A molecular phylogeny of the harriers (Circus, Accipitridae) indicate the role of long distance dispersal and migration in diversification

Accepted Manuscript A molecular phylogeny of the harriers (Circus, Accipitridae) indicate the role of long distance dispersal and migration in diversification Graeme Oatley, Robert E. Simmons, Jérôme Fuchs PII: DOI: Reference:

S1055-7903(15)00027-5 http://dx.doi.org/10.1016/j.ympev.2015.01.013 YMPEV 5111

To appear in:

Molecular Phylogenetics and Evolution

Received Date: Revised Date: Accepted Date:

22 September 2014 15 December 2014 29 January 2015

Please cite this article as: Oatley, G., Simmons, R.E., Fuchs, J., A molecular phylogeny of the harriers (Circus, Accipitridae) indicate the role of long distance dispersal and migration in diversification, Molecular Phylogenetics and Evolution (2015), doi: http://dx.doi.org/10.1016/j.ympev.2015.01.013

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

A molecular phylogeny of the harriers (Circus, Accipitridae) indicate the role of long distance dispersal and migration in diversification Graeme Oatleya *, Robert E. Simmonsb and Jérôme Fuchsc, d

a

Department of Zoology and Lab of Ornithology, Faculty of Science, Palacky University, Tř

17. Listopadu 50, 77146 Olomouc, Czech Republic b

DST-NRF Centre of Excellence, Percy FitzPatrick Institute, University of Cape Town,

Rondebosch 7701, South Africa c

UMR7205 Institut de Systématique, Evolution, Biodiversité CNRS MNHN UPMC EPHE,

Département Systématique et Evolution, Muséum National d’Histoire Naturelle, 55 Rue Buffon, 75005 Paris, France

d

UMS MNHN/CNRS 2700 Outils et Méthodes de la Systématique Intégrative (OMSI),

Muséum National d’Histoire Naturelle, 57 rue Cuvier, F-75231 Paris Cedex 05, France

*

Corresponding author. E-mail address: [email protected]

1

Abstract The monophyly of the raptorial Circus genus (harriers) has never been in question, but the specific status of many, often vulnerable island endemic, taxa remains uncertain. Here we utilise one mitochondrial and three nuclear loci from all currently recognised Circus taxa (species and subspecies) to infer a robust phylogeny, to estimate the divergence date and to reconstruct the biogeographic origins of the Circus group. Our phylogeny supports both the monophyly of Circus and polyphyly of the genus Accipiter. Depending on the rate of molecular clock used, the emergence of the harrier clade took place between 4.9 and 12.2 mya which coincides with the worldwide formation of open habitats which extant harriers now exploit. The sister relationship of the Northern Harrier C. cyaneus hudsonius and the Cinereous Harrier C. cinereus contradicts previous classifications that treated the former as conspecific with the Hen Harrier C. cyaneus cyaneus. Thus both should be elevated to species status: C. hudsonius and C. cyaneus. Further, the African Marsh C. ranivorus and the European Marsh C. aeruginosus Harriers emerge as sister species. The remaining marsh harriers exhibit very little genetic diversity, and are all recently diverged taxa that exhibit allopatric distributions. Considering their sister relationship and geographic proximity, we recommend treating C. approximans and C. spilonotus spilothorax as subspecies of C. approximans. For C. spilonotus spilonotus C. maillardi maillardi and C. maillardi macrosceles, their plumage and morphometric differences, phylogenetic relationship and geographic distributions make lumping of these taxa as a single species complicated. We thus propose to recognise as separate, recently evolved species: C. spilonotus, C. maillardi and C. macrosceles. Biogeographic inferences on the ancestral origin of harriers are uncertain, indicating that the harriers emerged in either the Neotropics, Palearctic or Australasia. We are, however, able to show that speciation within the harriers was driven by long range dispersal and migration events.

2

Keywords: Accipitridae; Circus; phylogenetics; biogeographic history; migration 1. Introduction The harriers (Accipitridae: Circus) represent one of the most distinctive lineages of raptorial birds due to their particular behaviour, flying slowly over grassland, steppe and marshland to detect prey (small birds, rodents), and their mating pattern of polygyny which is regularly found in at least six of the 16 putative species (Simmons, 2000). Harriers have always been considered to be a monophyletic group (Dickinson 2003) because of their distinctive shape as well as unique osteological features among Accipitridae such as facial ruff and asymmetric ears. The majority of northern hemisphere harrier species undergo long distance migration, while southern hemisphere breeding birds are usually short distance migrants in search of food during non-breeding dry periods (Thiollay, 1994). Harriers have traditionally been placed within their own subfamily, the Circinae (Peters, 1931). Recent molecular studies have revealed that harriers may be much more closely related to Accipiter (sparrowhawks and goshawks) than previously suspected. Earlier mitochondrial and nuclear datasets have shown Circus to be sister to Accipiter (Wink and Sauer-Gürth, 2004; Lerner and Mindell, 2005). More recent studies (Griffiths et al., 2007; Lerner et al., 2008; Ong et al., 2010; Barrowclough et al., 2014) with a more thorough sampling of Accipiter species concluded that Circus may be in fact nested within Accipiter and that Accipiter is a poly/paraphyletic genus. Two natural groups are recognized within the harriers: the marsh and steppe harriers (Thiollay, 1994; Simmons, 2000). The harrier taxa that make up the marsh harrier group are: European Marsh Harrier C. aeruginosus (and its dark morph C. a. harterti), African Marsh Harrier C. ranivorus, Eastern Marsh Harrier C. spilonotus spilonotus, Papuan Harrier C. s. spilothorax, Pacific [Australasian] Marsh Harrier C. approximans, Madagascar Marsh Harrier 3

C. macrosceles macrosceles, and Reunion Harrier C. m. maillardi. The steppe harriers includes the remaining harrier taxa: Long-winged Harrier C. buffoni, Spotted Harrier C. assimilis¸ Black Harrier C. maurus, Hen Harrier C. cyaneus cyaneus, Northern Harrier C. c. hudsonius, Cinereous Harrier C. cinereus, Pallid Harrier C. macrourus, Pied Harrier C. melanoleucos, and Montagu’s Harrier C. pygargus. The taxonomy of the marsh harrier group has changed substantially over the years, with initial classifications treating all marsh harriers (excluding C. ranivorus) as members of C. aeruginosus (Brown and Amadon, 1968; Sibley and Monroe, 1990). Subsequent classifications began to elevate races to species level (see Ferguson-Lees and Christie, 2005), but recently nearly all previously described marsh harrier subspecies have been elevated to species level (Simmons, 2000; Ferguson-Lees and Christie, 2005). A single subspecies of C. aeruginosus, however, remains: a darker morph restricted to North Africa C. a. harterti (Thiollay, 1994; Simmons, 2000; Dickinson, 2003). The relationship of the marsh harrier species to one another is less certain. A mitochondrial phylogeny with representatives of most Circus taxa provided support for the recognition of C. m. macrosceles as a separate species, even though it was surprisingly sister to C. aeruginosus and not C. m. maillardi. However, no bootstrap support was found for this relationship (Wink and Sauer-Gürth, 2004). Within the steppe harriers, the relationship between the C. c. cyaneus and C. c. hudsonius is also contentious (Dobson and Clark, 2011). At times these taxa have been regarded as subspecies (Sibley and Monroe, 1990; Thiollay, 1994; Amadon and Bull, 1998; Dickinson, 2003), while genetic results (Wink and Sauer-Gürth, 2004; Johnsen et al., 2010) have shown substantial differences between the two taxa to warrant species status (Simmons, 2000; Ferguson-Lees and Christie, 2005). The Pallid Harrier C. macrourus has also been inferred to be closely related to the Black Harrier C. maurus and the Cinereous Harrier C. cinereus, which emerge as sister species (Wink and Sauer-Gürth, 2004) even though their distributions 4

are separated by the Atlantic Ocean. Moreover, one (C. maurus) breeds in South Africa and the other (C. cinereus) in South America, a pattern that is at odds with current knowledge about bird dispersal patterns. Establishing a robust molecular phylogeny can be important for making informed conservation decisions (Moritz, 1995; Rolland et al., 2011), thus determining if taxa in already threatened areas are genetically distinct can go a long way to establishing timely conservation measures. Three harrier taxa (C. m. macrosceles, C. m. maillardi and C. maurus) are currently listed as either Vulnerable or Endangered (IUCN Red List 2014), while a fourth (C. macrourus) is listed as near threatened due to steep population declines (BirdLife International, 2004). In Madagascar and Reunion, habitat transformation and human disturbance are listed as one of the main threats to C. m. macrosceles (De Roland et al., 2009) and C. m. maillardi (Bretagnolle et al., 2000). The southern African endemic C. maurus is estimated to have fewer than 1000 mature individuals (Curtis et al., 2004), and in addition, their primary nesting habitat is threatened by agriculture (Curtis et al., 2004). While not listed as endangered or threatened, high fire frequency in New Guinea is thought to impact the Papuan Harrier C. s. spilothorax (Simmons and Legra, 2009) and population numbers for this taxon are also estimated to be particularly low (Simmons and Legra, 2009). Understanding the genetic status of these threatened and vulnerable taxa will enable authorities to make important recommendations regarding their conservation. On one hand, identifying discrete genetic entities will warrant often limited conservation resources to be directed towards their conservation. On the other hand, taxa that do not appear to form clear evolutionarily significant units (ESUs sensu Ryder, 1986; Moritz, 1994) will allow these limited conservation resources to be directed to taxa or species that are of more pressing conservation concern.

5

The number of recognised species within Circus ranges from 13 (Thiollay, 1994; Dickinson, 2003) to 16 species (Simmons, 2000), with some of the subspecies recognised by previous authors being elevated to species status. Some of the noteworthy elevated species include the Papuan Harrier C. spilothorax (also considered a subspecies of the Eastern Marsh Harrier C. spilonotus) and the Madagascar Marsh Harrier C. macrosceles (also considered a subspecies of the Reunion Harrier C. maillardi). If the species status of these two taxa are confirmed, this will add to the number of endemic bird species found on both these island biodiversity hotspots that are under increasing environmental decline (Brooks et al., 2006). In this study, we aim to infer the evolutionary relationship of all recognized Circus taxa. This will be achieved through the construction of a phylogeny based on multiple molecular markers from the mitochondrial and nuclear genome and the reconstruction of the biogeographic history of the genus using newly developed methods that take jump dispersal into account. A robust phylogeny will allow us to answer three evolutionary questions: 1) Does genetic data reflect the current Circus classification and taxonomy? It is expected that Circus is a monophyletic clade, and that they form two groups within this clade: the steppe and marsh harriers. 2) Does the diversification of the Circus clade coincide with the emergence of open habitats (i.e. C4 grass expansion) which harriers currently exploit? 3) What is the biogeographic history of the Circus group?

2. Material and methods 2.1. Sampling and laboratory procedure

6

Fresh tissue or toe pad material was obtained for each currently recognized harrier taxon (species or sub-species; Dickinson, 2003). Between one and four individuals were sampled for each taxon, covering as much of each taxon’s distribution as possible in an attempt to account for any genetic variation that may be present within each taxon (Table S1). The DNA extraction of fresh tissue samples were performed using the DNeasy Tissue Kit (Qiagen, Valencia, CA, USA) following the manufacturer’s protocol. DNA extractions from toe-pad samples were performed in a room specifically dedicated to work with historical DNA, following the fresh tissue protocol but with the addition of 20µl of dithiothreitol (DTT, 0.1M). We sequenced a continuous 1.2 kb fragment of mitochondrial DNA (tRNALeu to tRNAMeth) and a total of 2032 bp from three nuclear introns (Myoglobin intron-2 – MB, Beta Fibrinogen intron-5, FGB and TGFβ2 intron-5, TGFβ2). The primers used in this study are listed in Table S2 and S3. The PCR amplification comprised of an initial denaturation at 94°C for 5 min, followed by 35-40 cycles at 94°C for 30 s, 56 – 58°C for 30 s, and 72°C for 30 s – 1.30 min. The amplification was completed with a final extension of 72°C for 10 – 15 min. PCR products were visualised on a 1% agarose gel and cleaned using 3µl ExoSAP solution incubated for 30 min at 37°C, and then for 15 min at 80 °C. Purified PCR products were then cycle sequenced using Big Dye terminator chemistry (Applied Biosystems, Foster City, CA, USA). Sequenced products were purified using Sephadex columns and then run on an Applied Biosystems automated sequencer. Amplification and sequencing of the fresh and historical material were performed at different times of the year and all reagents used for amplification and sequencing of historical samples were new and specific to laboratory work with historical material. To avoid the construction of chimeric sequences, distinct intraspecific haplotypes were re-amplified and re-sequenced, as well as the contiguous fragments; this step allowed us to exclude sample mix up during the PCR or sequencing stages. Obtained 7

sequences were checked and edited in BIOEDIT (Hall, 1999). Heterozygous sites in nuclear loci (double peaks) were coded using the appropriate IUPAC code. All sequences have been deposited in Genbank (Accession Numbers: XXX–XXX). Alignment was performed by eye in BIOEDIT and was straightforward owing to the low number of insertion and deletion events.

2.2. Gene tree analyses Individual gene trees were inferred using Maximum Likelihood and Bayesian analyses, implemented in RAXML V7.0.4 (Stamatakis, 2006; Stamatakis et al., 2008) and MRBAYES 3.2 (Ronquist et al., 2011), respectively. We used PHASE

V2.1.1

(Stephens et al., 2001), as

implemented in DNASP 5.0 (Librado and Rozas, 2009), to infer the alleles for each nuclear locus. Three runs were performed and results were compared across runs. We used the recombination model and ran the iterations of the final run 10 times longer than for the other runs. Redundant alleles/haplotypes were identified and removed from the data set and the 95% minimum spanning networks were drawn using TCS (Clement et al., 2000). The best-fitting model of nucleotide substitution was selected, using Bayesian Information Criterion (BIC) and greedy algorithm, using PARTITIONFINDER (Lanfear et al., 2012). For the mitochondrial data set, the most supported number of partitions was determined from an initial set of six partitions. These were the three separate codon positions for ND1, tRNALEU, tRNAILE and tRNAGLN. As each of the nuclear DNA markers were non-coding, these alignments were also tested for the best fitting model and not for partitioning. For the Bayesian analyses, four Metropolis-coupled Markov chain Monte Carlo (MCMC) iterations (one cold and three heated) were run for 10 to 25 *106 iterations with trees sampled every 1000 iterations. We used default priors except for the branch length prior for which we used an exponential mean of 100 as this improved convergence of the mitochondrial data set 8

but also provided more realistic rate multipliers estimates. The ‘burn-in’ was set to 10% of the total iterations (i.e. between 1 and 2.5 *106, depending on the number of total iterations). Bayesian posterior probabilities were calculated from the remaining iterations. Using random starting trees, two independent runs were performed. We also ensured that the potential scale reduction factor (PSRF) approached 1 for all parameters and that the average standard deviation of split frequencies was less than 0.01. We also used TRACER V1.6 (Rambaut and Drummond, 2005) to ensure that convergence was reached for the posterior distributions of the parameter estimates, and ensured that the effective sample size (ESS) of these estimates was > 200.

2.3. Species tree analyses A species tree was reconstructed using the coalescent based model implemented in *BEAST V.1.8.0

(Drummond et al., 2012). Given that we could not amplify FGB for Melierax gabar

and C. s. spilonotus, and TGFβ2 for C. s. spilonotus, and that *BEAST requires at least one sequence per locus for each species, these two species were excluded from the data set for species tree reconstruction. We used the same substitution model partitioning model strategy as that used in the Bayesian analyses. We used a normal prior distribution for the TGFβ2 rate that corresponds to the rate obtained by Lerner et al. (2011). Runs were 10*10 8 iterations long. In an attempt to incorporate all taxa into our species tree analysis, a species tree was reconstructed from ND1 and MB, for which we have sequences from all species.

2.4. Divergence dating and biogeographic analyses Estimates of divergence time for our mitochondrial and nuclear trees were estimated using *BEAST v.1.8.0 (Drummond et al., 2012). For the mitochondrial divergence times, we implemented the ‘traditional rate’ (2.1% molecular clock corresponding to 1.05*10 -2

9

subs/site/lineage/myr; Weir and Schluter, 2008) and a faster rate of molecular evolution suggested by Lerner et al. (2011) (2.5*10 -2 subs/site/lineage/myr; 95% HPD 2.2/2.9*10 -2). For the FGB and MB data, divergence times were estimated using the average rate of autosomal

molecular

evolution

of

1.1*10 -3

subs/site/lineage/myr

(1.1*10 -3

subs/site/lineage/myr), as estimated from Lerner et al. (2011); these two loci were not analysed by Lerner et al. (2011). For TGFβ2, we used the substitution rate determined by Lerner et al. (2011) for this locus: 1.7*10 -3 subs/site/lineage/myr. Two Circus fossils have been described, the Wood Harrier (C. dossenus) from the Hawaiian Islands and the Eyles’ Harrier (C. eylesi) from New Zealand. However, the recent age of both these fossils (late Holocene and late Quaternary, respectively; Olsen and James, 1991; Holdaway and Worth, 1997) in conjunction with their taxonomic uncertainty within the Circus clade renders their use as calibration points for dating analyses speculative so did not attempt to use those two fossils as calibration points. We used the likelihood based methods implemented in the statistical package R (R Core Team 2014) package BIOGEOBEARS (Matzke, 2014) to determine the biogeographic history of Circus. The first analysis is based on the Dispersal-Extinction-Cladogenesis (DEC) model of LAGRANGE (Ree and Smith, 2008). This model has two free parameters, the rate of range expansion (dispersal, d) and the rate of range contraction (extinction, e) (Ree and Smith, 2008; Matzke, 2014). The second analysis (DEC + j) is similar to the DEC model, however, a third free parameter is added: founder event speciation, j. The best likelihood model (DEC vs. DEC + j) was selected using a Likelihood Ratio Test and AIC weights. We employed two slightly different biogeographic realm classifications to code presence or absence of the breeding distributions for each of the Circus taxa obtained from maps in Thiollay (1994) and Ferguson-Lees and Christie (2005). The first classification comprised seven biogeographic realms (Newton et al., 2003), namely: Neotropical, Nearctic, Palearctic,

10

Ethiopian, Madagascan, Indo-Malaya and Australian. The second classification comprised eleven realms (Holt et al., 2013): Neotropical, Panamanian, Nearctic, Palearctic, SaharoArabic, Afrotropical, Madagascan, Sino-Japanese, Oriental, Oceania and Australian. The purpose of using the two was to determine if any noticeable differences arose with the addition of realms put forward by Holt et al., (2013). We used the default settings in BIOGEOBEARS and ancestral areas were reconstructed using the ND1 and MB species tree as this tree contained all Circus taxa and did not differ from the combined marker species tree (see section 3.2. Species tree).

3. Results 3.1. Phylogenetic analyses The final mitochondrial DNA alignment from 61 individuals of 17 harrier taxa was 1193 bp long, consisting of 48 unique haplotypes. We excluded stop codons and intergenic regions because of difficulties in alignment due to the range of taxon groups in our sampling. Individual sequences ranged from 1189 (Pandion haliaetus) to 1193 (multiple taxa). The topologies resulting from the Bayesian (Fig. 1) and Maximum Likelihood (tree not shown) analyses were very similar. Both analyses indicate that Accipiter is polyphyletic and that Circus is monophyletic (Bayesian posterior probabilities BI = 1, Likelihood bootstrap BS = 100). Within the Circus clade, two primary sub-clades emerge. The first sub-clade contains six species that are part of the steppe harrier group: C. cyaneus cyaneus, C. cyaneus hudsonius, C. cinereus, C. assimilis, C. maurus and C. macrourus. Surprisingly, C. c. hudsonius and C. cinereus emerge as sister taxa with the Palearctic C. c. cyaneus sister to this group. Mitochondrial sequence divergence estimates between C. c. hudsonius and C. cinereus was 1.0%, while that between C. c. cyaneus and both of C. c. hudsonius and C. cinereus was

11

1.1% and 1.6%, respectively. Another noteworthy sister relationship is that of C. maurus and C. macrourus, these taxa showing average sequence divergence of 1.8% (Fig. 1). The second sub-clade, commonly referred to as the marsh harrier group, consists of C. aeruginosus aeruginosus, C. aeruginosus harterti, C. ranivorus, C. spilonotus spilonotus, C. spilonotus spilothorax, C. approximans, C. maillardi maillardi and C. maillardi macrosceles. Non-typical marsh harriers, C. melanoleucos and C. pygargus, are also found in this clade and are the first two species to split off. Five proposed species (C. approximans, C. s. spilonotus, C. s. spilothorax, C. m. maillardi and C. m. macrosceles; Simmons 2000) form a group with very little mitochondrial divergence (ranging from 0.1% to 0.2% sequence divergence). The placement of the last species, C. buffoni, does differ between the Bayesian and Maximum Likelihood trees. The Bayesian tree places C. buffoni as sister to the remaining Circus taxa with very good support (BI = 1; Fig. 1), but this taxon is sister to the extended marsh-harrier clade in the maximum likelihood tree (BS = 75). All individual nuclear gene trees were consistent between Bayesian and Maximum Likelihood analyses, and they also showed Circus to be monophyletic. The MB gene tree (Fig. S1) was inferred using 43 unique alleles from 59 individuals (722 bp). There is however, not much taxon structuring within the Circus clade (Fig. S1), with a high degree of allele sharing across taxa. There is more structure found in the FGB phylogeny, inferred using 42 unique alleles from 55 individuals (602 bp; Fig. S2). Although not supported by significant posterior probabilities (BI = 0.91, BS = 71), the marsh harrier taxa emerge as a monophyletic group. Good support is provided for the monophyly of C. pygargus (BI = 1, BS = 90), the sister relationship of C. maurus and C. macrourus (BI = 0.98, BS = 66) and the clade of C. c. cyaneus, C. c. hudsonius and C. cinereus (BI = 0.96, BS = 73; Fig. S2). The TGFβ2 gene tree (Fig. S3) showed similar genetic structure to that of the mitochondrial gene tree. This phylogeny was constructed using 49 unique alleles from 56 individuals (708

12

bp). Two well supported sub-clades emerge (BI = 0.97 and 0.98), corresponding to the steppe harrier and marsh harrier sub-clades recovered in the mitochondrial gene tree. While there is little genetic structure within the TGFβ2 Steppe harrier sub-clade, one difference emerging from the mitochondrial sub-clade is that C. c. hudsonius and C. c. cyaneus are more closely related with both taxa sharing a common allele (Fig. S3). Within the marsh harrier sub-clade, there is substantial allele sharing among seven taxa (C. a. aeruginosus, C. a. harterti, C. ranivorus, C. s. spilonotus, C. approximans, C. m. macrosceles and C. melanoleucos; Fig. S3). Once again, C. buffoni emerges as sister to all the other Circus taxa with good support (BI = 1, BS = 100).

3.2. Species tree The species tree was constructed for 39 of the taxa sequenced, including outgroup species. Within Circus, only C. s. spilonotus was excluded due to missing sequence data for the FGB and TGFβ2 loci. Overall, the inferred species tree topology (Fig. 2) is similar to the individual gene trees. Within the monophyletic Circus clade, the species tree topology is very similar to the mtDNA and TGFβ2 gene trees, especially in the recognition of the steppe and marsh harrier sub-clades (BI = 1). One major difference in comparison to these two trees is the positioning of C. buffoni, being nested within Circus and sister to the marsh harrier subclade (Fig. 2). Within the Steppe clade, C. cinereus and C. c. hudsonius emerge as sister, with C. c. cyaneus sister to this group. Additionally, C. maurus and C. macrourus are confirmed as sister taxa. In the marsh harrier clade, short branch lengths indicate close genetic relationships among C. approximans, C. s. spilothorax, C. m. macrosceles and C. m. maillardi. A species tree containing all Circus taxa (i.e. including C. s. spilonotus) was constructed using ND1 and MB sequence data (Fig. 2). The overall topology was very similar to that of

13

the species tree with all loci, however, support values for some nodes are lower. The position of C. s. spilonotus is within the marsh group and sister to C. m. macrosceles, and in turn sister to C. m. maillardi. The Australian C. approximans and Papua New Guinea C. s. spilothorax emerge as sister taxa (Fig. 2).

3.3. Divergence times and biogeographic reconstruction Two mitochondrial molecular clocks with different rates of evolution (Table 1) were used to infer divergence dates within the Circus clade. Across the different rates and molecular markers, the emergence date of Circus occurred between 2.7 mya (95% HPD = 2.1 – 3.3 mya) and 6.6 mya (95% HPD: 5.1 – 8.3 mya; Table 1, Fig. 2). Circus buffoni is predicted to have split from the marsh harrier group between 2.1 mya (95% HPD: 1.6 – 2.7 mya) and 5.4 mya (95% HPD: 4 – 7 mya), while the split between the Indo-Malayan and the Australasian marsh harriers is predicted to have occurred as recently as 0.1 mya (95% HPD: 0.05 – 0.2 mya) or 0.3 mya (95% HPD: 0.1 – 0.5 mya) (Table 1, Fig. 2). The diversification of the steppe harrier clade occurred between 2.2 mya (95% HPD: 1.7 – 2.8 mya) and 5.5 mya (95% HPD: 4.1 – 7.1 mya). For both the seven and eleven realm biogeographic ancestral reconstructions, the DEC + j model was favoured over the DEC model. For the seven realm reconstruction, DEC had lnL = -51.41 while the DEC + j was significantly better at lnL = -29.44 (Χ2 = 43.92, df = 1, P < 0.001). Similarly, the DEC + j model was favoured over the DEC model for the eleven realm reconstruction (lnL = -36.19 and -59.17, respectively; Χ2 = 47.67, df = 1, P < 0.001). The ancestral biogeographic state for Circus from both the seven and eleven realm analyses are similar (Fig. 3 and S4), with both suggesting a Neotropical or Palearctic origin. The eleven realm analysis also suggests the possibility of an Australian ancestor. Harrier fossil taxa, of very recent age, do not provide additional support for a potential Australian Circus

14

ancestor as they could represent very recent offshoots from extant species. Both fossil taxa are found on Pacific Ocean islands (New Zealand; Holdaway and Worth, 1997 and Hawaii; Olsen and James, 1991).

4. Discussion 4.1. Phylogeny and Circus species limits This is the first phylogeny of the genus Circus that included all described harrier species and sub-species. While our harrier phylogeny agrees broadly with a previous phylogeny (Simmons, 2000; Wink and Sauer-Gürth, 2004) inferred from cytochrome b sequence data in terms of Circus monophyly, species relationships within Circus do differ somewhat. Here we focus on these discrepancies and the implications of species relationships in the evolution of harrier species. Additionally, we address the species status proposed in previous work (i.e. Simmons, 2000; Ferguson-Lees and Christie, 2005) for certain taxa. Starting with the steppe harrier group, the previously classified subspecies, the Palearcticdistributed Hen Harrier C. c. cyaneus and Holarctic-distributed Northern Harrier C. c. hudsonius (Thiollay 1994, Dickinson 2003) are shown to be non-sister taxa. Instead, the South American distributed Cinereous Harrier C. cinereus is sister to C. c. hudsonius, with C. c. cyaneus sister to this group. The sub-species relationship between C. c. cyaneus and C. c. hudsonius has been previously questioned (Simmons et al., 1987; Johnsen et al., 2010; Dobson and Clark, 2011), while 1.7% mtDNA sequence divergence has also provided evidence of these taxa being distinct species (Wink and Sauer-Gürth, 2004). The level of mtDNA divergence between these two taxa presented here is slightly lower (1.1%) than in previous work, likely due to the fact that we sequenced a different mitochondrial locus. The phylogeny presented here provides evidence that C. c. cyaneus and C. c. hudsonius should be

15

elevated to species status (Simmons, 2000). It has been suggested that the earlier classifications that regarded C. c. cyaneus and C. c. hudsonius as conspecifics was due to erroneous illustrations of these two taxa (Dobson and Clark, 2011). These two taxa do, however, show differences in male and female adult plumage (Thiollay, 1994; Ferguson-Lees and Christie, 2005). Male C. c. hudsonius have rufous spots on the belly and breast while male C. c. cyaneus has a plain white belly and breast (Ferguson-Lees and Christie, 2005). In addition, female C. c. cyaneus is more similar to female C. macrourus and C. pygargus than it is to female C. c. hudsonius (Ferguson-Lees and Christie, 2005). The plumage of adult C. cinereous birds is also not dissimilar to that of C. c. hudsonius. Adult males share the same grey colour of the head and chest, with darker grey on the back and wings; there is a greater amount and bolder rufous streaking of C. cinereus in comparison to C. c. hudsonius (Ferguson-Lees and Christie, 2005). Female adults, while both being dark brown above, differ in the colouration and streaking on the underside. Our genetic data supports that the traditional conspecifics Northern and Hen Harriers be elevated to specific status, as previously suggested (Simmons, 2000). Staying in the steppe group, a further difference that emerges in comparison to previous phylogenies is the sister relationship between the southern African endemic Black Harrier C. maurus and eastern European and central Asian distributed Pallid Harrier C. macrourus. Previously, it was inferred that C. maurus and C. cinereus were sister taxa (Simmons, 2000; Wink and Sauer-Gürth, 2004) but this new relationship is biogeographically more plausible as the Pallid Harrier is an annual visitor to southern Africa (Thiollay, 1994). Differences to previous phylogenies also arise in the marsh harrier group. The African Marsh Harrier C. ranivorus emerges as sister to both sub-species of the European Marsh Harrier (C. a. aeruginosus and C. a. harterti), which was previously inferred to be sister to the Madagascar Marsh Harrier C. m. macrosceles. Our species tree places C. m. macrosceles 16

within a clade composed of the Reunion Marsh Harrier C. m. maillardi, the Papuan Harrier C. s. spilothorax and the Australian-distributed Pacific Marsh Harrier C. approximans. A species tree constructed using only ND1 and MB sequence data (Fig. 2) places the Eastern Marsh Harrier C. s. spilonotus within this clade, as sister to C. m. macrosceles. Up to five distinct species have been proposed from this group with convincing evidence for these classifications often lacking. At a first glance, our species trees indicate that there is limited genetic support for the recognition of C. approximans, C. s. spilothorax, C. s. spilonotus, C. m. maillardi or C. m. macrosceles as distinct species. The level of mitochondrial genetic differentiation is very low, and ranges from 0.08% (C. approximans and C. s. spilothorax; C. s. spilonotus and C. m. macrosceles) to 0.33% (C. s. spilothorax and C. m. maillardi), which corresponds to 1 - 5 mutations in our mitochondrial DNA alignment. Similar levels of mitochondrial divergence was also seen between some of our sampled Accipiter species (A. nisus/A. rufiventris: 0.08%; see also Breman et al., 2012). This pattern of mitochondrial divergence was also recovered using the same mitochondrial fragment in the genus Falco where several species pairs had the similar levels of divergence (F. newtoni/F. araea: 0.5%, F. cherrug/F. subniger: 0.3%; F. cuvieri/F. subbuteo: 0.3 - 0.6%; Fuchs et al., 2015). Accipitridae barcoding studies have shown that cytochrome c oxidase subunit I (COI) sequence divergence estimates among most good species within genera usually range from around 3% (Breman et al., 2012) to 6.6% (Ong et al., 2011) sequence divergence with some noticeable exceptions (e.g. A. nisus/A. rufiventris). Further research using cytochrome b (cyt b) showed that sequence divergence among recently diverged Gyps vulture species ranged from 0.6 – 2.6% (Arshad et al., 2008), while in hawk-eagles (Spizaetus spp.) mitochondrial control region (CR) and cyt b sequence divergence among species ranged from 3.2 – 12.8% and 1.6 – 11.3%, respectively (Haring et al., 2007). Even in this study, well accepted Accipiter species showed at least 1.4% mitochondrial sequence divergence (A. gentilis and A. 17

melanoleucos). Thus the small genetic differences found between the Indian Ocean island and Asian-Australasian harriers may not be a good reason to recognize them as distinct species. Plumage differences among the closely related C. m. macrosceles, C. s. spilothorax, C. s. spilonotus and C. approximans are also not substantial, with all taxa sharing similar plumage patterns with differences mainly found in upper part colouration (Ferguson-Lees and Christie, 2005). While substantial differences in plumage and morphometric measurements are reported for the Reunion and Madagascar Harriers (Bretagnolle et al., 1998; Simmons, 2000), and a reported 3% genetic divergence (Wink and Sauer-Gürth, 2004) have, however, prompted previous researchers to afford species status to both C. m. macrosceles and C. m. maillardi, the difference in size could easily be explained by the ‘island rule’ in which larger bodies organisms are smaller on islands (Foster, 1964; Case, 1978; Clegg and Owen, 2002). Additionally, our mitochondrial data shows that sequence divergence between C. m. macrosceles and C. m. maillardi is only 0.1% and not 3% as previously reported (Wink and Sauer-Gürth, 2004). It must be noted that these Cytochrome b sequences have not been released on any public database, i.e. Genbank). A thirty fold difference in substitution rates across mitochondrial coding loci appears very unlikely, as most mitochondrial coding genes evolve at comparable rates, with even Cytochrome b evolving two times slower than ND1 in songbirds (Lerner et al., 2011). The nuclear allele sharing among Circus taxa in our nuclear data can be seen as evidence of recent divergence of species with ancestral polymorphisms remaining within lineages (e.g. Corvus, Omland et al., 2006; Zosterops, Oatley et al., 2012; Oenanthe, Schweizer and Shirihai, 2013; Fuchs et al., 2015). Another option, although not necessarily exclusive from incomplete lineage sorting, could be the presence of gene flow among species thereby maintaining genetic cohesiveness (Mimus, Nietlisbach et al., 2013; Anas, Lavretsky et al., 2014). This explanation does, however, necessitate that taxon ranges are sympatric, or at the 18

least parapatric, for this to be possible: this does occur in among many taxa. Yet, the wide majority of these pairs are not directly related, suggesting that incomplete lineage sorting of ancestral polymorphism is a more likely explanation.

4.2. Biogeographic inferences From an evolutionary standpoint, many of the new Circus sister relations presented here make clear biogeographical sense. The species pairs highlighted above all consist of northern and southern hemisphere taxa breeding along similar longitudes (i.e. New World C. c. hudsonius/C. cinereus, Old World C. macrourus/C. maurus and C. aeruginosus/C. ranivorus). As most harriers are migratory, it is predicted that some individuals from northern hemisphere taxa became resident on the non-breeding grounds, breeding at different times to migratory individuals and ultimately leading to reproductive isolation between migratory and non-migratory populations. This hypothesis is in accord with recent studies that migration had a strong effect on avian diversification (Rolland et al., 2014), including in other raptors (falcons, Fuchs et al., 2015). This scenario is definitely more plausible than a trans-oceanic dispersal that would be needed to describe for example the sister-group relationship between C. maurus and C. cinereus inferred by previous phylogenies (Simmons 2000, Wink and Sauer-Gürth, 2004). Our sampling strategy, employed here with an emphasis on establishing a robust Circus phylogeny, is not suited to making many conclusions regarding intraspecific variation. Previous phylogeographic studies on the Black Harrier (Fuchs et al., 2014) and the Montagu’s Harrier (Garcia et al., 2011) have shown low (Garcia et al., 2011) to no (Fuchs et al., 2014) genetic variation among populations of both these species and very little to no geographic structure. Here, two harrier taxa were sampled from widespread localities of their 19

breeding distribution and they show similar trends. Although we obtained five samples of the European Marsh Harrier C. aeruginosus from North Africa, France, Switzerland, Sweden and Russia (Table S1), only two distinct mitochondrial haplotypes are found, the sample from Switzerland being distinct from the others and only weakly differentiated – one mutation). Multiple alleles were found in the nuclear sequence data but, many of these alleles were shared with other Circus taxa. Similarly, Australian Swamp Harriers from Australia and New Caledonia shared the same haplotype despite the presence of approximately 1400 km of open sea between the two landmasses. In the Northern Harrier C. c. hudsonius, four individuals from geographically distant (north east vs. south west) USA populations were sequenced and only one haplotype was found. The sampling localities for both taxa realistically fall on either side of migratory divides in both North America (e.g. Ruegg et al., 2014) and Europe (e.g. Møller et al., 2011; Mettler et al., 2013). Populations on either side of migratory divides are expected to undertake different migratory routes to reach wintering grounds but also maintain distinct breeding grounds. Individuals from distant populations are thus expected to exhibit genetic divergence given putative isolation between highly allopatric populations. However, phylogeographic and population genetic studies on each of these species, coupled with studies tracking harrier movements would provide further insight into levels of gene flow and connectivity among distant populations. Hence, it seems that it is characteristic for harriers to have very low to no mitochondrial diversity across broad distributions. Harriers are known to have large population fluctuations, likely due to predator-prey cycles (Korpimaki 1985, Simmons, 2000), with recurrent population reduction possibly slowly eroding genetic diversity over long time periods. Simulations have shown that slow range contractions, lead to more severe reductions in genetic diversity than abrupt contractions (Arenas et al., 2012). Thus slow but recurrent

20

decreases in population ranges linked to predator prey cycles could have resulted in the observed low levels of genetic diversity observed within some Circus taxa. Fuchs et al. (2014) suggested a recent origin or selective sweep as possible processes for the lack of genetic diversity observed. We suggest here that this trend of low genetic diversity, observed in both ‘old’ (e.g. C. hudsonius) and ‘young’ species (C. aeruginosus), is likely not due to the age of origin or a selective sweep. Thus, taking the results from this study and previous harrier molecular studies into account (Garcia et al., 2011; Fuchs et al., 2014), continuous fluctuations in prey availability leading to recurrent variations in harrier population sizes may be the driver behind low genetic diversity observed within Circus species. This notwithstanding, we suggest that increased sampling and population genetic analyses would provide further insights in the genetic dynamics in harrier species.

4.3. Circus divergence and biogeographic history The expansion of C4 grasses occurred globally during the Late Miocene and Pliocene, between 6 and 8 million years ago (Cerling et al., 1997; Edwards et al., 2010). Although changes in environment among various regions in the world differed in the timing and speed (Edwards et al., 2010), there was a global trend of closed environments transitioning to more open savannahs and woodland with associated C3 grasses, and then finally into fully open C4 grasslands. We used two rates of mitochondrial evolution to estimate the timing of origin of the Circus clade (Weir and Schluter, 2008; Lerner et al., 2011). The 2.1% rate of evolution for mitochondrial DNA is based only on cytochrome b sequence data (Weir and Schluter, (2008), whereas Lerner et al. (2011) provided rates of evolution for multiple loci within the mitochondrial genome (range 0.22 - 3.1*10-2 subs/site/lineage/myr) including ND1 and cytochrome b which display different rates of molecular evolution (1.4 vs 2.5 *10 -2 subs/site/lineage/myr). Hence, we expect that the dating accuracy of our calibrations to be

21

greatest from the Lerner et al. (2011) rate as this is specific for the mitochondrial locus we used in our study. This notwithstanding, we discuss the results obtained from both rates. The two mitochondrial rate estimates place the emergence of the Circus clade at 6.6 mya (5.1 – 8.3 mya; 95% HPD) and 2.7 mya (2.1 – 3.3 mya; 95% HPD), respectively. This timing coincides with the period during or after the formation of open habitats currently favoured and exploited by all Circus species (Thiollay, 1994; Simmons, 2000). Hence the most likely scenario is that Circus colonized the new forming open habitat, from a forest living ancestor (Accipiters are associated to forests or closed habitats). In further accord with this hypothesis, the Late Miocene was also a time of major turnover in mammalian communities, including the worldwide explosive radiation of rodents (Lecompte et al., 2008; Fabre et al., 2012), a group that most species of harriers feed on, whereas Accipiter are specialized bird predators. Interestingly, opening of the habitat during the Late Miocene has recently been shown to have favoured the radiation of the falcons (Fuchs et al., 2015), another lineage of raptors that feed on rodents and birds. These results suggest that different lineages of predatory birds reacted similarly to the appearance of savannah and steppes. There is some uncertainty regarding the ancestral distribution of the Circus lineage (Fig. 3 and S4), being either Neotropical, or Palearctic or Australian. The ancestor of the Marsh harrier/melanoleucos/pygargus/buffoni clade is recovered to be of Palearctic origin, followed by dispersal events to the Neotropics (C. buffoni), the Afrotropics (C. ranivorus) and Australasia. Subsequent dispersal events from Australasia account for the presence of Harrier taxa in Indo-Malaysia (C. s. spilothorax) and Madagascar (C. m. maillardi and C. m. macrosceles). The Steppe clade is proposed to have an Australasian origin (Fig. 3 and S4). There was a subsequent dispersal event to the Palearctic, followed by dispersal by independent ancestors to the Afrotropics (C. maurus) and to the Nearctic (C. c. hudsonius) and then to the Neotropics (C. cinereus). 22

4.4. Preliminary insights into the phylogeny of sparrowhawks and goshawks. Our results revealed that the genus Accipiter is polyphyletic, as two species are not related to the other Accipiter species. The first, A. superciliosus, is endemic to South America where it co-occurs with members of the Accipiter bicolor superspecies. With our sampling, our analyses could not clearly identify its closest relatives, although we have strong support for it being nested in the clade, based on previous studies (Lerner et al., 2008; Barrowclough et al., 2014), formed by the buteonine, sea eagles and kites, and a group of atypical hawks (Harpagus, Kaupifalco, Melierax). Kocum (2008) also suggested that A. superciliosus was genetically distinct from a monophyletic Accipiter-Circus clade, while morphological analyses of the procoracoid foramen indicates that this species differs from typical Accipiters (Olson 2006). A more thorough sampling of South America genera (Rosthramus, Ictinia, Busarellus, Geranospiza) will be necessary to definitely assess the relationships of A. superciliosus. The second species that is unrelated to other Accipiter species is the Indo-Malayan A. trivirgatus. This species is unusual among Accipiter in that it has a crest. Recent studies have showed it to be very distinct to other Accipiter species (Breman et al., 2012), although their sampling did not allow them to show that it is not directly related to other Accipiter (A. trivirgatus clustered close to the outgroup in most trees but only one non-Accipiter, Buteo, was used). Ong et al. (2011), in a barcoding survey of accipitrids from the Philippines, also included A. trivirgatus and could not identify its sister species as it was a single lineage at the base of the accipitrids in general. Clearly, assessing the relationships of the Crested Goshawks awaits further study using a broader taxonomic sampling of the basal Accipitridae.

23

Our results are also in accord with the topologies of Lerner et al. (2008) and Barrowclough et al. (2014) in that Accipiter sensu stricto is paraphyletic as Circus is nested within the genus. Our analyses identified two subgroups that are more closely related to Circus than to other Accipiter species; the A. nisus/A. rufiventris/A. striatus/A. chionogaster clade and the A. gentilis/A. melanoleucos/A. bicolor clade. We could not, however, identify with strong support which of these two lineages is the exact sister-group to the Circus clade. Another potential lineage that could be part of this group is the distinctive Doriae Goshwak (Megatriorchis doriae) (Barrowclough et al., 2014).

5. Conclusions Our inferred harrier phylogeny and dating analyses places the emergence of the Circus clade at, or just following, the origin of C4 grasses and associated spread of open habitats that harriers currently exploit. A Palearctic, Neotropic or Australasian ancestor of the harriers is supported, with long distance migration and dispersal contributing to the diversification of the genus. We suggest some taxonomic changes to the currently recognised classification. Our species level phylogeny supports the elevation of the Hen and Northern Harrier to full species status: C. hudsonius and C. cyaneus, respectively. The issue of specific status among the Asian and Indian Ocean marsh harriers is more complex. Our phylogeny infers a very recent radiation of this group, and that recognised sub-species are non-sister taxa. The geographic proximity of C. approximans and C. s. spilothorax, overall morphometric similarity, their sister relationship coupled with low genetic differentiation, indicate that these two taxa should be treated as subspecies of C. approximans Peale 1848: C. a. approximans and C. a. spilothorax. As the two Indian Ocean Marsh Harriers, C. m. maillardi and C. m. macroceles are polyphyletic and genetically similar to C. s. spilonotus it is tempting to also

24

consider them as different subspecies. However, the large distance between C. s. spilonotus and the Indian Ocean, the plumage, morphometric and behavioural differences exhibited by C. m. maillardi makes classifying these three taxa as one species difficult. We suggest treating these taxa as separate species: C. spilonotus Kaup 1847, C. maillardi Verreaux 1862 and C. macrosceles Newton 1863. It is apparent though, that further sampling and population genetic analyses testing for gene flow among the Asian and Indian Ocean Marsh Harriers are needed to confidently asses their species status.

Acknowledgements We thank the following people and institutions for the loan of tissue material: Leo Joseph and Robert Palmer (ANWC), Paul Sweet, George Barrowclough and Joel Cracraft (AMNH), Carla Cicero and Rauri Bowie (MVZ), Alice Cibois (MHNG), Jon Fjeldså, Jan-Bolding Kristensen (ZMUC), Darío Lijtmaer and Pablo Tubaro (MACN), Ulf Johansson and Per Ericson (NRM), Matthieu Le Corre and Colombe Valette (SEORF), Maëlle Kermabon, Lucie Yrles and Annika Wichert (Centre de soins LPO Villeveyrac), Ludovic Besson (Musée d’Histoire Naturelle, Bourges) and Marie-Sophie Garcia-Heras (Percy FitzPatrick Institute of African Ornithology, University of Cape Town). We also thank Nick Matzke for his help on the BioGeoBEARS analysis. Help during laboratory work was kindly provided by Céline Bonillo, Delphine Gey, Josie Lambourdière and Jose Utge. GO was funded by a European Union Research grant n.: CZ.1.07/2.2.00/28.0158. RES was partially supported by NRF funding.

25

References Amadon, D., Bull, J., 1998. Hawks and owls of the world, a distributional and taxonomic list. Proc. West. Found. Vertebr. Zool. 3: 295 – 327. Arenas, M., Ray, N., Currat, M., Excoffier, L., 2012. Consequences of range contractions and range shifts on molecular diversity. Mol. Biol. Evol. 29: 207 – 218. Arshad, M., Gonzalez, J., El-Sayed, A.A., Osborne, T., Wink, M., 2009. Phylogeny and phylogeography of critically endangered Gyps species based on nuclear and mitochondrial markers. J. Ornith. 150: 419 – 430. Barrowclough, G.F., Groth, J.G., Lai, J.E., Tsang, S.M., 2014. The phylogenetic relationships of the endemic genera of the Australo-Papuan hawks. J. Raptor Res. 48: 36 – 43. BirdLife International. 2004. Birds in Europe: population estimates, trends and conservation status. BirdLife International, Cambridge, U.K. Breman, F.C., Jordaens, K., Sonet, G., Nagy, Z.T., Van Houdt, J., Louette, M., 2013. DNA barcoding and evolutionary relationships in Accipiter Brisson, 1760 (Aves, Falconiformes: Accipitridae) with a focus on African and Eurasian representatives. J. Ornith. 154: 265 – 287. Bretagnolle, V., Thiollay, J., Attie, C., 1998. Status of Reunion Marsh Harrier Circus maillardi on Reunion Island. Proceedings of the V World Conference on Birds of Prey. Pretoria. Bretagnolle, V., Ghestemme, T., Thiollay, J.-M., Attié, C. 2000. Distribution, population size and habitat use of the Réunion Marsh Harrier, Circus m. maillardi. J. Raptor Res. 34: 8 – 17. Brooks, T.M., Mittermeier, R.A., da Fonseca, G.A.B., Gerlach, J., Hoffmann, M., Lamoreux, J.F., Mittermeier, C.G., Pilgrim, J.D., Rodrigues, A.S.L., 2006. Global Biodiversity Conservation Priorities. Science 313: 58 – 61. Brown, L.H., Amadon, D., 1968. Eagles, hawks and falcons of the world. Wellfleet, Secaucus, NJ.

26

Case, T.J., 1978. A general explanation for insular body size trends in terrestrial vertebrates. Ecology 59: 1 – 18. Cerling, T.E., Harris, J.M., MacFadden, B.J., Leakey, M.G., Quade, J., Eisenmann, V., Ehleringer, J.R., 1997. Global vegetation change through the Miocene/Pliocene boundary. Nature 389: 153 - 158. Clegg, S.M., Owen, I.P.F., 2002. The ‘island rule’ in birds: medium body size and its ecological explanation. Proc. R. Soc. Lond. B. Biol. Sci. 269: 1359 – 1365. Clement, M., Posada, D., Crandall, K.A., 2000. TCS: a computer program to estimate gene genealogies. Mol. Ecol. 9: 1657 – 1660. Curtis, O., Simmons, R.E., Jenkins, A.R. 2004. Black Harrier Circus maurus of the Fynbos Biome, South Africa: a threatened specialist or an adaptable survivor? Bird Cons. Int. 14: 233 – 245. De Roland, L.R., Thorstrom, R., Razafimanjato, G., Rakotondratsima, M.P.H., Andriamalala, T.R.A., Sam, T.S., 2009. Surveys, distribution and current status of the Madagascar Harrier Circus macrosceles in Madagascar. Bird Conser. Int. 19: 309 – 322. Dickinson E.C., 2003. The Howard and Moore complete checklist of the birds of the world, 3rd edition. Princeton: Princeton University Press, New Jersey. Dobson, A.D.M., Clark, M.L., 2011. Inconsistency in the taxonomy of Hen and Northern Harriers: causes and consequences. Br. Birds 104: 192 – 201. Drummond, A.J., Suchard, M.A., Xie, D., Rambaut, A., 2012. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Mol. Biol. Evol. 29: 1969 – 1973. Edwards, E.J., Osborne, C.P., Strömberg, C.A.E, Smith, S.A., C4 Grasses Consortium., 2010. The origins of C4 grasslands: Integrating evolutionary and ecosystem science. Science 328: 587-591. Fabre, P., Hautier, L., Dimitrov, D., Douzery, E.J.P., 2012. A glimpse on the pattern of rodent diversification: a phylogenetic approach. BMC Evol. Biol.12: 88.

27

Ferguson-Lees, J., Christie, D.A., 2005. Raptors of the World. Princeton: Princeton University Press, New Jersey. Foster, J.B., 1964. The evolution of mammals on islands. Nature 202: 234 – 235. Fuchs, F., Simmons, R.E., Mindell, D.P., Bowie, R.C.K., Oatley, G., 2014. Lack of mtDNA genetic diversity in the Black Harrier Circus maurus, a Southern African endemic. Ibis 156: 227 – 230. Fuchs, J., Johnson, J.A., Mindell, D.P., 2015 Rapid diversification of falcons (Aves: Falconidae) due to expansion of open habitats in the Late Miocene. Mol. Phylogenet. Evol. 82: 166 - 182. Garcia, J.T., Alda, F., Terraube, J., Mougeot, F., Sternalski, A., Bretagnolle, V., Arroyo, B., 2011. Demographic history, genetic structure and gene flow in a steppe associated raptor species. BMC Evol. Biol. 11: 333. Griffiths, C.S., Barrowclough, G.F., Groth, J.G., Mertz, L.A., 2007. Phylogeny, diversity, and classification of the Accipitridae based on DNA sequences of the RAG-1 exon. J. Avian Biol. 38: 587 – 602. Hall, T.A., 1999. BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucl. Acids Symp. Ser. 41: 95 – 98. Holdaway, R.N., Worth, T.H., 1997. A reappraisal of the late Quaternary fossil vertebrates of Pyramid Valley Swamp, North Canterbury, New Zealand. N. Z. J. Zool. 24: 69 – 121. Holt, B.G., Lessard, J., Borregaard, M.K., Fritz, S.A., Araújo, M.B., Dimitrov, D., Fabre, P., Graham, C. H., Graves, G.R., Jønsson, K.A., Nogués-Bravo, D., Wang, Z., Whittaker, R.J., Fjeldså, J., Rahbek, C., 2013. An Update of Wallace’s Zoogeographic Regions of the World. Science 339: 74 – 78. Johnsen, A., Rindal, E., Ericson, P.G.P., Zuccon, D., Kerr, K.C.R., Stoeckle, M.Y., Lifjeld, J.T., 2010. DNA barcoding of Scandinavian birds reveals divergent lineages in trans-Atlantic species. J. Ornith. 151: 565 - 578.

28

Kocum A: Phylogeny of the Accipitriformes (birds of prey) based on different nuclear and mitochondrial DNA-sequences. Vogelwarte 46: 141–143. Korpimaki E. 1985. Rapid tracking of microtine populations by their avian predators: evidence for stabilising selection. Oikos 45: 281-28. Lanfear R., Calcott, B., Ho, S.Y.W., Guindon, S., 2012. PartitionFinder: combined selection of partitioning schemes and substitution models for phylogenetic analyses. Mol. Biol. Evol. 29: 1695 - 1701. Lavretsky, P., McCracken, K.G., Peters, J.L., 2014. Phylogenetics of a recent radiation in the mallards and allies (Aves: Anas): Inferences from a genomic transect and the multispecies coalescent. Mol. Phylogenet. Evol. 70: 402 – 411. Lecompte, E., Aplin, K., Denys, C., Catzeflis, F., Chades, M., Chevret, P., 2008. Phylogeny and biogeography of African Murinae based on mitochondrial and nuclear gene sequences, with a new tribal classification of the subfamily. BMC Evol. Biol. 8: 199. Lerner, H.R.L., Mindell, D.P. 2005. Phylogeny of eagles, Old World vultures, and other Accipitridae based on nuclear and mitochondrial DNA. Mol. Phylogenet. Evol. 37: 327 – 346. Lerner, H.R.L., Klaver, M.C., Mindell, D.P., 2008. Molecular phylogenetics of the Buteonine birds of prey (Accipitridae). Auk 125: 304 – 315. Lerner, H.R.L., Meyer, M., James, H.F., Hofreiter, M., Fleischer, R.C., 2011. Multilocus resolution of phylogeny and timescale in the extant adaptive radiation of Hawaiian honeycreepers. Curr. Biol. 21: 1838 – 1844. Librado, P., Rozas, J., 2009. DnaSP v5: A software for comprehensive analysis of DNA polymorphism data. Bioinformatics 25: 1451 – 1452. Matzke, N.J., 2014. Model selection in historical biogeography reveals that founder-event speciation is a crucial process in island clades. Syst. Biol. 63: 951 – 970. Mettler, R., Schaefer, H.M., Chernetsov, N., Fiedler, W., Hobson, K.A., Ilieva, M., Imhof, E., Johnsen, A., Renner, S.C., Rolshausen, G., Serrano, D., Wesołowski, T., Segelbacher, G., 29

2013. Contrasting patterns of genetic differentiation among Blackcaps (Sylvia atricapilla) with divergent migratory orientations in Europe. PLoS One 11: e81365. Møller, A.P., Garamszegi, L.Z., Peralta-Sánchez, J.M., Soler, J.J., 2011. Migratory divides and their consequences for dispersal, population size and parasite–host interactions. J. Evol. Biol. 24: 1744 – 1755. Moritz, C., 1994. Defining ‘Evolutionarily Significant Units’ for conservation. Trends Ecol. Evol. 9: 373 – 375. Moritz, C., 1995. Uses of molecular phylogenies for conservation. Philos. Trans. R. Soc. Lond. Ser. B 349: 113 - 118. Newton, I., 2003. The speciation and biogeography of birds. Academic Press, London. Nietlisbach, P., Wndeler, P., Parker, P.G., Grant, P.R., Grant, B.R., Keller, L.F., Hoeck, P.E.A., 2013. Hybrid ancestry of an island subspecies of Galápagos mockingbird explains discordant gene trees. Mol. Phylogenet. Evol. 69: 581 – 592. Oatley, G., Voelker, G., Crowe, T.M., Bowie, R.C.K., 2012. A multi-locus phylogeny reveals a complex pattern of diversification related to climate and habitat heterogeneity in southern African white-eyes. Mol. Phylogenet. Evol. 64: 633 – 644. Olsen, S.L. 2006. Reflections on the systematics of Accipiter and the genus for Falco superciliosus Linnaeus. Bull. B.O.C. 126: 69 – 70. Olsen, S.L., James, H.F., 1991. Descriptions of thirty-two new species of birds from the Hawaiian Islands: Part 1. Non-passeriformes. In Ornithological Monographs No. 45 (Johnson, N.K. ed.). Allen Press Inc, Lawrence, Kansas, pp. 1 – 88. Omland, K.E., Baker, J.M., Peters, J.L., 2006. Genetic signatures of intermediate divergence: population history of Old and New World Holarctic ravens (Corvus corax). Mol. Ecol. 15: 795 – 808. Ong, P.S., Luczon, A.U., Quilang, J.P., Sumaya, A.M.T., Ibanez, J.C., Salvador, D.J., Fontanilla, I.K.N., 2011. DNA barcodes of Philippine accipitrids. Mol. Ecol. Res. 11: 245 – 254. 30

Peters, J.L., 1931. Check-list of Birds of the World. Harvard University Press, Cambridge. R Core Team., 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/. Rambaut, A., Suchard, M.A., Xie, D., Drummond, A.J., 2014. Tracer v1.6, Available from http://beast.bio.ed.ac.uk/Tracer. Ree, R.H., Smith, S.A., 2008. Maximum likelihood inference of geographic range evolution by dispersal, local extinction, and cladogenesis. Syst. Biol. 57: 4 – 14. Rolland, J., Cadotte, M.W., Davies, J., Devictor, V., Lavergne, S., Mouquet, N., Pavoine, S., Rodrigues, A., Thuiller, W., Turcati, L., Winter, M., Zupan, L., Jabot, F., Morlon, H., 2012. Using phylogenies in conservation: new perspectives. Biol. Lett. 8: 692 – 694. Rolland, J., Jiguet, F., Jønsson, K.A., Condamine, F.L., Morlon, H., 2014. Settling down of seasonal migrants promotes bird diversification. Proc. R. Soc. Lond. B. Biol. Sci. 281: 20140473 Ronquist, F., Teslenko, M., ven der Mark, P., Ayres, D.L., Darling, A., Hohna, S., Larget, B., Liu, L., Suchard, M.A., Huelsenbeck, J.P., 2012. MrBayes3.2: Efficient Bayesian phylogenetic inference and model choice across a large model space. Syst. Biol. 61: 539 – 542. Ruegg, K., Anderson, E.C., Boone, J., Pouls, J., Smith, T.B., 2014. A role for migrationlinked genes and genomic islands in divergence of a songbird. Mol. Ecol. 23: 4757- 4769. Ryder, O.A., 1986. Species conservation and systematics: the dilemma of subspecies. Trends Ecol. Evol. 1: 9 – 10. Schweizer, M., Shirihai, H., 2013. Phylogeny of the Oenanthe lugens complex (Aves, Muscicapidae: Saxicolinae): Paraphyly of a morphologically cohesive group within a recent radiation of open-habitat chats. Mol. Phylogenet. Evol. 69: 450 – 461. Sibley, C.G., Monroe, B.L., 1990. Distribution and Taxonomy of the Birds of the World. Yale University Press, New Haven, CT.

31

Simmons, R.E., 2000. Harriers of the World: their behaviour and ecology. Oxford University Press, Oxford. Simmons, R., Barnard, P., Smith, P.C., 1987. Reproductive behaviour of Circus cyaneus in North America and Europe: a comparison. Ornis Scand. 18: 33 – 41. Simmons, R.E., Legra, L.A.T., 2009. Is the Papuan Harrier Circus spilonotus spilothorax a globally threatened species? Ecology, climate change threats and first population estimates from Papua New Guinea. Bird Conserv. Int. 19: 379 – 391. Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22: 2688 – 2690. Stamatakis, A., Hoover, P., Rougemont, J., 2008. A fast bootstrapping algorithm for the RAxML web-servers. Syst. Biol. 57: 758–771. Stephens, M., Smith, N.J., Donnelly, P., 2001. A new statistical method for haplotype reconstruction from population data. Am. J. Hum. Genet. 68: 978 – 989. Thiollay, J.M., 1994. Accipitridae (hawks and eagles). In Handbook of the Birds of the World Volume 2: New World Vultures to Guineafowls. (Del Hoyo, J., Elliott, A. and Sargatal, J. eds) Barcelona: Lynx Edicions, pp. 52 – 205. Weir, J.T., Schluter, D., 2008. Calibrating the avian molecular clock. Mol. Ecol. 17: 2321 – 2328. Wink, M., Sauer-Gürth, H., 2004. Phylogenetic relationships in diurnal raptors based on nucleotide sequences of mitochondrial and nuclear marker genes. In: Raptors Worldwide (R.D. Chancelor, R.D. and Meyburg, B.-U. eds.), WWGBP, Berlin, pp. 483 – 498.

32

Figure Legends Figure 1: A 50% majority rule consensus tree from the Bayesian analysis of the ND1 mitochondrial data set. We present the posterior probabilities of the nodes with values ≥ 0.95. Haplotype diagrams within Circus species or closely related taxa are also presented. The size of circles represents the number of individual possessing a particular haplotype. The dots on branches connecting haplotypes represent missing haplotypes. Only unique haplotypes were included in analyses. The position of the marsh harrier clade is highlighted by a box with dashed borders.

Figure 2: Species tree of Circus inferred from four loci (ND1, MB, FGB and TGFβ2). The values on the nodes are the posterior probabilities. Letters on nodes represent timing events, and correspond to Table 1. Also included is the species tree inferred from only ND1 and MB, but which allowed the inclusion of C. s. spilonotus (position indicated by *). The Circus clade is enclosed by a box with dashed borders.

Figure 3: The reconstruction of Circus ancestral areas based on the seven biogeographic areas from Newton (2003; see text for details) and on DEC + j as implemented in BIOGEOBEARS. Pie charts on nodes show the relative probabilities of each area being ancestral. The colour codes are as follows: dark blue (NT) = Neotropics, light blue (N) = Nearctic, green (PA) = Palearctic, red (E) = Ethiopia, purple (M) = Madagascar, yellow (IM) = Indo-Malaysia and Orange (A) = Australia. For ease of reference, the colours used here are the same as those in Fig. S5 which shows the extent of the realms used for this analysis.

33

Table 1: Molecular dating estimates (95% confidence intervals) of nodes from the species tree estimated from two different rates for the mitochondrial ND1 (locus specific from Lerner et al 2011, ‘traditional rate’ Weir and Schluter 2008) evolution, and nuclear clock average (MB and FGB) and original (TGFβ2) rates. Node letters correspond to those on the species tree (Figure 2) and dates are in millions of years.

Node A B C D E F G H I J K L

Weir and Schluter (2008) 1.05*10-2 s/s/l/m 12.2 (9.3-15.5) 6.6 (5.1-8.3) 5.4 (4.0-7.0) 4.5 (3.2-5.8) 2.1 (1.4-2.9) 1.2 (0.7-1.8) 0.6 (0.3-0.9) 0.3 (0.1-0.5) 5.5 (4.1-7.1) 1.7 (1.0-2.3) 1.3 (0.8-1.9) 0.9 (0.5-1.3)

Lerner et al. (2011) 2.5*10-2 s/s/l/m 4.9 (3.8-6.1) 2.7 (2.1-3.3) 2.1 (1.6-2.7) 1.8 (1.3-2.3) 0.9 (0.6-1.2) 0.5 (0.3-0.7) 0.3 (0.1-0.3) 0.1 (0.05-0.2) 2.2 (1.7-2.8) 0.7 (0.5-1.0) 0.5 (0.3-0.7) 0.4 (0.2-0.5)

MB clock 1.1*10-3 s/s/l/m 3.6 (1.6 - 6.2) 3 (1.2 - 5.3) NA NA NA NA NA NA NA NA NA NA

34

FGB clock 1.1*10-3 s/s/l/m 11.2 (5.5 - 18.3) 4.5 (1.8 - 7.8) NA NA NA NA NA NA NA 1.5 (0.2-3.3) NA NA

TGFβ2 clock 1.7*10-3 s/s/l/m 11.9 (8.0 - 16.4) 6.7 (4.2 - 9.5) NA NA NA NA NA NA NA NA NA NA

Figure 1

Figure 2

Figure 3

• • • •

Circus is monophyletic and Accipiter is polyphyletic Harrier diversification coincides with C4 grass expansion Recent divergence of Indo-Asian Marsh Harriers Migration a driver of Circus diversification

35