Movement patterns and habitat preference of Oilbirds (Steatornis caripensis) in the southern Andes of Colombia

Habitat use and movement patterns in animals are usually determined by the availability of food, morphological and physiological traits, social systems, and safe sites to complete vital cycles. The objective of this study was to estimate the home-range area, daily traveled distance, and patterns of space use by Oilbirds (Steatornis caripensis), and the factors that can affect them in Cueva de Los Guacharos National Park (Huila, Colombia). We attached GPS devices to five individuals and recovered information from three. Home ranges were estimated by minimum convex polygon (MCP), 95 and 50% kernel methods. Traveled distances were estimated as the sum of the linear distances between sampling points (every 30 min), and patterns of habitat use were evaluated through ecologicalniche factor analysis (ENFA), including forest cover, altitude, and degree of fragmentation. Analysis of habitat preference was done using 25 vegetation plots established in frequently used areas and species richness, and the proportion of consumed plants was compared with other locations within their distribution range. Home-range area was estimated to be 4517 km2, possibly an underestimation due to reduced sample size. The average traveled distance per night was 55 km (range: 0-112 km). Analyses of habitat use showed a clear preference for areas with higher forest cover. Height above sea level was also largely associated with frequencies of habitat use, indicating a low use of areas above 3000 m, while the degree of fragmentation was not a strong predictor of habitat use. Finally, vegetation plots in frequently used areas showed a higher abundance of fruiting trees consumed by Oilbirds than control plots. Overall, Oilbirds have amazing movement capacity, being able to use distant landscape elements, and showing a clear preference for areas with forest cover, lower altitude, and areas with a high representation of the plant species they consume. Habitudes de déplacement et préférence en matière d'habitat des Guacharos des cavernes (Steatornis caripensis) dans les Andes du sud de la Colombie RÉSUMÉ. L'utilisation de l'habitat et les habitudes de déplacement chez les animaux sont généralement dictées par la disponibilité de ressources alimentaires, les traits morphologiques et physiologiques, les systèmes sociaux et les sites sécuritaires pour compléter les cycles de vie. L'objectif de cette étude était de calculer la superficie du domaine vital, les distances parcourues quotidiennement et les profils d'utilisation spatiale des Guacharos des cavernes (Steatornis caripensis), et cerner les facteurs qui peuvent agir sur eux dans le parc national Cueva de Los Guacharos (Huila, Colombie). Nous avons posé des unités GPS sur cinq individus et avons obtenu des informations sur trois d'entre eux. Les domaines vitaux ont été calculés au moyen des méthodes du polygone convexe minimum (MCP) et des kernels à 95 et à 50 %. Les distances parcourues ont été calculées en sommant les distances linéaires entre les points d'échantillonnage (toutes les 30 min), et les profils d'utilisation de l'habitat ont été établis par l'analyse des facteurs de niche écologique (ENFA), y compris le couvert forestier, l'altitude et le degré de fragmentation. L'analyse des préférences en matière d'habitat a été réalisée en inventoriant les espèces végétales dans 25 parcelles réparties dans des endroits fréquemment utilisés, et la proportion des plantes y ayant été consommées a été comparée avec celle d'autres endroits situés dans leur aire de répartition. Le domaine vital a été établi à 4 517 km2, sans doute une sous-estimation en raison de la faible taille de l'échantillonnage. La distance moyenne parcourue par nuit était de 55 km (étendue : 0-112 km). Les analyses de l'utilisation de l'habitat ont montré une nette préférence pour les endroits avec un couvert forestier dense. L'altitude au-dessus du niveau de la mer était aussi grandement associée avec la fréquence d'utilisation de l'habitat, les endroits au-dessus de 3 000 m étant peu utilisés, alors que le degré de fragmentation ne s'est pas avéré une bonne variable explicative de l'utilisation de l'habitat. Finalement, les parcelles de végétation dans les endroits fréquemment utilisés avaient une abondance plus grande d'arbres fruitiers consommés par les guacharos que les parcelles témoin. Dans l'ensemble, les Guacharos des cavernes ont une capacité de déplacement incroyable, étant capables d'utiliser des éléments distants du paysage, et montrent une préférence certaine pour les endroits offrant couvert forestier, faible altitude, et abondance des espèces végétales qu'ils consomment.

Key Words: birds; daily travel distance; habitat use; home range; movement ecology

INTRODUCTION
For animals with the capacity of individual movement, their location in space usually depends on habitat characteristics, including a wide range of ecological factors (Nathan 2006, Scharf et al. 2018). Among the most important is the availability of nutritional resources, adequate zones to carry out reproductive events, and places with low predation risk (Fahrig 2007, Gillies andSt. Clair 2010). Other parameters included in the study of movement consider topographic elements along the trajectory, travel length, and the ability to use different habitat types (Andrén 1994, Fritz et al. 2003, Frair et al. 2005. The relevance of these limiting factors is variable depending on the species (Villard et al. 1999, Johnson et al. 2002, Fritz et al. 2003 because movement capacity in birds depends on traits such as morphology, diet, foraging behaviors, body size, and habitat specialization (Neuschulz et al. 2013, Díaz Vélez et al. 2015. The most extreme cases of long-distance movement have been reported for birds that migrate from pole-to-pole (Gill et al. 2008). Still, some frugivorous species such as hornbirds, quetzals, and toucans travel long distances to find food resources that are unevenly distributed (Symes and Marsden 2007, Mueller et al. 2014, Díaz Vélez et al. 2015. On the other hand, species such as some passerine birds, from the Furnariidae family, are incapable of crossing between forest fragments or barriers such as rivers because their movement is strongly limited by open areas (Hayes and Sewlal 2004). Therefore, anthropogenic factors such as fragmentation, edge-effects, and changes in the matrix have a heavy impact on some species (Hobbs and Yates 2003), and their movement patterns can be affected depending on factors such as fragment connectivity and size (Gobeil and Villard 2002, Gillies and St. Clair 2010, Herrmann et al. 2016). This could lead, in a cascading effect, to alterations in some ecosystem processes that depend on movements, such as seed dispersal and the movement of nutrients (Howe and Smallwood 1982, Holbrook and Smith 2000, Gosper et al. 2005, Levey et al. 2005, Stevenson and Guzmán-Caro 2010. Therefore, knowledge on how landscape changes can affect these processes is important for the conservation of populations, their interactions within communities, and ecosystem processes (Holland et al. 2009, Gillies and St. Clair 2010, Maniguaje et al. 2011. Oilbirds (Steatornis caripensis) are nocturnal frugivorous birds distributed in the Neotropics (southern Central America, northern South America, and the tropical Andes). During the day, these birds usually sleep inside caves and canyons of both lowland and montane forests, and during the night, they travel long distances (Roca 1994, Thomas 1999, Bosque 2002, Tello et al. 2008, Holland et al. 2009). Foraging patterns in Oilbirds have only been studied in Venezuela, showing that the farthest distances from the cave are on average 44 ± 10.7 km (Holland et al. 2009). Oilbirds feed on fruits of numerous plant species (Tannenbaum and Wrege 1978, Bosque and de Parra 1992, Bosque 2002, Rojas-Lizarazo 2016, Stevenson et al. 2017, swallowing the entire fruit and digesting the pulp to later expel whole seeds in feces or through regurgitation (Bosque andde Parra 1992, Amico andAizen 2005). Therefore, these birds have the potential of being important vectors for long-distance seed dispersal between isolated fragments (Holland et al. 2009, Karubian et al. 2012a, Stevenson et al. 2017) and could be keystone species for regeneration processes in fragmented forests. However, little is known about this species' habitat requirements outside their caves, or about the availability of appropriate areas for them within the landscape.
The objective of this study was to determine which factors, at a landscape level such as cover types, degree of fragmentation, plant diversity, and tree composition are associated with habitat use by Oilbirds. Additionally, the home range, daily traveled distance, and flight speed of three individuals were estimated. The information on habitat use was used to understand which landscape configuration can lead to the maintenance of the Oilbirds' movement patterns (Holland et al. 2009, Rojas-Lizarazo 2016. It was hypothesized that the Oilbirds would prefer the use of coverage that provides food resources and places to complete their life cycles. We also expected that at the landscape scale, they would use more often elevations near the main cave (1500-2500 masl), places with greater forest cover to find feeding trees, and would be indifferent to fragmentation given the long flight distances so far reported (Holland et al. 2009). On a local scale, it was expected that the abundance of plants consumed would be higher in frequently used areas than in control areas (i.e., where they have not been reported).

Study area
The study area was defined as the minimum rectangle (ca. 8000 km²) enclosing all known movements of the colony of Oilbirds in the southern Andes of Colombia (Fig. 1). The studied Oilbirds based their activities in the main cave at Cueva de Los Guacharos National Park (NP), but their moving ranges included three additional national parks (Alto Fragua Indi Wasi, Serranía de los Churumbelos, and Puracé), regional reserves, private farms, and some urban areas. The main cave, in which nesting takes place, is located at 2000 masl ( Fig. 1). The forests at Guacharos NP are classified into three groups: sub-Andean forests located between 1100 and 2400 masl, which include primary and secondary forests; Andean forests, located from 2400 to 2700 masl; and subparamo above 2700 masl (MINAMBIANTE 2005). At least 74 families of woody plants have been found in the park (Prada and Stevenson 2016). Also, there are forests dominated by oak trees, which are distributed in the upper limit of sub-Andean forests and the lower limit of Andean forests (MINAMBIENTE 2005).

Data collection
GPS telemetry devices of approximately 23 gm (e-obs GPS-Tags WGS84-Height, Global Positioning System of e-obs digital telemetry ©) were attached to five individuals Oilbirds captured at the main cave of Cueva de Los Guacharos NP. The devices generate radio wave transmissions at a specific radio frequency that were received with a Yagi-Uda-antenna (provided with the devices for that specific task) in the cave, allowing the downloading of the data. GPS tags corresponded to approximately 6% of the individual's body weight (mean body weight = 419 gm; Holland et al. 2009) and were attached to the back of the birds through a harness-type system using an elastic thread of 0.6 mm wide, which surrounded the individual's wings and the chest in a T form (Roshier and Asmus 2009). Devices were located close to the individual's center of gravity, so that the installation would not affect their mobility. Captures were performed using mist nets in the cave's entrance during two different seasons. Birds of unknown sex greater than 200 gm were selected. Devices labeled G29 and G30 were installed in the nonbreeding season, beginning data collection on December 2015; and devices G31, G32, and G33 were placed during the breeding season, beginning data collection on March 2016. Data were recovered for 3 out of 5 individuals: 56 and 31 nights were sampled for devices G32 and G33, respectively, and 17 nights were sampled for device number G29. Two Oilbirds in the nonbreeding period provided no data because they did not return to the cave, at least when we downloaded data at proximity. Even with low sample sizes for both the nonbreeding and the breeding season, we present some data separately. Information on location (WGS-84), flight speed, elevation, and flight direction were collected for each bird at 30-minute intervals, from 18:00 to 6:00 h COT. A total of 2266 records were obtained for the 3 individuals, of which 29% reported coordinates and speeds of zero. Because the GPS signal is lost within the cave, coordinates of the cave were assigned when the trajectory showed an approach to that point. All location data generated by the telemetry devices were converted to UTM (universal transverse mercator) coordinates.

Fig. 1. Study area and the location of Cueva de Los Guácharos
National Park in which the Oilbird colony under study was located.

Home range and movements
The home range was estimated for each individual and all three, using 100% minimum convex polygon (MCP) because it allows comparisons with other studies. Also, within the home range established for each individual, a grid (5 x 5 km) was built to estimate the number of times a quadrat was used and thus determine if the cumulative sampling effort per individual was sufficient. Furthermore, Kernel UD analyses were performed to estimate the home-range size of the studied individuals, using 95% and 50% of probability of spatial occurrence as indicators of the total range and core areas, respectively. The home ranges were estimated and visualized using the extension Spatial Analysis for ArcGIS 10.3 and R software, using the package adehabitatHR (Calenge 2006, R Development Core Team 2012. Daily straight-line distances were calculated as the sum of the linear distances between sampling points every 30 minutes. Flight-speed data were downloaded as instantaneous samples (not calculated from a distance moved over time between consecutive GPS locations) to establish at which time of the night individuals used different movement patterns, indicating the hours when the Oilbirds engaged in fast or slow movements. Kruskal-Wallis tests were used to examine differences in flight speed between individuals and night hours.

Landscape elements and habitat use
Geographic information analyses were performed with the software QGis, ArcGis, and the packages raster, rgdal, sp, and gdistance in R (Pebesma and Bivand 2005, Hijmans and van Etten 2014, Bivand et al. 2016, OSGEO 2018. A vegetation map was generated to define the different habitat types available for the Oilbirds, including those known to be used by the species in previous studies (Snow 1961, Tannenbaum and Wrege 1978, Roca 1994, Bosque et al. 1995, Bosque 2002, Holland et al. 2009, Karubian et al. 2012a). Following CORINE land-cover classification, habitat types were grouped as (1) dense forest, (2) fragmented or disturbed forest, (3) coffee plantations, (4) mosaic of crops and natural forest, (5) mosaic of grass with natural forest, (6) pastures and crops, (7) burnt zones, naked, and degraded lands, (8) rivers and lagoons, (9) secondary vegetation, (10) shrublands, (11) forest plantations, and (12) urban zones. Additionally, the central cave was included as a cover category to avoid overestimating the use of dense forest, in which the main cave is located.
To identify which variables affect the frequency of habitat use, an ecological-niche factor analysis (ENFA) was performed using the package dehabitatHS in R 3.4.1, which estimates the degree of marginality and specialization using data on presence relative to the habitat availability in the study area (7.961 km²). Marginality is defined as the environmental distance between the species' optimum and available habitat, and specialization is the relationship between the ecological variance of the habitat and that of the species. In other words, specialization is opposite to tolerance and suggests how specialized individuals can be to specific variables (Hirzel et al. 2002). The following factors were used as explanatory variables: forest cover (percentage of trees per pixel, 30 x 30 m, in raster format; Hansen et al. 2013), height above sea level from a digital model of elevation (Amatulli et al. 2018), and degree of fragmentation, associated to the degree of continuity between the forest fragments in the study area, obtained from the forest cover map. To quantify the degree of fragmentation (F) four parameters were used: n = the quantity of cells with forest cover within the analysis window; c = the number of cells that have at least another contiguous cell with forest cover, meaning that it shares one of its sides completely; g = the number of groups of cells of the forest cover that are formed within the window; and N = the total number of cells. From these data, a fragmentation index was estimated according to González-Caro and Vasquez (2017) with values of zero when fragmentation is high and values close to one when fragmentation is low.
From these analyses, a raster is obtained, indicating the index of fragmentation for every pixel, and then used as a niche factor. The analysis was performed with 999 random data replicates. To test for spatial pseudoreplication, an estimation of Moran's Index 1 was included (Dutilleul et al. 1993).

Influence of vegetation
To evaluate if plant diversity and the proportion of species consumed by Oilbirds are associated with habitat use frequency, a comparison was made between the characteristics of the most frequently used areas by the individuals equipped with the tracking devices, and other plots in which Oilbirds were not reported. Frequently used areas (N = 30) were selected based on a high number of independent visits (more than 6). The minimum distance between sites was 300 m to avoid spatial pseudoreplication. A total of 25 locations were surveyed because of accessibility and security.
At each of the frequently used locations, 0.1 ha circular vegetation plots were established in which all trees, palms, and lianas with a diameter at breast height (DBH) > 10 cm were marked, identified, and measured. Because GPS points can have between 2-5 m of error, circular plots were established to increase the chance of including feeding trees. A GPS fix was taken as the center of the plot, and vegetation within 20 m was sampled. These plots in the frequently used locations were compared with plots within the PNN Cueva de Los Guacharos, other places in the south of Huila, and in the foothills of Caquetá, that were within the sampling window ( Fig. 1), but with no reports of visits by Oilbirds. These plots were taken as controls (N = 46), even though it cannot be guaranteed that these are zones are not used by the Oilbirds, and their location was determined to represent the floristic composition of different forest types in the region (e.g., Prada and Stevenson 2016 (Hubbell et al. 1999), which was highly variable . In fact, fewer individuals were found in frequently used areas than in control plots (ANOVA F 1,69 = 27, p < 0.001; Appendix 1, Table  A1.1). The number of species per stem, the number of individuals of consumed species, their proportion, and basal area were compared between frequently used and control sites using ANOVA tests, given the parametric nature of the data.

Home range and movements
According to the MCP, the three Oilbirds occupied a home range of at least 4517 km² (580.6 km² for individual G29, 1866.8 km² for individual G32, and 2295.5 km² for individual G33), whereas the 95 and 50% kernel were 2567.9 km² and 481.7 km², respectively. However, for all three individuals, rarefaction curves showed that the sampling effort was not enough to obtain a reliable estimation of the home-range area (Fig. 2), which suggests more extensive ranges. The individual sampled in the nonbreeding season spent more days outside the cave (58%) than those tracked during the breeding season (6 to 17.6% of days outside the cave), which also tended to remain closer to the cave than the other birds. Core areas included the region around the cave and grouped points in foraging areas separated by a maximum distance of 40.7 km (Fig.  3), which agrees with the Moran Index indicating spatial clustering. Values for G29 are shown in red (nonbreeding season), for G32 in green, and for G33 in blue.  The mean overall movement speed was 9.2 km/h (including travel and foraging speeds), but the maximum speeds registered were between 54.9 and 59.7 km/h (Appendix 1, Table A1.3). In general, higher movement speeds were recorded at the beginning and at the end of the night (X 2 = 197.9, df = 11, P < 0.0001; Fig. 4). The Oilbird sampled in the nonbreeding period showed the lowest speeds (Appendix 1, Table A1.3).

Landscape elements and habitat use
The most frequently used area was the cave (28% of the records), followed by dense forest (21%), a mosaic of crops and natural vegetation (15%), and coffee plantations (12%). The available places that did not show records included urban zones, exotic tree plantations, shrublands, and degraded, burnt, and exposed soils.
According to ENFA, there was a marginality of 0.3 (axis X in Fig. 5A), indicating that the monitored Oilbirds used specific places relative to the available habitat within the study window. Specialization (axis Y in Fig. 5A) was 4.09, showing that the variance of the available habitat was four times higher than the variance of used habitat. Considering the computation of the marginality and the specialization, there was a significant difference between the areas used and the available habitat (P = 0.001, Fig 5A). Results showed that the most influential variables of habitat use were elevation and tree cover, which showed high levels of specialization and showed tolerance to fragmentation (axis Y in Fig. 5A). There was a preference for the use of intermediate elevations of 1000-2000 m, compared to the 200-3500 m range available in the study area (Fig. 5Ba), showing intolerance to altitudes with no birds visiting elevations above 3000 m. Similarly, they chose areas with high tree cover (Fig. 5Bb), which does not imply that they do not travel across pastures, agricultural, or urban zones, but indicates a higher permanence in areas with forest cover. Oilbirds were detected in places with all possible degrees of fragmentation, but showed a peak at relatively low fragmentation scores (Fig. 5Bc). Several of the frequently used areas were located within transformed matrices, including coffee plantations and visits to isolated trees in pastures. Based on this information, we could suggest that the current degree of fragmentation is not a determining factor in the distribution and landscape use by the Oilbirds.

Influence of the vegetation
When controlling for the number of plant individuals, the proportion of basal area (BA) of plants consumed by Oilbirds was higher in frequently used areas. When examining the index of species per stem, no differences in diversity were found between plots in frequently used areas and control plots (F 1,69 = 0.11, P = 0.74). On the other hand, the proportion of individuals that were included in the Oilbirds' diet was higher in the frequently visited plots (F 1,69 = 9.8, P = 0.002; Fig. 6a), as well as the proportion of BA of consumed plant species (F 1,69 = 4.23, P = 0.04; Fig. 6b). Moreover, to evaluate if this difference was due to the presence of particular forest types, the same comparison was made removing plots from forests that do not have many elements consumed by the Oilbirds, such as those dominated by species of white oak (Quercus humboldtii), black oak (Trigonobalanus excelsa), and Alfaroa colombiana. These species are not part of the Oilbirds' diet and in 6 control plots represented between 60 and 90% of the total BA, which when removed from the analyses, the difference in the proportion of BA of consumed individuals was no longer significant (F 1,63 = 2.35, df = 63, P = 0.13). This suggests the avoidance of this highly dominated type of forest by Oilbirds because none of the frequently used plots presented dominance of these plants.   Smith 2000, Mueller et al. 2014), have home ranges between 9.25-44.72 km² (Holbrook and Smith 2000) and traveled distances of up to 15 km for Bycanistes bucinator (Mueller et al. 2014), which are also lower than the values reported for the Oilbirds.

DISCUSSION
Another interesting fact is the confirmation that these birds can spend several days outside the cave, as has been previously reported in Venezuela (Holland et al. 2009). Given their broad home range, they may play an important role as long-distance seed dispersers, even more so when they do not return to the cave every night. If this is so, Oilbirds would be dispersing seeds far from their parent trees, allowing them to escape possible pathogens, to avoid competition, and to colonize new habitats, thus giving them an advantage for establishment (Howe andSmallwood 1982, Kellner andHubbell 2018).
Considering the low number of Oilbirds providing data from GPS devices, it is difficult to infer definitive conclusions. However, the results suggest that movement patterns might differ between breeding and nonbreeding seasons. During the reproductive season, birds showed higher flight speeds, longer daily traveled distances, and more extensive home ranges compared to the individual monitored outside the reproductive season. At the same time, they spent a lower proportion of days outside the cave, compared to the individual monitored in the nonbreeding season. This could be related to the fact that during the reproductive season birds are forced to return to their nests in the cave to take care of the chicks, whereas outside the reproductive season, birds can spend several days outside the cave near foraging sites, as reported by Holland and collaborators (2009). This would also explain patterns of high flight speeds at the beginning of the night, when they leave the cave, and early in the morning when they must return to the cave, especially during the breeding period. The median overall speed was low, probably because Oilbirds tend to forage for fruits in flight, and these foraging movements require low velocity.
Our results indicate that caves were the most frequently used habitat, which is vital to complete reproductive cycles, and habitats with high tree cover were the second most important, allowing Oilbirds to find feeding sources. However, high levels of fragmentation did not seem to limit their movements because they could overfly above crop and pasture matrices, using the whole range of fragmentation scores in the matrix studied. It is known that for several bird species, long distances between desired covers (such as forests), mountains, rivers, roads, and even behavioral limitations can restrict their movement (Bélisle and St. Clair 2002, Harris andReed 2002, Laurance et al. 2004). For the Oilbirds we tracked, long distances were traveled both over continuous forest cover, and over severely fragmented areas (at least 10 km flying over pastures), suggesting that they may not require stepping stones to cover long distances. Long-wattled Umbrellabirds (Cephalopterus penduliger) can fly over open zones in Ecuador, allowing seed dispersal and gene flow between plant populations of different forest fragments (Karubian et al. 2012b). Oilbirds could be playing similar roles in their distribution range, even at broader spatial scales, and generating opportunities for plants to colonize new habitats, taking seeds to isolated places. Even though they were detected mainly in areas with high forest cover, they also forage in pastures and crops, in which they may find fruit sources (for instance, Nectandra acutifolia was found in open habitats). It is not clear how they know where these fruit trees are located, but their relatively unidirectional and nonerratic movement patterns suggest a previous knowledge of their location. This could be explained by mental maps and/or the sharing of information among colony members, which could be a topic for future studies.
At the scale of plots, it was evident that feeding trees are associated with the habitat-use frequency. The number, proportion, and basal area of consumed species were higher in the frequently used areas than in control plots (Appendix 1, Fig. A1.1), showing a prevalent role of fruit productivity rather than diversity. The diet of Oilbirds includes species with high contents of lipids such as laurels and some palms (Stevenson et al. 2017) and the sites that they visit present an abundance of consumed species. We also evidenced that frequently used sites had a high density of feeding trees, and within the study area, 58% of the frequently used plots had at least one large individual of consumed species (which possibly determined use). However, these species are also present in some control plots that might also be visited by the Oilbirds. A more considerable sampling effort would be required to confirm that Oilbirds would visit control sites when these species are fruiting. The analysis of vegetation differences between frequently used and control plots showed that when plots placed in oak forests were excluded (dominated by Trigonobalanus excelsa, Quercus humboldtii from the Fagaceae family and Alfaroa colombiana from the Juglandaceae family), differences were not significant. This suggests that Oilbirds avoid these zones in which there are fewer food resources for them.
Given the large home-range size found, the conservation of this colony very likely depends on the protection of forests in the whole region. Fortunately, in this region of Colombia, there is a network of national (Cueva de Los Guacharos, Churumbelos, Alto Fragua-Indi Wasi, and Puracé) and regional natural parks (e.g., Corredor Oilbirds in Puracé) that favor the maintenance of forest ecosystems. However, it was estimated that only 39.5% of the original cover remained by 2000 in the Andean territory of Colombia (Rodríguez et al. 2006) and annual deforestation in a recent year was 26,014 ha (Galindo et al. 2014). Although Oilbirds can tolerate open zones because of their wide home ranges, it is necessary to know which fragmentation threshold sustains appropriate resources and how far they can travel between forest patches. To overcome this and given the reports of individuals of this species using isolated trees, we suggest that the maintenance of forest cover in other landscape elements is necessary (as long as they include trees of the species consumed by the Oilbirds). It is also important to outline that not all types of forest cover are adequate for the species because commonly occurring vegetation types in the study area such as oak forests, and the forested plantations do not offer many resources.
Our results indicate that the elevation is also a variable that restricts habitat use by the Oilbirds, keeping their activities within an altitudinal range of 200-3000 masl, but with a higher frequency between 1000-2000 masl. Avoiding high elevations is likely a product of the fact that in this region at high altitudes, there is a preponderance of oak forests (Prada and Stevenson 2016). Besides, most flight paths to lowlands in Caquetá went through a narrow, low altitude area in the edge of the Eastern Cordillera, which could be associated with a preference for routes that minimize climbs and maintain relatively straight movements.

CONCLUSION
Landscape composition such as cover types, elevation, and plant composition play significant roles in habitat selection and movement patterns in Oilbirds. Oilbirds use caves very frequently, nesting in conditions that may minimize predation rates, and they prefer forest cover, in which they visit a particular set of fruiting plants. Our findings show that they move widely, visiting distant fragments or even isolated trees to find feeding resources in areas of at least 4517 km², avoiding high elevation areas, where their preferred food plants are scarce. Although Oilbirds are capable of inhabiting fragmented landscapes, our results highlight the importance of forests to maintain their nutritional needs. Because they use different habitats such as plantations, crops, and grass mosaics (as long as they have available resources), it is essential to maintain their feeding trees in the landscape to assure Oilbirds' visits, and thus, their ecosystem services.    (1) and where oilbirds have not been reported (0), in plots established in the Huila and Caquetá departments.