Habitat preferences of the forest birds on the island of Príncipe, Gulf of Guinea


*Biodiversity and Macroecology Group (BIOME), Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, U.K. Tel.: +44 114 2220037; Fax: +44 114 2220002; E-mail: m.dallimer@sheffield.ac.uk


The forests of southwest Príncipe are recognized as important for biodiversity conservation. Here, we study the distribution and density of fifteen of the island’s endemic bird species (including three Globally Threatened and one Near-Threatened) within the National Park. The endemic subspecies of the Near-Threatened Gulf of Guinea thrush occurred at a density of 0.08 birds ha−1, which is far lower than its sister subspecies on the neighbouring island of São Tomé. We also present evidence that the effects of access routes around the protected area exert an influence on bird species within the National Park. The grey parrot, which still suffers from illegal hunting, was more likely to be encountered further away from roads, with 25% of variation in its occurrence explained by distance to the nearest road. In addition, the occurrence of all four threatened species was more likely further from access routes. Careful consideration must therefore be given to the potential impacts within the National Park of developments outside the currently protected area.


Les forêts du sud-ouest de Principe sont reconnues comme importantes pour la conservation de la biodiversité. Nous étudions ici la distribution et la densité de 15 espèces d’oiseaux endémiques de l’île (y compris trois espèces « menacées au niveau mondial » et une « quasi menacée ») dans le parc national. Le merle de Principe qui est une sous-espèce endémique quasi menacée y est présent avec une densité de 0,08 oiseau/ha, ce qui est très inférieur à la sous-espèce correspondante sur l’île de Sao Tome. Nous apportons aussi des preuves que les routes qui entourent l’aire protégée ont une influence sur les espèces d’oiseaux dans le parc national. Le perroquet gris, qui souffre toujours de la chasse illégale, était plus susceptible d’être observé loin des routes, 25% de la variation de sa présence pouvant s’expliquer par la distance jusqu’à la route la plus proche. De plus, la présence des quatre espèces menacées était plus probable loin des routes d’accès. Il faut donc accorder une attention particulière aux impacts potentiels, au sein même du parc, des développements réalisés à l’extérieur de l’actuelle aire protégée.


Islands are recognized as being important for biodiversity and feature prominently in lists of the world’s most valuable areas in need of conservation action. Some 29% of Conservation International’s biodiversity hotspots (Myers et al., 2000) and 48% of Endemic Bird Areas (EBAs) are islands (Stattersfield et al., 1998), while more than 600 globally threatened bird species are restricted to islands (Ricketts et al., 2005). Príncipe, an oceanic island in the Gulf of Guinea, is home to many endemic and threatened species from various taxa (Jones, 1994). Of the 33 species of breeding landbirds on Príncipe, there is one monotypic genus (Horizorhinus) and five other single-island endemic species. A further five species are shared only with the neighbouring islands of São Tomé and Annobón, while seven African mainland species have endemic subspecies on the island (Table 1; Jones & Tye, 2006). This high level of avian endemism has led to the inclusion of Príncipe, São Tomé and Annobón in the 218 EBAs worldwide (Bibby et al., 1992), and São Tomé and Príncipe in the Important Bird Areas (IBAs) of Africa (Fishpool & Evans, 2001).

Table 1.   Species names, level of endemism and IUCN status of the fifteen forest-dwelling birds surveyed in Príncipe (Jones & Tye, 2006)
SpeciesLevel of endemismIUCN threat levela
  1. aIUCN Globally Threatened categories: VU, Vulnerable; NT, Near-Threatened; LC, Least Concern.

African green pigeon (Treron calva virescens Amadon 1953)Subspecies endemic to PríncipeLC
Sao Tomé bronze-naped pigeon (Columba malherbii Verreaux and Verreaux 1851)Species endemic to Príncipe, São Tomé and AnnobónLC
Lemon dove (Columba larvata principalis Hartlaub 1866)Subspecies endemic to PríncipeLC
Grey parrot (Psittacus erithacus Linneaus 1758)NoneLC
Blue-breasted kingfisher (Halcyon malimbica dryas Hartlaub 1854)Subspecies endemic to PríncipeLC
White-bellied kingfisher (Corythornis leucogaster nais Kaup 1848)Subspecies endemic to PríncipeLC
Gulf of Guinea thrush (Turdus olivaceofuscus xanthorhynchus Salvadori 1901)Species endemic to Príncipe and São Tomé with separate subspecies on each islandNT
Dohrn’s thrush-babbler (Horizorhinus dohrni Hartlaub 1866)Genus endemic to PríncipeLC
Príncipe sunbird (Anabathmis hartlaubii J. Verreaux 1857)Species endemic to PríncipeLC
Príncipe white-eye (Zosterops ficedulinus ficedulinus Hartlaub 1866)Species endemic to Príncipe and São Tomé with a separate subspecies on each islandVU
Príncipe speirops (Speirops leucophaeus Hartlaub 1857)Species endemic to PríncipeNT
Príncipe drongo (Dicrurus modestus Hartlaub 1849)Species endemic to PríncipeNT
Príncipe glossy starling (Lamprotornis ornatus Daudin 1800)Species endemic to PríncipeLC
Príncipe golden weaver (Ploceus princeps Bonapart 1850)Species endemic to PríncipeLC
Príncipe seedeater (Serinus rufobrunneus rufobrunneus Gray 1862)Species endemic to Príncipe and São Tomé with three subspeciesLC

Despite the presence of one Vulnerable and three Near-Threatened species on Príncipe (IUCN, 2006), there has been limited work conducted to quantify the population densities and habitat preferences of any of the endemic species. The most detailed research to date offers only broad habitat observations on forest types and bird occurrence (Christy & Clarke, 1998; Jones & Tye, 2006). For instance, the scarce island subspecies of the Gulf of Guinea thrush (Turdus olivaceofuscus xanthorhynchus Salvadori 1901) and Príncipe white-eye (Zosterops ficedulinus ficedulinus Hartlaub 1866) are believed to be restricted to the remaining primary forest (Jones & Tye, 2006), while other endemic species, such as Dohrn’s Thrush-babbler (Horizorhinus dohrni Hartlaub 1866) and the Príncipe sunbird (Anabathmis hartlaubii J. Verreaux 1859) are regularly found outside forested areas. This includes extensive areas of shade planted coffee and cocoa those are recognized as good quality habitat for some forest birds (Donald, 2004; Komar, 2006; Raman, 2006).

Príncipe’s southwest forests form part of the Parque Natural Ôbo de São Tomé e Príncipe, the creation of which was ratified by the islands’ government in 2003. However, the Park receives little active conservation. Hence, the forest is still subject to many of the pressures associated with unprotected areas such as illegal hunting. The Democratic Republic of São Tomé and Príncipe remains poor, with a per capita GDP of US$1200 and 54% of the population living below the poverty line (CIA, 2007). A proposed Free Trade Zone (Ministério da Economia, República Democrática de São Tomé e Príncipe, 2006) and recent oil discoveries (Chevron, 2006; CIA, 2007) could inject significant amounts of money into the island. Increasing development pressures, and rising demands for land and natural resources may then impact negatively on an under-managed or protected National Park. Here, we use the endemic forest birds of Príncipe to highlight the importance to biodiversity conservation of the existing (but unmanaged) National Park using distance sampling (Buckland et al., 1993) and habitat preference modelling. The results are discussed in relation to the future conservation of the remaining forest on the island.

Materials and methods

Study site

Príncipe, with an area of 139 km2, lies 220 km from the African coast and 146 km north of São Tomé. The north of the island is flatter and contains the majority of the human population and agriculture. To the south and centre, the island is rugged and includes the highest mountain, Pico de Príncipe (948 m). Significant areas of primary rainforest remain in the south and west, including the slopes of the mountainous central area. The study area covered a low-lying area (altitudinal range 80–210 m) of forest in the south of Príncipe covering 3 km by 2 km (Fig. 1). Surveys across 59 point transect locations were carried out in the morning and late afternoon between 18th and 27th January 2002.

Figure 1.

 Map of Príncipe, Gulf of Guinea, showing altitudinal bands and the location of the study area (Jones & Tye, 2006). Inset shows regional location of Príncipe in the Gulf of Guinea

Distance sampling

The theory and assumptions of distance sampling and practical aspects of survey design are described by Buckland et al. (2001, 2004). Point transect locations were placed on a grid that was positioned independently of the local topography and intended to cover the range of available habitat in that area. Each location was at least 100 m apart, a distance that was considered adequate to avoid double sampling the same areas of forest. It was often necessary to make trails to access point transect locations. To minimize disturbance while undertaking the actual surveys, trails were marked on a different day to collecting count data. Where maintained paths were used for access to survey locations, points were located at up to 50 m from the path.

On arrival at a point transect location, an initial 10-min search period was used to note locations, identities and cluster size of birds present. Thereafter, 2 min of actual survey time were allowed to measure distances, confirm locations and identifications. The short survey time associated with this method ensured that biases associated with birds moving in response to the observer were minimized. The initial 10-min search period both allowed birds to resume normal behaviour and ensured that cryptic or skulking species close to the point were detected with as much certainty as possible in a forest environment. In a closed forest environment, many registrations are likely to be purely aural (Scott, Ramsey & Kepler, 1981; DeJong & Emlen, 1985; Alldredge et al., 2007). Locating singing birds in dense vegetated habitats is known to be subject to error (Kepler & Scott, 1981; Alldredge, Simons & Pollock, 2007). Here, errors were minimized by using a laser rangefinder (Rangemaster LRF 800, Leica Camera AG, Solms, Germany; ±1 m over 100 m) for distance measurements and undergoing a training period to familiarize the observers with the surveyed species and habitats (Scott et al., 1981; Buckland et al., 2001). In addition, where contact with a bird was purely aural, if it were not possible to reliably locate the singing bird, the species was still recorded as present at that point transect location, but not included in any distance analysis.

Data analysis

Prior to analysis, distances were pooled and detections furthest from the point transect location removed. In general, it is appropriate to truncate (i.e. not to include) the largest 5–10% of recorded distances, with the higher figure being recommended for point transect studies (Buckland et al., 2001). This approach was adopted for the more common species. However, although truncating the larger distances reduces the complexity of the modelled detection function and leads to a better description of the remaining data, it also reduces the sample size and can decrease precision in the resulting density estimates. This effect is particularly apparent with small sample sizes where each data point removed represents a larger proportion of the overall number of registrations. Therefore, for the species with fewest registrations, truncation was either not performed, or only the most distant detections were removed. Subsequently, candidate models of the detection function were chosen and tested against the data. Model selection was based on minimum Akaike Information Criteria (AIC) and chi-squared goodness of fit tests. Distance data were analysed using Distance 4.1 release 2 (Thomas et al., 2004).

Detection functions were generated for individual species with more than 70 registrations (Buckland et al., 2001). For the remainder, densities were calculated by pooling registrations together with those of similar, often congeneric, species from the neighbouring island of São Tomé (M. Dallimer, unpublished data). Where no congeneric species were surveyed on both islands, registrations were pooled with a species that had similar detectability and behaviour. Species-specific densities, for the birds of Príncipe alone, were then extracted by poststratification by species within the ‘Distance’ software. Hence, density estimates were possible for the majority of species encountered.

Habitat associations

To assess the habitat associations of the forest birds, the following topographical and vegetation variables were measured within a 15-m radius of each survey point:

  • • Number of trees making up the canopy (canopy count).
  • • Percentage canopy cover, estimated by eye (canopy cover).
  • • Maximum height of the canopy measured using a rangefinder (canopy height).
  • • Average number of stems in three 1 m-radius circles, 1 m above ground level (ground cover).
  • • The presence of fruiting or flowering trees located using binoculars (fruit/flower).
  • • Abundance of climbers and epiphytes recorded on a scale of 0 (none) to 10 (dense).
  • • Inclination of the terrain (slope).
  • • Ridge position (0 – valley floor to 5 – ridge top).

Where possible, the position in the field was recorded using a GPS unit. This allowed all point locations to be plotted on a 1 : 25000 map (Ministério do Ultramar, 1958). Altitude, the distance of the point location to the nearest marked road, settlement and watercourse were taken from this map. Settlements were defined as any current known permanent locations of human habitation, none of which were within the study area.

Data analysis

Generalized linear models were used to assess the effects of the habitat and topographical variables on the occurrence of the forest bird species. Presence/absence was modelled with a binomial error term and logit link. All statistical analysis was carried out in R v 2.2.1 (R Core Development Team, 2005) following Crawley (2002). All explanatory variables were initially included in a maximal model. A step-down deletion process was used to remove the least significant variable in turn until only significant variables remained and the removal of further explanatory variables led to a significant reduction in deviance. In addition to modelling the habitat preferences of individual species, the occurrence of all IUCN listed species (Near-Threatened or Vulnerable) were pooled and modelled as a single response variable. A trend surface term was used to control for spatial structure and neighbourhood effects (Legendre & Legendre, 1998; Guisan & Zimmermann, 2000;Quevedo, Banuelos & Obeso, 2006). This involved adding spatial terms to the maximal models as the third degree polynomial of standardized UTM coordinates and retaining these spatial terms in the Minimum Adequate Models, regardless of their significance.


Distance sampling

In all, 256 registrations of 15 species from 59 point locations were collected (Table 2). Dohrn’s thrush-babbler (90 registrations) was the most frequently recorded species. The lemon dove (Columba larvata Temminck 1810) and the white-bellied kingfisher (Corythornis leucogaster nais Kaup 1848) (with one registration each) were the least. Only one species had more than 70 registrations (Dohrn’s thrush-babbler) and hence species-specific detection functions were modelled for this species alone. Density estimates were not made for the grey parrot (Psittacus erithacus Linneaus 1758) as insufficient registrations were collected and no detection function for the species from another location was available. Densities were calculated for all other species encountered. Density varied among species (Table 2), with Dohrn’s thrush-babbler occurring at a density of 6.93 birds ha−1, while the two Near-Threatened species (the Gulf of Guinea thrush and the Príncipe speirops) occurred at densities below 0.5 birds ha−1.

Table 2.   Detection function details and density estimates for forest birds, on Príncipe, Gulf of Guinea
SpeciesSurrogate species registrationsaRegistrationsTruncation distance (m)Included registrationsKey termExpansion termDensity (birds ha−1)Coefficient of variation 95% CI
  1. aIndicates the species surveyed on São Tomé (M. Dallimer, unpublished data) with which data were combined to model an appropriate detection function: BNP, São Tomé bronze-naped pigeon; Thrush, Gulf of Guinea thrush; Sunbird, São Tomé sunbird (Anabathmis newtonii Bocage 1887); Starling; Chestnut-winged starling (Onychognathus fulgidis fulgidus Hartlaub 1849); Speirops, São Tomé speirops (Speirops lugubris Hartlaub 1848); Weaver, São Tomé weaver (Ploceus sanctithomae Hartlaub 1848); Seedeater, Príncipe seedeater (Serinus rufobrunneus thomensis Bocage 1848).

African green pigeonBNP3453Half-normalCosine0.070.580.02–0.20
São Tomé bronze-naped pigeonBNP174514Half-normalCosine0.230.330.12–0.44
Lemon doveBNP1451Half-normalCosine0.021.010.00–0.12
Gulf of Guinea thrushThrush6306Half-normalCosine0.360.430.16–0.80
Blue-breasted kingfisherThrush10509Half-normalCosine0.190.480.07–0.47
White-bellied kingfisherThrush1301Half-normalCosine0.021.060.00–0.11
Dohrn’s thrush-babblerNA903582Hazard-rateCosine6.930.214.62–10.38
Príncipe sunbirdSunbird113511Hazard-rateCosine0.770.380.37–1.58
Príncipe white-eyeSpeirops2252Hazard-rateCosine0.240.750.06–0.88
Príncipe speiropsSpeirops5254Hazard-rateCosine0.480.550.17–1.32
Príncipe drongoThrush3303Half-normalCosine0.120.720.03–0.43
Príncipe glossy starlingStarling343530Half-normalCosine1.370.290.78–2.39
Príncipe golden weaverWeaver6206Half-normalCosine0.330.420.15–0.73
Príncipe seedeaterSeedeater353032Half-normalCosine1.660.251.02–2.70

Habitat associations

Fifteen species were recorded across 59 locations. The most commonly encountered species, observed at 57 locations, was Dohrn’s thrush-babbler. The Vulnerable Príncipe white-eye was recorded at just two locations. Habitat modelling was only performed for the seven species that occurred at more than ten locations, as well as for the pooled observations of Vulnerable and Near-Threatened species.

Eight of the measured variables explained significant amounts of the variation in occurrence for five species (Table 3). No single variable explained the occurrence of more than one species. Topographical features (distance to nearest road and altitude) explained 25.3% of the variation in occurrence of the grey parrot, with parrots significantly more likely to occur further away from roads and at higher elevations. The Near-Threatened Gulf of Guinea thrush preferred dense forested areas, while pooled observations of Near-Threatened and Vulnerable species combined were more likely at greater distances from the nearest road. Blue-breasted kingfishers were more likely to occur in areas with an open ground layer, whilst the Príncipe sunbird tended to be found in valleys where epiphytes and fruiting or flowering trees were not present. No terms were significant for either the São Tomé bronze-naped pigeon (Columba malherbii Verreaux and Verreaux 1851) or the Príncipe seedeater.

Table 3.   Significant variables included in the Minimum Adequate Models for each species of forest bird surveyed on the island of Príncipe, Gulf of Guinea
SpeciesPresenceaExplained variationVariableEstimateVariableEstimateVariableEstimate
  1. Explained variation = (total deviance − residual deviance)/total deviance, giving an indication of the proportion of the variance explained by the model.

  2. *P < 0.05, **P < 0.01, ***P < 0.001.

  3. aIndicates the number of points at which each species was detected.

African green pigeon5Not modelled      
Sao Tomé bronze-naped pigeon500.00None     
Lemon dove4Not modelled      
Grey parrot330.25Road657.20***Altitude0.04*  
Blue-breasted kingfisher310.11Ground cover−0.46**    
White-bellied kingfisher2Not modelled      
Gulf of Guinea thrush120.13Canopy count2.16*    
Dohrn’s thrush-babbler57Not modelled      
Príncipe sunbird130.22Ridge−2.19*Epiphyte2.49*Fruit/flower−1.94*
Príncipe white-eye2Not modelled      
Príncipe speirops7Not modelled      
Príncipe drongo6Not modelled      
Príncipe glossy starling300.08River1626.00*    
Príncipe golden weaver8Not modelled      
Príncipe seedeater350.01None     
VU and NT species combined240.06Road193.90*    


The lowland primary forest surveyed here is part of the recently gazetted Parque Natural Ôbo de São Tomé e Príncipe, whose importance for biodiversity conservation in Africa is well known (Collar & Stuart, 1988). In total, 15 species were recorded; including all four IUCN listed species (the Vulnerable Príncipe white-eye, speirops and drongo, the Near-Threatened Gulf of Guinea thrush), none of which were observed more than twelve times during the survey. Despite their rarity, by making use of data from similar species on the nearby island of São Tomé, density estimation was possible for the Gulf of Guinea thrush (0.36 birds ha−1) and Príncipe speirops (0.48 birds ha−1). Both occur at densities substantially lower than equivalent species on São Tomé, a pattern that is true for all the endemic species on Príncipe that have sister species on São Tomé (M. Dallimer, unpublished data). Recent work has highlighted the taxonomic differences between the Príncipe and São Tomé subspecies of the Gulf of Guinea thrush and has recommended that the Príncipe subspecies be considered a separate species (Melo, 2006). Any newly described species would deserve a higher threat level listing than its current Near-Threatened status.

Many of the other endemic species appear to survive well in the remaining forest, with Dohrn’s thrush-babbler (density 6.93 birds ha−1) the most common. Species such as the Príncipe golden weaver (Ploceus princeps Bonaparte 1850), African green pigeon and São Tomé bronze-naped pigeon occur at much lower densities than this, and lower than similar species on São Tomé. Hence, whilst the remaining forest is undoubtedly an important habitat for these species, further work based in the surrounding areas of secondary growth and abandoned plantations is required to quantify the value of the various habitats for the endemic species.

Habitat associations were modelled for seven of the fifteen encountered species. Thirteen habitat and topographical variables were measured, but only eight explained any of the variation in occurrence for these species, and none was important for more than one species. This indicates that the forest birds of Príncipe have a range of habitat requirements, which are unlikely to be universally met by concentrating on the needs of a single species. Any conservation management should therefore be focussed on maintaining the broad range of forest habitats.

The Gulf of Guinea thrush was more likely to occur where there was a high number of canopy trees present, confirming previous reports that the species prefers dense forest (reviewed in Jones & Tye, 2006). No other IUCN listed species were observed often enough for their habitat associations to be modelled, which must in itself be cause for concern in relation to their population size on the island. However, modelling the combined observations of the threatened species suggested that they are sensitive to disturbance as they were more likely to occur at greater distances from roads. A similar pattern was found for the grey parrot, which, historically has been hunted in large numbers for the pet trade (Juste, 1996; Melo & Ryan, 2007). Although such trade is now illegal in São Tomé and Príncipe, its impacts remain apparent. The parrot was more likely to be encountered at greater distances from the nearest road, implying the species remains susceptible to easy access to the forest by hunters. Hence on Príncipe, the effects of access routes near the National Park extend well into the nominally protected area, effectively reducing the amount of suitable habitat for certain disturbance sensitive species.

As infrastructure projects improve the efficiency of the transport network on Príncipe (UNDP, 2006), such species are likely to come under greater pressure. It is likely that quality of forest required by the Near-Threatened and Vulnerable species will be negatively affected. This could be mediated through increased forest resource use or a rise in the numbers of invasive organisms that might be associated with a larger area of forest being accessed more regularly. Easier access, more frequently by more people will probably increase the hunting pressure experienced by the grey parrot Without further controls in place, even within the existing protected area, economic development on Príncipe will very likely lead to declines in the grey parrot and Near-Threatened and Vulnerable species.

Implications for conservation

Currently Príncipe retains a relatively large area of native primary rainforest, much of which is covered by a National Park. Areas of secondary growth and extensively produced shade crops buffer the forest from more intensive agriculture and densely populated areas of the island. The recent legal recognition of National Park status for the forest is a step forward, even though conservation resources remain limited. However, a large portion of the island’s west coast is currently the subject of a proposal to develop a Free Trade Zone (Ministério da Economia, República Democrática de São Tomé e Príncipe, 2006) and a recent UNDP transport development strategy for São Tomé and Príncipe makes no mention of any potential threats to biodiversity (UNDP, 2006). The proposed Free Trade Zone covers none of the National Park on Príncipe. Nonetheless, as demonstrated here, its effects are likely to be felt well inside the protected area. Increases in forest use, hunting, illegal logging, further introductions of non-native species or higher levels of disturbance are all possible with potentially serious implications for the status of the grey parrot and the four IUCN listed species. In addition, many of the endemic species of Príncipe are believed to be common in abandoned plantation and secondary growth that characterize the area covered by the proposed development. The potential loss of habitat on the island could therefore have severe detrimental consequences.


We thank the staff of ECOFAC São Tomé and Príncipe for their help in carrying out the fieldwork. Pedro Nobre helped with contacts on Príncipe, and Manuel Borge Jiana was our indispensable guide in the southern forests. We also thank Rachel Atkinson, Angus Gascoigne, Peter Jones, Rob Smallshire and Gordon Brown for help in planning the study. Zoe Davies and Peter Jones provided useful discussions and Zoe Davies commented on earlier drafts of this manuscript. The work was funded by the Davis Expedition Fund, the British Ecological Society and the John Ray Trust. Additional support was provided by Garmin (Europe) Ltd and Berghaus Ltd.