Bushmeat biogeochemistry: hunting tropical mammals alters ecosystem phosphorus budgets

Abstract

Wild meat (or ‘bushmeat’) hunting is nearly ubiquitous across the tropics and is very often unsustainable—driving declines and extirpation of numerous mammal populations. Loss of these animals can alter the transport of nutrients within and between ecosystems. But whether the physical removal of vertebrate carcasses and the nutrients that they store can reduce overall nutrient availability in ecosystems has been little explored. At 32 sites on three continents, we show that annual phosphorus (P) loss via mammal exploitation was low relative to the rate of atmospheric P deposition. But at four sites in Africa and Southeast Asia, removal of P in the skeletons of hunted mammals exceeded the atmospheric input of this nutrient by 10-fold or more. Because P is the growth-limiting nutrient for many tropical terrestrial ecosystems and certain large mammals, the imbalance created by the removal of mammal biomass under very high hunting scenarios could reduce ecosystem carrying capacity if no compensatory P additions occur in the system. This biogeochemical perspective on bushmeat exploitation raises further concerns about harvest sustainability and human food security in areas where hunting rates are high and ecosystem P inputs low.

1. Introduction

The harvest of wild mammals, birds and reptiles for human food, known collectively as wild meat or bushmeat, constitutes a conservation crisis across the tropics. Unsustainable hunting is ubiquitous in many developing countries and has led to widespread extirpation of vertebrates [1–3]. Indeed ‘defaunation’, or the loss of animals from natural ecosystems, is now recognized as a leading threat to global biodiversity [4,5]. Beyond the population- and community-level impacts of overhunting, recent research suggests that defaunation can also affect ecosystem processes. For example, following overhunting of seed-dispersing tropical vertebrates, tree communities can shift towards smaller species or those with lower wood density, thereby reducing carbon storage [6–9].

Vertebrates play important roles in ecosystem biogeochemical cycling under some conditions [10]. Many major nutrient fluxes are mediated by microbial and geological processes [11], but consumption of plants by large herbivores and the defecation of the nutrients elsewhere can be one of the major ways by which nutrients such as phosphorus (P) are transported within, and even between, ecosystems [12]. For example, the extinction of large-bodied consumers in the Pleistocene may have reduced the local transport of P in Amazonia by 98% and limited crop fertility in new agricultural areas of Africa and Asia at the dawn of human civilization [13,14]. The presence of the largest terrestrial herbivores ever, the sauropod dinosaurs, on landscapes around the world is estimated to have increased P transport by 22% and concentrations of available P by 136% [15].

Aside from the reducing the transport of nutrients by living animals, however, unsustainable hunting could potentially constitute an actual loss of P from the ecosystem via the physical removal of carcasses [16], though such effects have been little explored. While the biomass of plants is much higher than that of animals in terrestrial systems, P is a key macronutrient for animals, and investment in phosphorus-rich skeletons elevates tissue P concentrations of mammals to be approximately 10⁴× higher than in plants [11]. If intensive harvests of mammals were to deplete P from portions of the landscape, this would constitute a hitherto unappreciated flux in ecosystem P budgets. Such a flux would be particularly important for biogeochemical cycles in the tropics, where soil P is generally scarce because the rocks are very old and were not subject to Pleistocene glacial scouring [17].

Here we ask whether widespread bushmeat hunting in tropical ecosystems has the potential to drive important effects on local nutrient budgets through the direct removal of the P stored in mammal skeletons. To interpret the ecosystem significance of P losses via harvest of mammals, we compared loss rates to natural P inputs via wet and dry atmospheric deposition [18]. Our goal was not to place the impacts of hunting into the entire P cycle, for example by assessing levels of ‘natural’ ecosystem export as well as input, but simply to measure P loss rates from hunting against P input rates from atmospheric deposition. Assuming even an approximately steady state in tropical forests, where natural input and output levels of P are roughly equal, then additional losses of P through hunting that are comparable in magnitude to input rates could be an important biogeochemical and conservation issue. Natural P inputs through other sources, such as weathering of rocks, is relatively minimal in most tropical forests [11]. P input from river transport can be higher [19], but mainly affects only narrow riparian zones. Anthropogenic P inputs are increasing in ecosystems around the world, so hunting-induced P loss would only be expected to be biogeochemically important in areas distant from fertilization or agricultural runoff.

We identified published studies from 1978 to 2014, presenting information from 36 sites spanning tropical Asia, Africa and Latin America (electronic supplementary material, table S1; figure 1), where mammal extraction rates could be quantified in kg km⁻² yr⁻¹. We calculated the P export that this hunting represented using general allometric relationships between body mass and skeletal mass of mammals [20], thereby accounting for disproportionate increases in skeletal investment with the size of harvested species (equation (2.1)). Many of the studies that we compiled were from the anthropological literature, and thus the study sites where chosen independently of hunting intensity. Nevertheless, we brought together relatively few studies, and site selection in studies that were focused on hunting may have been intended to document high extraction rates. Therefore, our hunting estimates in some areas could be higher than what would be representative of the respective regions.

Figure 1.

Map of global atmospheric deposition of phosphorus (P) [18] from 0.0 (black) to 100 (white) kg km⁻² yr⁻¹. The ratio of hunting-induced P loss to atmospheric P deposition shown across sites with annual vertebrate biomass removal data. Note that the point off the coast of West Africa is on Bioko Island, Equatorial Guinea.

2. Material and methods

We mined the literature for wild meat harvest rates using Web of Science search terms ‘ts = ((bushmeat or wild meat) and (harvest))’, published books, and articles referenced therein. We filtered the findings to only include studies where the authors directly presented harvest rates in units of mass, area and time (i.e. kg km⁻² yr⁻¹), or where we could calculate aggregate harvest rates in those units based on the information provided (electronic supplementary material, table S1).

We estimated the amount of skeletal mass represented by each of the available mammal harvest rates. The proportion (s) of an individual animal's live body mass (m; in kg) made up of skeleton was calculated using the following allometric equation, derived from mammals ranging in size from mice to elephants [20]:

$$s=\hspace{0.17em}\frac{0.61\hspace{0.17em}(m^{0.09})}{m}.$$ 2.1

We calculated the amount of P that this removal of animal skeletons represented by assuming that bone is made up of 12.3% P, the fraction of P by dry mass in the National Institute of Standards and Technology bone meal standard (https://www-s.nist.gov/srmors/view_detail.cfm?srm=1486; accessed August 2017). Of the wild meat hunting studies with usable harvest rate data, 16 of 36 (44.4%) sites had species-specific harvest rates, enabling us to use per capita body mass reported for each species in the PanTheria database [21] to estimate its average skeletal mass based on equation (2.1). For the remaining studies that lacked species-level harvest data, we approximated the aggregate proportion of mass as skeletal material by using the average from the 16 sites with species-specific harvest data. Because of the enormous logistical and cultural challenges involved in measuring wild meat harvest rates, there are no spatial or temporal data available to quantify uncertainty in these statistics. We consider the harvest rates and skeletal P content used in this study to be reliable but imperfect approximations that represent the best available information for analysing the potential biogeochemical implications of wild meat harvests. We did not account for P in soft tissues, such as nucleic acids; these organic compounds represent a small but non-trivial proportion of total body P in vertebrates [11], and soft tissue P is unequivocally lost from the local ecosystem when animals are removed and consumed by humans. The reported wild meat harvests were variable in their inclusion of small mammals, birds and herpetofauna. In some cases our estimates even of large mammal offtake rates are biased low due to opportunistic sampling or sampling of only a subset of the taxa at a given site (e.g. [22,23]).

Because P lacks a gaseous phase, long-distance transport of P occurs primarily via small particles. We used the results of a recent, second-generation global model of particulate P deposition rates [18] to estimate the input of ‘new’ P to the terrestrial ecosystem at each hunting site. This model offers estimated P deposition for both small (less than 10 µm) particles, which can be transported long distances, and large (10–100 µm), particles such as pollen, that tend to only cycle locally (i.e. within the same grid cell of the model). Because our focus is on the input–output balance of terrestrial ecosystems rather than internal cycling rates within the same ecosystem, we compared wild meat harvest fluxes to inputs of small particles (less than 10 µm). As context, deposition of P in the form of larger, locally sourced particles is estimated to be approximately one order of magnitude higher [18].

To help visualize the relative magnitudes of natural P inputs and hunting-induced P losses, we calculated how many individual mammals would have to be added to an ecosystem to provide the equivalent amounts of P delivered by atmospheric deposition of fine particles. Large animals have disproportionately larger skeletons than smaller animals do, so we calculated P ‘mammal equivalents’ in units of a large hunted species, the Asian elephant (Elephas maximus), and a small hunted species, the African giant pouched rat (Cricetomys gambianus). Body size information for these species was gleaned from the PanTheria database [21]: 3269.8 kg for the Asian elephant and 1.28 kg for the giant rat. Skeletal mass was calculated as 12.6% of live weight for elephants and 6.2% for pouched rats based on the allometric relationship in equation (2.1).

3. Results

The sites that we assessed differed substantially in atmospheric P deposition rates. Such rates are highly heterogeneous around the world [18], and most tropical areas receive relatively low P inputs (figure 1). In Thailand, for example, average annual P deposition is 2.3–4 kg km⁻², which is the P equivalent of adding approximately 235–410 giant rats (Cricetomys gambianus) or 0.04–0.07 Asian elephants (Elephas maximus) per square kilometre per year. By comparison, parts of Nigeria receive aeolian dust inputs from the Sahara that provide up to 80.2 kg P km⁻² annually, the equivalent of 8170 giant rats or 1.58 Asian elephants km⁻² yr⁻¹. The Thailand example is equivalent to a plausible level of wild meat harvest, while the Nigeria one is highly unlikely.

For 86% of our tropical sites, annual atmospheric P deposition exceeds (often greatly) the hunting-mediated removal of P (figure 1; electronic supplementary material, table S1). But in two regions and time periods (Southeast Asia in the 1990s and East Africa in the 1970s), P removal in the form of hunted animals—assuming that the entire animal carcasses were removed—exceeded atmospheric inputs by factors of 11.0–25.0 (Indonesia) and 1.7–10.4 (Kenya). At all of these sites, atmospheric P deposition was below average and mammal biomass removal was relatively high (electronic supplementary material, table S1 and figure S1). Across the studies we assembled, hunting-induced P losses exceeded atmospheric deposition when bushmeat harvest rates were over approximately 350–460 kg km⁻² yr⁻¹ and deposition rates below approximately 0.3 mg m⁻² yr⁻¹. We made similar comparisons for skeletal calcium (Ca), potassium (K) and magnesium (Mg; electronic supplementary material, table S2), but hunting-mediated flux rates appeared trivial in light of the high Ca : P, K : P and Mg : P ratios of tropical vegetation [24].

We had species- and body-mass-specific hunting removal rates for 16 of the 36 sites in our analysis, and used an average skeletal mass (calculated from the studies presenting species-specific information) to estimate P loss in the remainder. To assess whether using average skeletal masses biased our results, we performed a second analysis where we calculated P loss based on average skeletal mass instead of species-specific masses for those studies that provided the latter. This changed our estimated P deposition : removal ratios by an average of 6.0% (s.e. = 1.5%), and did not change the qualitative results (e.g. Figure 1) of the study.

To assess whether biogeochemical impacts varied with body size of the hunted animals, we generated accumulation curves of P loss with increasing body size for studies that provided species-specific hunting rates. Across most sites, P loss was driven by the removal of mammals lighter than 50 kg (electronic supplementary material, figure S2). At the two sites (both in Indonesia) with the highest ratios of hunting-induced P loss to atmospheric P deposition, small species (less than 1 kg) accounted for approximately 84% of total mammal harvest while large species (more than 50 kg) accounted for only 1–5% (electronic supplementary material, table S3).

4. Discussion

When might hunting-induced P loss be a conservation problem? In general, and across most of the tropics, such P losses appear to be small relative to atmospheric nutrient inputs. The few exceptions to this trend were on two continents representing different biomes, African savannah and Southeast Asian rainforest. In both of these cases, atmospheric deposition rates were low, large mammals were present and hunting levels were high. The African study [25] was an ecological and anthropological assessment of the Boni people of northeast Kenya who had originally been largely nomadic hunter–gatherers but who were, at the time of research, based in villages. It was not clear how many individuals of different species they were killing, but hunting may have focused on larger ungulates. The Asian study [26] was in the Minahasa and Bolang Mongondow Districts of Sulawesi (Indonesia), examining subsistence hunting by communities near protected areas, mostly focused on smaller-bodied species (electronic supplementary material, table S3). The people in this area are mostly Christian, and more catholic in their wild meat diet breadths than Muslims. Overall, the hunting-induced nutrient losses observed in Kenya and Indonesia contrast with the general picture of anthropogenic increases in nutrient availability and cycling rates globally [27]. Indeed, in the presence of agricultural runoff or other anthropogenic pollution [28], hunting is extremely unlikely to remove enough animals to offset P inputs. Identifying regions where wild meat serves as a dietary staple, human populations are growing rapidly and ecosystem P inputs are low could direct enhanced management efforts towards areas with critically unsustainable hunting.

Hunting is by no means the sole cause of wildlife declines, though it is a very strong driver in much of the world. While many nutrient cycles are driven by abiotic processes and microbial action [11], vertebrates are known to provide important transport of P both within and between ecosystems [13,14,29]. We build on this by demonstrating that over-hunting of vertebrates can, in a few instances and under certain conditions, affect P cycling, not just through reducing transport of the nutrient by live animals, but by physically removing the P stored in animal skeletons from the ecosystem. The ecological consequences of P removal by hunting depend on the magnitude of other fluxes. On the one hand, the amount of P being removed by hunting is small relative to the amount stored in soils and vegetation in tropical forests [24], though tropical savannahs have much lower plant-based P pools [30]. Indeed, most fluxes in ecosystem P cycles are small relative to the overall pool of P [11]. On the other hand, primary productivity in both tropical forests and savannahs is known to be sensitive to P availability [17,24,31,32], suggesting that high levels of hunting in areas that receive low P inputs could have ecological repercussions by exporting large quantities of this key macronutrient.

In the few situations where continual hunting erodes ecosystem P pools, the capacity of the ecosystem to produce large mammals with heavy skeletons may be reduced. Ungulates can be P-limited in their growth [33,34], suggesting that both individual- and population-level feedbacks from depleting ecosystem P stocks are possible. These relationships may be trivial from a biogeochemical perspective when P inputs are high, but long-term P depletion by sustained, intensive hunting could reduce the carrying capacity for large mammals in regions where P-poor soils and low atmospheric P inputs prevail. This could, in turn, amplify the demographic impacts of hunting by depressing potential population growth rates for large species. For example, bones from dead mammals provide critical nutrients for numerous animals in P-limited systems [35], including ungulates, rodents, carnivores and reptiles. Such osteophagy may be the only way by which large species such as giraffes (Giraffa camelopardalis) can acquire the P needed to build and maintain their skeletons on a low-P diet [34]. Thus, ecosystem P depletion could undercut the capacity for demographic rebounds of mammal populations even if hunting were curtailed.

Assessing the magnitude of nutrient export via wild meat hunting depends on determining the fate of the animal bones [35], where a large majority of P resides in vertebrates [11]. For most species, much or all of the carcass is taken to a human settlement for processing or sale [36,37], where the bones are generally discarded as refuse. In that scenario, hunting can be seen as gathering the P sequestered by mammals from the surrounding landscape, then concentrating it in a narrow halo around population centres, where it is burned or buried, or becomes otherwise inaccessible from the point of view of the broader landscape (i.e. the hundreds or thousands of km² over which mammal exploitation may be occurring). Over ecologically relevant time scales, this is functionally equivalent to a true loss of ecosystem P, as this nutrient tends to be actively recycled only over small distances [24,32]. But for larger animals such as elephants or giraffes, carcasses tend to be butchered on-site, with meat and only some bones (e.g. tusks) removed [36]. In these cases, hunting would have less effect on ecosystem P pools via direct removal, though it could still alter P cycling by removing the prodigious processing of food resources by these large grazers [13,14,29,35].

While wild meat harvests are often unsustainable demographically [1,2], and factors like habitat degradation [38,39] and competition with livestock can further reduce populations of large animals, the potential for hunting itself to undercut ecosystem production capacity for targeted animals has not been considered. Our results suggest that intensive hunting not only depresses populations of tropical mammals but can also—albeit quite rarely—have biogeochemical consequences. Unlike many biogeochemical fluxes, hunting-induced P loss can be directly regulated. Striving to ensure hunting sustainability—incorporating population viability, species interaction and biogeochemical impacts (e.g. [40])—is critical around the world.

Data accessibility

The data used herein are presented in the electronic supplementary material (table S1).

Authors' contributions

Both authors conceived of the study, collected and analysed data, and wrote the manuscript.

Competing interests

We have no competing interests to declare.

Funding

Funding and support were provided by the University of Montana, the University of Wisconsin—Madison, the David H. Smith Postdoctoral Research Fellowship, and a Packard Fellowship in Science and Engineering.