The Auk 122(4):1210–1224, 2005 © The American Ornithologists’ Union, 2005. Printed in USA.

POPULATION GENETICS OF THE GALÁPAGOS HAWK ( GALAPAGOENSIS): GENETIC MONOMORPHISM WITHIN ISOLATED POPULATIONS J L. B,1,2,6 N K. W,1 M D. C,3,4 J C. B,3 T V,5 P G. P1,2 1Department of Biology, University of Missouri-St. Louis, 8001 Natural Bridge Road, St. Louis, Missouri 63121, USA; 2Department of Evolution, Ecology, and Organismal Biology, The Ohio State University, Columbus, Ohio 43210, USA; Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 3Department of Biology, Arkansas State University, P.O. Box 599, Jonesboro, Arkansas 72467, USA; 4Department of Forest Science, Oregon State University, 321 Richardson Hall, Corvallis, Oregon 97331, USA; and 5Departamento de Biología, Pontifícia Universidad Católica del Ecuador, Quito, Ecuador

A.—Because of their smaller size and isolation, island populations tend to be more divergent and less genetically variable than mainland populations. We collected DNA samples from nine Galápagos Hawk (Buteo galapagoensis) island populations, covering the ’ entire range. Neutral minisatellite DNA markers were used to calculate within-island genetic diversity and between-island genetic

diff erentiation (FST). Typically, these markers mutate too quickly to be informative in such studies. However, in very small, isolated populations, concerns about high mutational rate are obviated by the relative force of genetic driV . Individuals within islands had the highest levels of reported genetic uniformity of any natural population, with mean within-population band-sharing similarity values ranging from 0.693 to 0.956, increasing with decreasing island size. Galápagos Hawks exhibit cooperative polyandry to varying degrees across islands; however, we did not fi nd an association between degree of polyandry and genetic variability. Between-island

FST values ranged from 0.017 to 0.896, with an overall archipelago value of 0.538; thus, most populations were genetically distinct. Also, we documented higher levels of genetic similarity between nearby populations. Our results indicated negligible gene fl ow among most Galápagos Hawk populations, and genetic driV has played a strong role in determining structure at these minisatellite loci. Received 16 April 2004, accepted 5 April 2005.

Key words: Buteo galapagoensis, cooperative polyandry, Galápagos Hawk, Galápagos Islands, genetic driV , minisatellites.

Genética de Poblaciones de Buteo galapagoensis: Monomorfi smo Genético dentro de Poblaciones Aisladas

Rz.—Debido al tamaño más pequeño y al aislamiento de las poblaciones de las islas, éstas tienden a ser más divergentes y menos variables genéticamente que las poblaciones continentales. Recolectamos muestras de ADN de nueve poblaciones isleñas de Buteo galapagoensis, cubriendo el rango total de distribución de la especie. Usamos marcadores neutrales de ADN minisatelital para calcular la

diversidad genética dentro de las islas y la diferenciación genética (FST) entre islas. Típicamente, estos marcadores mutan demasiado rápido como para ser relevantes en estos estudios. Sin embargo, en poblaciones aisladas muy pequeñas, las preocupaciones sobre altas tasas de mutación son descartadas por la fuerza relativa

6E-mail: [email protected]

1210 October 2005] Population Genetics of the Galápagos Hawk 1211

de la deriva génica. Los individuos dentro de las islas presentaron los niveles más altos de uniformidad genética que se conocen actualmente para poblaciones naturales de aves, con valores de similitud promedio de bandas compartidas dentro de las poblaciones que fl uctuaron entre 0.693 y 0.956. Estos valores se incrementaron al disminuir el tamaño de la isla. B. galapagoensis exhibe poliandría cooperativa de varios niveles en distintas islas; sin embargo, no encontramos una asociación entre

el grado de poliandría y la variabilidad genética. Los valores de FST entre las islas fl uctuaron entre 0.017 y 0.896, con un valor global para el archipiélago de 0.538. Por lo tanto, la mayoría de las poblaciones fueron genéticamente distintas. Además, documentamos niveles más altos de similitud genética entre poblaciones vecinas. Nuestros resultados indican que el fl ujo génico entre la mayoría de las poblaciones Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 de B. galapagoensis es despreciable, y que la deriva génica ha jugado un rol importante determinando la estructura en estos loci minisatelitales.

Puz l zz refl ects a et al. 2002), long-term genetic driV working in number of processes, including mutation small isolated populations (Baker et al. 1990, rate, genetic driV , gene fl ow, natural selection, Mundy et al. 1997), or a combination of the two. and phylogeographic history (Bohonak 1999, The Galápagos Hawk (Falconiformes: Buteo Ouborg et al. 1999). Genetic variability is lost galapagoensis) is endemic to the Galápagos archi- through genetic driV and selection against pelago located ~1,000 km west of South America. some genotypes. Generally, genetic driV has a The islands are volcanic in origin, having arisen stronger eff ect in smaller populations; thus, a from a mantle hotspot (Morgan 1971), and they positive relationship between population size have never been connected to the mainland. and genetic variation is expected (Nevo et al. The oldest of the present islands is ~4 million 1984, Frankham 1996). Populations may diverge years old (White et al. 1993). However, older, because of random fi xation of diff erent alleles, now-submerged seamounts to the southeast of diff erences in selective pressures, or addition of the archipelago indicate that islands have been novel mutations. Gene fl ow, however, can have present over the hotspot for at least 17 million a homogenizing eff ect among populations and years and probably for much longer (Christie et mitigate the loss of intrapopulation variation al. 1992, Werner and Hoernle 2003). by adding new alleles or replacing alleles lost Galápagos Hawks are presently found through driV (Slatkin 1985). on nine islands: Santa Fe, Española, Pinzón, Populations on islands oV en have lower Santiago, Santa Cruz, Isabela, Fernandina, levels of genetic variation than those on the Pinta, and Marchena (Fig. 1). Historically, mainland (Frankham 1997). Populations of humans have shot them, and the species has on island archipelagos tend to be more strongly been extirpated from two human-inhabited diff erentiated than geographically separate islands, San Cristóbal and Floreana. The popu- mainland populations, because water is an eff ec- lation on Santa Cruz (another human-inhabited tive barrier against gene fl ow in many species island) may also have been extirpated; no (Williamson 1981, Boag 1988, Baker et al. 1990). adults have been seen on the island in recent These pa erns of decreased genetic variation years, but juveniles are seen periodically. and increased diff erentiation may result from Distances of <5 km to ~240 km separate islands founder events that occurred at the time of colo- with Galápagos Hawk populations (Fig. 1). The nization. In many cases, though, founding fl ock migration rate between islands is unknown but sizes may be large enough that founder eff ects presumed to be low (de Vries 1975), given that are negligible (e.g. Clegg et al. 2002). Even when most Buteo species are reluctant to cross large the number of founders is known to be small, bodies of water (Kerlinger 1985). Swainson’s subsequent arrival of additional immigrants Hawks (B. swainsoni) are the Galápagos Hawk’s may prevent a measurable founder eff ect (Grant closest mainland relatives (Riesing et al. 2003), et al. 2001). Alternatively, lower variability and and they migrate long distances over land (from increased diff erentiation on islands may be North America to Argentina) but avoid fl ying a ributable to sequential founder events (Clegg over water (Fuller et al. 1998). 1212 B . [Auk, Vol. 122 Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021

Fl. 1. Distribution of the Galápagos Hawk on the Galápagos Islands. All labeled islands cur- rently have Galápagos Hawk populations, except for the three islands that are shaded. Genovesa has never supported a Galápagos Hawk population, and the populations on San Cristóbal and Floreana have been extirpated by humans.

Morphological and behavioral variation in from 2.5 to 4.5 birds (de Vries 1975, Faaborg et al. Galápagos Hawk populations also suggests that 1980, Bollmer et al. 2003). The factors contribut- they are genetically isolated. Galápagos Hawks ing to this variation in mating system (e.g. sex diff er in overall body size, and in allometry to ratio, survivorship) are unstudied but are likely a lesser degree, among islands (de Vries 1973, associated with diff erences in habitat structure Bollmer et al. 2003). The species exhibits coop- and resource availability. erative polyandry; territorial groups consist of Here, we describe the genetic structure of one female and up to eight (usually two or three) all nine populations of Galápagos Hawks (the unrelated males (Faaborg and Pa erson 1981, entire range of the species) using multilocus Faaborg et al. 1995). Paternity is shared within minisatellite DNA markers. Minisatellites— and among broods, though there are oV en more hypervariable regions of DNA consisting of males in a group than the number of chicks pro- tandem repetitions of short units of nucleotides duced per brood (one or two); all birds in the (Jeff reys et al. 1985)—have been used in other group defend the communal territory and care studies to characterize population structure for the brood, including males that are not the (e.g. Freeman-Gallant 1996, Carneiro da Silva genetic sires of the off spring (Faaborg et al. 1995, and Granadeiro 1999, Gullberg et al. 1999, Tarr DeLay et al. 1996). One Galápagos Hawk popu- and Fleischer 1999). We describe the amount lation appears to be monogamous (Española), of genetic variation present in populations and but the rest exhibit cooperative polyandry to measure the degree of diff erentiation among varying degrees, with mean group sizes ranging populations using Wright’s FST , the standardized October 2005] Population Genetics of the Galápagos Hawk 1213 variance in allele frequencies among popula- these loci (Flint et al. 1999). Moreover, Flint et tions (Wright 1951, 1978). We test the prediction al. (1999) cautioned that calculating FST values that genetic variation increases with population between human populations using minisatel- size by using total island area and total area of lites yielded an underestimate of genetic dif- appropriate habitat as indices of population ferentiation as compared with the level found size. Variation in mating system also is pre- using other markers. Therefore, their use in dicted to partly determine genetic variability characterizing population genetic diff erentia- by infl uencing eff ective population size, mostly tion, at least in light of this fi nding, is a statis- through biased sex ratios and variance in repro- tically conservative methodology. However, in ductive success (Nunney 1993, Parker and Waite special cases, such as those involving isolated 1997). In the more polyandrous populations of island vertebrate populations, “the fi xation of Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 Galápagos Hawks, there may be increased vari- restriction-fragment polymorphisms can out- ance in reproductive success and more skewed pace the generation of fragment-length variabil- sex ratios, which would lead to decreased eff ec- ity through recombination” (Gilbert et al. 1990: tive population sizes in relation to total popula- 764). This claim was bu ressed by the fi nding tion size and a more rapid loss in variation. We that all bands were fi xed within one population test for an eff ect of mating system (degree of of the Channel Island fox (Urocyon li oralis), polyandry) on genetic variability aV er control- and that individual foxes within each island ling for island area. Finally, we ask whether geo- had diagnostic, island-specifi c bands. Clearly, graphically closer populations are more similar in this and analogous special cases, “diff erences genetically because of increased gene fl ow or among hypervariable restriction-fragment pro- more recent separation (isolation by distance). fi les can be used to estimate relative genetic variability and to reconstruct the evolutionary M relationships of natural populations” (Gilbert et al. 1990:764), because concerns related to a high Field methods.—We visited the Galápagos mutational rate are largely obviated by the rela- Islands for two to three months between May tive force of genetic driV in small populations. and August each year from 1998 to 2003. In the present study, we extracted DNA and Galápagos Hawks (n = 541) were captured on performed multilocus minisatellite DNA fi n- nine islands: 25 individuals from Santa Fe, 23 gerprinting using the restriction endonuclease from three sites on Española (Gardner Bay, HaeIII and Jeff reys’ probe 33.15 (Jeff reys et al. Punta Suarez, and Punta Cevallos), 287 from 1985), following procedures described in Parker three sites on Santiago (James Bay, Sullivan Bay, et al. (1995). AV er hybridization, we used a Storm and the highlands), 93 from Volcan Alcedo on 820 Phosphorimager (Amersham Biosciences, Isabela, 41 from Pinta, 26 from Marchena, 10 Buckinghamshire, United Kingdom) to visual- from Pinzón, 32 from Fernandina, and 4 from ize fi ngerprints. For most populations, we used Santa Cruz. Birds were caught using two meth- only a subset of the samples (n = 163) for genetic ods: a bal-chatri trap baited with a live prey analyses: 15 from Santa Fe, 15 from Española, 37 such as a rat (Berger and Mueller 1959), from Santiago, 22 from Isabela, 20 from Pinta, or a rope noose on a stick to capture perched 20 from Marchena, and 20 from Fernandina. birds (Faaborg et al. 1980). We banded each From Pinzón and Santa Cruz, we used all birds bird with an aluminum or anodized color band sampled (10 and 4, respectively), and they were (or both) and took two 50-µL blood samples all juveniles. For the other populations, we via venipuncture of the brachial vein. Samples randomly selected individuals from the pool were immediately put into 500 µL of lysis buff er of sampled territorial adults (the class most (100 mM Tris, pH 8.0, 100 mM EDTA, 10 mM likely to consist of nonrelatives). We did not NaCl, 0.5% SDS; Longmire et al. 1988), shaken, run all samples; however, fewer individuals are and stored at ambient temperature. necessary to get a representative sample when, Minisatellite DNA markers.—Use of hypervari- as here, populations are lacking in genetic vari- able multilocus minisatellite profi les (VNTRs) ability. We ran nine gels with 17 to 26 lanes in studies of population genetic diff erentiation each. We ran samples in alternating blocks of is typically problematic, because of constraints three to seven individuals from each island, imposed, in part, by a high mutational rate at so that multiple islands were represented on 1214 B . [Auk, Vol. 122 each gel. We chose four individuals from diff er- regression using the log of total vegetated ent islands as ladders and ran them on each of area (excluding lava and beaches). We tested the gels. From the banding pa erns, we created for an eff ect of mating system with a general a presence–absence matrix of bands (alleles) linear model, using band-sharing values as the encompassing all individuals. Because of high dependent variable, mean group size as a fi xed within-population genetic uniformity, the pres- factor, and log of total island area as a covariate. ence of a number of bands fi xed across popula- Because of the nonindependence of minisatellite tions, and the ladders on each of the gels, we band-sharing values, we fi rst randomly selected were able to reliably score across gels. a subset of independent values (using each indi- We assumed that bands were assorting inde- vidual once) from each population. For mating pendently and calculated within- and between- system, we classifi ed each island as having a Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 island similarity indices as S = 2SAB/(2SAB + NA + mean group size of less than two males or more

NB), where S is the proportion of bands shared, than two males, using published data from de

SAB is the number of bands shared by individu- Vries (1975) and Bollmer et al. (2003) and new als A and B, NA is the number of bands unique data collected from Fernandina in 2003 (1.4 ± 0.5 to individual A, and NB is the number of bands males per group, n = 10 groups). Thus, we classi- unique to individual B (We on et al. 1987; fi ed Española, Santa Fe, Pinzón, and Fernandina Lynch 1988, 1990). We calculated these from our as less polyandrous (mean group sizes of 1–1.5 presence–absence matrix using GELSTATS, ver- males), and Isabela, Santiago, Marchena, and sion 2.6 (Rogstad and Pelikan 1996). Pinta as more polyandrous (mean group sizes In fi ngerprinting, individuals oV en are used of 2.3–3.5 males). We used a Mantel (1967) test in multiple pairwise comparisons, which results to examine isolation by distance (Slatkin 1993), in nonindependence of band-sharing values testing the prediction that genetic diff erentiation

(Danforth and Freeman-Gallant 1996, Call et al. among populations (FST) should increase with 1998, Leonard et al. 1999). We used the p-dif test increasing geographic distance between them. (Bertorelle et al. 1999) in the program WATSON We log-transformed the distance between (Bucchini et al. 1999), a test that permutes indi- islands as measured between nearest points. viduals, not band-sharing values, to ask whether We performed these analyses in SPSS, version within-island band-sharing values signifi cantly 10.0.5 for Windows (SPSS, Chicago, Illinois) and diff ered from between-island values. We calcu- IBDWS, version 2.0 beta (Bohonak 2002). We lated FST values for each pairwise comparison of excluded Santa Cruz from the above analyses islands, as well as an overall archipelago value, because of its small sample size. according to Lynch (1990, 1991). The maximum Because there does not appear to be a breed- value of FST is 1 when two subpopulations are fi xed ing population on Santa Cruz, we performed for diff erent alleles (complete diff erentiation), and an assignment test to see whether the juveniles 0 when alleles are distributed randomly among we captured on Santa Cruz closely matched any subpopulations (no diff erentiation). of the other populations, which would indicate We used a linear regression to test the pre- that they could be migrants. Although there are diction that population genetic uniformity (as no tests designed for codominant minisatellite measured by within-island similarity indices) data, the online program DOH (Brzustowski decreases with increasing island area. We cal- 2002), as fi rst described in Paetkau et al. (1995), culated total island area in ARCMAP, version can accommodate data from dominant markers 9.0, using digitized vegetation coverage maps by treating each band as a separate locus. We obtained from the Research performed a segregation analysis by tallying, Station. Projections were in decimal degrees, within each population, the co-occurrences of so we converted the areas to square kilometers each band with every other band to note cases of (1 degree ≈ 111 km) and used the log of island linkage (bands always appearing together within area in the regression. Large portions of some of individuals) and allelism (individuals always these islands (≤75% of total island area) are bar- having one or the other band but never both, ren of vegetation, which makes them less suit- indicating that they belong to the same locus). able for Galápagos Hawk territories. Total island We found no cases of linkage, and we eliminated area may, therefore, overestimate population all cases of allelism (most a ributable to rare size in some cases, so we performed a second bands) by removing the less-frequent band from October 2005] Population Genetics of the Galápagos Hawk 1215 each allelic dyad. We entered the remaining 23 area (r = –0.844, df = 7, P = 0.008; Fig. 3) and veg- independent bands into the DOH program as etated area (r = –0.846, df = 7, P = 0.008), though presence–absence data for each individual. The there was no substantial diff erence between the program assigns each individual into the popu- two measures. A general linear model showed lation in which its genotype has the highest prob- no eff ect of degree of polyandry on genetic vari- ability of occurring. ability aV er controlling for island area (F = 0.537, P = 0.466, n = 78), whereas there was still a strong Rz island area eff ect aV er controlling for mating system (F = 32.1, P < 0.0001, n = 78). Within-population similarity.—We scored an Population diff erentiation.—Between-island Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 average (± SD) of 14.1 ± 1.42 bands for each indi- FST values ranged from 0.017 to 0.896 (Table 2), vidual. Within-island similarity indices were with an overall archipelago value of 0.538. We high, ranging from 0.693 for Isabela to 0.956 for performed pairwise permutation tests to test Santa Fe (Table 1 and Fig. 2). The mean simi- whether populations were signifi cantly distinct larity index for Santa Cruz was slightly lower from each other. There were 28 pairwise com- (0.657), but this is based on only six pairwise parisons, so we used a Bonferroni correction to comparisons. Birds from Santa Fe were particu- avoid Type I errors, which brought our alpha larly lacking in genetic variation, having only level down to 0.002. Twenty-three of the 28 a few variable bands. Specifi cally, 13 of the 16 comparisons still showed signifi cant diff erences Santa Fe bands scored were fi xed in the popu- among populations (P < 0.001 for all). Four of lation. All 15 Santa Fe birds were identical to the fi ve nonsignifi cant values involved Pinzón two or three other birds, resulting in only four compared with Isabela (P = 0.058), Fernandina diff erent genotypes in that population. In addi- (P = 0.021), Santiago (P = 0.820), and Pinta (P = tion, 4 of the 10 birds on Pinzón were identical, 0.006). The remaining comparison, Isabela ver- whereas there were two sets of identical birds (2 sus Fernandina (P = 0.203), had the lowest FST and 3 birds each) out of 15 individuals sampled value (0.017; Table 2). Three of the fi ve nonsig- on Española and four sets of identical birds (2 or nifi cant values also represent the three smallest 3 birds each for 9 total) on Marchena. The other interisland distances. populations (Isabela, Fernandina, Santiago, and We had predicted that populations would Pinta) were more variable and had no identical exhibit isolation by distance. A Mantel test individuals. confi rmed this, showing a signifi cant pa ern of Regression analyses supported our prediction increasing genetic diff erentiation with increas- that genetic similarity among individuals in a ing distance between islands (r = 0.626, P ≤ population decreases with increasing total island 0.003; Fig. 4). Between-island dispersal.—Over the past few T 1. Mean within-island Galápagos Hawk decades, juveniles have occasionally been minisatellite band-sharing value (± SD), total seen on islands where there was no resident island area, and percentage of each island Galápagos Hawk population, but no individual that is vegetated (not lava or beach); islands banded on one island had ever been observed on are listed in order of increasing area as another island. In 2003, however, we observed calculated from digitized maps. two banded individuals on Fernandina, an island where Galápagos Hawks had not previ- Area Percentage ously been studied. One individual, a territo- Island Within-island S (km2) vegetated rial adult female, had been banded by us as a Pinzón 0.903 ± 0.067 18.1 95.2 second-year juvenile on Volcan Alcedo, Isabela, Santa Fe 0.956 ± 0.032 24.8 100.0 in 1998. The other bird was a territorial male Pinta 0.765 ± 0.083 59.4 62.0 whose band could not be read. It is very likely Española 0.900 ± 0.052 61.1 98.2 that he was also banded as a juvenile on Alcedo Marchena 0.891 ± 0.047 128.8 25.4 in 1998, given that 70 birds were caught there in Santiago 0.711 ± 0.086 577.5 68.6 two days, 64 of which were juveniles. Also, it is Fernandina 0.719 ± 0.101 647.6 30.5 unlikely that he came from an island other than Santa Cruz 0.657 ± 0.157 984.1 100.0 Isabela, because Isabela separates Fernandina Isabela 0.693 ± 0.086 4,710.7 66.5 from all the other islands (Fig. 1). 1216 B . [Auk, Vol. 122 Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021

Fl. 2. An example of a multilocus minisatellite DNA fingerprinting gel of Galápagos Hawks. Each lane represents the fingerprint of an individual randomly selected from those sampled on the four study islands named above the gel. Some of these populations exhibit the highest levels of monomor- phism at minisatellite loci of any natural bird population studied. Note that several bands are unique to or fixed in their respective island populations, highlighting the powerful effect that genetic drift has had in this system in limiting neutral genetic variance within islands and increasing it among islands.

In Table 3, we present the results of the assign- juveniles into the populations they most closely ment test for each population. The program matched. One of the four individuals caught on accurately assigned all the individuals from Santa Cruz had a banding pa ern identical to the more genetically monomorphic Española, that of one of the Santa Fe genotypes, and the Santa Fe, Pinzón, and Marchena populations assignment test placed it within the Santa Fe to their home islands, though there were misas- population. Another of the Santa Cruz individ- signments among the larger populations, likely uals had a banding pa ern very similar to those owing to their greater genetic variability. The on Pinzón (mean band-sharing between it and assignment test placed the four Santa Cruz the Pinzón individuals was 0.911 ± 0.03), and October 2005] Population Genetics of the Galápagos Hawk 1217 Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021

Fl. 3. Plot of mean genetic similarity (± SD) of Galápagos Hawk individuals within islands against the log of island area (km2). Data support our prediction that within-population genetic similarity should decrease with increasing island size. the assignment test placed it within the Pinzón even higher mean band-sharing values for population. The last two Santa Cruz individuals populations of Channel Island foxes, another matched Santiago best, though the chance for top predator, ranging from 0.75 up to 1.00. By an assignment error is higher for the more vari- contrast, unrelated birds in outbred mainland able populations. populations typically have band-sharing val- ues around 0.2 and 0.3 (Parker Rabenold et al. Dz 1991, Papangelou et al. 1998). Although there are no published studies using minisatellites in Genetic variation within populations.—We were other Buteo species, mean band-sharing within able to characterize population genetic structure a small sample of overwintering Swainson’s of nine Galápagos Hawk populations, covering Hawks was 0.374 ± 0.10 (n = 8; J. L. Bollmer the entire species range. The populations exhib- et al. unpubl. data). So the Galápagos Hawk’s ited very li le genetic variation, having within- ancestral mainland polymorphism was likely population similarity indices ranging from 0.6 much higher. to >0.9 at hypervariable minisatellite loci. To Extremely low genetic variability within this our knowledge, the smaller Galápagos Hawk species is probably the result of a single founder populations have the highest reported levels event coupled with long-term genetic driV . The of monomorphism at minisatellite loci of any Buteo phylogeny by Riesing et al. (2003) shows natural bird population, though some popula- a very recent divergence between Galápagos tions of New Zealand birds (reviewed in Miller and Swainson’s hawks, and mitochondrial DNA et al. 2003) and other endangered island bird (mtDNA) work underway on the Galápagos species (e.g. Rave 1995, Caparroz et al. 2001) Hawks indicates a single colonization event from are nearly as inbred. Gilbert et al. (1990) found the mainland (J. L. Bollmer et al. unpubl. data). 1218 B . [Auk, Vol. 122 values are reported below values Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 ST F erentiation in Galápagos Hawks. Mean between-island band-sharing values (± SD) are band-sharing values erentiation in Galápagos Hawks. Mean between-island the diagonal. above the diagonal, with total number and of independent pairwise comparisons scored in parentheses. above Española Santa Fe Pinzón Isabela Fernandina Santiago Marchena Pinta Marchena Santiago Fernandina Isabela Pinzón Fe 20) Santa Española Española (740, 0.08 Santa Fe Pinzón – 20) 0.896 ± 20) Isabela 0.306 ± 0.03 Fernandina 0.714 0.656 ± 0.04 – 0.872 0.583 0.421 0.472 0.393 Santiago 0.551 (740, 0.753 0.546 ± 0.08 0.591 – 0.752 0.862 Marchena (400, (225, 15) 0.534 ± 0.10 0.489 ± 0.04 0.659 0.522 0.737 0.341 0.264 0.708 0.291 0.213 0.304 Pinta – 0.593 ± 0.08 0.617 0.485 ± 0.08 (150, 10) – 0.579 ± 0.05 0.661 0.443 ± 0.08 0.322 0.335 0.563 ± 0.70 (330, 15) 0.509 ± 0.07 0.702 ± 0.08 (150, 10) 0.266 0.404 ± 0.05 (300, 15) 0.017 0.716 ± 0.09 – 0.470 ± 0.07 (330, 15) 0.737 ± 0.07 (555, 15) 0.100 0.753 ± 0.05 0.701 ± 0.09 (300, 15) – (300, 15) 0.748 ± 0.07 0.669 ± 0.09 (220, 10) (555, 15) 0.123 0.641 ± 0.08 (300, 15) 0.675 ± 0.09 (200, 10) 0.632 ± 0.09 (300, 15) 0.631 ± 0.08 – (370, 10) 0.636 ± 0.10 (300, 15) (440, 20) (200, 10) 0.672 ± 0.07 (814, 22) 0.667 ± 0.08 (200, 10) (440, 20) (740, 20) (440, 20) (400, 20) (400, 20) T 2. Pairwise comparisons of between-island diff comparisons of between-island 2. Pairwise T October 2005] Population Genetics of the Galápagos Hawk 1219 Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021

Fl. 4. Plot of pairwise interisland FST values against the log of geographic distances (km) between islands for Galápagos Hawks. Degree of genetic differentiation between populations increases with increasing geographic distance.

T 3. Results of Galápagos Hawk assignment test using minisatellite data. Rows represent the populations in which we sampled the individuals, and columns represent the populations to which DOH assigned the individuals. Santa Cruz is listed only as an island of capture, because there is no resident Galápagos Hawk population there with which possible migrants could be compared.

Española Santa Fe Pinzón Isabela Fernandina Santiago Marchena Pinta Española 15 0 0 0 0 0 0 0 Santa Fe 0 15 0 0 0 0 0 0 Pinzón 0 0 10 0 0 0 0 0 Isabela 0 0 2 10 8 2 0 0 Fernandina 0 0 0 5 13 2 0 0 Santiago 0 0 1 5 5 23 0 3 Marchena 0 0 0 0 0 0 20 0 Pinta 0 0 1 1 0 0 5 13 Santa Cruz 0 1 1 0 0 2 0 0

Although there is evidence that island coloni- a severe bo leneck occurred. The high mean zations may not always result in a signifi cant interisland band-sharing (0.617) and the pres- decrease in genetic diversity (Clegg et al. 2002, ence of bands that are fi xed across all populations Grant 2002), in the present case, the founding (even though most populations are currently population may have been small enough that genetically isolated) suggest that Galápagos 1220 B . [Auk, Vol. 122

Hawks became inbred early in their colonization migratory behavior of their closest mainland of the islands. The close relationship between relatives (Fuller et al. 1998). Most Galápagos island area and genetic variation across popula- Hawk populations appear to be signifi cantly tions indicates that long-term genetic driV has genetically diff erent from each other, with the also been an important factor infl uencing the level exception of the interaction between Isabela and of variability in Galápagos Hawks. The smallest Fernandina and four comparisons involving populations have become fi xed or nearly fi xed for Pinzón. The comparisons involving Pinzón are many of their bands, with diff erent bands being more suspect, given that we sampled only 10 common in diff erent populations. individuals on Pinzón, all of which were fl oater Within-island genetic uniformity decreased juveniles instead of territorial adults. Also, the signifi cantly with increasing population size, use of the Bonferroni correction increased the Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 as approximated by total island area and vege- probability of Type II errors, especially for the tated area. Although total island area explained two comparisons with P-values of 0.006 (Pinzón a large portion of the variance in genetic simi- vs. Pinta) and 0.021 (Pinzón vs. Fernandina). larity (r = –0.844), we had supposed that popu- These two comparisons are also the most geo- lation size (and thus genetic variability) would graphically distant of the nonsignifi cant values. correlate even more strongly with vegetated The populations were divergent to varying area, given the presence of large tracts of bar- degrees, as indicated by the pa ern of isolation ren lava on some islands. Using only vegetated by distance. Lower FST values between nearby area, however, did not substantially improve populations may be the result of ongoing (albeit the correlation (r = –0.846), even though fi ve relatively rare in most cases) gene fl ow between of the islands are <70% vegetated, two greatly them, more recent population separation, or a less. We excluded Santa Cruz from this analysis combination of the two. Española and Santa Fe because it diff ers from the other islands, in that were the most divergent from the rest of the human activities have made the population on archipelago, with FST values between them and this large island artifi cially small. Even though the other islands ranging from 0.5 to 0.9. Their the Santa Cruz population is almost certainly relatively extreme divergence (especially from the smallest in the archipelago, the four juvenile each other) is likely a ributable to the random Galápagos Hawks sampled there exhibited the fi xation of alleles in these populations that are lowest mean similarity of any of the popula- not common on other islands. tions, probably because of interisland move- Fernandina and Isabela were indistinguish- ments of birds, which we will discuss below. able at these minisatellite loci. Of all island We found that there was no eff ect of mat- pairs, they are separated by the shortest dis- ing system on genetic variability of Galápagos tance (<5 km), and we observed a bird banded Hawk populations. We had predicted that on Isabela residing in a territory on Fernandina. increased polyandry might result in lowered The lack of diff erentiation between these two eff ective population sizes in relation to total populations, therefore, may be a ributable to population size because of increased variance ongoing gene fl ow. Alternatively, their similar- in male reproductive success or more strongly ity may have resulted from more recent separa- biased sex ratios. The lack of diff erence between tion or from driV acting more slowly in larger low- and high-polyandry populations shows populations. With the present data, we are that mating system is not a strong determinant unable to distinguish among these scenarios. of genetic variability in the Galápagos Hawk; The four juveniles we captured on Santa shared paternity may mitigate the eff ects of Cruz are likely migrants from neighboring increased polyandry. Also, population size islands. When fl edglings leave their territo- accounts for such a large portion of the variance ries, they spend at least three or four years in in within-island genetic similarity that there a nonterritorial fl oater population, roaming all is li le remaining variability on which other over their native island and occupying areas forces might act. not used by territorial birds (de Vries 1975). Genetic divergence among populations.— Because of this nomadic behavior, we sug-

Overall, the high FST values indicate that gest that juveniles are much more likely than Galápagos Hawks are reluctant to cross large adults to move between islands. Dispersal of stretches of water, which is consistent with the juveniles to Santa Cruz may be more probable October 2005] Population Genetics of the Galápagos Hawk 1221 than movement to other islands, because Santa to the Charles Darwin Research Station and the Cruz is mostly or entirely uninhabited by a Galápagos National Park for allowing us to do territorial adult population, which means that this project and for logistical support in the fi eld. suitable habitat is vacant, and juveniles are not Also, TAME, the Ecuadorian airline, provided likely to be harassed and driven away by adults. discounted fl ights to Galápagos. The Parker lab The assignment test placed two of the birds into group made helpful comments on an earlier ver- the Santa Fe and Pinzón populations with high sion of the manuscript. This work was funded degrees of probability. The other two were most by the National Science Foundation (grant similar to Santiago, though there is more likely no. INT-9722735; Dissertation Enhancement to be a misassignment when dealing with more grant no. INT-030759 to P.G.P. and N.K.W.), variable populations. Santiago is a likely source the National Geographic Society (grant no. Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 population because it supports a large fl oater 6821-00), Sigma Xi, the International Center population and is an adjacent island. We can- for Tropical Ecology, and the Saint Louis Zoo’s not eliminate the possibility that one or more of Field Research for Conservation Program. these birds was born on Santa Cruz, because we could not compare them to a sample of resident Lz C Santa Cruz territorial birds, given the lack of known breeding adults there. B, A. J., M. D. D, A. L~, Archipelagoes are well known as arenas G. L G. 1990. Genetic divergence for species radiations (e.g. Darwin’s fi nches, in peripherally isolated populations Hawaiian honeycreepers). Although we have of Chaffi nches in the Atlantic islands. described morphological and behavioral dif- Evolution 44:981–999. ferences (Bollmer et al. 2003) and genetic B, N. H. 1998. Natural selection and diff erentiation (present study) among popula- random genetic driV as causes of evolution tions of Galápagos Hawks, these diff erences on islands. Pages 102–123 in Evolution on are on a microevolutionary scale. Presumably, Islands (P. R. Grant, Ed.). Oxford University Galápagos Hawks are one of the more recent Press, Oxford. arrivals to the archipelago, and have not been Bl, D. D., H. C. Mz. 1959. The there long enough to diverge into subspecies or bal-chatri: A trap for the birds of prey. Bird new species. DriV has had a strong infl uence on Banding 30:18–26. divergence at these neutral minisatellite mark- B, G., L. Bz, A. P, ers, but the importance of driV in speciation C. M. 1999. DNA fi ngerprinting data is debatable (Barton 1998). Given the genetic and the analysis of population genetic struc- isolation of many of these populations, the ture by comparing band-sharing pa erns. Galápagos Hawk may one day match the pat- Molecular Ecology 8:1851–1866. terns seen in other sedentary species groups Bl, P. T. 1988. The genetics of island birds. in the archipelago (e.g. Galápagos Pages 1550–1563 in Acta XIX Congressus [Geochelone elephantopus subspp.], lava Internationalis Ornithologici (H. Ouellet, [Microlophus spp.]), with multiple subspecies or Ed.). National Museum of Natural Sciences, species restricted to one or a few islands. O awa. B, A. J. 1999. Dispersal, gene fl ow, and A|l population structure. Quarterly Review of Biology 74:21–45. We thank T. Sanchez, D. Sanchez, S. Struve, B, A. J. 2002. IBD (isolation by distance): B. Cannon, K. Huyvaert, K. Levenstein, G. A program for analyses of isolation by dis- Jimenez, P. Jimenez, A. Lara, P. Castillo, G. tance. Journal of Heredity 93:153–154. Scacco, D. Santiago, and P. Sanchez for their B, J. L., T. S, M. Dl~ C, invaluable help in collecting blood samples. D. S, B. C, J. C. B, T. J. C. Bollmer produced the map for the fi rst fi g- V, M. S. Sz{, P. G. P. 2003. ure. J. Faaborg aided in the initial formation of Variation in morphology and mating system this project and in securing funds. We thank D. among island populations of Galápagos Wiedenfeld, M. Soria, C. Duffi e, and S. Siers for Hawks. Condor 105:428–438. help with the GIS portion. We are very grateful Bz, L., A. L, C. M. 1999. 1222 B . [Auk, Vol. 122

WATSON: A tool to analyse DNA fi n- polyandry in the Galápagos Hawk (Buteo gerprinting data. [Online.] Available at galapagoensis). Auk 97:581–590. www.igm.cnr.it/watson/manualwatson.html. Fl, J., P. G. P, L. DL~, T. V, Bz|, J. 2002. DOH assignment J. C. B, S. M. P, J. N, test calculator. [Online.] Available at T. A. W. 1995. Confi rmation of coop- www2.biology.ualberta.ca/jbrzusto/ erative polyandry in the Galápagos Hawk Doh.php. (Buteo galapagoensis). C, D. R., J. G. H, S. G. M, M. and Sociobiology 36:83–90. E{. 1998. Considerations for measuring Fl, J., C. B. P. 1981. The genetic variation and population structure characteristics and occurrence of coopera- with multilocus fi ngerprinting. Molecular tive polyandry. Ibis 123:477–484. Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 Ecology 7:1337–1346. F, J., J. B, D. C. R, A. J. B~, J. M. Cu, R., C. Y. M~, M. I. Bu, R-T, L. E}, J. B. Cll, A. W. 2001. Analysis of the genetic M. A. Bz, R. A. N, R. M. variability in a sample of the remaining Hl. 1999. Minisatellite mutational

group of Spix’s Macaw (Cyanopsi a spixii, processes reduce Fst estimates. Human Psi aciformes: Aves) by DNA fi ngerprint- Genetics 105:567–576. ing. Biological Conservation 99:307–311. F, R. 1996. Relationship of genetic C S{, M., J. P. G. variation to population size in wildlife. 1999. Genetic variability and isolation of Conservation Biology 10:1500–1508. Cory’s Shearwater colonies in the Northeast F, R. 1997. Do island populations have Atlantic. Condor 101:174–179. less genetic variation than mainland popu- C, D. M., R. A. Dz, A. R. MB~, lations? Heredity 78:311–327. M. A. R, W. M. W, K. S. Huu, F-G, C. R. 1996. Microgeographic C. G. F}. 1992. Drowned islands pa erns of genetic and morphological downstream from the Galápagos hotspot variation in Savannah Sparrows (Passerculus imply extended speciation times. Nature sandwichensis). Evolution 50:1631–1637. 355:246–248. Fz, M. R., W. S. Sl, L. S. Sz. Cll, S. M., S. M. Dl, J. K|, C. 1998. Routes and travel rates of migrat- M, A. Ezu, I. P. F. O|. ing Peregrine Falcons Falco peregrinus and 2002. Genetic consequences of sequential Swainson’s Hawks Buteo swainsoni in the founder events by an island-colonizing bird. Western Hemisphere. Journal of Avian Proceedings of the National Academy of Biology 29:433–440. Sciences USA 99:8127–8132. G, D. A., N. L, S. J. O’B, R. K. D, B. N., C. R. F-G. W~. 1990. Genetic fi ngerprinting refl ects 1996. DNA fi ngerprinting data and the population diff erentiation in the California problem of non-independence among pair- Channel Island fox. Nature 344:764–767. wise comparisons. Molecular Ecology 5: G, P. R. 2002. Founder eff ects and silver- 221–227. eyes. Proceedings of the National Academy DL~, L. S., J. Fl, J. N, S. M. P, of Sciences USA 99:7818–7820. T. V, P. G. P. 1996. Paternal G, P. R., B. R. G, K. P. 2001. care in the cooperatively polyandrous A population founded by a single pair of Galápagos Hawk. Condor 98:300–311. individuals: Establishment, expansion, and V, T. 1973. The Galápagos Hawk, an eco- evolution. Genetica 112–113:359–382. geographical study with special reference to Gzl, A., M. O, H. Tl¾. its systematic position. Ph.D. dissertation, 1999. Evolution in populations of Swedish Vrij e University, Amsterdam. sand lizards: Genetic diff erentiation and V, T. 1975. The breeding biology of the loss of variability revealed by multilocus Galápagos Hawk, Buteo galapagoensis. Le DNA fi ngerprinting. Journal of Evolutionary Gerfaut 65:29–57. Biology 12:17–26. Fl, J., T. V, C. B. P, J~, A. J., V. W, S. L. T. 1985. C. R. G. 1980. Preliminary observa- Hypervariable ‘minisatellite’ regions in tions on the occurrence and evolution of human DNA. Nature 314:67–73. October 2005] Population Genetics of the Galápagos Hawk 1223

Kl, P. 1985. Water-crossing behavior of Ozl, N. J., Y. Pvz, J. M. V raptors during migration. Wilson Bulletin G. 1999. Population genetics, 97:109–113. molecular markers and the study of disper- L, A. C., S. E. F, V. S. Hl, sal in plants. Journal of Ecology 87:551–568. M. K. S, G. P. T. 1999. Hypothesis Pz, D., W. C{, I. Sl, C. testing with the similarity index. Molecular S. 1995. Microsatellite analysis of Ecology 8:2105–2114. population structure in Canadian polar Ll, J. L., A. K. L|, N. C. B|, J. M. bears. Molecular Ecology 4:347–354. Bzl, L. M. C, M. D. J, L. J. Pulz, A., M. H. H, M. M. M, J. M~, R. L. R, F. A. R~, M~. 1998. Variation of multilocus . 1988. Isolation and molecular minisatellite DNA fi ngerprints in avian Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 characterization of a highly polymorphic populations. Zoological Studies 37:161–168. centromeric tandem repeat in the family P, P. G., T. A. W. 1997. Mating Falconidae. Genomics 2:14–24. systems, eff ective population size, and L~, M. 1988. Estimation of relatedness by conservation of natural populations. Pages DNA fi ngerprinting. Molecular Biology and 244–262 in Behavioral Approaches to Evolution 5:584–599. Conservation in the Wild (J. Clemmons and L~, M. 1990. The similarity index and DNA R. Bucholz, Eds.). Cambridge University fi ngerprinting. Molecular Biology and Press, Cambridge, United Kingdom. Evolution 7:478–484. P, P. G., T. A. W, M. D. D. L~, M. 1991. Analysis of population genetic 1995. Kinship and association in com- structure by DNA fi ngerprinting. Pages munally roosting Black Vultures. Animal 113–126 in DNA Fingerprinting: Approaches Behaviour 49:395–401. and Applications (T. Burke, G. Dolf, A. J. P R, P., K. N. R, W. H. Jeff reys, and R. Wolff , Eds.). Birkhäuser Pu, M. D. D, J. H~. 1991 Verlag, Basel, Switzerland. Using DNA fi ngerprinting to assess kinship M, N. 1967. The detection of disease and genetic structure in avian populations. clustering and a generalized regression Pages 611–620 in The Unity of Evolutionary approach. Cancer Research 27:209–220. Biology (E. C. Dudley, Ed.). Dioscorides, M, H. C., D. M. L, C. D. M, Portland, Oregon. B. C. R, E. O. M. 2003. R{, E. H. 1995. Genetic analyses of wild Minisatellite DNA profi ling detects lineages populations of Hawaiian Geese using DNA and parentage in the endangered Kakapo fi ngerprinting. Condor 97:82–90. (Strigops habroptilus) despite low microsatel- Rl, M. J., L. Kzz, A. Gz, lite DNA variation. Conservation Genetics 4: E. Hl. 2003. Molecular phylog- 265–274. eny of the Buteo (Aves: ) Ml, W. J. 1971. Convection plumes in the based on mitochondrial marker sequences. lower mantle. Nature 230:42–43. Molecular Phylogenetics and Evolution 27: Mz~, N. I., C. S. W, T. Bz, D. S. 328–342. Wz. 1997. Microsatellite variation and Rl, S. H., S. P. 1996. GELSTATS: microevolution in the critically endangered San A computer program for population genet- Clemente Island Loggerhead Shrike (Lanius ics analyses using VNTR multilocus probe ludovicianus mearnsi). Proceedings of the Royal data. BioTechniques 21:1128–1131. Society of London, Series B 264:869–875. S, M. 1985. Gene fl ow in natural popu- N{, E., A. B, R. B-S. 1984. lations. Annual Review of Ecology and The evolutionary signifi cance of genetic- Systematics 16:393–430. diversity: Ecological, demographic and S, M. 1993. Isolation by distance in life history correlates. Pages 13–213 in equilibrium and non-equilibrium popula- Evolutionary Dynamics of Genetic Diversity tions. Evolution 47:264–279. (G. S. Mani, Ed.). Springer-Verlag, Berlin. T, C. L., R. C. F. 1999. Nz~, L. 1993. The infl uence of mating system Population boundaries and genetic diversity and overlapping generations on eff ective in the endangered Mariana Crow (Corvus population size. Evolution 47:1329–1341. kubaryi). Molecular Ecology 8:941–949. 1224 B . [Auk, Vol. 122

W, R., K. H. 2003. New volca- a pathological mantle plume. Journal of nological and volatile data provide strong Geophysical Research B, Solid Earth 98: support for the continuous existence of 19533–19563. Galápagos Islands over the past 17 mil- W, M. 1981. Island Populations. lion years. International Journal of Earth Oxford University Press, Oxford. Sciences 92:904–911. Wl, S. 1951. The genetical structure of pop- W, J. H., R. E. C, D. T. P, ulations. Annals of Eugenics 15:323–354. D. W. 1987. Demographic study of a Wl, S. 1978. Evolution and the Genetics of wild House Sparrow population by DNA Populations, vol. 4: Variability within and fi ngerprinting. Nature 327:147–149. among Natural Populations. University of W, W. M., A. R. MB~, R. A. Chicago Press, Chicago. Downloaded from https://academic.oup.com/auk/article/122/4/1210/5562562 by guest on 02 October 2021 Dz. 1993. Petrology and geochem- istry of the Galápagos Islands: Portrait of Associate Editor: K. Steenhof