dispersal in all clades of Coeligena that influenced, and are still influencing, the current distribution and differentiation of the extant taxa.
Figure 3.2. Phylogenetic trees of the three examined Coeligena clades. The numbers at the top of the branches represent the species, the slashes the ancestors. 1: C. coeligena; 2: C. wilsoni; 3: C. prunellei; 4: C. iris; 5:
C.aurora; 6: C. violifer; 12: C. phalerata; 13: C. helianthea; 14: C. (=Pterophanes). cyanopterus; 15: C.
bonapartei; 16: C. eos; 17: C. lutetiae; 22: C. conradii; 23: C. insectivora; 24: C. torquata; 25: C. inca.
1. (C. violifer, (Coeligena prunellei, (C. coeligena, C.wilsoni),(C. iris, C. aurora));
2. ((Coeligena (=Pterophanes) cyanopterus, (Coeligena phalerata, C. helianthea)), (C.
lutetiae, (C. eos, C. bonapartei)); and
3. Coeligena inca, (C. torquata, (C. insectivora, C. conradii))).
7 9 18 20a
8 19 20
10
11 21
22 23 24
26
27
28
25
14 15 16 17
5 6 12 13
1 2 3 4
Figure 3.3.South America map showing the areas employed by the BPA analysis (taken from Simpson 1975).
The polygons represent the geological units considered in this study. The codification used for the area cladogram is: A1: Sierra Nevada de Santa Marta, A2: East Cordillera of Colombia, A3: Central Andes of Colombia and East Andes of Ecuador, A4: Western Cordillera of Colombia and Ecuador, A5: Cordillera Occidental of Peru, A6: Cordillera Oriental of Peru, and A7: Bolivian Altiplano.
Coeligena (=Ensifera) ensifera was excluded from the analysis because it is not part of any monphyletic sub-clade, appearing as the outgroup of the first and second clades. The genus Patagona was also excluded since it does not belong to the major monophyletic group. These exclusions did not affect the results due to the wide distribution of both genera (actually two species, being both monotypic), found in all the areas examined. This redundant occurrence would be signalled by the analysis as homoplasy on every single branch of the area cladogram.
Figure 3.4. shows the same cladogram with the areas examined, constructed on the phylogenetic relationships of the species included in the three clades of Coeligena. The internal branches are numbered, the number at the top of each terminal branch corresponding to one species. Tab. 3.2 lists the binary codes for members of each clade for each area.
Table 3.2. Primary matrix listing the geographical distribution of three clades of Coeligena – Pterophanes taxa, along with the binary codes representing the phylogenetic relationships for all three clades. ‘?’ = species missing from the area. For areas explanation see Fig. 3.2.
The primary BPA analysis for the three clades (based on the area-taxa matrix from Tab. 3.2) produced only one tree, with a consistency of 80% (tree-length = 35 steps, Fig. 3.6). The primary BPA provides strong support for a vicariance relationship between the areas A5 (Cordillera Occidental of Peru), A4 (Western Cordillera of Colombia and Ecuador) and A1 (Sierra Nevada de Santa Marta) in a south – north direction, moderate support for those areas plus A3 (Central Andes of Colombia and Eastern Andes of Ecuador) and A2 (Eastern Cordillera of Colombia, including the Perija and Merida ranges).
Area Taxa Nodes Binary code
A1 12 18,19,21 ?????????? ?100000110 01???????
A2 1,3,13,14,15,16,22,24 7,8,10,11,18,19,20,20a,21,26,27,28 1010001101 1011110110 111010111 A3 1,14,17,24 7,8,10,11,19,21,20,27,28 0000001101 1001001111 010010011 A4 2,24,15,17 7,8,10,11,20,20a,21,27,28 0100001101 1000101001 010010011
A5 4,5 9,10,11 0000100011 1????????? ?????????
A6 1,4,6,9,14,23,25 7,8,10,11,19,21,26,27,28 1001011111 1001000110 010101111
A7 1,6,25 7,8,10,11,28 1000011101 1????????? ??0001001
Figure 3.5. Phylogenetic trees with the areas. The areas correspond to the geographic range of the species that were originally at the top of each branch in Fig. 3.2. For area coding see Fig. 3.2.
No support a vicariance relationship was found between the areas A6 (Cordillera Oriental of Peru) and A7 (Bolivian Altiplano) suggesting the occurrence of a differentiation process, other than vicariance.
Examination of the homoplasy in the primary BPA cladogram area (numbers with asterisk in Figs. 3.6. and 3.7.) resulted in one case for A5 and A1, and four cases for A2, A3, A4, A6, and A7.
When these latter areas were duplicated for a secondary BPA (Tab. 3.3), it was discovered that replications of A4 were superfluous, the only required replicates being of A2, A3, A6, and A7 to fully explain the homoplasy on the tree. The secondary BPA produced three equally parsimonious area cladograms (89.6%, tree-length = 29), which collapsed to one consensus cladogram (Fig. 3.7).
I found that the ‘ancestor’ forms for clade 1 and 3 are present at the root of the area cladogram, and the differentiation of the third lineage (clade 2) appears later, along the main tree branch. Homoplasy and reverses in the area cladogram were interpreted either as dispersal events of the extant taxa (as well as of their ancestors, depicted as the nodes in the phylogenetic reconstruction) from their ancestral areas, or as later parapatric speciation (see
1 2 15
4 5 12 13 16
3 17
7 14 20a
18
8 9 6
20 19
10
11 21
A A62-A7
22 23
24 25
26
27 28
A21-A4 A21
A3-A4
A0- A21- A22- A3-
A61-A62 A1 A22
A22- A3-A61
A21 A61
A4 A22
A5-A61 A5
A61- A62-A7
Table 3.3. Secondary matrix listing the geographical distribution of three clades of Coeligena – former
Pterophanes taxa, along with the binary codes representing the phylogenetic relationships for all three clades. ‘?’
= species missing from the area.
Area Taxa Nodes Binary code
A1 12 18,19,21 ?????????? ?100000110 01???????
A2 1,3,14,15,16,24 7,8,10,11,18,19,20,20a,21,27,28 1000001101 1001110011 110010011
A21 13,22 18,19,21,26,27,28 ?????????? ?010000110 011000111
A3 1,17,24 7,8,10,11,21,20,27,28 1000001101 1000001001 010010011
A31 14 19,21 ?????????? ?001000010 01???????
A4 2,24,15,17 7,8,10,11,20,20a,21,27,28 0100001101 1000001001 010010011
A5 4,5 9,10,11 0000100011 1????????? ?????????
A6 1,4,6,14,23,25 7,8,10,11,19,21,26,27,28 1000001101 1001000010 010100111
A61 9 10,11 0000000011 1????????? ??0001001
A7 6,25 11,28 1000011101 1????????? ??0001001
A71 1 7,8,10,11 ?????????? ?001000010 ?1????????
From the area cladogram can be inferred that clades 1 and 3 are older than of clade 2, their radiation history being parallel, with their origin in the southern Andes of Peru and Bolivia (taxa 11 and 28 at the root of the cladogram). Considering that this was the first part of the current Andes to be uplifted (Simpson 1975), the divergece of the Coeligena lineage from the Trochilinae stem line can be dated to an older age than the Bleiweiss et al. (1994) Mid-Miocene estimate. It can be speculated that the lineage ancestor was present in the Pre-Andes of the Eocene (c. 50 Ma.), based on evidence that the extant Trochilidae group had already differentiated from the stem ‘swift-like Apodiformes’ in the Mid Eocene (Mayr 2003).
The divergece of the clade 2 can be estimated to have occurred later (taxon 21 before the second branch of the area cladogram), presumably after the last uplift of the East Ecuadorian and Central Colombian Andes, in the Paleocene. For this reason, extant members of each clade (species) reached their current geographical distribution independently, and in order to avoid confusion the individual speciation scenarios for each clade will be treated separately, the uplift events and their temporal scale are depicted in the Fig. 3.6.
CLADE 1 (Fig. 3.1) had its origin in the southern Andes of Peru and extended northwards, along with the uplift of the Cordilleras Oriental and Occidental of Peru (taxon 10), and southwards to the Bolivian Altiplano during the Miocene. The formation of the deep and arid River Marañon Valley (late Miocene) caused the diverging of the species Coeligena violifer (terminal taxon 6), which later, due to the geological changes during the Pliocene-Pleistocene glaciation, divided into four sub-forms in a south-to-north direction, each corresponding to the currently recognised subspecies (see Species accounts).
Figure 3.6. Parsimonious area cladogram produced from primary BPA of three clades of genus Coeligena and former Pterophanes; points of congruence (homoplasies) are marked with an asterisk. The numbers 11, 21, and 28 represent the ‘ancestral taxon’ for each clade. Remainder numbers accompanying slash marks refer to species code (from Fig. 3.2 and Tab. 3.3).
The uplift of the Cordillera Occidental of Peru occurred in several periods, whereas the final uplift can be dated to the first and second Pleistocene ice advances. These events isolated part of the ancestral population, which dispersed along both Peruvian Cordilleras (homoplasy of taxon 9), producing the differentiation of the species Coeligena iris and C. aurora (terminal taxa 4 and 5, respectively). The current range of these taxa is very patchy, the product of several isolations events during the Quaternary, causing a pronounced polymorphism between the populations. Probably the sympatry observed between C. iris and C. aurora is secondary, being caused by a recent dispersal of the north Peruvian C. iris populations (C. iris iris).
The widely distributed ancestral population (taxon 8) followed the uplift of the Andes in a northerly direction, occupying the newly formed mountain ranges from Ecuador to Venezuela and yielding a form adapted to lower altitudes (above 2000 m) that rapidly dispersed northwards (taxon 7). The high-altitude relict ancestral form differentiated into the species C.
prunellei (terminal taxon 3), today with a very restricted distribution range. During the Miocene, the re-uplift of the Western Cordillera of Colombia and Ecuador split the ancestral
X-1 X-14 23 12 13,22 3 2,x1 1,X-15
4*,6* 14*,16 4*,5 25
18 19* 17,24
14* X-1
15 6*,25
19*,26
20,27
1 7,8,27
21 9
28
9,10 11 28
A31 A71 A6 A1 A21 A2 A4 A3 A7 A5 A61
C. coeligena (terminal taxon 1). Populations of the latter species on the eastern slopes of the Colombian Andes dispersed very quickly southwards and northwards, occupying a lower altitude zone along the Andes (homoplasy of the taxon 1).
Later geological events during the Tertiary and early Quaternary produced fragmentations of the early continuous mountain range, e.g., the separation from the East Andes of Colombia of the Cordillera of Perija by the Maracaibo Depression (Tertiary) and the Cordillera de la Costa in Venezuela by the Barquisimeto Depression (Quaternary), which consecutively produced subspecific differentiation. If it can be proved that the more isolated forms (C. coeligena coeligena in Venezuela, C. c. ferruginea in the West Cordillera of Colombia and Ecuador, and C. c. boliviana in Bolivia) already have reached the reproductive isolation, a parapatric speciation could be invoked. The absences on some of the area cladogram branches are better explained as unsuccessful dispersal into these areas (to assume extinction would be the less parsimonious reason, thus resulting in a violation of the Henning Auxiliary Principle).
CLADE 2 (Fig. 3.2, taxon 21) had a younger origin farther north than the other two. The area of origin can be allocated to the Andes of Colombia and Ecuador, in the Mid-Miocene (16 Ma.), being isolated from the Coeligena stem line during the erosion and volcanic episodes of the late Tertiary. The ancestral group diverged in two lineages after the uplift of the Eastern Cordillera of Colombia: one north (taxon 19) and the other south of the Colombian and Ecuadorian Andes (taxon 20). Both lines experienced a very complex differentiation process, influenced by climatic and geological changes during the Quaternary. It is most likely that the already differentiated taxon 20 dispersed northwards, existing in sympatry with taxon 19, while the population that stayed in the Central Andes of Colombia differentiated, forming the species C. lutetiae (terminal taxon 17), which eventually dispersed northwards.
After the uplifting of the West Cordillera of Colombia and Ecuador, the northern taxon 20 probably had an ancestrally wide distribution, reaching the Perija Cordillera in northern Colombia and the Venezuelan Andes, but the separation of these components of the Andes in the Tertiary (Maracaibo Depression) produced a contraction of the ranges, leaving one relict in the Venezuelan Andes that differentiated in the species C. eos, and another in the species C.
bonapartei, with a relict population on the Perija mountain range. The populations of C.
lutetiae suffered the same split, giving rise to two subspecies on both slopes of the Ecuadorian Andes (see Species Accounts). The lineage of taxon 19 experienced another form of
included in the genus Pterophanes, see above), which successfully dispersed southwards, and is now found along the eastern slopes of the Andes from the Eastern Cordillera of Colombia to southern Peru (homoplasy on the area cladogram). The absence of this taxon can be solved with the same argumentation as with C. coeligena (e.g.,absence in the Bolivian Altiplano, see clade 2 explanation).
Figure 3.7. Parsimonious area cladogram produced with the secondary BPA with three clades of genus Coeligena and former Pterophanes; points of congruence (homoplasies) are marked with an asterisk. The numbers 11, 21, and 27 represent the ‘ancestral taxon’ of each clade. Remainder numbers accompanying slash marks refer to species code (from Fig. 3.2 and Tab. 3.3).
The northernmost differentiated group (taxon 18) reached north-eastern Colombia (Sierra de Perija) and the southernmost the Venezuelan Andes. The abrupt elevation of the Sierra de Santa Marta during the Pliocene (c. 1000 m uplift) isolated its populations, giving rise to the species C. phalerata (terminal taxon 12, defined by several exclusive derived characters, see previous section and Species Accounts). The other branch formed the species C. helianthea, in north-eastern Colombia, with an isolated relict population on the Paramo de Tama, probably the product of migration and subsequent isolation during the Pleistocene glaciation.
4*,14* 3,13,14*,15*,16 12
23 22,24*,26 14*,17*,24* 18* 2,15*,17*,24* 4*,5
9*,26* 18,20*
20* 20* 9
6,25
1
19
7,8,28 21 27
10 11
A1 A4 A5
A6 A7 A2 A3
Figure 3.6. Andes uplift progression, the different geological events, and the time scale are depicted (taken from Simpson 1975).
CLADE 3 (Fig. 3.2) is the oldest lineage of the Coeligena clade, originating in the southern Andes of Peru and northern Bolivia. The clade is composed of four allopatric polymorphic species (with one exception); its speciation is very well-supported by vicariance (absence of homoplasy on the area cladogram). The first species differentiated very early (late Eocene) in the form C. inca, which dispersed southwards to the Bolivian Altiplano, forming two subspecies; the northern, ancestral fraction settled widely from southern Peru to the Venezuelan Andes, following the progressive uplift of the Andes.
The separation of the Merida Andes from the remaining range during the Tertiary and Quaternary caused the differentiation of the species C. conradii, which lost several of the distinct characters common to the clade, and the remaining population diverged in the species C. torquata, which during the uplift of the West Cordillera of Ecuador was divided into two populations. This separation must have be relatively recent, judging by the poor differentiation of the subspecies on both slopes of the Andes. Later the formation and enlargement of the river Marañon valley divided the southernmost population of C. torquata, initiating the parapatric speciation of this population in the form C. insectivora, which still shares many features with the parental species, whereas the separation is old enough to have produced polytypes within the group (three geographically isolated subspecies, see Species
The Coeligena group, as other members of the Andean biota, showed a direct response to the progressive sequential south-to-north uplifting of the mountain range as proposed by Doan (2003). The more decisive speciation process is the vicariance, being more finely modulated by other geological events, such as the Pleistocene ice expansions. Simultaneous and comparative study of other Andean elements (trochilids as well as other groups) would afford the most accurate and perhaps final verification of the model proposed here.
Based on this hypothetical speciation model, constructed on the phylogenetic relationships between taxa and areas, it is possible to make predictions on how the speciation process proceeds in a closely related taxon. In this sense, considering that Coeligena (Ensifera) ensifera is a basal group in the Coeligena clade, being actually the sister group of clades 1 and 2 (this implies it would be somewhat younger than clade 3). I am confident in affirming that its origin was also in the extreme south of the Andes, but that it did not respond to the vicariance process as the other groups did, and was capable of dispersing through all the Andes yet presenting no important geographic variation between populations, although many are apparently isolated.
The resulting hypothesis of this study differs in some aspects from the speciation models proposed for other Andean taxa. Many studies on geographic variation and biogeography of Andean trochilids have been carried out with close related taxa such as Eriocnemis (Schuchmann et al. 2001a), Haplophaedia (Schuchmann et al. 2000), Ramphomicron (Weller
& Schuchmann 2002), Boissonneaua (Schuchmann et al. 2001b), and the Heliodoxa genera group, including Sephanoides (Renner 2000) All these works propose Pleistocene climatic fluctuations and glacial ice-front advances as the main modelling factors of the speciation and distribution patterns of the extant taxa, with subsequent colonisation events (dispersal).
In these hypotheses a lowland origin for the Andean fauna is proposed, with subsequent migrations to the surrounding highlands that emerged in the late Pleistocene. This agrees in general with Bleiweiss’ (1998) reconstruction of hummingbird origins, who based his conclusions on phylogenetic reconstructions of trochilids lineages (Bleiweiss et al. 1994, 1997). In addition, a reconstruction of the ancestral characters at the genus level was presented. The group diversification and basal separation of the genera Eriocnemis and Heliodoxa from the sister group Coeligena within the ‘brilliants’ lineage would have taken place at the end of the Tertiary (Late Miocene). Coeligena species radiation would have begun later, in the Mid-Pliocene (before the Quaternary), as a byproduct of coevolution and
Some crucial differences are found if the biogeographic hypothesis for Coeligena species presented in this study is compared with those obtained in studies on closely related Andean taxa like Eriocnemis, Heliodoxa (Bleiweiss et al. 1997), Sephanoides (Renner 2000, Schuchmann 1999), and Haplophaeia (Schuchmann et al. 2000).
Schuchmann et al. (2000) state that Haplophaeia species, contrary to other hummingbird taxa, have their centre of origin and evolution in the northern Andes, subsequently expanding from this region. The phylogenetic relationships within the genus were based on morphological and behavioural synapomorphic characters. Phylogentic relationships with the taxa Eriocnemis and Urosticte were suggested. The initial separation of a ‘proto-Haplophaeia’ population was probably completed before the Pleistocene but after the final rise of the High-Andean crest.
The colonisation of the current range occurred as gradual invasions initiated during each glacial period.
Renner (2000) proposes a complicated hypothetic scenario to explain the radiation of the genera Heliodoxa and Sephanoides. In his study, the phylogeny of the group is not completely solved, Sephanoides being presented - instead of Heliodoxa - as the sister group of Coeligena.
Renner proposes a lowland origin for Heliodoxa, in western Brazil, and explains the current Andean ranges as the result of possible colonisations via the eastern slopes in southern Peru, which later continued northwards reaching the northern Andes. Sephanoides species show a southern origin, their dispersion being southwards reaching Patagonia, and westwards to the Juan Fernández Island.
Schuchmann et al. (2001a) analyse the biogeography and speciation pattern of Eriocnemis.
The biogeographic hypothesis for the origin and radiation of Eriocnemis species is similar to that proposed for Haplophaeia (see above, Schuchmann et al. 2000), with the centre of origin located in the northern Andes of Colombia, thus agreeing with the hypothesis postulated for other Andean vertebrates. In this hypothesis, as well as in the Haplophaeia study, Pleistocene climatic fluctuations are the main factor modelling radiation and current distribution patterns, and the conclusions are based on the occurrence or frequency of determined characters (plesiomorphies or apomorphies) in extant populations regardless of the earlier geological history of the region. In both analyses, dispersal is the principal process invoked to explain current geographic ranges.
The biogeographic hypothesis presented in this study differs in many aspects from those mentioned above. These differences not only encompass the conclusions but also the
speciation. Putative speciation models, like those described above, are based upon the interpretation of the geographic distribution, the phenotypic characteristics of populations, and the geo-historical events. These are models difficult or impossible to test (Graves 1982).
For this study, the phylogenetic reconstruction was obtained first of all because it was required for biogeographic analyses (Shapiro 1989, Brooks & McLemman 1991). Once the relatedness between species was known, it was possible to deduce their diversification history. The determination of the common history (monophyly) and the phylogenetic relationships within one group allow us to couple these with the evolution of the area occupied by the taxa. Because no current phylogenetic reconstruction for the genus Coeligena existed, this study closed the gap, using the phylogenetic hypothesis here proposed for the biogeographic analysis.
In this study, an early origin for the Coeligena species is assumed, earlier than the Pleistocene, contemporary with the Andean uplift progression. This assumption does not discard the effects of Pleistocene events on the differentiation of Coeligena populations, but it is considered to have more probably affected speciation at the subspecific level (almost all the evidence indicating a Pleistocene origin of the Andean avifauna was obtained from Passeriformes taxa, which are known to be the most recently differentiated bird order (see Fjeldså et al. 1999). Here vicariance is considered the main modelling factor of the current range and speciation of Coeligena species. The importance of vicariance is supported by other studies on Andean small mammals, where a vicariant (allopatric) speciation model is favoured in order to explain patterns observed in extant taxa (Patton et al. 1989, Patton et al. 1990).
Based on the geological history of the Andes, a southern origin is proposed for Coeligena, located on the south-western slopes of the present Peruvian Andes. The colonisation of new habitats was determined by the progressive Andes uplift and followed a general south-to-north direction, with some exceptions, caused by occasional further uplift events in the south.
This model coincides with the general speciation proposed for several hummingbird genera such as Aglaiocercus (Schuchmann & Duffner 1993), Chalcostigma (Schuchmann & Heindl 1997), and Metallura (Heindl & Schuchmann 1998). Nevertheless it is important to note that the radiation scenario for these hummingbirds has been proposed for the Pleistocene.
Moreover, I prefered to compare my scenario with phylogenetically nearer taxa (member of the same lineage) because of high probability that they shared speciation processes. That is
Sephanoides species show an exclusive southern Andes range, with some representatives on the Juan Fernández Islands. In my phylogenetic reconstruction, Sephanoides is more basal than Heliodoxa and the monophyletic Coeligena supporting the southern origin of the
‘brilliant’ lineage. Once the last elevation of the northern Andes (late Pliocene and early Pleistocene) was reached, isolation, a consequence of glacial advances and the formation of new geographic barriers, led to parapatric as well as allopatric differentiation, which still are in progress and will probably lead to the complete differentiation of isolated populations.
The other taxon studied here was Patagona gigas, which is also an Andean component. The phylogenetic reconstruction excluded it from the monophyletic Coeligena group, being placed basally, together with Sephanoides. The speciation model would predict for Patagona subspecies a very similar differentiation process as the one observed in the other taxa studied (a southern origin and successful dispersal in a northerly direction).
In fact, Patagona populations show a south-to-north divergence at a subspecific level, probably caused by vicariance during the uplift of the Cordillera Oriental and Occidental of Peru (before the Miocene), but the isolation was not complete; contact zones were detected near the River Madre de Dios, in Peru (see Species Accounts), caused in part by the migrational behaviour of the group. The curious division of the southernmost population could originate in the relatively recent uplift of the Principal Cordillera (northern Chile and Argentina), but as in the case of Ensifera ensifera, this taxon did not respond to the vicariance process with diversification, being still in dispersal to the north as shown by the recent evidence (Ortiz-Crespo 1974), reaching Colombia (Fjeldså & Barbosa 1983, this observation could be a wrong identification, Schuchmann, pers. comm.).