 180 Ma. In fact, Dillon and Robinson (2009) concluded from > 25 years of careful study that certain living freshwater faunal elements of the Appalachians might even be relicts of Palaeozoic era watersheds.
On the other hand, Zygoptera in NYS demonstrate the more typical post-Eocene tropical conservatism pattern (Weins & Donoghue, 2004; Weins et al., 2006) and these odonates are relative newcomers to Eastern/central NA compared to many of the much older endemic dragonflies (Figs 2 and 3). This alternate subordinal pattern might not occur in Eurasia, however (Heiser & Schmitt, 2013), where the signature of historical biogeographical patterning appeared to be weak at broad regional (100 000s km2) scales (Rakosy et al., 2012). Despite their tiny size, certain damselflies' remarkable colonisation abilities (e.g. Sherratt & Beatty, 2005) seem to obscure the fact that although they are primarily tropical-originated insects, certain groups readily colonise and rapidly radiate in temperate and even boreal regions (Pritchard, 2008). Nevertheless, many of these younger lineages might well be ‘evolutionary dead ends’ as it is the pruning of these (often more lentic) extinction-prone clades that yields monotypics having larger range sizes and favouring long-term persistence (Jansson & Dynesius, 2002; Hadly et al., 2009).
Recent post-glacial recolonisation, cryptic northern refugia, and sexual selection
We found the same recent origins as Beatty and Beatty (1968) and NYS has clearly been repeatedly recolonised along pre-Pleistocene invasion routes (Carle, 1995), primarily from refugia in unglaciated nearby forested realms just to the south. A broad north(east)ward recolonisation pattern along either side of the Appalachian uplands is evident (Fig. 1) being parallel to the modern prevailing wind direction – yet these are also well-known pathways for many organisms in Eastern/central NA (Soltis et al., 2006). This strong Pleistocene imprint is also found across Eurasia (Sternberg, 1998; Kosterin, 2005) and post-glacial mixing of separate temperate, tropical, and boreal faunas has also been identified in other Odonata hotspots (Heiser & Schmitt, 2010; Morrone, 2010). At the same time, species' responses to repeated glacial cycles have been idiosyncratic, resulting in the highly complex distribution patterns seen today (Soltis et al., 2006; Stewart et al., 2010).
Our fine-scale distribution maps in White et al. (2010) depict this same intra-specific pattern. For example, many populations of the same putative species have recolonised NYS since the LGM from separate glacial refugia primarily along two different pathways: an eastern coastal route up the Hudson River valley (Corser, 2010) continuing northward into New England and eastern Canada, and another from the southwest, up the Ohio River valley into southwestern Ontario and NYS and around the Great Lakes (Fig. 1). This has created a suture zone where hybridisation and hidden cryptic diversity would be expected (Hewitt, 2004; Swenson & Howard, 2005), and indeed ongoing odonate hybridisation is well-documented in the vicinity of this broad contact zone (Donnelly, 2003, 2004c).
This landscape has thus played a pivotal role in serving as both a conduit, as well as generator of diversity, where odonate communities were continuously disassembled and reassembled into new collections of glacial races (McPeek & Gavrilets, 2006). McPeek and colleagues' extensive study on Enallagma and Lestes damselflies in northeastern NA pointed to relatively recent (100 000 years) explosive radiations, incipient species formation with numerous examples of morphologically distinct subspecies, hybrid zones, and secondary contact among differentiated forms showing demographic signatures of major bottlenecks, range fragmentation, and range expansions caused by repeated Pleistocene glacial cycles (Turgeon et al., 2005).
New species that came into being during these glacial cycles colonised newly opened, well-watered terrain, chasing the retreating glacier north (Siepielski et al., 2010), some all the way to Eurasia. In eastern NA, the transition from a temperate to a boreal forest-dominated odonate assemblage lies near the terminus of the Appalachian Mountains in the Canadian Maritimes (Larson & Colbo, 1983). NYS lies along the rear edge of this assemblage, with these boreal species mainly occurring in regions of the state with the shortest growing seasons. Likewise, most of the truly boreal NYS odonates on the rear edge of their often large ranges well north into Canada (i.e. many Somatochlora) are often ranked highly for conservation concern in northeastern NA and elsewhere (Bried & Mazzacano, 2010; DeKnijf et al., 2011).
Surprisingly, Beringia appears not to have not played a significant role as a LGM odonate refugia; Cannings and Cannings (1994) fingered just one species (Somatochlora sahlbergi) as a Beringian resident during the LGM finding instead that the boreal elements dominating the fauna were mostly recent arrivals from the southeast. This implies that the boreal assemblage might be derived from the temperate component (Belyshev & Kharitonov, 1978; Cannings & Cannings, 1994), and repeated isolation of temperate species in cryptic northern refugia during repeated glacial cycles can play an important role in generating certain boreal species over a couple of million generations (Barraclough & Vogler, 2002; Weir & Schluter, 2004; Carstens & Knowles, 2007). Our data support this view because there is a preponderance of NA range centres for several groups that clusters just south of the approximate latitude of the LGM in NYS (Fig. 4).
McPeek et al. (2010) provide an overview of how sexual selection on male genitalia has promoted the recent diversification of Enallagma damselflies in northeastern NA, and Svensson (2012) used a damselfly example to highlight the predominance of non-ecological speciation mechanisms, such as sexual selection and thermal adaptation, suggesting that reproductive isolation often precedes ecological (habitat) differentiation in these insects. Our dated phylogenetic approach (Fig. 2) also uncovered a role for Tertiary vicariance in the deeper diversification of Zygoptera in NA. For example, the opening of the North Atlantic ocean ~ 30 Ma facilitated generic-level diversity in many insects (Noonan, 1988), including the establishment of the Coenagrioninae in Eastern/central NA (Fig. 2; Guan et al., 2013).
Although much understudied compared to Zygoptera, incipient species formation arising from sexual selection in the Anisoptera is also rampant (Misof, 2002). For most dragonflies, however, diversification patterns are much older than Plio–Pleistocene because even the most derived groups like Leucorrhinia and Libellula show complex histories of repeated episodes of dispersal, vicariance, and radiation across the Holarctic, back and forth between NA and Eurasia over the past 10 to 100 Ma (Fig. 3; Artiss, 2004; Kosterin, 2005). Even more ancient supercontinental patterns of vicariance and dispersal define the biogeographical patterns for these old Holarctic groups, involving land bridges, widespread climate change, and repeated patterns of dispersal, radiation, and subsequent extinction of lineages (Sanmartin et al., 2001). Nevertheless, certain groups such as the Sympetrum dragonflies demonstrate many of the same more recent post-glacial fragmentation patterns in the vicinity of NYS as do some of the damselflies (e.g. Donnelly, 2004b; Paulson, 2011).
Evolutionary role of temperate forested ecosystems and conservation implications
Continental-scale distributions and dragonfly richness patterns are known to conform to the water–energy hypothesis (Keil et al., 2008), highlighting the role of forested ecosystems in promoting high levels of both temperate and tropical odonate diversity (Paulson, 2006). Forested ecosystems provide the landscape matrix within which both lentic and lotic aquatic oviposition sites are embedded, and the keystone role that forests play in the maintenance of healthy aquatic systems where significant forest regrowth has occurred like NYS is plain (Huntington et al., 2009). Intact mature forests are vital to the viability of odonate populations because they serve as thermoregulatory refugia where pre-reproductive adults feed and attain sexual maturity, greatly enhancing their survival rates (Corbet, 2006). Rith-Najarian (1998) empirically demonstrated how intact older forests foster both larval and adult life histories resulting in greater dragonfly species abundance and diversity in the landscape manifesting as regional patterns of species density.
Arid subtropical climates, even at significantly lower latitudes in NA (Stevens & Bailowitz, 2009) support an order of magnitude lower odonate diversity than NYS. Similarly, over the past few million years as the tropical forests of Africa experienced increasing aridification, the African odonate fauna became impoverished, favouring the evolution of widespread lentic generalists (Damm et al., 2010). On the other hand, in neotropical forests, site- and landscape-scale diversity of Odonata can reach astonishingly high levels (Paulson, 2006; Gonzalez Soriano et al., 2011). At broader scales though, two of tropical Mexico's best surveyed and highest diversity states – Chiapas and Veracruz – that together are a bit larger than NYS, contain nearly identical concentrations of species richness as NYS (Gonzalez Soriano & Novelo Gutierrez, 2007).
Does the evolution of both Odonata and temperate forested ecosystems bear this out? All the modern families of Odonata have been in existence since the Mesozoic era (Grimaldi & Engel, 2005), and the basal family members of our estimated tree are all much older than 100 Ma (Fig. 3). Carle and Cook (1984) and Carle (1995) hypothesised that they had originated in cool groundwater-fed seeps and streams in the Gymnosperm forests of the mid-latitudes of eastern NA. Pangaea was breaking apart at this time and the NA land mass had just broken away from EurAfrica, and NYS would have been at a slightly lower latitude (~ 35°N) than today (Noonan, 1988; Sanmartin et al., 2001) with fluctuating temperate palaeoclimates (Dera et al., 2011).
This predates the rise of Angiosperms during the Cretaceous period (< 100 Ma) and the deciduousness that is the hallmark of these forests is a reflection of the increasing seasonality of the climate (Wolf, 1987). Their pre-adapted cold-hardy aquatic larval stage must have buffered certain odonates from ever more extreme temperature fluctuations post-Eocene (Archibald et al., 2010), and these detritus-based groundwater-fed ecosystems are known to offer an unparalleled buffering effect even during extreme environmental deterioration leading to mass extinctions (Robertson et al., 2013). Although fossil preservation is low in the humid erosional forests of Eastern/central NA (Dillon & Robinson, 2009), there are very few stem group Odonata (extinct lineages) in NA compared to other parts of the globe and the oldest Odonata fossil evidence from NA that we could find dates to around the Palaeocene ~ 60 Ma (Wighton & Wilson, 1986). Thus, we cannot rule out relatively lower extinction rates as a contributing factor to the high levels of forested freshwater diversity maintained here for aeons.
Our 5-year atlas detected range margin shifts for a number of odonates, but range contractions leading to species' losses from a region (extirpations) are notoriously difficult to document. Where sufficient fine-scale distribution data are available, the trend in Europe is clearly towards range expansion of warm-adapted tropical odonates and retraction of the more cold-adapted boreal species as the climate warms (Hassall & Thompson, 2008; Ott, 2010). This pattern has also been well-documented for other European insects, but comparable evidence in NA had until recently been lacking. New trend data for 100 butterfly species gathered by citizen scientists in neighbouring Massachusetts between 1992 and 2010 have found this same process of large-scale climate-driven insect species reshuffling in northeastern NA (Breed et al., 2013); we anticipate seeing similar results emerge in the near future.