OUP user menu

Habitat Selection at Two Spatial Scales and Diurnal Activity Patterns of Adult Female Forest Buffalo

Lisa M. Korte
DOI: http://dx.doi.org/10.1644/06-MAMM-A-423.1 115-125 First published online: 19 February 2008


Forest buffalo (Syncerus caffer nanus) occur throughout the Congo Basin forest region of central Africa. Unlike for the well-studied Cape (or savanna) buffalo (Syncerus caffer caffer), few data exist for forest buffalo. I tracked 7 radiocollared adult female forest buffalo at Lopé National Park, Gabon, over a 2-year period (2002–2004) to examine home ranges, habitat selection, and diurnal activity patterns. Home ranges of female forest buffalo averaged 4.55 km2 in area (mean number of locations per animal = 205); the percent of home-range overlap between individual radiocollared buffalo was small. Home ranges remained the same size and in the same locations over both study years. Distance analysis of habitat use from radiotracking data was used to assess forest buffalo habitat selection at 2 spatial scales. At the landscape scale, buffalo selected savanna and marsh habitat over forest habitat within a 72-km2 study area. Thus, forest buffalo home ranges were savanna-dominated despite the greater amount of forest habitat available in the overall landscape. At the scale of the home range (2.30–7.64 km2), habitat selection within home ranges varied with season. Adult female forest buffalo preferred forest habitat between March and August but preferred marsh to forest between September and February. Forest buffalo dwell in forest habitat, feed in savannas, and wallow in marshes, utilizing all habitat types in the landscape. Although the subspecies is forest-dwelling, forest buffalo depend on open habitat adjacent to continuous forest.

Key words
  • activity pattern
  • animal movement
  • distance analysis
  • Gabon
  • habitat selection
  • local convex-hull (LoCoH) home-range estimation
  • Lopé National Park
  • minimum convex polygon (MCP)
  • space use
  • Syncerus caffer nanus

Ungulates are the focus of many studies of mammalian ecology and space use. Antelope are of particular interest because of their considerable variation in morphology, geographic range, and behavior among the species and the number of species (i.e., 143 species within family Bovidae). Because wild populations are observable and captive populations are easily managed, the group is well studied. Early studies highlighted important relationships between feeding style and habitat use by African antelopes (Gwynne and Bell 1968; Jarman 1974). Subsequently, advanced statistical techniques and a phylogeny of extant species have been used to verify the significant correlation between habitat use and feeding styles in ungulates (Perez-Barbería et al. 2001). However, few studies have examined habitat use for single species that range across a variety of habitats (Brashares and Arcese 2002; Brashares et al. 2000; Jarman 1974). Data are lacking for analysis of habitat selection for those species that may provide the most insight on the relationship between feeding style and habitat use. Here, I investigate space use of the African buffalo (Syncerus coffer). This large-bodied grazer (Halley and Minagawa 2005; Hofman and Stewart 1972; Janis 1988; Jarman 1974) has both savanna and forest-dwelling populations (Haltenorth and Diller 1980; Kingdon 1997; Sinclair 1977).

Differences in morphology, geographic range, and behavior are evident among subspecies of Syncerus. The 2 most widely recognized subspecies are the savanna or Cape buffalo (5. coffer coffer) and the forest buffalo (S. coffer nanusGrubb 1972; Wilson and Reeder 2005). The more familiar Cape buffalo are massive animals, weighing 400–800 kg, with downward curved horns (Haltenorth and Diller 1980; Sinclair 1977). At weights of 250–320 kg, forest buffalo are approximately one-half the size of Cape buffalo (Haltenorth and Diller 1980). Forest buffalo have small, swept-back horns without the lateral extension characteristic of Cape buffalo.

Cape buffalo have been studied in most detail in the savannas of eastern and southern Africa where vast, open grasslands support herds averaging 350 animals, with large groups reaching into the thousands (Conybeare 1980; Funston et al. 1994; Halley and Mari 2004; Halley and Minagawa 2005; Halley et al. 2002; Hunter 1996; Mloszewski 1983; Prins 1996; Ryan et al. 2006; Sinclair 1977; Taolo 2003). In contrast, the geographic range of S. c.nanus is the Congo Basin forest region of central Africa (Haltenorth and Diller 1980; Kingdon 1997; Sinclair 1977). Because of the forest habitat and their elusive lifestyle, few data exist for S. c. nanus (Blake 2002; Melletti et al. 2007). Relative to Cape buffalo, much less is known about space use of forest buffalo.

Although some studies of forest buffalo have addressed space use, conclusions are limited because of short study periods and small sample sizes. At Lopé National Park (Lopé NP), Gabon, Molloy (1997) found that forest buffalo fed on the flush of new grass on savannas, but data collection was limited to one 6-month period and home-range estimates relied solely on visual observations of only 3 groups. Although Abernethy (K. A. Abernethy, Station des Etudes des Gorilles et Chimpanzés, Gabon, in litt.) later tracked radiocollared forest buffalo over a 2-year period at Lope NP, the sample size (i.e., 1 adult female and 1 adult male) was too small for analysis of habitat selection (Aebischer et al. 1993). At Dzanga-Ndoki National Park, Central African Republic, forest buffalo were highly dependent on forest clearings, but observations were of a single group (Melletti et al. 2007). For analysis of habitat selection, home-range data with larger sample sizes are needed.

Kingdon (1982) suggested that small open areas found within forests could support forest buffalo because the humid climate provides sufficient food resources for this large grazer throughout the year. For mammals, habitat productivity and body mass influence areas of home ranges (Harestad and Bunnel 1979; McNab 1963). Home-range size increases with body mass (McNab 1963); however, smaller home ranges are found in more productive habitats (Fisher and Owens 2000; Harestad and Bunnel 1979). Although field surveys indicate that forest buffalo use open habitat within the forest landscape (Blake 2002; Prins and Reitsma 1989; Tutin et al. 1997; White 1994), data on space use within home ranges are needed to test if forest buffalo select open areas.

In this study, I used radiotelemetry to examine diurnal space use of adult female forest buffalo at Lope NP, Gabon. My objectives were to estimate home-range areas, to examine ranging patterns, to assess habitat use at 2 spatial scales, and to record diurnal activity of forest buffalo. I compare my findings for forest buffalo with observations of Cape buffalo and analyze how forest buffalo use habitat within the forest landscape.

Materials and Methods

Study area.—I observed buffalo in Lopé NP, Gabon. This 5,000-km2 park is located in the center of Gabon just south of the equator (00°12′04″S, 11°36′05″E; Fig. 1). White (1983) classifies the park as lowland tropical rain forest. Average total annual rainfall is 1,500 mm, unevenly distributed over 2 dry seasons (December-February and June-August) and 2 wet seasons (March-May and September-November). Temperatures fluctuate little throughout the year with maxima ranging from 27°C to 31°C and minima from 20°C to 22°C (White and Abernethy 1997).

Fig. 1

Geographie location of the study area in Lope National Park, Gabon.

The ungulate community at Lopé NP includes forest buffalo, forest elephant (Loxodonta cyclotis), sitatunga (Tragelaphus spekii), bushbuck (Tragelaphus scriptus), red river hog (Potamochoerus poreus), water chevrotain (Hyemoschus aquaticus), and 7 species of duikers (CephalophusTutin et al. 1997). Leopards (Panthera pardus), the only large carnivore at Lopé NP, are a natural predator of the ungulate community. Forest buffalo were 1 of the main prey items found in leopard scat; however, based on analysis of bones in the scat, leopards apparently prey only on juvenile buffalo (Henschel et al. 2005). Hunting pressure by humans on the population of forest buffalo is low because the park headquarters is adjacent to the study area and because tourists and researchers use the area on a daily basis.

The research site included 3 habitat types: marsh, savanna, and forest. Digital maps of the northeastern area of the park are available to researchers at the Lopé field station, Station d'Etudes des Gorilles et Chimpanzés, in the form of Arc View 3.x (Environmental Systems Research Institute 1999) shape files, and include the locations of roads, rivers, marsh, savanna, and forest habitat types. For marsh habitat that had not been previously mapped, I measured the perimeter of marsh habitat, created shape files, and added the marsh shape files to the digital map collection at Station d'Etudes des Gorilles et Chimpanzés. Thus, maps including marsh, savanna, and forest habitat were available for analyses.

Observations of buffalo were restricted to the mosaic of savanna, marsh, and forest habitat in the northeastern corner of the park. The majority of the park is forest habitat and large mammal surveys by White (1994) suggested there were few buffalo in the forest adjacent to the mosaic habitat. When present, large mammals are clearly visible in the savanna habitat. Sixty kilometers of road traverse this mosaic and are used by tourists and researchers. Park management staff annually burn the savannas, maintaining the open areas that would otherwise be colonized by forest (Peyrot et al. 2003). These open areas facilitate the viewing of large mammals on safari drives, and provide scenic vistas for tourists. Savanna areas are burned in small sections based on natural fire breaks and roads in the landscape. Small patches are burned as soon as the rains have stopped each year, when grasses are sufficiently dry to burn well. A few small areas are burned each week during the 2nd dry season through the end of September or beginning of October, depending on rain and burning conditions. This patchy burning creates flushes of new grass, and the number of buffalo observed in the savanna areas is low in July but increases in August and remains high through December (based on observations between July and December 1996— Molloy 1997).

Observations of animals and data sampling.—The first 3 months of observation (September-December 2002) were used to identify groups of forest buffalo. Three fixed road circuits were established to monitor buffalo, so that individuals from different groups could be selected for radiocollaring. These circuits covered all open savanna areas visible from roads and each circuit was surveyed at least 4 times each month (twice during the morning, 0630–0930 h, and twice during the late afternoon, 1530–1830 h). Date, time, location, group size, and activity of individuals were recorded. Age class and sex were determined for the majority of individuals in each group. If there was a clear view and buffalo were within 50 m of the observer, a Sony digital video camera (model DCR-TRV140 NTSC; Sony, Tokyo, Japan) was used to record buffalo. Videos were reviewed to note additional unique characteristics that could be used to recognize individuals. Distinct pelage characteristics and horn morphology were used to differentiate individuals. Using easily recognizable buffalo as indicator animals, it was possible to distinguish groups.

Radiocollars were placed on 8 adult female buffalo (AF1-AF8) during the 1st week of December 2002. Four collars were manufactured by HABIT, Inc. (Victoria, British Columbia, Canada) and 4 by Telonics, Inc. (Mesa, Arizona); receivers from both companies were used to track buffalo. Females were from 8 different groups, which represented about one-half of the groups using the study area. Ideally, Börger et al. (2006) recommend that telemetry studies strive to maximize the number of animals with radiocollars; however, I was limited by the number of collars. Therefore, I chose to sample animals within the same sex and age class (i.e., adult females) because core group members in Cape buffalo are adult females (Mloszewski 1983; Prins 1996; Sinclair 1977). Sampling additional sex and age classes during this study would have severely reduced sample size (Aebischer et al. 1993).

Two collars malfunctioned, with 1 failure in July 2003 and a 2nd in October 2003. In early December 2003, the collar on AF3 that failed in October 2003 was replaced. A 2nd collar was placed on a new adult female (AF9) in the same group as the remaining buffalo (AF5) with the malfunctioning collar. The 2nd collar failed immediately resulting in 1 group with 2 collared buffalo (AF5 and AF9), but neither collar emitting a signal. Without the signal, it was not possible to locate animals when they were in forest habitat. Observations were therefore limited to visual observations of these 2 animals for the 2nd year of the study. Hence, data for adult females AF5 and AF9 were used only for the analysis of home-range overlap and were excluded from the analyses of habitat selection and activity patterns. I therefore report results for 7 adult females over a 2-year period for the analyses of habitat selection and diurnal activity patterns.

During 3 study periods (December 2002-April 2003, July-December 2003, and July-December 2004), radiocollared animals were located twice per week on different days. To ensure that animals were tracked throughout the day, daily observations were blocked into four 3-h observations periods (i.e., 0630–0930, 0930–1230, 1230–1530, and 1530–1830 h). Each animal was tracked at least twice during each of 4 daily observation periods over the course of a month. However, observations were reduced to 1 location per week between 0900 and 1500 h during May-June 2003 and January-June 2004. Opportunistic observations of radiocollared animals outside the search periods also were recorded. For each observation, date, time, location, group size, and activity state of individuals were recorded.

Activity data were collected during observations when radiocollared animals were visible. Focal animal surveys on radiocollared individuals lasted 20 min with an activity recorded every minute. A group scan was conducted if the collared animal was not alone; the activity of each animal in the group was recorded every 5 min for 20 min. Observations were only recorded during daylight hours. Visual observations also were limited to open savanna or marsh areas because observing animals in the forest resulted in animals either running from or toward the observer. In both cases, behavior was changed because of the observer.

All research was conducted under a permit from the Gabon Ministry of Water and Forests and under the direction of Station d'Etudes des Gorilles et Chimpanzés, the field station of the Centre Internationale de Recherche Médicales de Franceville. Work with live animals was carried out in a humane manner and in accordance with guidelines of the American Society of Mammalogists (Gannon et al. 2007) and Michigan State University. Veterinarians from the Wildlife Conservation Society's Field Veterinary Program handled buffalo sedation for radiocollar placement.

Determination of home ranges.—I plotted the locations of radiocollared animals as points in an Arc View 3.x (Environmental Systems Research Institute 1999) shape file. Each collared animal had a point shape file, with date, time seen, time of 1st and last telemetry fix, activity, number of buffalo in group, notes, departure time, and type of observation (i.e., visual or fix) recorded into the attribute table.

I used the local nearest-neighbor convex-hull construction (LoCoH—Getz and Wilmers 2004) to estimate the area of the home range for each radiocollared buffalo, as well as minimum convex polygons (MCP) for comparison with previous studies. Locations used for the estimations included visual observations of the animals and telemetry fixes and were statistically independent because observations were recorded during different time periods if recorded on the same day (an infrequent occurrence) or different days for the majority of observations. I used the animal movement analysis Arc View extension (Hooge and Eichenlaub 1997) to determine the MCP and to calculate its area. Although the MCP method is an accepted standard for calculating ranges and is widely used because of its simplicity (Burgman and Fox 2003; White and Garrott 1990), it has been criticized because presence of outliers can dramatically overestimate the home-range area (Burgman and Fox 2003). Thus, LoCoH has been advanced by Getz and Wilmers (2004) and was 1 st used to estimate ranges for African buffalo in South Africa (Ryan et al. 2006).

I used the LoCoH Homerange Generate Arc View extension to estimate area of the home range for each radiocollared buffalo (see Getz and Wilmers [2004] and Ryan et al. [2006] for details). This extension uses the locations to create the convex hull with each location and its k nearest neighbors. Because the k parameter is user-selected, I ran this method for k values from 2 to 40 to identify the plateau that gives stable-area values across a range of k values that represent the estimated area of the home range. When several plateaus occurred, I chose k values that eliminated the unused areas within the range because the topology of the study area did not include lakes, mountains, or inhospitable habitats that may be avoided by buffalo. This selection process followed the “minimum spurious hole covering” rule (Getz and Wilmers 2004) and I report k values for estimated LoCoH home-range areas. I used a Mann-Whitney U-test to compare the estimated areas of the 2 methods (LoCoH and MCP).

I also examined home-range overlap between individuals and between years for each individual. To determine whether radiocollared buffalo maintained discrete home ranges, I calculated the percentage of home-range overlap between radiocollared buffalo based on areas calculated using the LoCoH method. Home-range overlap between years for individual radiocollared buffalo was examined with the Mann-Whitney U-test. This test also was used to test for differences in the area of home ranges between years.

Habitat measurements and landscape variables.—To measure available habitat in the overall landscape, a 72-km2 study area was delineated based on the merged LoCoH home-range areas of the radiocollared buffalo with a 1-km buffer. The habitat composition of the study area was 2.45% marsh, 43.20% forest, and 54.35% savanna. Forest habitat within the study area included forest fragments, which have been described by Tutin et al. (1997): gallery forest (narrow strips of forest along watercourses continuing at one end to join the main forest), “corridor” (a narrow gallery linked at both ends to forest), and bosquets (small forest blocks completely surrounded by savanna; Fig. 2). Forest habitat in the study area also included continuous forest.

Fig. 2

Locations (points), 100% minimum convex polygon (MCP; black polygons), and local convex-hull (LoCoH; color polygons) home-range estimates for 9 adult female forest buffalo (AF1-AF9; Syncerus caffer nanus) with radiocollars at Lope National Park, Gabon, December 2002-December 2004.

Habitat selection and distance analysis.—Several methods exist to compare resource use and availability (Thomas and Taylor 2006); I used Euclidean distances to assess nonrandom habitat use and to rank the habitat types (Conner and Plowman 2001). The distance analysis is robust to telemetry error and unlike compositional analysis can include zero-use areas without influencing type I error (Bingham and Brennan 2004). Each radiocollared buffalo was treated as 1 sample.

I used the Nearest Features Arc View extension (version 3.8a—Jenness 2004) to calculate distances between animal locations and the nearest representative of each habitat type; distances for each animal were averaged for the analysis. I examined both 2nd-order selection of habitat within the landscape and 3rd-order selection of habitat within the home range (Johnson 1980). For comparison at the landscape level, 2,500 random points were generated in the study area and distances between each habitat type and each random location were calculated and averaged to create an average distance for each habitat type. I calculated ratios for each animal by dividing the distance associated with buffalo locations by distances derived from random locations within the study area. I used a multivariate analysis of variance (MANOVA) to determine if the mean ratio vectors differed from a vector of 1, which indicates nonrandom use. To test for significant differences between individual ratios for each habitat type and the available habitat in the study area, univariate t-ests were used. Habitats were ranked using pairwise habitat comparisons to construct a ranking matrix of t-statistics.

For analysis within the home range, locations of buffalo were paired with a random location within each home range and distances to each habitat for the random locations were calculated. For each season, the ratios for each animal were calculated by dividing the buffalo locations by distances derived from random locations within the home range. Then I used the distance analysis as described above to determine if use was significantly different from random locations and to rank habitats. All statistics for the distance analysis were performed in SAS (version 9.1—SAS Institute Inc. 2002) using code provided by Conner and Plowman (2001) with a = 0.05.

Diurnal activity pattern.—I used MANOVA in SAS (version 9.1—SAS Institute Inc. 2002) to examine diurnal activity patterns for the 7 adult female forest buffalo with functioning radiocollars, testing for overall effects of habitat, year, season, and time of day on activity (feeding, active, and inactive). I tested for significant differences attributable to habitat and time of day among the means for each of the 3 behavioral categories. I used analysis of variance (proc GLM in SAS) followed by least-square difference to test for significant differences in activity categories of buffalo between savanna and marsh habitat type and among time periods.


Ranging patterns.—The MCP and LoCoH methods for calculating home-range areas yielded similar results with no significant difference between the estimated home-range areas for the 2 methods (MCP n = 7, LoCoH n = 7, U = 29, P = 0.31). However, because the MCP method is more likely to overestimate home-range areas (e.g., for buffalo AF4 and AF6; Fig. 2), LoCoH area estimates are reported and used for analyses.

Radiocollared adult female forest buffalo maintained home ranges that were <8 km2 and of the same size and in the same location from year to year. Mean (± SE) home-range area was 4.55 ± 0.72 km2, with individual areas ranging from 2.30 km2 to 7.64 km2 for the 7 collared buffalo included in the analysis (Table 1; Fig. 2). The total number of locations was not correlated with estimated home-range area (r2 = 0.23, P > 0.05, n = 7), indicating that increased sample size would not have resulted in larger home ranges. Individual home-range areas were not significantly different between year 1 (December 2002-November 2003) and year 2 (December 2003-November 2004; Mann-Whitney U-test; year 1 n = 7, year 2 n = 7, U = 26, P = 0.45), with the percent of home-range overlap between years ranging from 50% to 91% for the 7 individuals.

View this table:
Table 1

Home-range estimates based on 100% minimum convex polygon (MCP) and local convex-hull (LoCoH) methods for 9 radiocollared adult female (AF) forest buffalo (Syncerus caffer nanus) at Lopé National Park, Gabon, during December 2002-December 2004, including number of locations, k, and habitat composition of home ranges (ranked by home-range size in ascending order).

% area for each habitat types based on LoCoH area estimates
AnimalNo. locationsArea (km2) 100% MCPArea (km2) LoCoHkMarshSavanna Forest
X̄(SE)205 (11)5.38 (0.99)4.55 (0.72)4(1)65 (5)31 (4)
  • a Analysis of AF5 and AF9 limited to overlap of home ranges.

Overlap in home ranges between radiocollared buffalo.— The percent of home-range overlap between individual radiocollared buffalo was small (Table 2). The 1 exception was 99% overlap between AF5 and AF9, which were members of the same group. Four buffalo (AF1, AF2, AF4, and AF7) had ranges that overlapped with AF6. She used a long and narrow range, running north and south in the center of the study area (Fig. 2). Thus, the perimeter of her range had the greatest opportunity to overlap with others, and the highest percentages of overlap (26%, 17%, and 16%) involved her home range. Excluding AF5, AF6, and AF9, the percent of home-range overlap between any 2 individuals was ≤7%, including adjacent ranges where overlap was most often zero or <1%. The number of locations in areas of overlap was small and there were few occasions when 2 radiocollared buffalo were present in the overlapping area at the same time. Overlap occurred primarily in savanna areas with almost no home-range overlap in forest habitat, especially forest galleries.

View this table:
Table 2

Range overlap between individual adult female (AF) forest buffalo based on local convex-hull home-range area estimates. Values are the percentage of the range of a buffalo (row) shared with another buffalo (column).


Habitat selection based on distance analysis.—The average distance between the location of a buffalo and the nearest marsh, savanna, and forest habitat was 141.45 m, 24.49 m, and 66.32 m, respectively. In the study area, the average distance between randomly selected locations and the nearest marsh, savanna, and forest habitat was 390.98 m, 115.73 m, and 56.15 m, respectively. At the landscape level, the analysis of distance ratios indicated that locations of forest buffalo differed from random locations (F = 268.82, d.f. = 3, 4, P < 0.0001). Buffalo were found closer to marsh (t = −13.31, d.f. = 6, P < 0.0001) and to savanna (t = ℒ27.09, d.f = 6, P < 0.0001) than expected based on the relative availability of these habitats in the study area. There was no difference between locations of buffalo and random points with regard to distance to forest (t = 0.87, d.f. = 6, P = 0.4179). A ranking of habitats indicated that buffalo were found significantly closer to savanna than marsh habitat and significantly closer to marsh habitat than to forest habitat within the study area (Table 3).

View this table:
Table 3

The ranks and pairwise habitat comparisons based on a) mean distance of a forest buffalo from habitat A/mean random distance to habitat A within the study area, and b) mean distance of a forest buffalo from habitat A/mean random distance to habitat A within the animal's local convex-hull (LoCoH) home-range area by season.

Pairwise comparisonsb
Level of analysisRankaMarshSavanna
a) Habitat selection within study areaSavanna > marsh > forestSavannat = 2.72 (P = 0.0347)
Forestt = −4.10 (P = 0.0063)t= −4.84 (P = 0.0029)
b) Seasonal habitat selection within
estimated home ranges (LoCoH)
March—May, wet seasonForest > marsh > savannaSavannat = −3.85 (P = 0.0023)
Forestt = 4.67 (P = 0.0005)t = 5.33 (P = 0.0002)
June—August, dry seasonForest > marsh > savannaSavannat = −3.74 (P = 0.0025)
Forestt = 3.95 (P = 0.0017)t = 4.09 (P = 0.0013)
September-November, wet seasonMarsh = savanna, marsh > forest,Savannat = 0.96 (P = 0.3542)
savanna = forestForestt = −3.09 (P = 0.0085)t= -1.58 (P = 0.1389)
December—February, dry seasonMarsh = savanna, marsh > forest,Savannat = −2.10 (P = 0.0554)
savanna = forestForestt = −3.38 (P = 0.0049)t = -0.33 (P = 0.7437)
  • a Greater than symbol (>) denotes a significant difference between variables.

  • b The t-statistic testing the null hypothesis that (mean distance of a forest buffalo from habitat A/mean random distance to habitat A) — (mean distance of a forest buffalo from habitat b/ mean random distance to habitat b) = 0.

The proportion of locations within each habitat type varied with season (Fig. 3); therefore, habitat selection within home ranges was examined by season. The analysis of distance ratios indicated that locations of buffalo differed from random locations for all seasons: March-May (F = 68.49, d.f. = 3, 10, P < 0.0001), June-August (F = 5.89, d.f. = 3, 11, P = 0.0120), September-November (F = 7.96, d.f. = 3, 11, P = 0.0042), and December-February (F = 5.89, d.f = 3, 11, P = 0.0120). During March-May, buffalo were found closer to forest (t = −7.66, d.f = 12, P < 0.0001) than expected and associated less with savanna (t = 4.13, d.f. = 12, P = 0.0014) than expected; there was no difference between locations of buffalo and random points with regard to marsh (t = 1.85, d.f. = 12, P = 0.0890). Buffalo were found closer to marsh than expected during September-November (t = −4.53, d.f. = 13, P = 0.0006) and December-February (t = −4.12, d.f. = 13, P = 0.0012), but associated less with marsh habitat than expected during June-August (t = −4.12, df. = 13, P = 0.0012). During June-February, there were no differences between locations of buffalo and random points with regard to distance to savanna (June-August t = 0.92, df. = 13, P = 0.3755; September-November t = −0.67, d.f = 13, P = 0.5168; December-February t = 0.92, d.f = 13, P = 0.3755) or to forest (June-August t = 1.93, d.f. = 13, P = 0.0761; September-November t = 1.90, df. = 13, P = 0.0805; December-February t = 1.93, d.f. = 13, P = 0.0761).

Fig. 3

Mean proportion of locations (± SE) during daylight hours in each habitat by season for 7 radiocollared adult female forest buffalo at Lopé National Park, Gabon, December 2002-December 2004.

Within home ranges, a ranking of habitats indicated that forest was proportionally used most during March-August when buffalo were found significantly closer to forest than marsh or savanna (Table 3). During September-February, buffalo were located significantly closer to marsh than forest, whereas there was no significant difference between marsh and savanna or between forest and savanna.

Diurnal activity patterns and behavior.—Buffalo tended to use savannas for feeding and marsh areas for resting during daylight hours (Fig. 4). The mean proportion of time spent feeding varied with habitat (F = 29.55, d.f. = 1, 441, P < 0.0001) and time of day (F = 5.57, d.f. = 3, 441, P = 0.0009); buffalo most often fed on savannas in the early morning. During daylight hours, buffalo spent >30% of their time feeding. Active behavior also varied with habitat (F = 9.03, df. = 1, 441, P = 0.0028) and time of day (F = 4.48, d.f = 3, 441, P = 0.0041). Buffalo were most active in savanna habitat and in the late morning (0930–1230 h). Buffalo spent <15% of daylight hours in active behaviors. As would be expected, periods of inactivity also varied with habitat(F = 46.30, d.f. = 1, 441, P < 0.0001) and time of day (F = 9.99, d.f = 3, 441, P < 0.0001). Inactive behaviors were most often observed in marshes in the late afternoon. During daylight hours, buffalo spent >38% of their time inactive.

Fig. 4

Mean proportion of time (± SE) spent in each activity category by habitat type A) marsh and B) savanna, and time of day for 7 radiocollared adult female forest buffalo at Lopé National Park, Gabon, December 2002-December 2004.

Proportions of time spent in different behaviors did not differ between years (F = 0.85, d.f. = 6, 874, A = 0.99, P = 0.5309) or among seasons (F = 0.34, d.f. = 9, 1,061.3, A = 0.99, P = 0.9604). However, behavior varied significantly among individuals (F = 2.02, d.f. = 18, 1,225.2, A = 0.92, P = 0.0070). This variation could be attributed to 1 individual (AF8) that spent a larger proportion of daylight hours feeding (56%) than inactive (38%). The other 6 buffalo spent a larger proportion of daylight hours inactive (54%) than feeding (37%). When AF8 was removed from the model, there was no overall effect of individuals.

Although it was not possible to visually monitor animals when they were tracked to forest areas, it was possible to monitor the signal emitted from the collar. Over the first 18 months of the study, animals were monitored for 20 min after pinpointing their location in a forest area. During this time, it was possible to determine if animals remained stationary or moved away. Based on 303 observations, buffalo remained in place during the 20-min observation period and were most likely resting in 1 location. Buffalo were relatively inactive during daylight hours between March and August, when >60% of their locations were in forest habitat (Fig. 3).


Home-range size.—Home-range size for adult female forest buffalo at Lopé NP (2.30-7.64 km2) was considerably smaller than home-range sizes for Cape buffalo, which have been reported at 10.50–296.3 km2 in eastern Africa (Prins 1996; Sinclair 1977), and > 1,000 km2 in southern Africa (Hunter 1996). Small home-range size relative to that of Cape buffalo was expected based on the much smaller body size of forest buffalo and the higher productivity of their habitat, 2 factors that can influence home-range size (Harestad and Bunnel 1979; McNab 1963).

Larger home ranges typically occur in less-productive habitats (Harestad and Bunnel 1979). Although habitat productivity is usually expressed in terms of mean annual net primary productivity (Murphy 1975), a more available measure, annual rainfall, was used here, following Harestad and Bunnell (1979) and Fisher and Owens (2000). Based on available net primary productivity data, regions with high annual rainfall typically have higher net primary productivity (Murphy 1975, 1977). In eastern Africa, mean home-range size of Cape buffalo is <200 km2 (mean annual rainfall > 600 mm—Prins 1996; Sinclair 1977); however, home-range size notably increases to > 1,000 km2 (a 5-fold increase) in the arid region along the border of Botswana and Zimbabwe (mean annual rainfall ∼ 500 mm—Hunter 1996). Where net primary productivity data are available, tropical forest has a higher net primary productivity than savanna (Murphy 1977, 1975), indicating the potential for patches of grass in a forest opening to provide buffalo a constant food source.

The forest buffalo at Lopé NP, an area with high annual mean rainfall (1,500 mm), inhabit a much more productive area than do most Cape buffalo (Conybeare 1980; Hunter 1996; Prins 1996; Sinclair 1977). In addition, the annual burning of savanna at Lopé NP creates open grass areas that provide food for buffalo. Home-range size of forest buffalo in my study was similar to that of 2 radiocollared forest buffalo examined in a pilot study at Lopé NP between December 1998 and June 2000; in that case, home-range size for a solitary adult male was 4.94 km2 (n = 82 locations) and 3.70 km2 for an adult female (n = 88 locations—K. A. Abernethy, in litt.). Despite the small size of these areas, the humid climate provides sufficient food resources for this large grazer throughout the year. Thus, home-range sizes observed in Lopé NP support Kingdon (1982) hypothesis that small open areas found within rain forest could support forest buffalo, and follow the predicted relationship of home-range size to body size and habitat productivity.

Ranging patterns.—Examination of my data demonstrates that groups of forest buffalo at Lopé NP occupy separate home ranges. Adjacent home ranges showed very little overlap (typically <1%) and radiocollared buffalo from different groups were rarely present in the overlapping area at the same time. Large areas of overlap were in savanna areas, but buffalo tended to use forest galleries within their home ranges separately. One individual (AF6) that overlapped considerably in space with 4 other radiocollared buffalo spent little time in the areas of home-range overlap and moved to the southern portion of her home range via forest galleries. In contrast, the percent of overlap between the other 4 pairs with adjacent home ranges was small. Despite distinct home ranges and occasional encounters, territorial defense was not observed among forest buffalo at Lopé NP. Similar discrete home ranges with little evidence of territorial defense have been described for Cape buffalo (Prins 1996; Sinclair 1977). The finding of exclusive forest buffalo home ranges at Lopé NP is consistent with the hypothesis that a single group of forest buffalo occupies a forest clearing (Blake 2002).

Patterns of home-range use by radiocollared buffalo suggest that forest buffalo are more sedentary than Cape buffalo, for which seasonal movement has been observed (Funston et al. 1994; Halley and Mari 2004; Halley et al. 2002; Ryan et al. 2006; Taolo 2003). Adult female forest buffalo were consistently and reliably located in the same areas throughout the 2-year study period, whereas Cape buffalo can migrate 10–40 km between seasons in Botswana (Halley et al. 2002). Cape buffalo return to the same dry-season home ranges (Halley and Mari 2004), indicating site fidelity, which I also observed for forest buffalo. The home range of 1 forest buffalo (AF1) in my study included 96% of the home range of a single radiocollared adult female buffalo tracked at Lopé NP between 1998 and 2000 (K. A. Abernethy, in litt.). This same area also was used by a group that I studied in 1996 (Molloy 1997), suggesting that occupation of an area by groups of buffalo probably extends over longer periods than the 2-year time span of my study. Home ranges of forest buffalo in Lopé NP appeared remarkably stable in both time and space.

Habitat selection at the landscape level.—Examination of my data suggests that forest buffalo select proportionally more open habitat (marsh and savanna) and use continuous forest less than would be expected based on available habitat at the landscape level (Fig. 2; Table 3). Surveys of large mammals in forests (Blake 2002; Prins and Reitsma 1989; White 1994) have found that forest buffalo are absent or present at low densities in continuous forest, which similarly suggests that buffalo avoid continuous forest. For example, in the continuous forest area adjacent to my study area, mammal surveys over two 14-month periods covering 2 annual cycles reported few observations of forest buffalo and low estimates of biomass (White 1994). During March and April, 2 graduate students repeated 1 of the White (1994) transects and found that dung of forest buffalo was absent from the continuous forest (Bankert and Schenk 2003). Low estimates of biomass or few signs of forest buffalo in continuous forest have been reported from other forest sites in Gabon and the Republic of Congo (Blake 2002; Prins and Reitsma 1989). My landscape-level results and work on forest buffalo in Central African Republic (Melletti et al. 2007) suggest that forest buffalo select open habitat for their home ranges, where their main food source, grass, is plentiful, rather than continuous forest.

Habitat selection within home ranges.—Forest buffalo had stable, savanna-dominated home ranges, but habitat use within home ranges varied with season (Fig. 3), and for at least a portion of the year buffalo preferred forest habitat during daylight hours (Table 3). Between March and August >60% of locations of forest buffalo were in forest habitats; however, forest habitat represented <50% of the area of home ranges of forest buffalo (Table 1). Continual 12-h monitoring during daylight hours of 3 of the radiocollared forest buffalo confirmed high use of forest between February and May (Bankert and Schenk 2003). In addition, buffalo dung was commonly found in forest fragments in April and May 2003; significantly less dung was present in savanna, marsh, and continuous forest during these months (Bankert and Schenk 2003). Tourists visiting the park between March and August sometimes leave with an impression that the park has no buffalo because buffalo primarily use forest habitat during the daylight hours.

Forest fragments, galleries, and corridors are used by forest buffalo more often than continuous forest and most individuals based their home ranges in a single forest fragment or gallery (i.e., 10–33% of the locations for AFI, AF2, AF3, AF4, and AF6). Adjacent home ranges of radiocollared buffalo had minimal overlap in savanna areas and no, or smaller areas of overlap, in forest fragments or galleries. In addition, buffalo remained close to the savanna areas and did not penetrate deep into the forest. Even animals with home ranges on the edge of the continuous forest remained near the forest edge. This is consistent with the low biomass of buffalo reported for continuous forest (White 1994) and the high biomass of buffalo found in forest fragments at Lopé NP (Tutin et al. 1997).

The March-August period of high forest use includes wet and dry periods with similar temperature ranges, which suggests that forest buffalo use forest habitat for reasons unrelated to rainfall or temperature alone. Small streams run throughout the landscape at Lopé NP, which eliminates the need for migration of forest buffalo during the dry season based on limited water resources. Similarly, Cape buffalo in the Serengeti show little movement within home ranges, remain near watercourses, and maintain smaller home ranges than those of sympatric migrating ungulates during the dry season (Sinclair 1977). In areas dominated by forest or bush, Cape buffalo feed exclusively on grasses and forbs and stable isotope analysis has confirmed that they are not consuming the leaves of trees or bushes (Halley and Minagawa 2005). Buffalo at Lopé NP moved into the savannas to feed on the flush of new grass when the savannas were burned at the end of the June-August dry season and were highly visible on the savanna areas between September and February. Although they dwell in forest habitat, forest buffalo depend on open habitat for food.

Marshes also were an important open habitat between September and February. Home ranges of forest buffalo were 4% marsh and 19% of locations of buffalo were in marsh habitat, despite the small percentage of marsh habitat in the study area (only 2.45% marsh). The mud wallows within marsh habitat contributed to the preference for marsh habitat, especially between September and February. After early morning feeding bouts in the savanna areas, buffalo at Lopé NP often retire to a wallow and rest in the marsh until sunset. Groups tended to frequent 1 or 2 wallows within their range and it appeared that repeated use of the same wallow enhanced water and mud conditions. On the Ishasha Plateau of Virunga National Park, Democratic Republic of Congo, African buffalo use wallows for thermoregulation and drinking (Mugangu et al. 1995) and, among temperate bovids, bison (Bison bison) wallow to gain relief from biting insects (McMillan et al. 2000).

Because my analyses were limited to data from diurnal observations of adult female buffalo, some differences may be observed for male buffalo and nocturnal observations. Males may differ in home-range size based on observations of Cape buffalo, where adult males wander more widely (Halley and Mari 2004; Prins 1996). However, the 4.94-km2 home-range area estimated for the solitary adult male monitored in a pilot study (K. A. Abernethy, in litt.) fell within the range of areas for radiocollared adult females reported in my study. In addition, adult females with radiocollars were in groups that included an adult male, which was regularly with the group (L. M. Korte, in litt.). For the majority of visual observations, the adult male was with the group, and I had no observations of the identified males from these groups in locations outside the home-range areas of the groups. For a definitive result, it would be necessary to track adult males.

Nocturnal locations of radiocollared forest buffalo would most likely not significantly change the location of the home ranges or estimated home-range area. Given the stability of the ranges over the 2 years of the study, it seems unlikely that animals would leave their ranges only at night, resulting in a location or size change. But, nocturnal observations may influence activity patterns and habitat selection within home ranges. For the majority of Cape buffalo, grazing is done during the nighttime hours (Prins 1996; Sinclair 1977; Taolo 2003), which suggests that the early morning peak of feeding observed at Lopé NP may be a continuation of nocturnal feeding. If forest buffalo at Lopé NP spend more time feeding at night, then use of savanna habitat would increase, whereas use of forest and marsh would decrease. Therefore, nocturnal observations may change conclusions about habitat preferences within home ranges, especially if forest buffalo are using the savanna areas at night during the peak of diurnal use of forest, that is, March through August. Regardless of how the overall patterns of habitat use may change with nocturnal data for activity patterns, the diurnal observations of adult female buffalo demonstrate that habitat preference changes with season based on food availability.


I thank the government of Gabon for permission to work at Lopé NP and for the permits to radiocollar animals. Station d'Etudes des Gorilles et Chimpanzés provided me with a place to live and work while collecting data, field assistants, vehicle use, and facilitated logistics on site. Wildlife Conservation Society field veterinarians, B. Karesh and M. Kock, handled the sedation of buffalo for radiocollar placement with commendable technical expertise and respect for the animals. I would also like to thank the parent organization of SEGC, the Centre International de Recherches Médicales de Franceville, for their core support for the research facilities and their acceptance of my study within their research program. I am indebted to B. Kiene Boussoughou, my field assistant, for her efforts to collect telemetry data. I thank my academic committee members, B. Lundrigan, K. Holekamp, C. Lindell, and S. Winterstein, for their comments on this manuscript and guidance throughout my dissertation. S. Ryan and 1 anonymous reviewer provided useful comments on the manuscript. This study was conducted in partial fulfillment of my Ph.D. and funding was provided by the United States Student Fulbright Program, the Wildlife Conservation Society, and the Department of Zoology and Center for African Studies at Michigan State University.


  • Associate Editor was Jane M. Waterman.

Literature Cited

View Abstract