Divergence of Adaptive Phenotypic Traits and Mate

DIVERGENCE OF ADAPTIVE PHENOTYPIC TRAITS AND MATE RECOGNITION SYSTEMS (SONG AND COLOR) IN RESPONSE TO RECENT ANTHROPOGENIC HABITAT CHANGES IN AN OCEANIC ISLAND (SÃO TOMÉ, GULF OF GUINEA) ENDEMIC PASSERINE (SPEIROPS LUGUBRIS).

Final project report submitted in candidacy for the degree of Biologist

ANDREA CAROLINA BAQUERO LOZANO

Director: MARTIM PINHEIRO DE MELO PhD. Biology

Co-directors: CARLOS DANIEL CADENA PhD. Biology

CLAIRE DOUTRELANT PhD. Biology

UNIVERSIDAD DE LOS ANDES FACULTAD DE CIENCIAS DEPARTAMENTO DE CIENCIAS BIOLÓGICAS BOGOTÁ D.C., COLOMBIA 2008 ABSTRACT

The impact of human activities on natural ecosystems is evident and its consequences are now more profound. The changes in direction and strength of natural and sexual selection that arise in human altered environments due to changes in ecological factors might promote divergence among populations even if these are not isolated, provided selection is sufficiently strong. In birds, such novel selective pressures may lead to divergence in phenotypic traits affecting both ecological adaptation and mate recognition. We examine the effects of recent anthropogenic change in the divergence of some phenotypic patterns of Speirops lugubris, an endemic passerine of Sao Tome Island. Data for morphology, plumage color and song were collected from individuals on both primary forest and shade forest plantations, a habitat less than 200 years-old. Results showed that birds in shade forest plantations were significantly smaller than birds in undisturbed forest however, no plumage color differences between habitats was found. The physical properties of song related to vegetation structure characteristics that differed between habitats and to morphological variation found. This observed change shows that mate recognition signals, a fundamental trait for fitness, are able to respond quickly to habitat changes. This study suggests that the endemic Speirops lugubris is experiencing an adaptive and potentially evolutionary response to diverging habitat structures caused by anthropogenic destruction of the original environment.

1 ACKNOWLEDGMENTS

This project has been a great experience both academically and personally. For that I would like to thank Martim Melo, my principal supervisor for this project, who has been supportive and enthusiastic from day one. He suggested and guided this project every step of the way, leading me in the lab and in the field (and getting to the field all the way in São Tomé!). It was a great pleasure working with Martim and learning from him. He was always willing to help and patiently answer questions, even when it took adjusting time schedules for telephone meetings or finding the only computer with working internet in Principe to answer my emails. Martim always was a great motivator and mentor, again I am very grateful.

Many thanks go to Claire Doutrelant, who along with Martim Melo, blindly took me in as an intern in the CEFE and gave me the great opportunity of starting this project. She taught and supervised me in all the plumage color analyses and was a great help in the revision of this paper. She was always very welcoming and helped me around the CEFE.

I would like to thank Daniel Cadena for willing to co-direct this project, for his great patience, very helpful corrections and guidance. His work and courses have been a big influence in this project and will be a guideline for the future.

Special thanks go to Rita Covas was always helpful with suggestions and who along with Martim and Francisca, was a great friend and company in São Tomé, and thereafter. Fieldwork would not have been the same without Rita’s Portuguese lessons, the nice talks and playing around with Francisca.

I am grateful to Pablo Stevenson for motivating my interest in ecology and giving me the chance to learn from him in the field several times. I also thank Esteban Payan for helping me with the grant application processes and for his advice.

The work in São Tomé was made possible with the help of the Associação Monte Pico, especially my guides Gabriel, Antonio and Señor Pedro. Guillermino, Luis Mario, Lagoas and Georgina were a great help around the island and made me feel at home, as well as Octavio and his family who opened up their home to me. Steffan Andersson and Maria Prager for letting me stay with them when I was ill and kindly providing their house for the last weeks.

I am thankful for all the help from people of the Laboratorio de Biología Evolutiva de Vertebrados, Universidad de los Andes in Bogota, Colombia and the Centre d’Ecologie Fontionnelle et Evolutive, CNRS in Montpellier, France.

2 To my friends in the Laboratorio de Biologia Evolutiva de Vertebrados, especially to Angela, Juan Camilo and Carlos P for all their help when writing this paper. Several people helped me along the way. I thank Helena and Manuel, Carmine and Melba, and Jenny Bravo for letting me into their homes on my way to the island.

In Montpellier, I would like to thank my dear friends for helping me in several ways. Luisa, Julio, Jenny, Sarah, Douniah, Stephanie, Romain, Laetitia, and Sebastien. I thank Alex Courtiol for his great patience teaching R and helping me with the initial statistics of this project.

To my friends in Bogota for being supportive, giving advice and just being there for me. Caya, Camila L, Andres H, Valeria, Lina Q, Lina V, Vicky, Ana G, Ana P, Laura, Alex y Andres P (Mario). Also to Camilo for always helping me out and being there when I needed a little push.

Greatest thanks go to my family for always being there for me every step of the way from the very start, even when I was determined to go all the way to São Tomé, which seemed a crazy idea at the time. For understanding and supporting me with my ideas and goals and for bearing with me in times of stress. They endlessly listed to me talk about the project, and motivated me even when they didn’t understand what the excitement was about. Simply for their love and belief in me, thanks.

I am very thankful for the financial support for this project, provided by The Rufford Small Grants Foundation for Nature Conservation and The British Ecological Society.

3 TABLE OF CONTENTS

Abstract 1 Acknowledgments 2 Contents 4

Introduction 5

Methods 8 Study site 8 Study specie 10 Field sampling 10 Habitat structure measurements 13 Morphological analyses 14 Plumage color analyses 15 Song analyses 16

Results 17 Habitat structure measurements 17 Morphological analyses 18 Plumage color analyses 21 Song analyses 23

Discussion 25 Morphological analyses 25 Plumage color analyses 26 Song analyses 27

Conclusions 28 Future directions 29 References 29

INTRODUCTION

The impact of human activities on natural ecosystems is ubiquitous. Most research on this fundamental issue has focused on ecological impacts such as describing changes in community composition or in functional disruptions resulting from the loss or gain of particular species (e.g., predators, pathogens, invasive species), and also on following or predicting demographic changes with the goal of preventing species extinctions (Myers et al 2000, Foley et al 2005). Until recently, little attention had been paid to the influences of human action on evolutionary processes, but a growing interest has led to studies evidencing that human-driven evolutionary change is having a widespread impact, posing threats to the natural evolutionary course of species (Seehausen et al 1997, Smith et al 2008, Smith and Bernatchez 2008).

Humans can drive phenotypic change in contemporary populations by exposing them to dramatic environmental perturbations that exert novel and strong selective forces driving adaptive divergence over short timescales (Hendry and Kinnison 1999, Reznick and Ghalambor 2001, Palumbi 2001, Stockwell et al 2003, Smith et al 2005, Bell and Collins 2008, Smith et al 2008). Under novel conditions, individuals might survive and reproduce through phenotypic plasticity, and populations might eventually evolve to adapt to such conditions if they harbor sufficient genetic variation to respond to selection (Stockwell et al 2003, Charmantier et al 2008, Hendry et al 2007). Alternatively, environmental perturbations may be too strong for individuals to survive or reproduce successfully, which leads to population declines (Bell and Collins 2008). Accordingly, attention should be directed to understanding and predicting how and under which conditions would populations persist and respond adaptively to the increasingly rapid and strong changes in selective pressures brought about by human activities (Bell and Collins 2008, Hendry et al 2007, Smith and Bernatchez 2008).

5 The changes in direction and strength of natural and sexual selection that arise in human altered environments due to changes in ecological factors might promote divergence among populations even if these are not isolated, provided selection is sufficiently strong (Smith et al 1997, Smith et al 2005, Smith et al 2008). In birds, such novel selective pressures may lead to divergence in phenotypic traits affecting both ecological adaptation and mate recognition (Gibbs and Grant 1987, Badyaev and Leaf 1997, Smith et al 1997, McNaught and Owens 2002, Seddon 2005, Seehausen 2006a). Specifically, morphological characters are often related to variation in feeding ecology, flight performance, and fitness and because these characters are highly heritable in birds (Schluter and Smith 1986, Gibbs & Grant 1987, Smith 1990) they are amenable to diverge even in the face of gene flow (Smith et al 1997). On the other hand, changes in the signaling environment may result in the divergence of traits involved in mate recognition (Endler 1992, Seehausen et al 1997, Boughman 2002) and may eventually lead to the evolution of pre-mating barriers, one of the most important mechanisms in reproductive isolation (Coyne and Orr 2004).

Mate recognition systems in birds are commonly based on plumage coloration and acoustic signaling. Plumage color differences among closely related species are often considered to have evolved as a way of minimizing the risk of hybridization; this species isolation hypothesis is supported by cases of reproductive character displacement between taxa in sympatry (McNaught and Owens 2005). An alternative explanation to this variation in color is the light environment hypothesis in which divergence in coloration presents itself due to the variation of light between their environments (McNaught and Owens 2005). Environmental factors play a major role in color perception given that color perceived by the observer is a product of four main factors: i) reflectance, a property of the sender; ii) receiver sensitivity; iii) transmittance and iv) ambient light. These last two factors are closely related to environmental conditions as the reflected radiance (product of ambient light and reflectance) is the actual signal emitted from the surface of the sender and the fraction of this radiance actually arriving to the receiver depends on transmittance of the medium (Endler 1992; Andersson and Prager, 2006). Different light environments are therefore expected to generate divergent selection on plumage coloration as a

6 mating signal. For example, it is predicted that in “closed” habitats such as forests, a higher hue in plumage coloration is favored, given that reds and oranges contrast well with vegation and have a better reflection of the long wavelength light that characterizes these habitats. However, a higher overall reflectance (brightness) is expected in “open” habitats, as plantations, where a higher brightness would favor communication over long distances by maximizing contrast (Endler 1993, McNaught and Owens 2002).

Song plays a major role in species recognition, territory defense and mate choice in birds (Payne 1986, Slabbekoorn and Smith 2002, Podos et al 2004), and as with color, environmental factors are thought to influence the evolution of song. Divergence of song may arise directly as an adaptation to the signaling environment or indirectly as a by product of morphological adaptation. For instance, a direct influence of the environment on song traits is that since sound propagation is limited in closed forest habitats because dense foliage creates barriers, songs with low- frequency sounds and slower repetition rates are expected to be favored because they travel best in such environments (Doutrelant et al 1999, Boughman 2002, Slabbekoorn and Smith 2002, Badyaev and Leaf 1997, Seddon 2005). An example of indirect influences involves the relationship between morphological traits that can by influenced by selection for reasons unrelated to song, which in turn alter song production: for instance, selection for feeding efficiency leads to adaptive divergence in beak morphology, and birds with larger beaks are unable to produce fast trills and to sing at high frequencies (Podos 2001, Podos et al 2004, Huber and Podos 2006). Body mass also has a strong relationship with song properties as a result of the size of the syringeal membrane: larger birds with larger membranes are unable to generate high frequency sounds (Ryan and Brenowitz 1985, Badyaev and Leaf 1997, Tubaro and Mahler 1998).

In this study, I assess the divergence of morphological and mate recognition traits in response to habitat differentiation caused by human alteration of the environment in the São Tomé Speirops (Speirops lugubris), a passerine bird that is endemic to an

7 oceanic island in the Gulf of Guinea, Africa. I compare morphometrics, plumage coloration and song between individuals occurring in primary forest and in shaded plantations established c. 500 years ago, in order to determine if microevolutionary changes can take place over such a small spatial and temporal scale, and in the presence of gene flow. Results of this study will be of direct relevance for understanding the evolutionary responses of small island populations to human- altered environments.

METHODS

Study Site

The island system that bisects the Gulf of Guinea is made up of four volcanic islands; Bioko, Príncipe, São Tomé and Annobón, from north to south (Fig 1). These islands exhibit high levels of endemism across many groups, making them a major African biodiversity hotspot (Gascoigne 2004). At least 176 endemic angiosperm species have been reported; surprisingly for oceanic islands, eight endemic amphibians have been described, together with eight endemic reptiles, one endemic shrew, and several endemic bats. Avian endemism is extremely high, with the three oceanic islands supporting c. 28 endemic species, most of which are single-island endemics (Jones & Tye 2006).

The oceanic island of São Tóme (0º25’N-0º01’S, 6º28’E-6º45’E), with an area of 857 km2, lies 255 km W from the mainland of Africa (Gabon) and 150 km SSW from the island of Príncipe; Pico de São Tomé represents its highest elevation point (2,024 m ASL). The island is characterized by an oceanic equatorial climate, with a dominant rainforest vegetation; temperatures average around 25ºC and decline with elevation. Small mangroves are found on the coast and rainforest vegetation is thought to have dominated the island originally, with dry woodland vegetation patches near the coast. The original rainforest, which is still present in the southern half of the island, is stratified in: lowland forest, ranging from sea level to 800 m; montane forest, from

8 800m to 1400m, with a dense canopy; and mist forest, above 1400 m, with low temperatures and reduced light caused by high mist levels (Exell 1944). São Tomé supports up to three monospecific endemic genera, 15 single-island endemic species, and it shares five endemic species with Príncipe (Jones and Tye, 2006).

Figure 1. Gulf of Guinea Island System, West Africa

São Tomé was discovered by the Portuguese in 1470 and it is thought that it was uninhabited at the time. Soon after colonization, the island became the prime producer of sugar cane in the world, and cultivation triggered the destruction of the lowland forest at a very fast pace (Jones and Tye, 2006). Further destruction of the forest took place in the early 19th century as a result of the introduction of coffee and cocoa plantations under shade trees, creating “shade forest” which resembles the 9 original rainforest in structure, but differs greatly in vegetation composition (Jones and Tye, 2006). Despite the destruction of large expanses of the original habitat, many endemic bird species are still common in both shade and secondary forests (Jones and Tye, 2006).

Study Species

The passerine genus Speirops is endemic to the Gulf of Guinea. It comprises the four species of phenotypically ‘aberrant’ white-eyes (Zosteropidae), each co-occuring with a “typical” white-eye in the genus Zosterops. Recent genetic analyses have revealed that Speirops is not monophyletic, and that all its members belong in the genus Zosterops (Melo 2007). The species occurring on São Tomé and Principe are a result of very recent speciation events characterized by a very fast rate of phenotypic evolution (Melo, 2007).

Speirops lugubris is endemic to the island of São Tomé, where it is a very common species occurring widely throughout the different ecosystems of the island (Jones & Tye 2006). This species exploits all vegetation strata, from few cm off the ground up to the edge of the forest canopy; it is considered both frugivorous and insectivorous, although nectarivory on Erythrina flowers has also been recorded (Christy and Clarke, 1998). Birds are often seen in couples or family groups and frequently emit contact calls; flocking with other species has also been reported (Christy and Clarke, 1998; Jones and Tye, 2006). Plumage is similar in males and females; both sexes exhibit a general grayish coloration with a marked white line along the body below the wing. A notorious white orbital ring resembles that of the “typical” white-eyed Zosterops ficedulinus feae, with which it co-occurs on the island (Christy and Clarke, 1998).

Field Sampling

10 Martim Melo collected morphological measurements and feathers during two sampling seasons; the first one took place from October 2002 to March 2003 and the second from November 2003 to January 2004. Individuals were captured with mist- nets and ringed on the right leg with individually numbered rings (AFRING, Avian Demography Unit, University of Cape Town). Feathers and morphological measurements were collected from different sites, from 99 and 276 individuals, respectively. However, because some measurements were missing for some individuals, morphological analyses are based on data from only 136 individuals (Table 1, Fig 2). Blood samples were also taken and individuals were sexed following a molecular protocol (Melo, 2007).

I collected song recordings and habitat structure measurements on three different forest sites and four plantation sites on the island from January to mid February, 2008. I used a Marantz PMD222 tape recorder with Type II 60 min tapes, connected to a Sennheiser ME66K6 directional microphone to record songs. Recording took place mainly at dawn, when birds sang the most; however, some recording sessions continued throughout the day until dusk.

Figure 2. São Tomé Island. Sampling locations. Forest sites in green, Plantations sites in Orange. Map adapted from Jones and Tye (2006)

Table 1. Sampling locations and sample sizes. H: Habitat measurements, F: Feathers, M: Morphological measurements, S: Song recordings

ELEVATION SAMPLE SIZE HABITAT SITE LOCATION COORDINATES (m) H F M S

0°16.98’N 1 Lagoa Amelia 1498 13 26 31 24 06°35.426 'E 0°09.200' N 2 Umbumgu 300 - 6 7 - 006°33.860’E 0°14.465' N 3 Formoso 467 10 - - 10 006°37.574’E FOREST 0°16' N 4 Pico Calvário 1587 - 10 9 - 006°34’E 0°09.200' N 5 Quija 658 10 7 16 6 006°33.860’E 0°10' N 6 São Miguel 400 - - 5 - 006°30’E Bom 0°17.396' N 7 1158 10 23 23 8 Sucesso 006°36.793’E 0°17.873' N 8 Monte Café 723 10 17 7 18 006°37.685’E 0°20.194' N 9 Generosa 177 10 - - 7 006°32.401’E PLANTATION 0°14.534 N 10 Bombaim 312 10 - - 6 006°38.296’E 0°11.919' N 11 Alto Douro 150 - 10 26 - 006°42,263’E Contador 0°19 N 12 600 - - 12 - Dam 006°36’E TOTAL 73 99 136 79

Habitat Structure Measurements

In order to obtain a quantitative rather than just a qualitative description of forest and plantation habitat types, I collected vegetation structure measurements at each area point from where I performed song recordings, being careful not to “re-collect” data when recording points were the same (several individuals singing in same point) or very near to each other.

Vegetation structure measurements, adapted from the methodology used by Dallimer and King in 2002, were taken in a 15-meter radius of each recording point. These were:

• Maximum height of the canopy, using a Leica LRF800 Laser Ranger Master (Canopy height). • Number of trees making up the canopy, classified into three size categories according to their diameter at breast height: small (10 – 40 cm), medium (40 cm to 1m) and large (1m and larger). • Percentage of canopy cover, calculated at three different points using a mirror with a 8 x 5 grid of 1 cm2 squares held at waist height and parallel to the ground. The total cover was the number of squares that were at least 90% obscured. (Canopy cover). • Number of stems counted at three different 1m-radius circles, 50 cm above ground (Ground cover). • Abundance of climbers and epiphytes on a scale of 0 (none) to 3 (dense). • Geographical coordinates and elevation were recorded using a GPS unit.

13 A discriminant function analysis was carried out to determine whether forest and plantation habitats could be distinguished based on habitat structure .

Morphological analyses

Morphological analyses comprised a total of 136 individuals that were measured for seven morphological traits: mass, wing, tail and tarsus length; bill length, width and height. Of these, 68 individuals came from forest habitat and the other 68 from plantation habitat. Mass was measured to the nearest 0.5 g with a 50-g Pesola spring balance; for wing and tail length were measured with a standard wing ruler to the nearest 0.5 mm and bill and tarsus measurements were taken with a digital caliper to the nearest 0.1 mm (Melo, 2007).

Morphological variables were tested for normality and non-normal data were log- transformed; for data that could not be normalized, I conducted non-parametric tests. A principal component analysis was done in order to visualize morphological differences between forest and plantation populations and the resulting first component, was used to run an analysis of covariance (ANCOVA) in order to test for difference between habitat types using data from the eight sampling sites, and correcting for elevation by including it as a covariable. To test for differences in the seven morphological characters between habitat types a multivariate analysis of variance (MANOVA) was employed. Differences between males and females were also tested. A Mantel test was used to evaluate whether morphological variation was related to geographic distance between sampling sites, by testing for a correlation between a geographic distance matrix and a morphological distance matrix (Euclidean distances). The Mantel test was performed with SPSS; all other tests were run using R 2.6.2 for Mac.

Plumage Color Analyses

To evaluate the divergence of populations between habitats in plumage coloration, I obtained quantitative measurements of plumage color using reflectance spectrometry. Analyses were carried out with an AvaSpec-2048 spectrometer and an “Avantes” Avalight–DHS lamp that uses both a Halogen and a Deuterium bulb as UV and visible light sources. White and black standards were used as reflectance references (100% and 0%, respectively). The fiber optic probe used was a 200 µm Oceanoptics probe. It was held 2 mm away from the feathers at a 90° angle and excluding external light with a black PVC tube fixed to the probe tip. Forty scans with an integration time of 22 ms were averaged for each spectrum, and were analyzed with AVICOL © 2006 (Doris Gomez).

Analyses were carried out for a total of 99 individuals, 50 of which were from plantation and 49 from forest sites. Color of the head, breast and mantle was measured for all individuals, whereas color of the sub-rump was measured for 72 individuals. Feathers from each individual and distinct body part were grouped into two stacks of five each, to simulate the normal thickness of the bird’s plumage; these feathers were superimposed with the “sun” side of the feather facing up. Whenever the number of feathers was not enough, I measured at least one complete stack and the other with the number of feathers available, with a minimum of three feathers. Five readings were taken from each stack from different points in the feather’s apex, on the basis that it is the part of the feather that shows naturally.

The reflectance spectra obtained were analyzed as three colorimetric variables: spectral location, purity and intensity, referred to here as hue, chroma and brightness,

15 respectively (Box 1). There were no significant color differences between sexes (P>0.05) and thus all samples were analyzed together. Differences between habitat types for all three colorimetric variables in all of the four feather patches were tested with analyses of variance (ANOVAs).

Repeatability of color measurements was calculated to assess the reliability of the multiple measurements taken on each individual and it is given by:

2 2 2 R = S A / (S + S A)

2 2 Where S A is the among-groups variance component and S is the within groups variance component, both derived from a one-way ANOVA (Lessells and Boag 1987).

Box 1. Colorimetric variables extracted from reflectance measurements.

Adapted from Andersson & Prager (2006). R: reflectance; λ: wavelength.

Human Colorimetric Definition Reflectance property perception

Brightness Rsum R summed over λ interval Spectral Intensity

Hue λR50 λ halfway between Rmax and Rmin Spectral Location

Chroma Cmax (Rmax - Rmin) / Ravg Spectral Purity

Song analyses

Analog song recordings were converted into digital format with an Edirol UA-25 audio capture interface with 16 bit accuracy and a sampling rate of 20,500 Hz using AVISOFT-SASLAB PRO version 4.3 (R. Specht, Berlin) on a PC. Song recordings were then analysed on a Mac using Raven Pro sound analysis software (version 1.3;

16 Cornell Laboratory of Ornithology, Ithaca, NY). Spectrograms were analyzed with a fast Fourier transformation-size of 256 points and a temporal resolution overlap of 50%, giving a frequency resolution of 86.1 Hz and a temporal resolution of 5.6 msec. Of the 114 individuals recorded at the 8 sampling sites, 79 recordings were of sufficient quality to be analyzed, of which 40 came from forest sites and 39 from plantation sites.

In order to analyze song structure, a one-minute length sample was taken from each recording, starting at the beginning of the first full vocalization. From each vocalization in that one-minute length sample, temporal and frequency characteristics were measured. Temporal features included: length of vocalization, and of interval time between vocalizations (both in seconds); number of notes per vocalization, and pace, defined as number of notes per second. For spectral features, I measured: lowest and highest frequency; delta frequency, defined as the bandwidth or the high frequency minus the low frequency; and peak frequency, which is the frequency with the highest amplitude in the vocalization (all in kHz) (Price et al, 2007). All eight song variables were averaged for each individual across all vocalizations heard in the interval.

A discriminant function analysis was carried out to visualize song structure differences between forest and plantation populations and to determine whether individuals could be correctly assigned to their habitat according to a linear combination of the acoustic variables. The significance of differentiation between habitat types was tested with a MANOVA.

RESULTS

Habitat Structure Analyses

17 Discriminant function analysis showed total differentiation between forest and plantation habitats, which were mainly separated by Function 1. This function represented 99.3% of the variation and reflects variation in canopy height, canopy cover, number of canopy trees in each size category and epiphyte abundance. DFA classification tests resulted in 100% correct classification of data into the corresponding habitat (λ-W: 0.017, DF: 16, P= 0.000*).

Morphological analyses

Multivariate analysis of variance (MANOVA) analyses run on the seven morphological traits revealed no significant differences between males and females (d.f.: 1, F: 2.785, P>0.05). Therefore, measurements of all individuals were considered as one set. Significant differences were found between habitat types for most characters except for tail length and beak width. The multivariate response variable (MRV) generated from the seven morphological characters, also presented significant difference between habitat types (Table 2). Principal Component Analysis plots showed differentiation for individuals of forest and plantation habitats (Fig. 3), mainly separated by the first principal component (PC1), which explained 42% of the variation. PC1 is a general index of body size axis as it reflects variation all morphological characters except for tail length, which loads heavily on PC2. PC1 was used to run an analysis of covariance (ANCOVA) in order to test for difference between habitat types and along the elevation variance for all nine sampling sites (Fig. 4). Difference was found to be highly significant between habitat types (t-value: - 7.473, P < 0.0001****) and for elevation (t-value: -3.765, P<0.005***). Geographic and morphological distance were weakly but significantly correlated (Mantel Test, ρ 0.1105, p-value < 2.2e-16) (Fig. 5)

Table 2 MANOVA for seven morphological characters between habitat types. Significant values in bold. (****P<0.0001)

Habitat type Character d.f. F P

Mass 1 13.4243 0.0003

Tarsus length 1 11.9208 0.0007

Tail lengthNP 1 0.1497 0.6988

Wing length NP 1 7.7057 0.0055

Beak Length 1 35.4342 ****

Beak Height 1 15.6266 0.0002

Beak WidthNP 1 0.0708 0.7902

MRV 1 8.2280 ****

NP:Non parametric characters for which a Kruskal test was used, F value column, then corresponds to Kruskals’s chi-squared value. MRV: multivariate response variable.

Figure 3. Principal component analysis plot for seven morphological characters. N:136 individuals, Forests N:68, Plantations N:68. PC1 and PC2 explained 61% of the variance. All characters except tail length represent PC1. PC2 is represented mainly by tail length.

Figure 3. Box whisker plots for morphological characters of individuals for forest and plantation habitats. N:136 individuals, Forests N:68, Plantations N:68. MANOVA ****P<0.0001, ***P<0.001, **P<0.005, *P<0.01

Figure 5. Morphological variation (PC1) along an elevational gradient for both habitat types. Blue: Plantation habitat, Green: Forest habitat N:136 individuals, Forests N:68, Plantations N:68.

Plumage Color Analyses There were no significant differences in coloration between sexes for any of the body parts (ANOVA, P>0.05); therefore all data were pooled. ANOVA tests for colorimetric variables in all feather patches resulted in non-significant differences between habitat types (P>0.05) (Fig 6). Measurement repeatabilities for all three colorimetric variables in all of the four feather patches were low: the mean repeatablity across variables was only 0.41 +/- 0.15, with a maximum of 0.67 for the mantle’s chroma and a minimum of 0.05 for the head’s brightness. This low repeatability is possibly a consequence of the high sensitivity of these measurements and the difficulty due to the low reflective signal produced by the brown-grayish color of S. lugubris feathers.

Figure 6. Box whisker plots for colorimetric variables (Brightness, Hue and Chroma) of body feather patches (Breast, Head, Mantle and Subrump). Comparing Forests (N=49) and Plantation (N=50) populations.

Song analyses 22

Discriminant function analysis (DFA) showed differentiation for individuals of forest and plantation habitats (Fig.7), which were mainly separated by Function 1. This function represented 95.6% of the variation and reflects variation in pace, number of notes per song, and peak frequency. DFA classification tests resulted in 88.5% of forest and 63% of plantation individuals correctly classified into their corresponding habitat based on their songs. ANOVA for univariate analyses showed significant differences between habitat types for pace, peak frequency and the number of notes in each vocalization (Fig 8).

Figure 7. Plot of discriminant function scores for Forest (green) and Plantation (Red) habitats. Function 1: Peak Frequency, Number of Notes, and Pace (95.6% of variation). Function 2: Length of song, Low Frequency, Delta Frequency (4.4% of variation). Group centroids are represented by black symbols. λ-W: 0.557, DF: 14, P= 0.014*

Figure 8. Box whisker plots for song variables for forest and plantation. N:79 individuals, Forests N:40, Plantations N:39. ANOVA ****P<0.0001, ***P<0.001, **P<0.005, *P<0.01

24 DISCUSSION

In this study, I examined the effects of human-driven habitat change on the phenotype of an endemic oceanic island passerine, the São Tomé Speirops considering both adaptive morphological traits (body and bill dimensions) and mate recognition traits (color and song). Although the habitat created by anthropogenic action (shade forest plantations of coffee and cocoa) dates back to less than 150 years ago, I detected significant differences in morphology and song could between birds occurring in the original rainforest and in the plantations.

Morphological analyses

Little is known about how morphological characteristics vary according to habitat types. In addition, there is little information on the ecology of Speirops lugubris in forest and plantation habitats, in relation to diet composition and foraging behavior. In spite that there are no clear predictions, studies have shown that these morphological traits have a heritable base and are generally under strong selection (Schluter and Smith 1986; Gibbs & Grant 1987).

I found that individuals from forest habitats are larger overall than individuals from plantation habitats: significant differences between habitat types were found for mass, wing and tarsus length, and bill length and height. This suggests a strong environmental influence on morphological variation, but it is important to consider potential confounding variables aside from the differences in habitat structure between forests and plantations. For example, body size is known to increase with elevation in several organisms, including African Zosteropidae (Moreau 1957), and because most of the remaining forest in Sao Tomé is restricted to high elevation areas, the body size increase could be an artifact of the elevational distribution of forests and plantations. However, although I did find a significant correlation between elevation and morphology, the differences between habitats remained significant after

25 controlling for elevation. In addition, whereas body size increased with elevation in plantations, it decreased with elevation in forests. This suggests that there is a strong influence of habitat structure on morphological variation beyond the influence of elevation.

The weak correlation between geographical and morphological distance I observed, together with the significant differences between habitats, suggests that differentiation is not being driven by random factors, such as genetic drift in isolation, and supports divergence as a result of varying selection pressures. Also, the apparent existence of gene flow between forests and plantations suggested by variation at six microsatellite loci (M. Melo unpublished data), suggests that divergence by genetic drift is unlikely. However, it is important to note that habitat conversion from primary forest to plantations is very recent, and it would be very unlikely to detect microsatellite differentiation even if movement between habitats is restricted (M. Melo, pers. comm.).

Plumage color analyses

Although plumage coloration is expected to vary as a result of different signaling light environments (McNaught and Owens 2005), I found no significant differences between habitat types for any of the colorimetric variables in all feather patches, despite the great difference in light that enters each habitat, related to canopy cover measurements.

The lack of plumage color differentiation likely reflects that differences in selection pressures acting on plumage between habitats might be low. Because results from the habitat structure analyses showed that forests had a higher percentage of canopy cover than plantations (i.e. much less incoming light), the lack of a varying impact of selection cannot be attributed to a lack of differences in environmental conditions for optimal signal transmission. Alternatively, the result may reflect that plumage

26 signalling may play a limited role in mate selection in this species, as suggested by the lack of sexual dichromatism that I found. Also, Speirops lugubris has lost traits such as the yellow and green colors present in closely related species such as Zosterops ficedulinus feae. Loss of color is a common among island birds including many species from the Gulf of Guinea (Amadon 1953), and is generally explained by reduction of sexual selection. This supports the idea that plumage color is possibly not used as a mate recognition signal in this species.

Theory suggests that sexual selection favors traits that are most conspicuous, less costly or are a more accurate signal of the individual’s fitness condition in a particular environment (Endler 1992, Price et al 1993, Badyaev et al 2002). Thus, if there is a constraint on sexual chromatism, alternative traits such as song may replace chromatism as the trait under sexual selection (Darwin 1871, Shutler and Weatherhead 1990). This could be possible in Speirops lugubris because its song is very elaborate and complex, and was found to diverge according to habitat types (see below).

Song Analyses

Acoustic properties of birdsongs have been found to be adapted to habitat characteristics that impose different barriers and restrictions, and hence drive divergence in order to optimize communication (Morton 1975, Seddon 2005, Badyaev and Leaf 1997). Plantations are open habitats, with almost no barriers to sound transmission, contrary to the high number of barriers present in closed forest habitats. These limitations constrain temporal patterning due to reverberations occurring with high paced songs (Badyaev and Leag 1997, Seddon 2005). Frequency structure is also restricted because of barriers imposing sound attenuation, where high frequencies are easily lost with obstacles. I found that the Speirops lugubris song varied between habitat types in temporal (notes per song, pace) and frequency structure (peak frequency) patterns. Song pace was found to be lower in forest, as

27 predicted by closed habitat restrictions. Peak frequency of song also diverged in the direction predicted by the sound environment, being higher in plantations.

No direct correlation could be tested between morphological measurements and song features because data were collected in different years and mostly at different sites. However, the coincidence of lower frequency songs and slower singing pace in environments where birds exhibit larger bills and greater body mass, is also consistent with the hypothesis that differences in song features may arise as a result of morphological adaptation.

CONCLUSIONS

My results suggest that differences in the structure of habitats are associated with divergence in morphological and song traits. Moreover, divergence of morphological and song variables in this study responded as predicted according to environmental conditions and followed the direction found in other studies (e.g. Smith et al 2007, Seddon 2005). Although phenotypic plasticity cannot be completely ruled out, various authors have found the morphological traits I analyzed are highly heritable and are under selection (Schluter and Smith 1986). Additionally, plasticity can contribute to adaptive evolution when populations are exposed to novel environmental conditions. This adaptive plasticity should favor persistence in the new environment if the response is close to the new favored phenotypic optimum (Ghalambor et al 2007, Charmantier et al 2008). In conclusion, these results suggest that the endemic Speirops lugubris is experiencing an adaptive and potentially evolutionary response to diverging habitat structures caused by anthropogenic destruction of the original environment. Furthermore, this study supports the hypothesis that ecological adaptation to human modified habitats may drive signal evolution even in the face of gene flow.

28 FUTURE DIRECTIONS

Morphological, color and song structure analyses, although providing interesting insights on their response to different environmental pressures, must be further examined in order to provide clear predictions and additional support.

Plasticity is a strong factor limiting assertive conclusion on adaptation to habitat divergence, therefore, common garden experiments would be helpful to assess the heritability of morphological and song characters. Additionally, detailed qualitative song analysis is needed to further assess song variation and to test if cultural imprinting is taking place. Song structure adaptation to habitat conditions should be looked into with greater depth, performing sound transmission experiments and ambient noise measurements for each habitat, with the purpose of evaluating the adaptation of song characteristics to sound competition and habitat constraints.

REFERENCES

Amadon D. 1953. Avian systematics and evolution in the Gulf of Guinea. Bulletin of the American Museum of Natural History. 100: 394-451.

Andersson S, and Prager M, 2006. Quantification of Coloration. In: Bird Coloration. Vol. 1. Mechanisms and Measurements (Hill GE, McGraw KJ, eds). Cambridge, USA: Harvard University Press.

Badyaev A and Leaf E 1997. Habitat associations of song characteristics in Phylloscopus and Hippolais warblers. The Auk 114(1): 40-46

Bell G and Collins S. 2008. Adaptation, extinction and global change. Evolutionary Applications. 1: 3-16.

29 Boughman JW. 2002. How sensory drive can promote speciation. Trends in Ecology and Evolution. 17: 571-577

Charmantier A, McCleery RH, Cole LR, Perrins C, Kruuk LE and Sheldon BC. 2008. Adaptive Phenotypic Plasticity in Response to Climate Change in a Wild Brid Population. Science. 320: 800-803.

Coyne J.A. and Orr H.A, 2004. Speciation. Sinauer Associates, Sunderland, Massachusetts.

Christy, P and Clarke, W. 1998. Guide des Oiseaux de Sao Tomé et Príncipe. First Edition. ECOFAC, Multipress-Gabon, Libreville.

Dallimer M and King T. 2007. Habitat preferences of the forest birds on the island of Principe, Gulf of Guinea. African Journl of Ecology.

Doutrelant C, Leitao A, Giorgi M and Lambrechts M. 1999. Geographical variation in Blue Tit song, the result of an adjustment to vegetation type? Behaviour 136: 481-493.

Endler JA 1992. Signals, signal conditions and the direction of evolution. American Naturalist. 139 Supplement: S125-S153.

Endler, J. A. 1993. Some general comments on the evolution and design of animal communication systems. Philosophical Transactions of the Royal Society of London B. 340:215–225

Exell AW 1944. Catalogue of the Vascular Plants of S. Tomé (with Príncipe and Annobón). British Museum (Natural History), London.

Foley JA, DeFries R, Asner GP, Barfold C, Bonana G, Carpenter SR, Chapin FS, Coe MT, Daily GC, Gibbs HK, Helkowski JH, Holloway T, Howard EA,

30 Kucharik CJ, Monfreda C, Patz JA, Prentice IC, Ramankutty N and Snyder PK. 2005. Global Consequences of Land Use. Science, 309: 570-574.

Gascoigne A. 2004 São Tomé, Príncipe, and Annobon Moist Lowland Forests. In: Terrestrial Ecoregions of Africa and Madagascar: a Conservation Assessment (eds. Burgess N, D'Amico Hales J, Underwood E, et al.), pp. 236-238. Island Press, Washington.

Ghalambor CK, McKay JK, Carroll SP and Reznick DN. 2007. Adaptive versus non-adaptive phenotypic plasticity and the potential for contemporary adaptation in new environments. Functionl Ecology. 21: 394-407.

Gibbs HL and Grant PR 1987. Oscillating selection on Darwin’s Finches. Nature 327: 511-513

Hendry AP and Kinnison MT. 1999. The pace of modern life: measuring rates of contemporary microevolution. Evolution. 53: 1637 – 1653.

Hendry AP, Farrugia TJ and Michael Kinison. 2008. Human influences on rates of phenotypic change in wild animal populations. Molecular Ecology, 17:20- 29.

Huber S and Podos J. 2006. Beak morphology and song features covary in a population of Darwin’s finches (Geospiza frotis). Biological Journal of the Linnean Society. 88: 489-498.

Jones PJ and Tye A 1992. The status of endemic birds and their habitats in São Tomé and Príncipe. Proceedings of the 7th Pan-African Ornithological Congress, 453-459.

Lessells, C.M. and Boag, P.T. 1987. Unrepeatable repeatabilities: a common mistake. The Auk. 114: 116-121.

McNaught, M and Owens, I. 2002. Interspecific variation in plumage colour among birds: species recognition or light environment? Journal of Evolutionary Biology 15: 505-514.

Melo, M, 2007. Bird Speciation in the Gulf of Guinea. PhD Thesis. Institute of Evolutionary Biology, University of Edinburgh, Edinburgh.

Morton E. 1975. Ecological sources of selection on avian sounds. The American Naturalist. 109: 17-34.

Myers N, Mittenmeier R, Mittenmeier C, da Fonseca G, Kent J. 2000. Biodiversity hotspots for conservations priorities. Nature, 403: 853-858.

Palumbi SR. 2001. Humans as the World’s Greatest Evolutionary Force. Science. 293:1786-1790

Payne RB. 1986. Bird songs and avian systematics. Current Ornithology 3: 87- 126.

Podos J. 2001. Correlated evolution of morphology and vocal signal structure in Darwin's finches. Nature 409:185–188.

Podos J, Huber SK and Taft B. 2004. Bird song: the interface of evolution and mechanism. Annual Review of Ecology, Evolution and Systematics 35:55-87

Price J, Friedman N and Omland K. 2007. Song and Plumage evolution in the new world orioles (Icterus) show similar lability and convergence in patterns. Evolution 61-4: 850-863.

32 Reznick and Ghalambor. 2001. The population ecology of contemporary adaptations: what empirical studies reveal about the conditions that promote adaptive evolution. Genetica, 112-113: 183-198.

Ryan MJ and Brenowitz EA. 1985. The role of body size, phylogeny and ambient noise in the evolution of bird song. American Naturalist. 126: 87-100.

Schluter D and Smith TB. 1986. Natural selection on beak and body size in the song sparrow. Evolution 40: 221-231.

Seddon N 2005. Ecological adaptation and species recognition drives vocal evolution in neotropical suboscine birds. Evolution 59(1): 200-215.

Seehausen O, Takimoto G, Roy D and Jokela J. 2008. Speciation reversal and biodiversity dynaics with hybridization in changing environments. Molecular Ecology. 17: 30-44.

Slabbekoorn H and Smith TB. 2002. Bird song, ecology and speciation. Philosophical Transactions of the Royal Society of London B, 357:493-503.

Slabbekoorn H and Smith TB. 2002. Habitat-dependent song divergence in the little greenbul: an analysis of environmental selection pressures on acoustic signals. Evolution 56(9): 1849-1858.

Smith TB. 1990. Resource use by bill morphs of an Africa finch: evidence for intraspecific competition. Ecology, 71: 1246-1257.

Smith TB, Wayne RK, Girman DJ and Bruford MW. 1997. A Role for Ecotones in Generating Rainforest Biodiversity. Science 276: 1855-1857

33 Smith TB, Calsbeek R, Wayne RK, Holder KH, Pires D and Bardeleben C. 2005. Testing alternative mechanisms of evolutionary divergence in African rain forest passerine bird. Journal of Evolutionary Biology, 18: 257-268.

Smith TB and Bernatchez L. 2008. Evolutionary chance in human altered environmnets. Preface to the Special Issue. Molecular Ecology, 17: 1-8.

Smith TB, Milá B, Grether GF, Slabbekoorn H, Sepil I, Buermann W, Saatchi S and Pollinger JP. 2008. Evolutionary consequences of human disturbance in a rainforest bird species from Central Africa. Molecular Ecology, 17: 58-71.

Stockwell CA, Hendry AP and Kinnison MT. 2003. Contemporary evolution meets conservation biology. Trends in Ecology and Evolution.18:94-101.

Tubaro P and Mahler B. 1998. Acoustic frequencies and body mass in New World Doves. Condor 100:54-61.