**4. Discussion**

## *4.1. Biogeographic and Phylogenetic Interpretations*

We estimated the dates of divergence from mainland populations for Pacific Island anoles from molecular data. The range of estimated dates is consistent with past findings: the ancestors of the included species have a long history on the mainland [20,21,24], are derived from clades with disparate physical and temporal origins. The genus *Dactyloa* represents one of the first splits in the anole tree and arose in the northern South America, while *Norops* (+*Ctenonotus*) is one of the last clades to diverge and occurred approximately 20 million years later [24] in the Caribbean, followed by dispersal to Central America before dispersing south into South America [21].

We interpret our data as follows, beginning with the multi-species island of Gorgona. Isla Gorgona has a complex geologic history leading to its aerial emergence potentially by 27 Mya [50]. The island is close to the mainland (although currently separated by a deep channel) and because of previous contact with the mainland, it could have been easily colonized early by overland dispersal prior to its separation from the mainland, or via overwater dispersal when the sea levels were lower, causing the landmasses to be more easily reachable. Three of the five Isla Gorgona species do not appear to di ffer morphologically from their mainland counterparts. However, denser sampling in mainland Colombia and Ecuador might be able to elucidate closer relationships than our present sampling allows (see *N. urraoi* example, [41]). *Dactyloa chocorum* occurs both on mainland South America and on Isla Gorgona and there have been no identified morphological distinctions recorded to date. *Dactyloa gorgonae* only di ffers from mainland *D. chloris* in being solid blue. *Dactyloa princeps* was always believed to be closely related to *D. frenata* with some investigators debating whether any features distinguish the two, unpubl. per [35,79,80], although our phylogenetic estimates support evolutionary distinctness. The lack of obvious morphological di fferences from mainland relatives may be due to low selection pressure, suggesting the environment on the island is strongly similar to the mainland environment from which it is derived, or to a recent arrival on the island that has ye<sup>t</sup> to produce substantial change.

The two *Norops* species' colonization of Isla Gorgona are even more interesting considering the lack of solid hypotheses regarding the origin and biogeography of *Norops* as a whole. Early hypotheses suggested that South American *Norops* couldn't have invaded from Central America until after the closure of the Panamanian Portal which was believed to have occurred ~3.5 Mya [10]. However, the timing of this event is believed to have occurred multiple times with several estimates at 23–25 Mya ([81,82] with some controversy on this date reviewed in Elmer et al. [83]), 15 Mya [84], and 3.5 Mya [85,86]. More recent work has postulated that perhaps the *Norops* clade was widespread across Central America and South America before the continents were separated by the eastward movement of the proto-Greater Antillean land arc [21]. The estimated dates of divergence for *N. parvauritus* (3.2 Mya) and coalescence for *N. medemi* (6.6 Mya, Table 1) would lend support for the Nicholson et al. [21] biogeographic hypothesis, although more testing is required to investigate that hypothesis. Our sampling of Isla Gorgona was low because of the cryptic nature of some species combined with governmen<sup>t</sup> permitting restrictions, so these dating estimates are tentative and should be tested in future with additional data. However, our findings of habitat partitioning among three of the Isla Gorgona species is of interest and warrants further study.

The biogeographic scenarios for the two solitary anoles (*D. agassizi*—Isla Malpelo; *N. townsendi* — Isla Cocos) are more di fficult to interpret. In both cases, the estimated timing of divergence from mainland relatives (*D. agassizi:* 22.8 Mya; *N. townsendi*: 7.6 Mya) considerably predate the islands' estimated origin (Malpelo: 15–17 Mya; Cocos: 1.9–2.4 Mya). One explanation for both solitary anoles could be that each species originated on the mainland prior to dispersal to each island, followed by subsequent extinction of the species on the mainland, a conclusion also reached by Román-Palacios et al. [24]. The route of dispersal to the islands from the mainland appears to be fairly straightforward from the standpoint of currents, as currents move virtually straight to each

island from the mainland at regular intervals throughout the year [29,50], making over-water dispersal plausible. For many over-water dispersal scenarios it may seem improbable for anoles to disperse over grea<sup>t</sup> expanses of water, but other species made the same trip and colonized the same islands, and similar dispersal events have been observed and hypothesized for anoles and other lizard fauna in the Caribbean [19,67,87].

Our calculated divergence estimate for *D. agassizi* is similar to Román-Palacios et al. [24], who found that *D. agassizi* and its sister species, *D. insignis*, diverged during the Oligocene, 22–26 Ma. We recovered *D. agassizi* sister to a sister-pair of *D. insignis* and *D. microtus* with a mean divergence of 31 (fossil 95% 22–38) or 29 Ma (rate 95% 23–34). Román-Palacios et al. [24] did not include *D. microtus* in their phylogeny, so this is not a contradiction of their topology, although our dates sugges<sup>t</sup> an older divergence, which is concurrent with the assessment that rate-based methods can overestimate the ages of shallower nodes if gene saturation is in effect [72]. Poe et al. [20] placed *D. agassizi* as sister to a clade of 14 other species of *Dactyloa*, but did not include *D. insignis* in their analysis. They also included *N. townsendi* within a group of anoles aligned with *N. trachyderma*. We note that Poe et al. [20] also included other loci in their analysis, which might account for the differences seen here.

Prior to this study, no one had published an estimated age of colonization for *N. townsendi*. Our analyses sugges<sup>t</sup> a split within *N. townsendi* around 3.5 Mya that separates individuals on Isla Cocos from those on Manuelita Islet (Figure 4). This is problematic in that the divergence predates the origin of the island, but our limitations on sample collection might be influencing our results. A more fine-scale sampling effort could provide a more reliable estimate of the coalescence between the two lineages (incidentally, the Manuelita population was significantly larger than any of the three populations on Isla Cocos, Table 4). The phylogenetic placement of *N. townsendi* was unknown until recently but was presumed to be within the mainland *Norops* clade [88,89]. Our analyses also placed it sister to *N. poecilopus*, a semi-aquatic anole from Panamá (100% posterior probability support), concurring with other recent topologies [21]. This clade also contains two additional semi-aquatic species (*N. lionotus* and *N. oxylophus*). If *N. townsendi* evolved from a semi-aquatic ancestor, it may have had more opportunities for rafting, and also makes biogeographic sense, as the currents could have carried an ancestor from Panama to Isla Cocos. Other semi-aquatic anoles (*N. lynchi*, *N. macrolepis*, *N. rivalis*) occur in coastal South America that are missing from our phylogeny, providing other possible relatives. If one of these shares a more recent common ancestor with *N. townsendi*, it may result in a more recent estimate of colonization closer to the origin of Isla Cocos.

The oceanic currents may have played a significant role in the colonization of the islands in the eastern Pacific region by anoles and other taxa [35,90,91]. El Niño Southern Oscillation (ENSO) cycles are believed to have played an important role in island colonization in the past three to five million years [92,93], which could explain dispersal events to Isla Cocos and even the Galápagos Archipelago [3]. As suggested by Román-Palacios et al. [24], this is too recent to influence the divergence of *D. agassizi*, but the Colombian Current as part of the Humboldt Current has been active for over 100 million years [94], and likely was an important factor in the colonization of Isla Malpelo by all inhabitants. For both *D. agassizi* and *N. townsendi*, divergences predating island emergence strongly sugges<sup>t</sup> dispersal from the mainland (likely lower Central America given their closest ancestor as suggested by our phylogenetic reconstruction). This could have taken place via island hopping, possibly using seamounts that were previously aerial [24,95–97]. Recent work in the ETP suggests such a scenario for previously emergen<sup>t</sup> landmasses that are part of the Galápagos archipelago [98], therefore, we cannot discount the possibility that there are other seamounts in the ETP that were previously emergen<sup>t</sup> and could have served as the "stepping-stones" to Isla Cocos and/or Malpelo.

## *4.2. Taxonomic Implications*

Our findings support Armstead et al. [55] in their recognition of *Norops parvauritus* as a distinct species from *N. biporcatus*. The age of the divide between the two species is older than many other species pairs, so we conclude this taxonomic change was warranted. Our inclusion of a specimen from

Isla Gorgona (the type locality) was highly important to confirm the specific status of *N. parvauritus*. Fine-scale sampling throughout Mesoamerica also suggests geographic structure in *N. biporcatus*, with some division aligning with the continental divide in Costa Rica and the Nicaraguan highlands (Figure 4, Supplementary Figures S1 and S2).

Given the placement of the Isla Gorgona *N. parvauritus* as sister to the rest of its species cannot rule out the possibility of *N. parvauritus* first colonizing Isla Gorgona and then back-colonizing the mainland after ancestral *N. biporcatus* populations had been extirpated. In this way, Isla Gorgona may have served as a refugium for at least one species, but perhaps also for *N. medemi*. While *N. biporcatus* is still wide-ranging for Neotropical anoles, the recently elevated *N. parvauritus* represents what would have been a disjunct portion of its range. This pattern may carry over to other lineages of Neotropical reptiles as we note that there are also disjunct populations of the colubrid snake *Leptophis ahaetulla* in Northwest Ecuador and Isla Gorgona [99], which are ye<sup>t</sup> to be analyzed in a phylogeographic context.

## *4.3. Natural History Observations*

Both *D. agassizi* and *N. townsendi* can be described as ecological generalists unlike solitary Lesser Antillean anoles [100]. *Norops townsendi* appears to clearly conform to this generalization, given that is was found ubiquitously on Isla Cocos in terms of perch height, diameter, and type, as briefly noted by Carpenter [38]. *Dactyloa agassizi* has distinctly unique behavioral characteristics that challenge the original meaning of generalist as applied to the Lesser Antillean anoles. The food web of Isla Malpelo appears to be very tight given the lack of vegetation and scarcity of obvious food resources for all the island inhabitants. Everything seems to revolve around the birds that visit the island, particularly the Nazca Booby (*Sula granti*). Any food dropped by the birds, and all waste products were immediately consumed by all *D. agassizi*, as well as *Diploglossus millepunctatus* lizards and *Geocarcinus malpilensis* crabs which congregate en masse around these resources. The *Geocarcinus* and *Diploglossus* attempt to capture any *D. agassizi* that venture too close [34], pers. obs., and the crabs will also consume eggs of all species [34]. Rand et al. [34] and Wolda [31] reported that *D. agassizi* ate both insects and seeds, and that individuals would quickly consume potential food items revealed when rocks were turned over. All *D. agassizi* departed from typical anole behavior in being very curious, climbing on, and licking the observers (clearly investigating food possibilities), never displaying any defensive behaviors or territoriality. Therefore, insofar as their environment allowed them, *D. agassizi* appeared to be an ecological generalist. Isla Malpelo is unlike most other islands that are home to anoles, so it is not surprising that *D. agassizi* departs from the general patterns observed for solitary Caribbean anoles.

## *4.4. Conservation of Island Fauna*

While these Pacific Island anoles represent novel diversity unseen elsewhere, it is fortunate that all three islands are strictly controlled and monitored, virtually ensuring the conservation of the species. Tourists are somewhat rare to both Islas Cocos and Malpelo, and mostly consist of individuals seeking to scuba dive in the waters around the islands. It requires special permission to step onto the islands and is virtually impossible to be allowed onto Malpelo unless undertaking research. More tourists visit Isla Gorgona, which is also considerably closer to the mainland and faster to visit. Even then, they are only allowed to walk along trails accompanied by a guide and walk within the grounds of the former prison which is slowly being taken over by the forest. These strict controls are highly beneficial to the conservation of each island's flora and fauna and no development of any kind is allowed. Thus, the future seems cautiously bright for the conservation status of these islands.

Only one species (*D. agassizi*) has estimates of population size [32], suggesting 140,000–206,000 individuals. None of the seven species of anoles discussed in this study has received any IUCN designation. However, we currently have no reason to believe any of these species/populations is currently at a high risk of extinction/extirpation because of the restricted access of these islands by the Colombian and Costa Rican governments. It is important that these policies continue, as island species are generally more predisposed to extirpation because of high endemism and smaller population sizes. [6]. In particular, tropical islands should be a priority in conservation since they serve as hotspots for biodiversity in many clades [101,102]. Anthropogenic disturbances have contributed to the extinction of other oceanic taxa [103,104], so the protection of these islands and their native ecosystems needs to continue for the conservation of not just these anoles but all flora and fauna on these islands.
