Study area
We conducted our study in the estimated gorilla habitat (ca. 39 km²) located in the eastern north of the Ebo forest which is extended geographically between the longitudes 010 ° 02'59.2''E and 010 ° 38'30.9''E and the latitudes 04 ° 05'09.5''N and 04 ° 31'01.6''N (Morgan 2010; Mfossa et al. 2022). The area is characterized by a warm and humid tropical equatorial climate with dry and rainy seasons (Abwe 2018). The rainy season extends from March to November. The main annual rainfalls exceed 2,500 mm and the average elevation in the area is 850 m (Abwe 2018). The dominant vegetation in the study area is composed of mature forest, secondary forest, inundated forest or swampy zone, and grassland (Fig. 1, Mfossa et al. 2022). The Ebo forest harbors a rich biodiversity including plant species narrowly endemic to Ebo (Gosline et al. 2022), endangered threatened mammal species (Whytock et al. 2021), and a rich assemblage of diurnal primates including gorilla (Gorilla sp.), Nigeria-Cameroon chimpanzee (P. t. ellioti), drill (Mandrillus leucophaeus), Preuss’s red colobus (Piliocolobus preussi), and Preuss’s monkeys (Allochrocebus preussi) (Dunn et al. 2014; Abwe 2018). Common African wild predators of great apes such as cats (lion, leopard, etc.) no longer exist in Ebo. The gorilla’s habitat is surrounded by four villages depending on the forest resources for survival (Dunn et al 2014; Mfossa et al. 2022; Whytock and Morgan 2010).
Data collection
The gorilla population occurred only in the northern part of the Ebo forest with a distribution range restricted to about 22 km² (Mfossa et al. 2022). Due to the high anthropogenic pressure in the study area (Mfossa et al. 2022), we were not permitted to use a line transect method which can destroy vegetation and increase in return facility of poaching and hunting (Fedigan 2010; Zhou et al. 2013) and opted for alternative method with minimal environmental impact. To obtain a preliminary gorilla abundance index and data on nesting site characteristics, we used the recce survey method which was relatively low cost, caused minimal environmental disturbance, and allowed us to cover the entire study area while balancing the survey effort across the habitat types (Fig. 1) (Ismaila and Maloueki 2021; Kühl et al. 2008; White and Edwards 2000). Data collection was organized during trips of five to ten days every month from January 2013 to November 2017. The survey typically started at 7 am and ended at 4 pm. The surveyors, consisting in a team of three to four experimented biological data collectors, conducted random nest searches in the study area using GPS (Garmin GPSmap 60CSx) to navigate along recces (Kühl et al. 2008; Ross and Reeve 2011; White and Edwards 2000). Recces were not permanent, and the team randomly chose the survey direction. To reduce the risk of imperfect detection of gorilla nests in different habitat types, the team moved carefully and slowly (1–2 km/h), maintaining their direction regardless of the habitat type. When a gorilla nest was spotted, team members scanned the area within a 50 m radius to search for more nests (Morgan et al. 2006). We aimed to patrol the entire study area (~ 39km²) to describe environmental variables of sites where nests where present or absent. However, because recce surveys can provide biased samples as the encounter rates likely result from variation in vegetation density or topography (Kühl et al. 2008; Tagg and Willie 2013), we equalized survey efforts between habitat types proportionally to their coverage in the study area by calculating the total distance walked over each of them (Table 1). We recorded the entire track of the daily survey line using GPS tracklog option with record interval set at 1 minute and surveyed a total of 1077.5 km random tracks in 338 days (see Fig. 1 for the distribution of recce tracks). Thereafter, we extracted recce tracks in each habitat type in QGIS software (version 3.14), and controlled the survey effort per habitat type (Table 1).
Table 1
Effort survey in various habitat types within the Ebo forest
Habitat types | Coverage proportion (%) | Track distance (km) | Track Proportion (%) |
Grassland | 2.63 | 49.60 | 4.60 |
Mature forest | 87.04 | 921.97 | 86.99 |
Secondary forest | 7.90 | 58.58 | 5.53 |
Swampy area | 2.43 | 29.74 | 2.81 |
Gorilla nest abundance index
Since gorillas are sympatric to chimpanzees in this habitat (Dunn et al. 2014), we used the presence of feces, shed hair, signs of passage and other signs in nesting sites to attribute nests to either great ape species (Sanz et al. 2007). When these evidences were absent, we referred to Tutin and Fernandez’s criteria (1984) to distinguish between nests of the species by attributing (i) all sites with only ground nests to gorillas; (ii) all tree nests closely associated as nearby ground nests of the same age to gorillas; (iii) all sites with only tree nests not associated with ground nests to chimpanzees. We considered a nesting site as all nests of the same age within 50 m of one another (Morgan et al. 2006). We recorded the location of the nesting site with a GPS at the location of the identified central nest, and collected environmental variables for each nest and site. To avoid duplication of nest counts in subsequent surveys, we marked encountered nests with brightly colored flagging tapes (Abwe 2018) with date, age and other information including the group size (i.e. the number of similar-age nests within a nesting site). Finally, we calculated the Encounter Rate of nesting sites (ER) to estimate their abundance index in this habitat as follows: ER = N/Lt; where N = number of nesting sites, Lt = Total distance surveyed in km.
Nesting site characteristics: used versus available sites
To investigate the preference in nesting site selection by gorillas (i.e. if gorillas select specific nesting sites among those available in the Ebo forest), we collected environmental data from 178 nesting sites (nest present) and 703 available sites (nest absent) along recces. For both available sites and nesting sites, characteristics variables associated with environmental data included habitat types, canopy openness, horizontal visibility at the ground level, undergrowth composition, slope, and altitude.
To assess the habitat-use for nesting, we considered the four habitat types categories (i.e., mature forest, secondary forest, grassland and swampy area or inundated forest) identified by Mfossa et al. (2022). The altitude of the available sites and nesting sites was recorded at the estimated central location while taking the GPS coordinates of the site. Since the altitude records of the nest sites felt between 800 and 1300 m, we classified nests into five 100 m-altitude classes according to their value (Alt): Alt ≤ 900 m, 900 m < Alt ≤ 1000 m, 1000 m < Alt ≤ 1100 m, 1100 m < Alt ≤ 1200 m, and Alt > 1200 m. The description of other variables and their attributes is given in Table 2.
Table 2
Description of environmental variables used to investigate the preference in nesting site selection by Ebo gorillas
Environmental variables | Attributes | Description |
Canopy: the percentage of light available above the site | Very open (Voc) | No or little few tree branches or foliage above the site (76–100%) |
Open (Oc) | Site partially covered with few foliage or branches (51–75%) |
Closed (Cc) | Foliage and branches relatively covered the site (26–50%) |
Very closed (VCc) | Site totally covered with very few lights reaching the soil (0–25%) |
Undergrowth: the vegetative composition under trees | Ligneous (Li) | Exclusively with sapling and/or liana |
Herbaceous (He) | Exclusively with herbs |
Herbaceous/Ligneous (He/Li) | Combination dominated by herbs |
Ligneous/Herbaceous (Li/He) | Combination dominated by ligneous |
Visibility: the distance at eye level (1.7 m) beyond which objects can no longer be seena | Very open (VOv) | Visibility more than 15 m |
Open (Ov) | Visibility between 11 and 15 m |
Closed (Cv) | Visibility between 5 and 10 m |
Very closed (VCv) | Visibility less than 5 m |
Slope: the inclination of the terrain (angle) from team members eyes appreciation | Flat (Fl) | Angle ≤ 5% |
Gentle (Ge) | 5%< angle ≤ 15% |
Moderate (Mo) | 15%< angle ≤ 25% |
Steep (St) | Angle > 25% |
Source: aDupain et al. 2004 |
Characteristics of individual nests
To investigate the nesting behavior of gorillas, we recorded for each individual nest the size (diameter), age, height from the ground, nest type and nest materials.
We measured nest diameters at the lateral size of the nest considered as the size of the gorilla back at the seating position. We used the nest group size (i.e., the number of nests of similar-age age within a nesting site) as a proxy for the gorilla’s group size (excluding infants) in this population.
Following Tutin and Fernandez (1984) and Kühl et al. (2008), we categorized nest age into five classes : (1) fresh nest (a construction less than 5 days, all leaves green and fresh feces/urine under, in or close to nest); (2) recent nest (a construction between 5 and 15 days, leaves green mixed with few turning brown, no fresh feces/urine); (3) old nest (a construction between 15 days and 6 months remains intact, leaves mixed in color as brown/dried and green); (4) very old nest (a construction between 6 months and 1 year, leaves completely brown but nest still complete) and (5) rotten (a construction more than a year, leaves are gone and only the skeletal branch and twig structure remain).
We categorized nests into five construction types according to Tutin et al. (1995): (1) nests with no vegetative construction (zero material or bare ground nests), (2) nests exclusively built with herbs (herbaceous nests), (3) nests with both herbs and woody materials (mixed nests), (4) nests exclusively built with woody materials (woody nests), and (5) nests built in trees (tree nests or arboreal nests).
To assess key plant species used by Ebo gorillas to construct their nests, we enumerated and identified all the components used to build each nest. We identified plant materials used to construct the nests to the species or genus level whenever possible. When plants could not be identified in the field, we collected and coded specimens for later identification at the National Herbarium, Yaoundé.
Seasonal variation in nesting
To investigate nesting seasonality, we assessed habitat type used for nesting and nest types according to the seasons (March to November for the rainy season and December to February for the dry season). We used the nest age estimate to identify the month when the nest was constructed.
Nesting site reuse
Based on GPS coordinates, we considered all nesting sites close one to another inside 50 m radius as reused nesting site (i.e., a geographical site reused for nesting at different time by a group of gorillas). The difference in age and the number of nests in the group allowed us to separate different group nests in the site especially when the nests were fresh or recent.
Data analysis
We used R version 4.1.1 (R Development Core Team, 2021) for statistical analyses (α = 0.05).
Nesting site characteristics: used versus available sites
To test the influence of the site characteristics (set as explanatory variables) on the probability that a given site was selected by gorillas for nesting (set as the binary response variable), we ran binomial family generalized linear models (glm) with a logit link function. Logistic regression models were chosen due to their sensitivity to binary data in comparison to other model families (Lewis 2004). We ran two distinct models: 1) an ordinal model including ordinal categorical variables (i.e., canopy openness, horizontal visibility, slope, and altitude) with the reference intercept set as the lowest value category for each variable, and 2) a nominal model including non-ordinal categorical variables (i.e., habitat type and undergrowth composition) with the reference intercept set as the category with the highest sample size. We tested seasonal interactions between the variables in the ordinal and nominal models. However, models poorly predicted outcomes and variable significance were overly sensitive indicating Type I error. Therefore, interactive models were not taken into account within the analyses. To determine the best-fit model, we performed automated model selection with the dredge function from the R package MuMIn (Burnham and Anderson 2002) using the ΔAkaike Information Criterion (ΔAIC) < 2 values and the model weight (AICWt). To test for the effects of multicollinearity, we performed Variable Inflation Factor (VIF) checks on the global model using the R package car and vif function (Fox and Weisberg 2018), with VIF > 10 indicating that the multicollinear variables may be highly correlated and therefore significantly influencing the coefficients in model output (Hair et al. 2010; Fox 2016). Thereafter, we used the R package DHARMa (Hartig and Hartig 2017) to test for residual patterns of over- and under-dispersion and mis-specification problems in the best-fit models. Finally, for plotting the best-fit models, we ran the R package margins (Greene 2012) to calculate the Average Marginal Effects (AME). Marginal Effects use model predictions for interpretations and can be used as a way of presenting results as a difference in probabilities by centering the scale of all the covariates (Perraillon 2013).
Characteristics of individual nests
To identify the most used plant species for gorilla nesting, we calculated the frequency rate (i.e., number of nests carrying the species divided by the total number of nests) and the species occurrence value (i.e., number of nests with the species divided by the total number of species) of each plant species (Willie et al. 2014, Setiawan et al. 2021). We used the chi-square test to assess the nesting seasonality for gorillas in accordance with nest types and habitat use.