5.3.1 Direct Surveys of Population Density

In a review of sustainability indicators for wild meat hunting, Weinbaum et al. (Reference Brashares, Golden, Weinbaum, Barrett and Okello2013) proposed monitoring of harvested populations through time as one of the gold standards in sustainability monitoring. The surveys provide indications whether a target population is stable, increasing or decreasing. For an overview of techniques to estimate absolute and relative densities, see Millner-Gulland and Rowcliffe (Reference Kümpel, East, Keylock, Rowcliffe, Cowlishaw, Milner‐Gulland, Davies and Brown2007).

• Pros: The advantage is that it is the only method that directly estimates sustainability. The method is very powerful if monitoring is continuous and the results are fed into adaptive harvesting strategies.

• Cons: The major caution is that changes in population abundance are difficult to interpret if estimates of associated species (e.g. predator–prey systems), harvesting and external factors (e.g. habitat change or climate change) are not simultaneously estimated, or if spatial scales are too small to detect source–sink patterns. Time frames need to be sufficiently long to allow distinguishing stochastic change from systematic change, albeit sudden declines in population density can act as early warning systems to trigger more intensive monitoring. The major disadvantage is that it is time intensive and expensive, especially when remote, tropical locations are concerned. Whilst intensive monitoring is more likely for those species in logistically easy-to-monitor habitats (e.g. savannahs), where animals can be directly observed or trapped, monitoring prey populations in dense vegetation environments (e.g. tropical forests) is more difficult.

  Example: Weinbaum et al. (Reference Brashares, Golden, Weinbaum, Barrett and Okello2013) describe the earliest textbook example of this method as Larivière et al.’s (Reference Larivière, Jolicoeur and Crête2000) grey wolf monitoring study in southern Québec, Canada. Here, wolves are found in a mosaic of wildlife reserves, where hunting is controlled by quotas on hunting licenses, unlike public and private lands where harvests are less restricted. Over a 15-year time period, starting with the onset of wolf trapping in the reserves, the study monitored wolf densities in nine wildlife reserves by a combination of questionnaires distributed to moose hunters and the usage of radio-tracking wolves. Aerial surveys were used to monitor moose densities. Pelt sales and tanning records for each trapping district were employed to quantify wolf harvest. Over the study period, although wolf densities fluctuated widely in seven reserves, these showed no indication of long-term declines. By contrast, wolf populations in two reserves declined steadily. Without continuous monitoring, estimates of sustainability would have been highly biased. Population variability was negatively correlated with reserve size, indicating that wolf populations in smaller reserves were more unstable than those in larger reserves. In the two smallest reserves, however, harvesting frequently exceeded wolf densities but without population decline. This points to the presence of a source–sink system in which wolves from adjacent reserves repopulated the smaller reserves. Previously it was thought that these reserves acted as sources for the surrounding public and private lands, but the Larivière et al. (Reference Larivière, Jolicoeur and Crête2000) study showed the contrary. These results demonstrate the importance of investing in continuous population density surveys to provide information to local wildlife managers to ensure conservation of the target species alongside their exploitation. As mentioned above, population monitoring of species such as the grey wolf is possible because the target species is generally visible to the observer (and therefore can be counted using direct methods) or their numbers can be inferred indirectly from records of hunted animals.

5.3.2 Catch Per Unit Effort over Time

The yield or number of animals removed by hunting, H, depends on catchability, q, hunting effort, E, and population size, N:

$H=q·E·N$

The parameter q is a species-specific constant quantifying how difficult or easy it is to hunt the species. If the effort is independent of the yield and population density, then changes in H/E translate in variations in N. This is the catch per unit effort (CPUE) – the ratio of yield to the effort expended to achieve the yield. Hunting effort can be measured as duration of hunts that result in H hunted animals, number of hunted animals per trip, number of snares set or amount of ammunition used. Using interviews and hunting returns, Vickers (Reference Vickers1994) was the first to estimate population density trajectories, but without calling it CPUE, for Siona-Secoya hunters in the northwest Amazonian Peru. More recently, Rist et al. (Reference Rist, Milner-Gulland, Cowlishaw and Rowcliffe2010) investigated the methodology in more detail by comparing data collected by hunters with more direct information gathered by accompanying hunters on hunting trips. By applying simulations, Rist et al. (Reference Rist, Milner-Gulland, Cowlishaw and Rowcliffe2010) assessed the accuracy, power and resolution of the method.

• Pros: CPUE values can be obtained directly from hunters, which is easier and cheaper than monitoring populations in the field. Hunters require little training for data recording.

• Cons: Estimates resulting from CPUE data are unable to determine population density and thus yield. CPUE needs to be monitored over time to reliably identify whether it increases, decreases or stays stable. It requires an adequate and representative sample of hunters per area/region to account for differences in their hunting efficiencies and strategies and geographic substructure such as contrasts between villages. The sample must be sufficiently large to distinguish between measurement errors and stochasticity from ‘true’ changes in CPUE. Studies need to demonstrate this, for example by subsampling and modelling (Rist et al. Reference Rist, Milner-Gulland, Cowlishaw and Rowcliffe2010). Selecting a sufficiently large number of monitored hunters can be a challenge, especially where hunting is illegal (e.g. protected areas), or where specific hunting methods are prohibited (e.g. snares). Moreover, if CPUE is used for management of hunting quotas, reductions in quota will likely erode the hunters’ willingness to participate and compromise the trustworthiness of the data provided. Rist et al. (Reference Rist, Milner-Gulland, Cowlishaw and Rowcliffe2010) modelled CPUEs assuming a statistical power of only 80% and α = 0.05 reporting that information on 1,000 hunts had to be collected to allow the detection of a 20% density change. As many as 3,000 hunts were required for a detection of a 10% change. The method relies on trust between all participants where trust-building is time consuming. Reported values must be unbiased, but experience from fisheries has shown that there is over-reporting of both catch and effort in some fisheries (Lunn & Dearden Reference Lunn and Dearden2006). A crucial component of the equation is that yield is directly proportional to both effort and population size. CPUE assumes that no density-dependent changes occur in hunter effort, such as change in technology or strategy. However, hunters might change to night hunting with flashlights when densities decline, which would bias CPUE and might even result in a stable CPUE despite population declines (Bowler et al. Reference Bowler, Beirne and Tobler2019). Similarly, hunting yield needs to be proportional to density. Even the assumption that catchability is a constant might not hold for many species, for example, when animals respond behaviourally to the presence of humans as a reaction to hunting pressure and, thus, bias the yield (Keane et al. Reference Keane, Jones and Milner-Gulland2011; Papworth et al. Reference Papworth, Milner-Gulland and Slocombe2013a). Aggregation behaviour, the tendency for animals to group together in flocks or herds as seen in many bird and ungulate species can result in similar hunter effort independent of whether density is stable or declining. This is because it is easier to hunt gregarious species than solitary, territorial ones and because aggregations occur despite changes in density. Finally, the same caveats apply to the interpretation of inferred population decline as for the direct surveys of population density.

  Example: Hill et al. (Reference Hill, McMillan and Farina2003) recorded harvest-rate data for 5,526 Aché hunter days during seven years in the Mbaracayu Reserve, Paraguay. CPUE, expressed as animals killed per hunting day, was seen to decline in seven of the ten prey species, which jointly contributed 95% of all individuals and 96% of biomass harvested. Only the drop in capuchin monkeys was significant but high variability in monthly harvest rates may have masked the negative trends. To be able to interpret the complexities of the observed fluctuations in CPUE, 7,535 km of diurnal random line transect surveys were conducted by teams of five observers. Encounter rates from the line transects showed negative trends in nine of the ten species, with four species exhibiting significant declines (Fig. 5.2). However, these four species did not include the capuchin monkey. The 95% confidence interval (CI) of the estimated maximum harvest rate was lower than 1% of the standing stock for six species including the capuchin monkey and was lower than 3.7% for the remaining four species. Overall, the declining CPUE and encounter rates of most species caused concern, but there was little evidence that hunting pressure by Aché hunters was the main cause of these observed decreases. However, there was support for considerable poaching by non-Aché hunters, which could explain the observed patterns. Regarding the capuchin monkey, CPUE was possibly misleading as the observed significant declines may have been caused by changes in hunting effort E, but possibly not in N. During the study, Aché hunters appeared to refocus their attention to peccaries rather than capuchin monkey, thus changing E for the species involved. Finally, the results demonstrate that the applicability and interpretability of the CPUE index in the study was limited without the inclusion of direct surveys of population density.

Figure 5.2 Crude encounter rates/100 km of ten important hunted species. Encounter rates were calculated as total encounters divided by total kilometres of transect walked in each 12-month period of the study.

(From Hill et al. Reference Hill, McMillan and Farina2003; adapted with permission from John Wiley & Sons.)

5.4.1 Estimation of the Population Growth Rate

Another approach to using full demographic information is the modelling of net recruitment rates. Rather than modelling the likelihood of a population to persist as in the VORTEX software approach, the population growth rate lambda, λ, is estimated from the data including harvest whereby λ ≥ 1 implies sustainability and λ < 1 unsustainability. Simulations allow us to evaluate the sensitivity of λ to the effects of parameter variations (Combreau et al. Reference Combreau, Launay and Lawrence2001; Marboutin et al. Reference Marboutin, Bray, Péroux, Mauvy and Lartiges2003).

Figure 5.3 Frequency distribution of estimated wolverine population growth rate (λ) using simulation, British Columbia, Canada for 2007.

(from Lofroth and Ott Reference Lofroth and Ott2007; adapted with permission from John Wiley & Sons)

5.4.2 Population Viability Analysis and the Madingley General Ecosystem Model

Several software PVA packages are available, which can produce different results. These include GAPPS, INMAT, RAMAS Age, RAMAS Metapop, RAMAS Stage and VORTEX. The most often used software for PVA is VORTEX, an individual-based simulation of population demography (Lacy Reference Lacy1993, Reference Akçakaya and Sjögren-Gulve2000). VORTEX is the PVA model of choice for use to simulate the fate of small populations threatened by extinction vortices and for complex models that include individual variation, spatial and metapopulation structure, and complex feedback between demography and genetics (Lacy Reference Lacy2019). Although hunting can be incorporated alongside other additional types of mortality, PVA’s are essentially designed for single-species systems and are difficult to apply for multi-species ones, such as wild meat hunting. For multi-species systems, Barychka et al. (Reference Barychka, Mace and Purves2020a) recently used a new approach, the Madingley General Ecosystem model, which allows simulation of ecosystem dynamics with multi-species harvesting. In computer simulations for duikers, the most heavily hunted species in sub-Saharan Africa (Chapter 1), the model adequately predicts yields, species extinction rates and ecosystem-level harvesting impacts compared to single-species models. Barychka et al. (Reference Barychka, Mace and Purves2020a) suggest that this method should be used more widely for management, but so far it awaits implementation on the ground.

5.5.1 Maximum Sustainable Yield Model

The standard logistic growth rate model introduced in Section 5.1 allows the calculation of maximum sustainable yield (MSY) as

$\text{MSY}=\left(r\cdot K\right)/4$

and a maximum sustainable harvest rate, (MHR), of

$\text{MHR}=r/2$

Whenever an observed harvest is larger than MSY, it is considered unsustainable. However, a harvest that is smaller or equal to MSY is not necessarily sustainable as we do not know whether the harvest yield is on the left side or the right side of the parabola; in case of the first, it would be unsustainable (see Fig. 5.1). This model is mainly used in fisheries (Weinbaum et al. Reference Brashares, Golden, Weinbaum, Barrett and Okello2013).

5.5.2 Robinson and Redford Index

This algorithm has been the most widely used for wild meat species in tropical settings, despite some concerns being raised about its application (Weinbaum et al. Reference Brashares, Golden, Weinbaum, Barrett and Okello2013). The index is relatively easy to calculate according the equation:

$P=0.6\cdot K\cdot \left(r_{max}–1\right)\cdot F$

with an ad hoc mortality factor F dependent on the species’ life history (F = 0.2 for long-lived species whose age of last reproduction is over 10 years, F = 0.4 for short-lived species those whose age of last reproduction is between 5 and 10 years and F = 0.6 for short-lived species whose age of last reproduction is less than five years). The value of 0.6 K is merely a precautionary factor that stems from the assumption that the maximum produced would be achieved when population density (N_i) was at 60% of carrying capacity K. This is a subjective percentage and could be adjusted according to density estimates and knowledge of the population. To assess the species’ intrinsic rate of increase Robinson and Redford (Reference Robinson, Redford, Robinson and Redford1991b) use Cole’s (Reference Cole1954) equation to calculate r_max from the age at first reproduction, the age at last reproduction and the annual birth rate of female offspring, b:

$1=‐{\mathrm{e}}^{\text{‐rmax}}+\mathrm{b}\cdot {\mathrm{e}}^{\text{‐rmax}\left(\text{age}\text{at}\text{first reproduction}\right)}‐\mathrm{b}\cdot {\mathrm{e}}^{\text{‐rmax}\left(\text{age}\text{at}\text{last reproduction}+1\right)}$

This equation does not consider mortality, ‘which is a very strong assumption’’ (Milner-Gulland & Akçakaya Reference Milner-Gulland and Akçakaya2001). Estimates of these reproductive parameters are available for many commonly hunted forest species (e.g. Robinson & Redford Reference Robinson and Redford1986) but vary in their accuracy depending on their origin (see above). For example, in a comparison between published ex situ data and empirical in situ information for the ten most-hunted hunted Amazonian mammal species, the authors found concordance, underestimation and overestimation of species’ r_max values, resulting in different biases for those studies that used the various estimates (Mayor et al. Reference Mayor, El Bizri, Bodmer and Bowler2017). The discrepancies can be so wide that new assessments of these parameters which relate to various ecological conditions are urgently needed (Van Vliet & Nasi Reference Nasi2018). According to Milner-Gulland and Akçakaya’s (Reference Milner-Gulland and Akçakaya2001) the index performed rather badly under realistic conditions simulation experiments (Section 6.3). Population persistence of 200 individuals during a 50-year period was never achieved and harvest was less than half of that of the best performing model, the National Marine Fisheries Service algorithm. Weinbaum at al. (Reference Weinbaum, Brashares, Golden and Getz2013) identified five publications which calculated the Robinson and Redford index alongside at least one other index. From 86 population comparisons, 23 (27%) resulted in divergent conclusions of sustainability.

5.5.3 Bodmer A and B Indices

The equation for the two Bodmer indices (Bodmer Reference Bodmer, Western and Wright1994a), also called the ‘unified harvest model’, is the same:

$P=0.5\cdot N\cdot \phi \cdot s$

and requires information on the female part of the population, 0.5 ∙ N, female fecundity, φ, and female survival to the average reproductive age, s. The latter is either estimated as s = 0.2 for long-lived species and s = 0.6 for short-lived species in case of the Bodmer A model or estimated from actual data in case of the Bodmer B model. Fecundity is typically estimated from ex situ populations that have the same associated problems, as already discussed for the Robinson and Redford index. Sustainability is achieved if the observed harvest is smaller or equal to the estimated P (Weinbaum et al. Reference Brashares, Golden, Weinbaum, Barrett and Okello2013). In Milner-Gulland and Akçakaya’s (Reference Milner-Gulland and Akçakaya2001) simulation experiments, the Bodmer B model performed better in terms of yield, but led almost as often to a high risk (approx. 90%) of the population falling below 200 individuals during a 50-year period as the Bodmer A model (100%).

5.5.4 US National Marine Fisheries Service Index

The algorithm, also called the potential biological removal index (PBR), was developed for cetacean bycatch (Wade Reference Wade1998):

$\text{PBR}=N_{\text{min}}·0.5・r_{\text{max}}・F_{\text{R}}$

with N_min = minimum population estimate, F_R = recovery factor between 0.1 and 1, and r_max = maximum rate of population increase. The index specifically calls for the minimum population estimate rather a mean estimate, thus accounting for uncertainty of density estimates. This is one of the main strengths of the index. The recovery factor allows for the implementation of different management strategies but needs to be assigned with care. Importantly, current knowledge of the species in general and the targeted population in particular, conservation goals, harvesting requests and the feasibility of further monitoring the population, need to be taken into consideration. The smallest value of F_R = 0.1 allows a population to be maintained close to its carrying capacity, to minimize extinction risk for depleted and small populations, or to delay the recovery of a depleted population only slightly. The largest value of F_R = 1 allows a healthy population to be maintained at its maximum net productivity at the MSY density. A recovery factor of F_R = 0.5 accounts for unknown bias or estimation problems such as overestimating r_max or underestimating mortality.

The risk of extinction is low in simulations and the algorithm performs best amongst the indices based on surplus production models. In Milner-Gulland and Akçakaya’s (Reference Milner-Gulland and Akçakaya2001) simulation experiments, the algorithm performed best in its ability to have a no risk of the population falling below 200 individuals during a 50-year period. Total harvest was lower than the full demographic model (Section 6.3) because of its precautionary approach, which has been listed as a potential disadvantage since the algorithm is not designed to maximize yield (Weinbaum et al. Reference Brashares, Golden, Weinbaum, Barrett and Okello2013). Notwithstanding the favourable performance of this index and its popularity in fisheries and marine mammal, turtle and seabird bycatch studies, the approach remains very rarely used, despite its potential favourability over the Robinson and Redford and the Bodmer indices (Weinbaum et al. Reference Brashares, Golden, Weinbaum, Barrett and Okello2013).

5.5.5 Modelling Parameter Uncertainty

Barychka et al.’s (Reference Barychka, Purves, Milner-Gulland and Mace2020b) model allows the implementation of parameter uncertainty, which is pertinent in all field situations for the planning of sustainable hunting. The model centres on the Beverton–Holt population model which is widely used in fisheries (Beverton & Holt Reference Beverton and Holt1957):

$N_{t+1}=r_t{N}_t/\left(1+\left[\left(r_t-1\right)/K\right]N_t\right)$

with N_t and N_t+1 = the population densities at time t and the following time step, respectively; K = the equilibrium population size without harvesting; r_t = the density-independent intrinsic rate of natural increase (i.e., the balance of births and deaths) for year t. Uncertainty for the parameters r_t and K are modelled based on prior belief. The prior belief is ideally based on field data for the studies populations and, failing that, on expert judgement. Barychka et al. (Reference Barychka, Purves, Milner-Gulland and Mace2020b) implemented two harvesting strategies at a constant rate, set either as a quota or proportional to the population size, N_t.

Figure 5.5 Estimated yields (animals/km²/year) from quota-based harvesting of blue duiker without (a) and with (b) parameter uncertainty. Yields are estimated over 25 years in 5-year increments. The survival probabilities are shown in top-right corner of each rectangle.

(From Barychka et al. Reference Barychka, Purves, Milner-Gulland and Mace2020b; reprinted with permission from PLOS ONE.)

5.6.1 Comparing Populations between Sites

The comparison of population density and/or population structure, especially age and sex, between hunted and non-hunted or lightly hunted sites has been used to assess harvesting sustainability. It is assumed that significant differences in density or age and sex composition can be interpreted as the result of unsustainable harvest in the exploited area.

• Pro: Population density can be relatively easily estimated using line transects or camera traps for some species (Milner-Gulland & Rowcliffe Reference Kümpel, East, Keylock, Rowcliffe, Cowlishaw, Milner‐Gulland, Davies and Brown2007). Age and sex structure can be determined from direct observations of the hunted carcasses brought back by the hunters and requires little training. Differences can be statistically tested.

• Cons: In dense tropical and subtropical habitats, the estimation of population density and age and sex structure can be challenging. Sites must be ecologically comparable, but that can often be verified only by intensive field work. Even when differences are significant and large, the data indicate only local depletion, but not sustainability. As indicated in Fig. 5.1, sustainable harvesting is possible in a large range of population densities. Thus, differences in population density per se do not prove unsustainability. Also, differences in sex and age structure per se do not prove unsustainability but the absence of differences does not demonstrate sustainability either. For example, Fitzgibbon et al. (Reference Fitzgibbon, Mogaka and Fanshawe1995) observed significant differences in density estimates in hunted versus unhunted areas in four-toed elephant shrews, squirrels and Syke’s monkeys, but not for yellow baboons. Conversely, comparing current harvest levels reported by hunters with the estimated maximum potential sustainable harvest rates according to the Robinson and Redford (Reference Robinson, Redford, Robinson and Redford1991b) model (Section 5.4.2) indicated non-sustainability of yellow baboons and Syke’s monkeys but not the other species.

  Example: Hurtado-Gonzales and Bodmer (Reference Gonzalez, Silvius, Bodmer and Fragoso2004) assessed the sustainability of red and gray brocket deer hunting in unhunted, slightly hunted and heavily hunted sites in Peru. For the two species, gross productivity was higher in the heavily hunted site compared to the unhunted site as measured by the number of foetuses recorded per female. The heavily hunted area had a higher density of gray brocket deer compared to the non-hunted area (Fig. 5.6) but differences in age structure were not tested. No significant differences in density and age structure were found for red brocket deer. To solve this ambiguity, the Bodmer B algorithm (Section 5.4.2) was applied revealing that the harvest of both species was sustainable.

Figure 5.6 Densities of red and gray brocket deer according to different hunting pressures in the study site. Densities were estimated by line transect surveys.

(From Hurtado-Gonzales and Bodmer Reference Gonzalez, Silvius, Bodmer and Fragoso2004; adapted with permission from Elsevier.)

5.6.2 Differences in Harvest Characteristics

Changes in the characteristics of harvest data over time or differences between the characteristics of harvest data between ecologically similar sites might indicate depletion of populations or overharvesting. Such changes may encompass changes in hunting pressure, that is, number of animals killed per area, increasing distances required to reach profitable hunting grounds, and changes of species composition over time (Albrechtsen et al. Reference Albrechtsen, Macdonald, Johnson, Castelo and Fa2007; Hurtado-Gonzales & Bodmer Reference Gonzalez, Silvius, Bodmer and Fragoso2004; Smith Reference Smith2008).

• Pros: Data directly from hunters can be used, thus, relatively easy to obtain.

• Cons: As discussed for CPUE (Section 5.2.2) the application of this method requires an adequate and representative sample of hunters involved in data generation. This may be a challenge in some situations, especially if hunting is illegal. Moreover, the sample obtained may be biased if offtake differs by age or sex of the hunter, or if hunting is for subsistence rather than for trade, factors often overlooked in harvest studies (Ingram et al. Reference Ingram, Coad and Collen2015). Moreover, hunters may under-report or fail to report the hunting of protected species. A suite of different factors can impact the system, other than hunting. These could include biological factors, such as climate or vegetation structure, and anthropogenic impacts, such as logging or road development, hunting motivations, hunting technology, market supply and demand, or law enforcement. There are neither standardizations nor any quantitative or even qualitative generalized guidelines or agreement on how to accept or reject the hypothesis of sustainability.

  Example: Smith (Reference Smith2008) mapped the spatial patterns of hunting yields in Panama within a community of Indigenous Buglé and Ngöbe hunters. Kill sites were concentrated within just 2 km of the hunters’ homes; nearly 90% of the total harvest originated here (Fig. 5.7). Hunting at larger distances occurred less often and depends on the availability of firearms, whereas other methods including slingshots and bow and arrow are more common near the settlements. While most species were killed near settlements, other species, in particular the black-handed spider monkey, were hunted only further away. Smith (Reference Smith2008) argues that this pattern suggests that some degree of localized depletion may have occurred due to overhunting close to human settlements. Based on the local ecological knowledge of rural hunters, population depletion of hunted forest wildlife close to villages has been clearly demonstrated by Parry and Peres (Reference Parry and Peres2015) in the Brazilian Amazon. Similarly, mammal and bird population densities declined with their proximity to infrastructure such as roads (Benítez-López et al. Reference Benítez-López, Alkemade and Verweij2010). For example, abundance of several species including duikers, sitatungas and forest elephants in Central Africa were depressed by the presence of roads.

  (Laurance et al. Reference Elkan, Elkan, Moukassa, Malonga, Ngangoue, Smith, Peres and Laurance2006)

Figure 5.7 Hunting yields as a function of distance from hunters’ primary residences in 500-m intervals.

(From Smith Reference Smith2008; adapted with permission from Elsevier.)

The special case of duikers. Duikers (genera Philantomba and Cephalophus) are amongst the most hunted mammals for wild meat throughout the Congo Basin (Fa et al. Reference Bartlett, Williams and Prescott2016; Kingdon et al. Reference Kingdon and Hoffmann2013; Wilkie & Carpenter Reference Wilkie and Carpenter1999; Yasuoka Reference Yasuoka2006b). Thus, there is a much interest in ascertaining the hunting sustainability for these species. Yasuoka et al. (Reference Yasuoka, Hirai, Kamgaing, Dzefack, Kamdoum and Bobo2015) suggested that the catch ratio between the smaller blue duikers and the larger duikers, especially red duikers, could be used as an indicator of depletion in a site; higher numbers of red duikers denoting a less-affected system. The different duiker species respond distinctly to increased hunting pressure with the smaller blue duiker being less affected at a population level. This is because this species reaches reproductive age earlier than medium-sized duikers and have, thus, a higher reproductive output. The other reason is that they are more tolerant to and thrive in human-modified landscapes (Hart Reference Hart, Robinson and Bennett1999). These predictions were fulfilled in Yasuoka et al.’s (Reference Yasuoka, Hirai, Kamgaing, Dzefack, Kamdoum and Bobo2015) study on duiker densities in southeastern Cameroon. However, duiker densities in central African forests can vary from 3.5 to 59.8 individuals/km² for blue duikers and 2.6 to 64.5 individuals/km² for red duikers but explanations for these differences are likely to be related to a combination of habitat type, hunting history and hunting pressure factors (Breuer et al. Reference Breuer, Breuer-Ndoundou Hockemba, Opepa, Yoga and Mavinga2021). Since the interactions of these factors has not been adequately determined in African forests, unlike studies for the Amazon for other species (see Peres Reference Peres, Robinson and Bennett1999a), trends of the few studied duiker populations have been shown to decline with increases in hunting pressure (Grande-Vega et al. Reference Grande-Vega, Farfán, Ondo and Fa2016; Hart Reference Hart, Robinson and Bennett1999) but the impact of habitat or hunting history is unknown. Moreover, hunting methods can impact small- and medium-sized duikers differently with gun hunting, especially at night, making blue duiker easier prey than the medium-sized red duikers (Yasuoka et al. Reference Yasuoka, Hirai, Kamgaing, Dzefack, Kamdoum and Bobo2015). However, the large variance of the blue to red duiker ratios ranged from 0.39 to 22. 5 (n = 5, median = 1.33) in hunted areas and from 0.23 to 1.66 (n = 4, median = 1.4) for unhunted areas. Breuer et al. (Reference Breuer, Breuer-Ndoundou Hockemba, Opepa, Yoga and Mavinga2021) suggest to use them only with precaution as an indicator of hunting pressure or habitat type. Thus, the catch ratio as a means to estimate sustainability remains uncertain and for the time inapplicable.

5.6.3 Changes in Body Mass

A drop in the mean body mass of harvested species in a site can be used as an indicator of depletion. Trends in species composition and the average size of prey have been used in monitoring fish exploitation, a phenomenon known as ‘fishing down marine food webs’ (Pauly Reference Pauly1998) where fisheries increasingly rely on the smaller, short-lived fishes as the larger ones are depleted. This phenomenon has been measured by the large fish indicator (LFI), which captures changes over time in the contribution of biomass from large fish to the catch (Greenstreet et al. Reference Greenstreet, Rogers and Rice2011; Shephard et al. Reference Shephard, Reid and Greenstreet2011). The mean body mass of hunted terrestrial prey within each sample can also be used as a proxy of species composition, where a drop from larger to smaller species may indicate a process of defaunation of a habitat (Dirzo et al. Reference Dirzo, Young, Galetti, Ceballos, Isaac and Collen2014). In hunted terrestrial systems, the decrease in mean body mass of prey can reflect the increase in the proportion of small-bodied species over time, either because large-bodied species were extirpated, or more small-bodied species are being harvested. Based on these premises, Ingram et al. (Reference Ingram, Coad and Collen2015) used the mean body mass indicator (MBMI) to integrate taxonomically, spatially and temporally disparate data collated from multiple sources over a period of 40 years. The MBMI can be used to offer insights into wildlife exploitation dynamics and is useful in understanding trends in hunted wildlife. In addition to the MBMI, Ingram et al. (Reference Ingram, Coad and Collen2015) used an index of hunting pressure, the offtake pressure indicator (OPI) providing a measure of relative change in the number of harvested individuals indexed across multiple sites and species.

• Pros: As suggested by Ingram et al. (Reference Ingram, Coad and Collen2015), these two indicators offer potentially useful approaches to assess wildlife offtake in the absence of comprehensive monitoring schemes, especially once further calibrated. For example, because each index is calculated differently, with the former employing an arithmetic mean and the latter a geometric mean, MBMI will change more rapidly compared to the OPI. With more time series at multiple sites available, it would be possible to calculate both indicators for the same sites and compare them. However, identifying causal links between changes in pressure on and the state of wild animal populations is often difficult. Wild animal offtake indicators have the potential to establish such linkages when combined with indicators of state to potentially estimate sustainable exploitation.

• Cons: Long-term data are required, which are rarely collected. As discussed for CPUE (Section 5.2.2) and for monitoring changes in harvest characteristics (Section 5.2.2), recruitment of hunters can be a challenge, especially when long-term monitoring is involved, as required here.

  Example: Using data available for West and Central African mammals and birds, Ingram et al. (Reference Ingram, Coad and Collen2015) demonstrated the indexed number of individuals harvested, OPI, of both mammals and birds increased dramatically between 1998 and 2010, indicating increasing hunting pressure (Fig. 5.8). During the same time span average body mass of harvested mammals declined significantly between 1966 and 2010, whereas that of birds increased between 1975 and 2010, indicating changed compositions of hunting bags (Fig. 5.9). This difference may indicate that whilst mammal prey were being depleted, in birds larger species were progressively being targeted, reflecting either the change in the demand for larger birds and their bills, such as the black-casqued hornbill, or in response to the decline in mammalian prey, although these effects are difficult to determine. The latter difficulty emphasizes that the trends in MBMI and OPI need careful interpretation because species might be hunted because of differing demands, for example, subsistence versus trophy hunting.

Figure 5.8 Offtake pressure indicator for (a) mammals and (b) birds in Central Africa and (c) the distribution of time-series data at four sites The indicator is set to 1 in the first year for which data were available (dotted horizontal line). Shading (a and b) represents ±95% confidence intervals generated with 1,000 bootstrap replicates. Width of bars (c) represents the number of mammal (grey) and bird (black) species sampled at four sites.

(From Ingram et al. Reference Ingram, Coad and Collen2015; reprinted with permission from the Resilience Alliance.)

Figure 5.9 Mean body mass indicator for mammals (grey circles) and birds (black circles) in West and Central Africa. Circles represent offtake samples and are scaled by the number of species harvested within each sample; lines are fitted using linear mixed effects models. Samples are plotted on a logarithmic scale.

(From Ingram et al. Reference Ingram, Coad and Collen2015; reprinted with permission from the Resilience Alliance.)

5.6.4 Market Indices

Among all different types of data required for sustainability assessments, market data are the easiest to collect. Surveys of wild meat markets over time allow monitoring of different aspects that might indicate unsustainability: price trends, quantity of species available, species composition and wildlife source distance. It is considered unsustainable whenever prices increase, quantity of available species decrease, species composition change and the distance where wildlife is sourced increase (Albrechtsen et al. Reference Albrechtsen, Macdonald, Johnson, Castelo and Fa2007; Cowlishaw et al. Reference Cowlishaw, Mendelson and Rowcliffe2005; Damania et al. Reference Damania, Milner-Gulland and Crookes2005; Milner-Gulland & Clayton Reference Milner-Gulland and Clayton2002; Rowcliffe et al. Reference Rowcliffe, Cowlishaw and Long2003). The hypothesis is that the composition of species for sale in wild meat markets, as influenced by hunting history in their catchment areas (Cowlishaw et al. Reference Cowlishaw, Mendelson and Rowcliffe2005), will indicate the level of exploitation in the supply areas since vulnerable taxa (slow reproducers such as large ungulates and primates) are depleted first and are replaced by smaller-bodied robust taxa (fast reproducers), such as rodents and small antelopes. Consequently, there is increased prominence of species with high reproductive potential (as defined by their intrinsic rate of natural increase r_max) sold in markets characterizes heavily exploited catchments as shown in Dupain et al. (Reference Dupain, Nackoney and Mario Vargas2012).

• Pros: Requires monitoring of market stalls that sell wild meat, which is more easily conducted by local people than any other method of surveying sustainability. Changes can be statistically tested.

• Cons: Data collection is not advised for foreigners in many settings because of safety concerns. Supply and demand of traded wild meat are impacted by a complex mixture of factors, such as taste preferences, tradition, economic settings, supply, demand and price of domestic meat and fish, environmental changes or law enforcements. There are neither standardizations nor any quantitative or even qualitative generalized guidelines or agreement on how to accept or reject the hypothesis of sustainability. The difficulties in interpreting the data make these changes more an early warning system than a decision-making system on sustainability.

  Example: Albrechtsen et al. (Reference Albrechtsen, Macdonald, Johnson, Castelo and Fa2007) explored market data at the main wild meat market in Bioko Island, Equatorial Guinea, in 1996 and 1998. There was an increase in price, a significant decline in total and individual animal group carcass volumes, the species composition differed, and the diversity indices changed. Whilst the first three parameters indicate overhunting, the diversity index changed towards more diversity, which is surprising as the classic depletion model predicts the decrease of diversity. The authors argue that this is a transient phase and will be followed by a decrease of diversity once only the more resilient species are left as shown elsewhere in Bioko (Fa et al. Reference Cannon2000). Subsequent, independent market surveys showed that the numbers of these animals entering the market have not increased (BBPP Reference Balée and Erickson2006; Reid et al. Reference Reid, Morra, Bohome and Sobrado2005), thus indicating a sustained vulnerability of animal populations to extensive hunting.

The study by Fa et al. (Reference Fa, Olivero and Farfán2015a)shows a clear relationship between anthropogenic pressures in catchment areas and the profile of species appearing in wild meat markets from data for 79 markets (covering 30,000 km²) in the Cross-Sanaga region in Nigeria and Cameroon (Fa et al. Reference Fa, Seymour, Dupain, Amin, Albrechtsen and Macdonald2006; Macdonald et al. Reference Macdonald, Johnson and Albrechtsen2011, Reference Abernethy, Coad, Taylor, Lee and Maisels2012). The hypothesis tested was that the percentage composition of various mammal groups (ungulates, rodents, primates and carnivores) in a market can be a crude measure of depleted faunas of the areas supplying that market. Results indeed confirmed that markets in heavily exploited areas (defined by indirect anthropogenic metrics such as high human population densities), the percentage contribution of larger-bodied prey (slow-reproducing species), especially ungulates and primates, to markets was characteristically lower than the percentage contribution of smaller fast-reproducing prey, such as rodents. Carnivores (mostly smaller taxa) become more numerous in markets as areas become more depleted of wild meat. Research from other sites in Central Africa are consistent with these results, especially the observation that species hitherto unimportant in the wild meat trade gain prominence when ungulates become scarce (Ingram et al. Reference Ingram, Coad and Collen2015; Wilkie Reference Wilkie1989). In lightly hunted rural sites, duikers and other antelopes are the more common prey (Lahm Reference Lahm, Hladik, Hladik, Pagezy, Linares, Koppert and Froment1993; Noss Reference Noss1998a). Although factors, such as habitat quality and human pressures, will jointly impact the wildlife supplying markets (Fa & Brown Reference Fa and Brown2009), the most parsimonious interpretation of the composition of traded species in wild meat markets in the Fa et al. (Reference Fa, Olivero and Farfán2015a) study was the contrasting ability of ungulate and rodent populations to recover from hunting. Thus, at sites where larger species have been severely depleted, hunters would extract fewer of the preferred larger red and blue duikers and more of the smaller species such as the Emin’s and the African giant pouched rats or the cane rat. In most continental sites in Africa (unlike Bioko Island, see Fa et al. Reference Cannon2000), rodents only become important prey items in disturbed areas (Eves & Ruggiero Reference Eves, Ruggiero, Robinson and Bennett1999). Therefore, increases in rodents hunted suggest reductions in the availability of more preferred wild meat species. The relative proportions of ungulates and rodents in the offtake can be used as indicators of site over-exploitation. Fa et al. (Reference Fa, Olivero and Farfán2015a) showed that the relationship between ungulates and rodents was related to a number of prominent anthropogenic factors, rather than environmental variables per se. Higher road densities were linked to reduced abundance of a number of mammal species due to higher hunting pressure. Heavily populated and accessible areas have fewer duikers, forest buffalos and red river hogs (Laurance et al. Reference Elkan, Elkan, Moukassa, Malonga, Ngangoue, Smith, Peres and Laurance2006). Moreover, using an index of game depletion (GDI) for each market (the sum of the total number of carcasses traded per annum and species, weighted by the intrinsic rate of natural increase (r_max) of each species, divided by individuals traded in a market), Fa et al. (Reference Fa, Olivero and Farfán2015a) showed that this index increased as the proportion of fast-reproducing species (highest r_max) rose and as the representation of species with lowest r_max (slow-reproducing) declined. The GDI is akin to indicators used for fishery catches and, as noted for the MBMI and OPI, can be used as a framework for discerning the status of hunted prey (particularly mammals) in a catchment area from observations of the composition of species for sale in markets (Section 5.5.3).

5.7.1 Ecosystem-Based Fisheries Management

The harvesting of any wildlife resource, whether wild meat and fisheries, should strive for the sustainable use of exploited species on land and in the sea. Because of the much longer history of industrial exploitation of marine life, and accompanying research on how to achieve sustainable exploitation of marine fish and stop fisheries collapses (Pauly et al. Reference Pauly, Christensen and Guénette2002), lessons learnt in fisheries can be applied to wild meat (Section 5.4.4; Milner-Gulland & Akçakaya Reference Milner-Gulland and Akçakaya2001). Given this, we briefly describe some of the latest developments in fisheries management that are relevant to wild meat.

The failure of several fisheries, such as the notorious collapse of the Newfoundland Atlantic cod fishery (Bavington Reference Bavington2011), has led to the recognition that traditional management has failed to lead to sustainable exploitation of marine resources. These have caused a clear paradigm shift in fisheries research leading to a change in focus from a single species and MSY approach towards ecosystem-based management (EBM) (Lidström & Johnson Reference Johnson, Hitchens and Pandit2020; Pikitch et al. Reference Pikitch, Santora and Babcock2004; Townsend et al. Reference Townsend, Harvey and deReynier2019). EBM has been defined by the FAO as aiming ‘to balance diverse societal objectives, by taking into account the knowledge and uncertainties about biotic, abiotic and human components of ecosystems and their interactions and applying an integrated approach to fisheries within ecologically meaningful boundaries’ (FAO Fisheries Department 2003). Similarly, the United States’ National Oceanic and Atmospheric Administration (NOAA), has adopted ecosystem-based fisheries management (EBFM), as the agency’s strategic policy. NOAA defines EBFM as: ‘a systematic approach to fisheries management in a geographically specified area that contributes to the resilience and sustainability of the ecosystem; recognizes the physical, biological, economic, and social interactions among the affected fishery-related components of the ecosystem, including humans; and seeks to optimize benefits among a diverse set of societal goals’ (NOAA 2016).

EBM and EBFM differ from the conventional management of single species by describing management strategies for entire ecosystems, to explicitly achieve not only sustainability of target species but ensure the ecological sustainability of the species’ ecosystems, including economic and social sustainability. These holistic approaches take into account natural marine resources’ interactions with their environment as well as human interactions with these resources and the environment. The implementation of policies reflecting these new approaches must necessarily rely on the support of ecosystem science, continuing population monitoring, modelling and analysis and crucially the collaboration and consultation between scientists, policy makers, stakeholders and the public. Despite these encouraging innovations, there is still little practical advice on how to better select specific management measures to achieve EBFM goals. The main operational problems with implementing EBM/EBFM are: (1) defining proper long-term ecosystem related objectives, (2) identifying meaningful indicators and reference values for sustainable use and (3) developing appropriate data collection, analytical tools and models (Cury et al. Reference Cury, Mullon, Garcia and Shannon2005). As a response to these issues, Levin et al. (Reference Levin, Fogarty, Murawski and Fluharty2009) suggested the adoption of integrated ecosystem assessments (IEAs), as a framework of ‘formal synthesis and quantitative analysis of information on relevant natural and socioeconomic factors, in relation to specified ecosystem management objectives’. This framework is a looping workflow composed of scoping, indicator development, risk analysis, management strategy evaluation and ecosystem assessment repeated in an adaptive manner (Fig. 5.10). Importantly, the mechanistic indicators and estimators for population sustainability discussed in this chapter still remain important in indicator development, embedded in this adaptive management loop. Several example cases exist of successful implementation of the IEA framework under the EBFM policy, in particular outside the tropics (Townsend et al. Reference Townsend, Harvey and deReynier2019). In these cases, such as the Gulf of Alaska Pacific cod harvest, the use of ecosystem models in a management process has improved the health or status of particular fish stocks or habitats. These examples highlight the importance of collaboration between modellers, stakeholders and resource managers to ensure sustainable management. For fisheries in the tropics, however, an additional series of challenges and problems exist, resulting from undeveloped or inappropriate governance structures, poor science, lack of political will in many cases and often economic development overriding biodiversity protection (Aswani et al. Reference Aswani, Christie and Muthiga2012). Moreover, many developing countries have property rights which do not grant local coastal communities any legal rights to establish and enforce control over the coastal resources.

Figure 5.10 The Five-Step Process of Integrated Ecosystem Assessment. It begins with a scoping process to identify key management objectives and constraints of the ecosystem-based management, identifies appropriate indicators and management thresholds, determines the risk that indicators will fall below management targets and combines risk assessments of individual indicators into a determination of overall ecosystem status. The potential of different management strategies to alter ecosystem status is evaluated, and then management actions are implemented, and their effectiveness monitored. The cycle is repeated in an adaptive manner.

(From Levin et al. Reference Levin, Fogarty, Murawski and Fluharty2009; adapted with permission from PLOS Biology.)

Problems associated with the implementation of EBM approaches to fisheries in tropical countries also apply to wild meat hunting. A holistic system such as EBM can be the gold standard for wild meat, but we are still far from pursuing this goal. Data-deficiency for most wild-meat-producing areas is not necessarily a limiting factor in implementing EBM because in such situations natural history and general knowledge can be used to develop precautionary safety margins as a starting point for more comprehensive EBM in the future (Pikitch et al. Reference Pikitch, Santora and Babcock2004). However, economic underdevelopment and local poverty can prevent the implementation of more sophisticated monitoring, modelling and management systems. For example, solid frameworks of population assessments and monitoring which are the foundation of EBM, are exceedingly rare in the tropics and subtropics. In fact, in many tropical countries, monitoring programmes are often short-lived, largely research projects limited by funds and unable to be extended over longer time periods. Even the basic acknowledgement of the need for monitoring is often missing at local, regional and national levels. Nevertheless, the idea of managing wildlife populations within a clear ecosystem and social context has been raised by some authors (Van Vliet et al. Reference Van Vliet, Fa and Nasi2015b).

5.7.2 Shifting from Biological Indicators to Resilience Analysis for Wild Meat

In line with the EBM approach, Van Vliet et al. (Reference Van Vliet, Fa and Nasi2015b) have argued that wild meat hunting must be considered as a social-ecological system, and by so doing managers must move towards resilience analysis, adaptive resource management and participatory governance. Although this notion has been critiqued by Sirén (Reference Sirén2015) who argues that MSY is seldom, if ever, the goal, most approaches to resolving wild meat hunting systems have been caught up in assessing rather achieving sustainability, as we show in Section 5.8. Because hunting systems involve human actions and include social structures, as well as biological processes (prey species, ecosystems), the distinction between the social and the natural is arbitrary (Berkes et al. Reference Berkes, Colding and Folke2002); hence a social-ecological system. Van Vliet et al. (Reference Van Vliet, Fa and Nasi2015b) argue that the emphasis should move away from just assessing stocks of prey populations, to considering the complex and dynamic relationships between the hunting ground, its resources, the stakeholders in play and the different exogenous drivers of change affecting the system. This new way of looking at hunting systems incorporates uncertainty and stochasticity inherent to complex systems. It also recognizes that systems evolve over time, adapt and transform.

The main implication of changing the theoretical understanding of hunting systems is that one-off biological indicators are not useful for the estimation of sustainability. Other methodologies that integrate complexity, for example, theoretical models, such as agent-based models, companion modelling approaches, fuzzy cognitive mapping, and resilience analysis tools, among others, are advantageous. Spatially explicit multi-agent-based models are particularly adapted to understanding wild meat hunting sustainability (Bousquet et al. Reference Bousquet, LePage, Bakam and Takforyan2001; Iwamura et al. Reference Iwamura, Guisan, Wilson and Possingham2013; Van Vliet et al. Reference Van Vliet, Milner-Gulland, Bousquet, Saqalli and Nasi2010a). Agent-based models are a class of computational models for simulating the actions and interactions of autonomous agents, both individual and collective entities such as organizations or groups, with a view to assessing their effects on the system as a whole. These models combine elements of game theory, complex systems, emergence, computational sociology, multi-agent systems and evolutionary programming (Grimm et al. Reference Grimm, Revilla and Berger2005).

A number of methods are available for use in addressing the complexity of social-ecological systems. Firstly, collective decision-making processes in complex situations can be better understood with the application of participatory models (Barreteau et al. Reference Barreteau, Bousquet, Étienne, Souchère, d’Aquino and Étienne2014). This approach facilitates collective decision-making processes by making more explicit the various points of view and subjective criteria to which the different stakeholders refer implicitly. As demonstrated in past research (Funtowicz et al. Reference Funtowicz, Ravetz and O’Connor1998; Mermet Reference Mermet1992; Ostrom et al. Reference Ostrom, Gardner, Walker, Walker and Walker1994), when a complex situation exists, the decision-making process is evolving, iterative and continuous. This process always produces imperfect ‘decision acts’, but by following each iteration they are less imperfect and more shared. Participatory scenario planning allows the description of how the future might unfold on the basis of coherent assumptions about the relations among drivers of change and key aspects of the system. The method also allows the participation of a great diversity of stakeholders.

Graphical stock-and-flow modelling, such as fuzzy-logic cognitive mapping (FCM), is a simple and easy method that allows groups to share and negotiate knowledge collaboratively and build semi-quantitative conceptual models. Fuzzy-logic cognitive mapping facilitates the explicit representation of group assumptions about a system being modelled through parameterized cognitive mapping (Gray et al. Reference Gray, Zanre, Gray and Papageorgiou2014). Specifically, FCM allows cognitive maps to be constructed by defining the most relevant variables that constitute a system, the dynamic relationships between these variables and the degree of influence, either positive or negative, that one variable can have on another. In group settings, FCM models are constructed based on combining group beliefs in a similar format as individuals share their experiences and understanding (Gray et al. Reference Gray, Zanre, Gray and Papageorgiou2014). The strength of using FCM in this context is the ability to extract, combine and represent group knowledge in a sensitive situation for comparison between or among groups. Nyaki et al. (Reference Nyaki, Gray, Lepczyk, Skibins and Rentsch2014) used FCM to understand the drivers of wild meat trade in four Tanzanian villages bordering Serengeti National Park.

Resilience analysis explicitly allows for sustainable use options (Box 5.1). Traditional one-off biological models applied to hunting systems, which were based on a binary yes/no question, do not allow the system to be brought back to sustainability in case of a ‘no’ answer. In such cases, the response would be to ban hunting, reinforcing the protection of wildlife by prohibiting its use through legal prohibitions. In contrast, the resilience focus provides the opportunity for identifying strategies that strengthen resilience when the system is close to a given threshold. By recognizing the benefits that wild meat use generates for people, especially indigenous and local people who live with wildlife, and therefore bear associated costs (e.g. danger to life, damage to crops, restrictions on land use), the resilience approach can incorporate the diverse views and value systems of stakeholders, as well as different knowledge sources, including experimental or scientific knowledge and experiential or local ecological knowledge (Cooney & Abensperg-Traun Reference Cooney and Abensperg-Traun2013). As a result, resilience approaches recognize multiple objectives, design mechanisms for incorporating them, weigh trade-offs and establish conflict resolution mechanisms that are fair to all parties. Identifying areas of agreement and disagreement between actors helps in understanding and overcoming obstacles between them (Biggs et al. Reference Biggs, Abel, Knight, Leitch, Langston and Ban2011).