Wild meat hunting and use by sedentarised Baka Pygmies in southeastern Cameroon

Study area

This study was conducted in 10 Baka villages, located along the Djoum-Mintom road south of the Dja Faunal Reserve (DFR) and bordering the Dja Biosphere Reserve in southeast Cameroon (Fig. 1). Bemba II is the village nearest to the DFR and Belle-Ville is the village furthest away (shortest bee-line distances of around 17 km and 30 km, respectively). From population censuses conducted by us in all ten villages (Table S1), we counted a total of 237 dwellings (largely poto-poto houses but also mungulu huts), of which 172 (72.57%) were lived in during the study. There was a mean (±SD) of 4.33 ± 2.77 (range 1–17) persons (men, women and children of differing ages) living in the occupied dwellings. Vacant dwellings belonged to families that were in forest at the time of the study. Total village population sizes ranged from 25 persons in Meyos-Mintom to 111 in Akom. Villages inhabited by Bantu-speaking farmers were also found on the Djoum-Mintom road. The relationship between farmers and Pygmies in the study region is similar to that reported for other rainforest areas of central Africa (see Rupp, 2003; Van de Sandt, 1999; Matsuura, 2011; Oishi, 2012). These studies have shown that farmers and Pygmy hunter-gatherers share a mutually dependent relationship with regard to lifestyle and culture but have ambivalent attitudes about one another characterized by both discrimination and respect (Hattori, 2014).

Agriculture alongside the harvest and trade of non-timber forest products are practised by Baka and Bantu-speaking farmers but to different degrees. Baka have gradually become sedentarised since the 1950s (Althabe, 1965; Kitanishi, 2003; Leclerc, 2012). After relocation from the forest, Baka started growing crops such as plantain, banana, and cassava. Subsistence farming has increased in recent years in the studied villages, particularly as a result of agricultural programmes initiated by our partner Zerca y Lejos (ZyL) (Zercay Le jos, 2020a; Zercay Le jos, 2020b), a Spanish NGO working on development and health support to Baka communities in the region. The result of this has been that fewer Baka work directly for the Bantu-speaking farmers on “petit jobs” such as clearing forest for planting, planting and harvesting crops. Although we did not observe any major conflicts as a result of these changes, in western Cameroon Van de Sandt (1999) reported that community divisions between Bagyeli and Bantu-speaking farmers were exacerbated when the former group adopted farming, leading to the seizure of Bagyeli land by farmers. Baka, however, are socio-economically more advantaged, compared to other Pygmy groups such as the Bagyeli (Kitanishi, 2003).

The terrain of the region is sloping with gently rolling hills. Altitude varies from 250 to 800 m and averages 600 or 650 m. The climate of the region is characterized by a four-season equatorial climate. A major dry season is from December to March, a minor rainy season from March to June, a minor dry season in August, and the major rainy season from September to December. Rainfall recorded for Djoum averages 1,500–2,000 mm per year, and some precipitation is common even during the dry seasons (World Weather Online, 2020). Average temperature remains fairly steady year-round, averaging 25 °C, fluctuating slightly with the seasons. The long, rainy season is the coldest time of year, and the long dry season the warmest. Humidity is high and does not vary throughout the year.

The major vegetation type is a mixture of evergreen and semi-deciduous forests (Letouzey, 1985). The forest is degraded alongside the dirt roads and main road, due to pressure from housing and agriculture. Logging operations in the region started in the 1970s as far of the extension of operations into the last remote tracts of intact forest in the east and south of the country (Global Forest Watch, 2000). Logging roads has enabled the connection of villages to market areas and to the main towns of Mintom and Djoum (Fig. 1).

Hunts take place at a distance from the village (Fa et al., unpublished data, 2017–2018). Women also hunt but this is much more localized, around their farms. Baka farms are relatively small patches of land dedicated largely to the cultivation of subsistence crops such as cassava, plantain and bananas. Hunting territories extended up to the Dja River, but never overlapped with the DFR (Fa et al., unpublished data, 2017–2018).

Data collection

In following the principle of free, prior and informed consent (FPIC), allowing the studied communities to give or withhold consent to our project, we (E.A.M., G.R.B., R.O.) first organized several meetings with each village. During these meetings we asked the participants to highlight the main challenges faced in the daily lives; they identified key problems around agriculture and hunting. To understand the issues affecting hunting, we recruited hunters in each village to volunteer to self-report their daily hunting activities over a period of 5 months.

All hunters freely participated in our project without remuneration, gave verbal consent to do so, and could stop contributing to the project if they so wished. Hunters were informed that their identity would be kept anonymous and all information provided would be treated confidentially. No informant under the age of 18 (minors according to Cameroon law) were involved in our study. Importantly, we emphasised that our project was not focusing on the illegality of hunting (i.e., poaching) but to understand hunting as a subsistence activity. In each village, we recruited a reporter to oversee data collection from participating hunters. These village reporters (VR) were given a mobile telephone to allow them to communicate with the field team when needed and were paid 25,000 CFA (around 40USD) per month.

VR were all Baka schoolteachers (hence literate) who worked for ZyL’s education programme (Zercay Le jos, 2020b); they were full-time residents in each village. All VR were trained by R.O. in data collection methods in Fang; the Fang language being the lingua franca used between Baka and Bantu-speaking farmers in our study area. The VR visited every participating hunter’s household at the end of each day and hunters were asked to provide information for each returned carcass: 1) species, 2) hunting method used, and 4) whether the quarry was to be eaten at home, the complete carcass sold, or parts of the animal consumed and other parts sold. The unit of reporting was each hunting trip completed at the day of recall (rather than a 24 h recall), as some trips lasted for two or more days. The VR also noted the start and end of each hunting day by asking the hunter the time he/she left and returned to the house (if they possessed watches) or the approximate time taken as the mid-point in one of six pre-determined periods of the day (Grand matin—05.00–06.00, Matin—07.00–11.00, Midi—11.30–12.00, Après midi –12.00–15.00, Soir—15.00–20.00, Nuit—20.00 onwards). No-hunt days were also recorded. All VR communicated with hunters in Baka and information was recorded onto pre-prepared data sheets.

Species names were given in Baka by the hunters and scientific names were assigned according Kingdon et al. (2013) for mammals, Borrow & Demey (2008) for birds and Chirio & LeBreton (2007) for reptiles. Species hunted were classified according to their threatened status in the IUCN (2020), and according to the three protection classes (A, B, C) within the Cameroon Forestry & Wildlife Law (Republic of Cameroon, 1994).

Checks to ensure the correct application of methods, and to collect completed forms, were undertaken weekly by one of our field team members (R.O.), with supervisory visits by R.O.B. and E.A.M. All three team members were permanently based in the study area and were headquartered in Djoum.

Hunting practices were recorded every day by the VR in the 10 study villages over a complete period of five months each. Sampling was organised into two phases: a first Phase during Mar. to Jul. 2018 in which Bemba II, Abing-Nkolemboula, Doum, Assok and Belle-Ville were covered, and a Phase 2 (Oct. 2018 and Feb. 2019) in which the other five villages (Akonetyé, Adjab-Mintom, Meyos-Mintom, Odoumou, Akom) were visited. We selected villages to sample in each Phase so as to achieve a broad geographical spread of settlements i.e., not clustered, in each of the two phases.

Ethical approval was not required in this study, although it meets the guidelines of the Social Research Association (2003). Permission to undertake field work in our study area was granted by the Ministry of Scientific Research and Innovation (MINRESI), via the Center for International Forestry Research (CIFOR) in Cameroon. Authorisation to work with human subjects was covered by the Arrete No. 00034/A/MINATD/DAP/SDLP granted by the Ministere de L’Administration Territoriale et de La Decentralisation of the Government of Cameroon to ZyL.

Statistical analyses

We tested skewedness of the data using Spearman’s rank correlation tests and the Jarque–Bera normtest (Jarque & Bera, 1987; Gavrilov & Pusev, 2015).

To evaluate the variables impacting the proportion of hunting trips that failed to produce any carcasses we used a general linear model, GLM, with a binomial error distribution. Variables were: the field effort deployed by each hunter in terms of hunting effort in hours, the hunter’s overall experience as estimated by the total hunting effort and hunting return of carcasses over all hunting trips, and the village characteristics as estimated by the total hunting return of carcasses weights in kg by all village hunters, the average return of carcass weights by each village hunter, the number of carcasses returned by all village hunters and the percentages of hunting with rifles, traps and dogs. Carcass weights were estimated from published mean weights of mammals (Kingdon et al., 2013) and birds (Dunning, 2007), and reptiles (Chirio & LeBreton, 2007).

To evaluate the diversity of hunted animals by village, we applied the following diversity metrics (Magurran, 1988; for further details see Table S2): (1) Species richness; (2) Dominance; (3) Simpson’s index; (4) Shannon–Wiener index; (5) Evenness and (6) Chao 1. For all diversity metrics, we calculated the 95% upper and lower confidence intervals by 10,000 bootstraps, each with the same number of individuals as in each original sample. For generating bootstraps, the random samples were taken from the total of all columns, and the taxon was chosen with probabilities according to the original pooled abundances.

To compare animal biomass extracted in each study village with other hunting studies in the Congo Basin in Fa et al. (2016), we calculated: (1) extraction rate, as the mean number of kilograms of wild meat extracted by a hunter (H) in a year by multiplying the average amount of wild meat (kg) taken per hunter (kg H⁻¹ yr⁻¹) by the number of occupied households (since there was a hunter in each household) in the village; (2) harvest rate, as the total number of kilograms of wild meat extracted per inhabitant in each village per annum (kg P⁻¹ yr⁻¹) calculated from 1) divided by the total number of inhabitants in each village. This index measures per capita yield of game animals for the average person (P) in a community in one year. To calculate the harvest rate, we assumed, based on our field observations, that there is one hunter per household. Data on numbers of households and numbers of inhabitants per village stem from anthropological surveys (see above).

To better visualize the amount of meat extracted per village, we translated the weight of wild meat estimated for each into cattle meat equivalents, as used by Nasi, Taber & Van Vliet (2011) to highlight the level of substitution needed if wild meat was to be replaced by domestic meats. We choose adult UK dairy cattle, which weigh on average 617 kg, only slightly more than Argentinian, Canadian, Danish, Dutch, German and Swiss dairy cattle (600 kg), but less than French (650 kg), US (680 kg) but more than New Zealand (458 kg) and the assumed ‘standard average’ for all breeds across 10 European countries (500 kg) (Schubert et al., 2019).

Statistical analyses were conducted using the R statistical environment (R Foundation for Statistical Computing, 2016) and Past 3.0 (for diversity analyses; Hammer, 2019), with all tests being two-tailed. Alpha level was set at P = 0.05.

Hunting effort and hunting returns

Excluding 68 hunting trips with no information on hunting returns, we obtained data on 1,946 hunting trips by 121 hunters (118 men, 3 women) from the ten study villages (raw data in Table S3). Assuming that there was at least one male hunter in the 172 occupied houses (see Table S1), we then sampled 77% (N = 133) of hunters in the studied villages.

Hunting effort, expressed as the duration of hunting trips, varied between one and 180 h and was skewed with a mean of 13.17 h and a median of eight hours (Fig. 2A). For each hunting trip, a mean ± SD of 1.15 ± 1.11 carcasses (median 1, range 0–9) were returned (Fig. 2B). Total carcass weight returned from trips was skewed with a mean of 15.0 kg, median of 6.3 kg and range of 0.3 to 148.3 kg (Fig. 2C). Hunting effort and number of returned carcasses were positively correlated indicating both, a significant linear relationship and a large variance with only 10.8% of the variance explained (Spearman’s r = 0.33, P < 0.0001, N = 1, 946). The variance occurred between trips of each hunter and between hunters, as indicated by the larger proportion of variance explained (67.3%) by the correlation between total hunting effort and returned carcasses for all hunters (Spearman’s r = 0.82, P < 0.0001, N = 121). The average hunting effort versus the average number of returned carcasses across all hunters in the ten villages was positively correlated, explaining 73% of the total variance (Spearman’s r = 0.85, P < 0.004, N = 10). The average number of returned carcasses and the average time spent to hunt each carcass was negatively correlated across villages explaining 48.6% of the total variance (Spearman’s r = 0.70, P < 0.031, N = 10). In other words, the higher the number of animals successfully hunted the smaller the proportional time required hunting each animal.

About a quarter of trips (23.4%) failed to return a carcass. As expected, failure was negatively correlated with hunting effort (GLM coefficient = − 0.015, Standard error = ±0.005, P < 0.0001). Failure was also significantly negatively correlated with the hunter’s total return of carcasses (−0.030 ± 0.005, P < 0.0001) and significantly positively correlated with the total hunting effort (0.003 ± 0.0005, P < 0.0001). At a village level, failure was significantly negatively correlated with average village return weight (−0.191 ± 0.025, P < 0.0001), average village return rate (−0.871 ± 0.305, P = 0.004) and the proportion of dogs used in hunts (−0.135 ± 0.020, P < 0.0001), but significantly positively correlated with average hunter’s return weight (0.009 ± 0.001, P < 0.0001) and percentage gun hunting (0.022 ± 0.010, P = 0.030). The parameter of percentage snare hunting was not significant (0.003 ± 0.010, P = 0.74).

The distributions of total and average hunting effort, the average number of returned carcasses and the average effort required to successfully hunt one animal were all highly skewed (Fig. 3). The skew of the visually least skewed distribution of the average of returned carcasses per hunter was significant (Jarque–Bera normtest, T = 0.91, P < 0.0001).

Average return over hunting trips and hunters per village ranged from 0.61 to 2.06 (mean ± SD 1.15 ± 0.43) and was not skewed (Jarque–Bera normtest, T = 0.89, P = 0.12). Hunting effort per hunter was skewed (T = 1.49, P = 0.014) and ranged from 125.6 to 357.1 h (mean 213.7 h, median 180.12 h). The average time required to hunt one carcass was also skewed (T = 1.49, P = 0.014) and ranged from 6.6 to 26.9 h (mean 12.5 h, median 10.7 h). The average return was significantly correlated with the average time spent on hunting trips by hunters (Spearman’s r = 0.85, R² = 73%, P < 0.004, N = 10). Average return per hunter and average time per carcass were significantly negatively correlated (Spearman’s r =  − 0.70, R² = 48.6%, P < 0.031, N = 10).

Species composition

As many as 49 separate species (Mammals = 37; Birds = 6; Reptiles = 6) of 24 families (Mammals = 16, Reptiles = 5, Birds = 3) were hunted in the study; 2,245 carcasses in total, out of which 68 carcasses (3%) were not identified to species (Full species list, Table S3). No songbirds were hunted, but large birds, hornbills and guinea fowls, were taken. Seven species of the mongoose family, one cat, one mustelid and one palm civet were noted in the study. Amongst the primates, eight monkeys, one ape and a prosimian were hunted. As many as nine bovids, one member of the pig family and one Tragulid represented the hunted ungulates. Only three rodents and one unidentified squirrel were recorded.

By village, mammals contributed between 80% and 100% of all carcasses, but more reptiles were hunted than birds (Fig. 4A). In only two villages, reptiles and birds constituted about 20% of returns, in all others less than 10%. Amongst mammals, ungulates contributed the highest proportion of mammalian returns (χ² test on the data frequencies, P < 0.001), followed by rodents with two exceptions where ungulates and rodents were equal (Fig. 4B). In all villages, more ungulates were hunted than rodents (ungulate: rodent ratio, 2.2 ± 1.4), ranging from 0.98 in Assok to 5.74 in Akonetyé. Hunted ungulates ranged between 30% and 55% with one notable exception of about 80% in one village (Akonetyé). Spearman’s rank correlations of mammal orders, birds and reptiles versus hunting methods (rifle, traps, dogs and others) were significant in only two cases, namely for primates for rifle hunting (Spearman’s r = 0.65, P = 0.049) and snare hunting (Spearman’s r =  − 0.73, P = 0.02).

Species richness and diversity indices

The summary of the diversity metrics calculated for each community of hunted animals by village is presented in Table S5. Scrutiny of the diversity, evenness and dominance indices (and of their confidence intervals) at species groups and species clearly revealed that the hunted animals’ community characteristics were similar among villages, with however one exception: the village of Meyos-Mintom. In this village, the species diversity and evenness were higher, and the dominance index was lesser, than in all other villages. This result is fully confirmed by a diversity profile analysis (Figs. 5A, 5B), with the curve for Meyos-Mintom standing alone and significantly higher (P < 0.01 at ANCOVA) compared to the curves of all other villages. Chao-1 estimates (Table S5) showed that the number of hunted species recorded during our study was significantly lower than that expected for most of the surveyed villages (i.e., Akom, Akonetyé, Assok, Belle-Ville, Doum, Meyos-Mintom and Oudoumou), and a saturation curve analysis (Fig. 5B) established that in all villages a plateau phase in the number of hunted species was not reached. Even by this analysis, Meyos-Mintom appeared the village where the plateau phase was the furthest from being reached, and therefore where our sampling appeared the least accurate (given that the diversity of species is highest there).

Biomass extracted

A total of 53 carcass types, i.e., distinct identified or unidentified species, were reported by hunters. Of these, 44 could be determined to the species level, two only to the genus level (one duiker and one pangolin) and two to family level (Herpestidae and Phasianidae, respectively) (Table S3). Five reported carcass types (a “bird”, “monkey”, “squirrel”, “snake” and “animal”, respectively) could not be identified because the carcasses were sold prior to the VR’s visit. The mean weight of the 46 species could be estimated according to the literature, representing 95% of all carcass types, whereby the weights of the two possible Phasianidae species and the two possible pangolin species were averaged since the species could not distinguished separately (Table S3).

The accumulated total weight of the 46 species with weight estimates was 20,609 kg across the ten villages. The average return per hunter varied between 52.3 kg and 389.1 kg among villages (mean ± SD = 183.7 ± 115.6 kg, skew not significant with T = 0.369, P = 0.48). By species, just over half of the animal biomass extracted by all hunters in the studied villages was provided by four mammal species: the blue duiker Philantomba monticola (17.99%), brush-tailed porcupine Atherurus africanus (15.10%), bay duiker Cephalophus dorsalis (10.07%), and Emin’s pouched rat Cricetomys emini (7.93%).

Estimates of mean ± SD harvest and extraction rates per annum (from the average amounts per village per sample month multiplied by 1) for all villages pooled were 124.0 ± 118.1 kg P⁻¹ yr⁻¹ (median = 69.4) and 440.8  ± 277.4 kg H⁻¹ yr⁻¹ (median = 380.5), respectively. By village, a mean ± SD of 7,569.7 ± 6,103.4 kg yr⁻¹ of wild meat were extracted per annum, ranging from around 2,080.8 kg in Belle-Ville to 19,351.4 kg in Akom (Table 1). Mean body mass of animals hunted in all villages was 10.1 ± 4.0 kg, lowest (6.5 kg) in Abing-Nkolemboula and highest (16.9 kg) in Akonetyé (Table 1).

Hunting methods

Most animals were trapped (65.77% ± 16.63) using snares; these consist of a noose anchored to the ground, made usually of wire or a strong string. A lower proportion of animals were killed with single-barrelled shotguns (22.56% ± 17.72) but other methods (8.69% ± 6.96) such as hand collection (for tortoises in particular) and dogs (2.96% ± 4.49) were employed. By village, the highest percentage of animals trapped was in Akonetyé (99.4%), but highest percentage of animals shot was in Odoumou (50.9%). The highest percentage of animals hunted with dogs was in Bemba (37%), followed by Akom (34%); in the other villages only between 0–6%.

Calculated efficiency of the different hunting methods in terms of animals killed per hour hunting, taken from recorded hunting trip duration, shows that the average returns for trapping was considerably more efficient (15.15 ± 27.80 animals/hr) than other methods (4.80 ± 7.34 animals/hr), dogs (0.19 ± 0.36 animals/hr) or hunting with guns (0.10 ± 0.19 animals/hr). Differences in hunting returns by method varied between villages, with gun hunting being more efficient in Doum, trapping returns and hunting with dogs highest in Akom, and other methods in Nkolemboula (Table 2). Trapping returns correlated with gun hunting (Pearson’s r = 0.91), dogs (r = 0.64) and other methods (r = 0.13).

Destination of carcasses

Most animals hunted in all villages were consumed by the hunter and his family (48.07% ± 17.58) The mean number of carcasses sold in all villages was 32.73% ± 12.55, followed by a lower percentage of carcasses partially sold and consumed (19.21% ±17.02). The highest proportion of animals sold in a village was in Assok (50.5%) and lowest in Belle-Ville (14.3%). Most carcasses were sold to neighbours (41.9%) and resellers (31.3%), the rest to passersby. Animals consumed by the hunter were lowest in Akonetyé (6.9%) and highest in Bemba (69.0%).

Threatened and legal status of hunted species

Figure 6 presents the IUCN Red List categories and the Cameroonian legal status of hunted species. Between 60% and 80% of carcasses belonged to the “least concern” category, followed by “near threatened”, “vulnerable” and, rarely “endangered”. The “near threatened” category was represented between 5% and 20% with the exception of Akonetyé, where over 30% of carcasses belonged to this threat category. “Endangered” species were represented only in Odoumou where 1% of carcasses belonged to this group. The only endangered species was the central chimpanzee (Pan troglodytes). The most frequent Cameroonian legal status was “Class C”, followed by “Class B” with the exception of Akonetyé where more carcasses belonged to “Class B” than “Class C”. Meyos-Mintom and Odoumou had the highest proportions of “Class A” carcasses with 17% and 10%, respectively. All other villages had a frequency of 0% to 5% of “Class A” animals.