Among the wild animals that attack people and their livestock across Africa, crocodiles (the Nile crocodile Crocodylus niloticus and the West African crocodile Crocodylus suchus) are widely distributed and are claimed to be responsible for most attacks on people (e.g. Lamarque et al., Reference Lamarque, Anderson, Fergusson, Lagrange, Osei-Owusu and Bakker2009; Dunham et al., Reference Dunham, Ghiurghi, Cumbi and Urbano2010). Male Nile crocodiles may exceed 4 m in length, and grow up to 5 m in exceptional cases, taking large prey such as wildebeest Connochaetes spp. and buffalo Syncerus caffer. They are adaptable to local environmental conditions and occur in a wide range of natural and human-made aquatic habitats, such as canals and dams, where they increasingly come into contact with people and their livestock.
Increasing human populations and utilization of rivers, lakes, wetlands and dams (from small farm dams to large irrigation dams), as well as gillnetting (for fish), are resulting in an increasing number of human–crocodile interactions and a perception that adverse encounters between people and crocodiles are increasing (Aust et al., Reference Aust, Boyle, Fergusson and Coulson2009; Lamarque et al., Reference Lamarque, Anderson, Fergusson, Lagrange, Osei-Owusu and Bakker2009; Fergusson, Reference Fergusson, Manolis and Stevenson2010; Wallace et al., Reference Wallace, Leslie and Coulson2011; Zakayo, Reference Zakayo2014). The online database CrocBITE (2018) contains records of attacks in 29 African countries, and attacks are known to have occurred in one additional country (Kpéra et al., Reference Kpéra, Aarts, Tossou, Mensah, Saïdou and Kossou2014).
Research articles on crocodile attacks are revealing informative spatial and temporal patterns in attacks, and provide useful information about the demographics of attack victims (recent examples include Brien et al., Reference Brien, Gienger, Browne, Read, Joyce and Sullivan2017; Shaney et al., Reference Shaney, Hamidy, Walsh, Arida, Arimbi and Smith2017; Vyas & Stevenson, Reference Vyas and Stevenson2017; Das & Jana, Reference Das and Jana2018), but data for Nile crocodiles are inadequate. Published data of varying quality and quantity (most not peer reviewed) exist for 12 of the 30 African countries where attacks are known to occur (Pooley, Reference Pooley2018). More data, as well as reviews of mitigation efforts, are required urgently (Fergusson, Reference Fergusson, Manolis and Stevenson2010; Pooley, Reference Pooley2015a).
Here we present an analysis of 67 years of data on crocodile attacks on people in South Africa and the Kingdom of Swaziland (now eSwatini) during 1949–2016. We use the resulting generalizations to investigate some of the patterns and challenges identified in specific locations, in the context of the history and management of crocodile attacks in the study region. Drawing on attack data and historical evidence, we suggest ways forward for conservation policy and management of human–crocodile relations in the study region.
This study focuses on north-eastern South Africa, including the warmer, low-lying (lowveld) region of the interior confined mainly to Limpopo and Mpumalanga Provinces, and northern KwaZulu-Natal Province (KZN), and the lower-lying warmer areas of eSwatini (Fig. 1). Nile crocodile distribution in the region is limited to the warmer, summer rainfall regions of these countries, with the hot and wet season (minimum temperatures > 15 °C) during October–March (November–March in the interior of South Africa). Most of the rivers flow eastwards, from the central plateau and eastern escarpment to the Indian Ocean.
During c. 1949–1992 South Africa's black African majority was persecuted under the system of Apartheid, with resettlement in remote rural homelands with poor land and few jobs, and men working in cities as migrant labourers. This system kept two-thirds of the population rural (some of them more likely to encounter crocodiles) until the early 1980s, when Apartheid began to fail. Apartheid influx laws were defied, and urbanization accelerated, especially after an African National Congress-led government came to power in 1994 (Turok, Reference Turok2012). Employment in the agricultural sector is now low (5.6%) and declining, with unemployment much higher in rural areas (Turok, Reference Turok2012; UNDP, 2018).
Census data (decadal, from 1951) are of limited use for investigating relationships between human population densities and crocodile attacks in particular locales. Data from the Apartheid era are considered to be highly questionable (Christopher, Reference Christopher2011), the borders of magisterial districts varied across the study period (Giraut & Vacchiani-Marcuzoo, Reference Giraut and Vacchiani-Marcuzoo2009), and estimates of human population density exist only at a coarse scale (magisterial districts; Table 1).
eSwatini is a small, stable absolute monarchy with a largely rural population. The population increased sixfold during the study period. eSwatini is categorized as a lower-middle income country, and the majority of Swazis are poor, with an estimated 70% of the population employed in subsistence farming (CIA, 2017; UNDP, 2018). Many Swazis depend on rivers for water, for drinking, cooking and washing.
Crocodile populations and threats
There are naturally occurring wild Nile crocodiles as far south as the Zinkwazi River in South Africa but the major viable populations are restricted to three disjunct protected areas: the eight large seasonal and perennial rivers traversing Kruger National Park in Limpopo and Mpumalanga provinces; and in KwaZulu-Natal Province, in Ndumo Game Reserve and the Lake St Lucia estuarine system (Ferreira & Pienaar, Reference Ferreira and Pienaar2011; Combrink et al., Reference Combrink, Warner, Downs, Perissinotto, Stretch and Taylor2013; Calverley & Downs, Reference Calverley and Downs2014a).
Crocodile abundance in Kruger National Park peaked in the early 1990s and then declined during 1993–2000, but has since increased to an estimated 4,300 individuals > 1 m in length. This is despite die-offs since 2008, caused by the nutritional disease pansteatitis (Ferreira & Pienaar, Reference Ferreira and Pienaar2011). In Limpopo and Mpumalanga provinces, outside the Park just over 600 individuals were counted in the 1980s, most < 3 m in length, with breeding populations in the Olifants, Limpopo, Luvuvhu, Komati and Blyde rivers (Jacobsen, Reference Jacobsen1984). At the time of writing, only the 12.8 km2 Flag Boshielo Dam on the Olifants River retained a viable crocodile population outside Kruger National Park in Mpumalanga Province (Botha, Reference Botha2005). This population declined by 27% following the raising of the dam wall by 5 m in 2006 (Ashton, Reference Ashton2010).
In northern KwaZulu-Natal Province, populations declined after World War II as a result of hunting and snaring, as well as habitat destruction and water shortages caused by the expanding agriculture and forestry sectors. Tony (A.C.) Pooley started a crocodile restocking programme in 1966, which in combination with legal protection (effective from 1969) resulted in a significant recovery by the 1990s (Pooley, Reference Pooley1982; Calverley & Downs, Reference Calverley and Downs2014a; Harvey & Marais, Reference Harvey, Marais, Bates, Branch, Bauer, Burger, Marais and Alexander2014).
However, the Ndumo Game Reserve population decreased by 38% during 1993–2009, possibly because of an increase in illegal killings and disturbance facilitated by the removal of the eastern boundary fence in May 2008 (Calverley & Downs, Reference Calverley and Downs2014b); in 2009 the population comprised 516 crocodiles.
The first aerial survey of Lake St Lucia (1972) recorded 356 crocodiles > 1 m (Pooley, Reference Pooley1982). The lake was restocked with juvenile crocodiles during 1967–1976 (Pooley, Reference Pooley, Bruton and Cooper1980), and 975 individuals were counted in 1993. The population remained stable until 2008 but has since declined, possibly as a result of prolonged drought (Combrink, Reference Combrink2014).
With the exception of the 132 km2 Pongolapoort Dam in KwaZulu-Natal (Champion & Downs, Reference Champion and Downs2017), declines have been reported for all major crocodile populations in South Africa. As a result, Nile crocodiles are categorized as Vulnerable in the country (Harvey & Marais, Reference Harvey, Marais, Bates, Branch, Bauer, Burger, Marais and Alexander2014).
In eSwatini, extensive habitat has been converted for agriculture, and illegal hunting remained rife into the 1980s. In 1992 King Mswati III ordered a new draft of the Game Act (1953, as amended), passed in 1993, which introduced the first legal protection for crocodiles outside protected areas (Big Game Parks, 2017). No crocodile population data are available for eSwatini, but the species is considered to be Vulnerable there (Harvey & Marais, Reference Harvey, Marais, Bates, Branch, Bauer, Burger, Marais and Alexander2014).
Ezemvelo KZN Wildlife is the responsible authority in KwaZulu-Natal, South Africa. They remove rather than kill crocodiles whenever possible, and do not erect or maintain protective structures, or pay compensation for attacks outside protected areas. The Mpumalanga Tourism and Parks Agency and Limpopo Province's Department of Economic Development, Environment and Tourism deal with attacks in the interior. Crocodiles are protected under provincial conservation legislation.
In Mpumalanga problem crocodiles are trapped and released in either the 24 km2 Loskop Dam or Flag Boshielo Dam (H. Botha, pers. obs.), or sold to commercial farms. The Limpopo authorities have issued tenders licensing trophy hunters to control damage-causing crocodiles but few have been destroyed in this way. Fences have been built at some dams (Anthony et al., Reference Anthony, Scott and Antypas2010).
In eSwatini, Big Game Parks is mandated by the office of the King to manage wildlife in the royal parks and outside protected areas. Their policy is to capture and remove confirmed problem crocodiles. No protective structures are built, and compensation is not paid (Mick Reilly, Big Game Parks, pers. comm., 2014).
Information on attacks by Nile crocodiles was obtained from the personal archives of Tony (A.C.) Pooley and Ian Player, the St Lucia Crocodile Centre, and the Times of Swaziland archive in Mbabane, eSwatini. We searched newspaper reports (print and online), journals and popular magazines, using the search term ‘crocodile’ paired with ‘attack’, ‘bite’ or ‘victim’, in English and Afrikaans.
Only details of attacks by wild crocodiles that resulted in injury or death were included. Alleged attacks that were not witnessed or that lacked forensic proof were excluded. Fatal attacks include attacks from which victims died later as a result of injuries sustained. Demographic categories for age were child (< 16 years) and adult (≥ 16 years; sometimes exact age data were missing but victims were described as children or adults), and 5-year age categories were used for cases for which exact age data were available.
We excluded crocodile attacks prior to 1949 because of a paucity of reliable data. It is likely that during the study period some attacks involving minor injuries went unreported. In remote regions, particularly areas to which people were relocated by Apartheid authorities, some serious attacks may have gone unreported.
We tested for temporal trends by constructing Poisson generalized linear models of attack frequency as a function of year. Quasi-Poisson generalized linear models were constructed when the data were overdispersed. We tested for differences between victim demographic categories using χ 2 tests.
Literature searches returned 132 print newspaper stories and six magazine features for South Africa, and 15 print newspaper stories of attacks in eSwatini. Sixteen online stories were retrieved through Google searches and searches of digital archives of five South African newspapers (in English and Afrikaans), and nine stories from the digital archives of two Swazi newspapers. Tony Pooley's archive included personal records of 73 attacks in the study region, and Ian Player's archive included 15 newspaper reports of attacks.
The dataset comprises 214 crocodile attacks for the period 1949–2016: 185 attacks in South Africa and 29 in eSwatini. In South Africa, attacks have been recorded in 13 district municipalities but only five districts have more than five attacks recorded. Fig. 1 shows the spatial distribution of attacks, highlighting the provinces and water bodies with the highest numbers of recorded attacks.
The majority of crocodile attacks occurred in natural water bodies (Table 2), with 69% of attacks in rivers or streams (n = 148), 15% in lakes or pans (n = 33), 3% in the St Lucia estuary (n = 7) and 1% in wetlands (n = 2). Attacks have also been recorded in man-made water bodies, with 8% in dams of various sizes (n = 18) and 2% in canals or drains (n = 5). The exact location of one attack (< 1%) is unknown.
A Poisson generalized linear model indicates no significant trend in annual attack frequency across South Africa (eSwatini attack data for before 2000 are patchy) during 1949–2016 (estimate = 0.002, z = 0.529, P = 0.597). However, the records indicate some temporal trends in particular districts and locality types. For example, there has been a significant decrease in the number of attacks reported in the Umkhanyakude district since 1949 (estimate = −0.026, SE = 0.007, t = −3.996, P < 0.001), with only one attack recorded since 2010. In comparison, there has been a significant increase in the number of attacks reported in Limpopo Province's Mopani district (estimate = 0.072, SE = 0.024, z = 2.997, P = 0.003) and Vhembe district (estimate = 0.047, SE = 0.017, z = 2.756, P = 0.006). During 2006–2016 eight attacks were recorded in Mopani district and six in Vhembe district, accounting for 32% of all attacks recorded during this period (n = 44). Fig. 2 shows the number of attacks recorded by 5-year period for the five municipal districts in South Africa with the highest number of attacks. There have been temporal trends in attack frequency for different water body types. For example, there was a significant increase in the frequency of attacks reported in dams between 1949 and 2016 (estimate = 0.065, SE = 0.018, z = 3.703, P < 0.001), with 64% of all reported attacks in dams occurring post 2000, and a record high of six attacks recorded in dams in 2014.
Seasonality of attacks
One hundred and ninety records include the activity the victim was engaged in when the attack occurred. Of these, most victims (31%) were attacked while swimming or bathing (n = 59), followed by fishing (n = 41, 22%), doing domestic chores at the water's edge (n = 35, 18%), crossing the water (n = 30, 16%), or other (n = 25, 13%). There was a significant relationship between gender and activity when attacked (χ 2 = 59.363, df = 4, n = 190, P < 0.001). The data indicate a relationship between age (adult ≥ 16 > child) and activity but this was not significant (χ 2 = 8.625, df = 4, n = 124, P = 0.071). Table 3 summarizes the number and per cent of attacks for each activity by age and gender. Of the reports including exact age information (n = 139), 68 attacks (49%) were on adults (≥ 16 years), and 71 attacks (51%) were on children (< 16 years). A greater proportion of the attacks on children were fatal (54%), compared with adults (35%), and this difference is significant (χ 2 = 3.962, df = 1, n = 139, P = 0.047). Fig. 4 shows the distribution of victim ages, subset by fatal and non-fatal attacks.
The analysis and interpretation of long-term data on crocodile attacks provide valuable information on the seasonality of attacks, locations of attacks and demographics of attack victims. Outcomes from this research could help focus mitigation efforts, provided that local contexts are taken into account, as outlined below.
Seasonality of attacks
Three possible explanations for the seasonality of crocodile attacks have been offered: increased dispersal and encounter rates resulting from high rainfall and water levels (wet season), temperature (crocodiles are ectothermic and thus more active when it is warmer), and increased aggression during the breeding season (Pooley et al., Reference Pooley, Hines, Shield and Ross1992; Pooley, Reference Pooley2015a). Crocodile attack incidence tracks high mean water levels (where data exist) and high monthly mean rainfall (particularly in the interior of South Africa). However, preliminary studies indicate there is no significant relationship between individual attacks in the study region and high rainfall and water-level conditions recorded for dates of attacks only (Potter, Reference Potter2014; Powell et al., Reference Powell, Versluys, Williams, Tiedt and Pooley2019). In neighbouring Mozambique the short-term data (1997–2003) of Le Bel et al. (Reference Le Bel, Murwira, Mukamuri, Czudek, Taylor, La Grange and Lapez-Pujol2011) indicate that most attacks occur in the dry season.
Monthly mean daily temperature is the strongest environmental predictor, with most attacks occurring at temperatures of ≥ 16 °C (Potter, Reference Potter2014; Powell et al., Reference Powell, Versluys, Williams, Tiedt and Pooley2019; see Lance, Reference Lance2003, on American alligators). This effect of temperature could be explained by crocodiles’ decreased physiological maintenance costs under cooler conditions and, conversely, increased activity levels and food requirements under warmer conditions, as suggested for saltwater crocodiles Crocodylus porosus in Australia (Manolis & Webb, Reference Manolis and Webb2013, p. 100).
The seasonality of crocodile attacks cannot be explained based on biophysical variables and crocodile behaviour alone because of the overlap between human and crocodile activity (e.g. the seasonality of aquatic activity of both crocodiles and people). Nearly half of attacks in the study region occurred on weekends and holidays, suggesting human activity patterns are influential. Although the climate varies slightly between the interior and the coastal regions where crocodiles occur, the peak attack season is the same: December–March (Fig. 3). More data on local behaviour patterns of crocodiles and people in hotspots for crocodile attacks would contribute to more effective mitigation measures. For instance, it is known that crocodiles congregate in lakes in Ndumo Game Reserve and on the eastern shores of Lake St Lucia in winter. Larger individuals disperse outside the protected areas or around the lake system in the summer. Thus in recreational areas around the Lake St Lucia system, notably the estuary, there are seasonal overlaps between the distributions of larger crocodiles and people (Pooley, Reference Pooley1982; Combrink, Reference Combrink2014).
Spatial distribution of attacks
Our data indicate that historically most attacks occurred in waterways linked with major crocodile populations, namely the St Lucia Lake system, Ndumo Game Reserve and Kruger National Park. This situation was exacerbated in South Africa by the Apartheid Homeland or Bantustan system under which Africans were relocated to remote rural regions with little infrastructure (Beinart, Reference Beinart2001). Wildlife conservation areas persisted where land was undesirable for farming and settlement (McCracken, Reference McCracken2008). Most crocodile attacks occurred where so-called native reserves bordered or were crossed by rivers linked with protected areas. In KwaZulu-Natal Province this includes former native reserves on the Hluhluwe, Nyalazi and Umfolozi rivers. In Mpumalanga, attack hotspots persisted where the former homeland of Gazankulu was located on the western border of Kruger National Park. In Limpopo Province, attacks persisted where the former homeland of Venda straddled rivers flowing into the Park.
Crocodile attack incidence does not track fluctuations in crocodile numbers (decline followed by a small recovery during 1957–1972, the period of peak attack incidence, and rapid recovery and stabilization during the period of reduced attacks, in the late 1980s–1990s). Shifts in the distributions of crocodiles as a result of environmental events (e.g. droughts, floods) and anthropogenic interventions (dam building, pollution, habitat transformation, direct persecution), and rapid urbanization of South Africa's human population since the 1980s seem to be of greater consequence (Pooley, Reference Pooley, Beinart, Middleton and Pooley2013).
An upwards trend in attacks in the interior since 2000 may be the result of encounters with crocodiles in unexpected locations, notably dams, where they have moved in response to the drying up of perennial rivers, disturbances along the riverbanks, or pollution in rivers (Botha et al., Reference Botha, van Hoven and Guillette2011). Some have been displaced through habitat loss caused by the widespread raising of the walls of larger dams in South Africa, notably Flag Boshielo Dam, and the 150 km2 Massingir Dam in Mozambique, both on the Olifants River (Harvey & Marais, Reference Harvey, Marais, Bates, Branch, Bauer, Burger, Marais and Alexander2014, p. 88).
Commercial and subsistence fishing on a number of dams inhabited by crocodiles are an ongoing safety concern (Tapela et al., Reference Tapela, Britz and Rouhani2015). The key dams include Flag Boshielo Dam and the 0.75 km2 Makuleke Dam, and possibly Middle Letaba Dam (18.79 km2, in Limpopo Province) and Driekoppies Dam (18.7 km2, in Mpumalanga).
Some of South Africa's historically most problematic rivers for crocodile attacks (including the Usutu, the Pongola and its pans, and the Mkuzi River) have had few attacks since 2000, possibly offering proxy data that few crocodiles survive outside protected areas in these river systems (Table 2). Since 2000 there have continued to be attacks in the rivers listed in Table 4 and in the St Lucia system. Attacks since 2000 are listed separately because they are of more relevance for managers, as they reflect more recent trends in attack incidence. Dams in or near waterways listed in Table 4, and in addition the Limpopo River (where crocodile populations may be recovering), should be regarded as higher-risk areas, as they have more recent records of attacks. Of the 17 attacks recorded in eSwatini during 2000–2016, 65% occurred in the Usutu (or Lusutfu) River, and 29% in the Mbuluzi River.
1The Nseleni River is linked to the Goedertrouw Dam.
The finding that it is mostly males (65%) that have been attacked in this region contradicts the assumption that in Africa women and girls are disproportionately at risk because of their domestic tasks at the water's edge (e.g. Lamarque et al., Reference Lamarque, Anderson, Fergusson, Lagrange, Osei-Owusu and Bakker2009, p. 19). The numerous attacks on females, most performing domestic chores, along the Pongola floodplain system in the 1960s and 1970s are atypical. Census data reveal a higher proportion of women than men resident in this region in this period, with men away working as migrant labourers (Smedley & Ribeiro-Tôrres, Reference Smedley and Ribeiro-Tôrres1979). Our data show that domestic chores have been a less important factor in the wider region since c. 2000, reflecting both the crocodile's contracting range and improved water provision in some rural areas.
A key finding is that 51% of victims were aged 0–15 years. That 62% of victims were aged 0–20 and the largest adult category was 21–30 (19%) may simply reflect the demography of the country (median age 26). Nevertheless, the high proportion of children, especially aged 11–15 years, 72.5% of whom were boys, suggests this should be a focus for concern and education.
The overall fatality rate from attacks was 49% (1949–2017), comparable to the findings of Thomas (Reference Thomas2006) for the Okavango Swamps (55%) and Maheritafika et al. (Reference Maheritafika, Robsomanitrandrasana, Rabesihanaka, Rafenomanana, Ravaoarimalala and Andrianjaratina2016) for Madagascar (56%), but notably lower than the 63% estimated by Fergusson (Reference Fergusson2004) for Africa in general. However, 57% of attacks on children (0–15 years) were fatal (n = 65) and 54% of victims aged 0–20 years were killed (n = 79), in comparison with 40% of attacks on those aged ≥ 21 years (n = 45). Fatality rates were influenced by whether the victim was accompanied or alone, and the size (length) of crocodile involved, as well as the size of the victim. Smaller victims (children) are more vulnerable to fatal attacks, as found in an analysis of factors affecting the survival of victims of attacks by saltwater crocodiles (Fukuda et al., Reference Fukuda, Manolis, Saalfeld and Zuur2015). We found that, of those adults who escaped death, 57% (29) escaped without help and 43% (22) were rescued, whereas only 35% (11) of children escaped unaided and 65% (20) were rescued.
Only 15 crocodiles involved in attacks were measured accurately, and therefore size data could not be used as an accurate variable. Furthermore, most crocodile counts have been made from fixed-wing aircraft (X. Combrink, pers. obs.), so there are no general data on the size of crocodiles to facilitate comparison of the number of fatal attacks with the proportion of large crocodiles in wild populations. Comparing the length of crocodiles with fatality/non-fatality outcomes is complicated by age and size of victim, and whether there were rescuers present. Better data would be required to assess the relationship between size and deliberate attacks on people by crocodiles in this region, although data from alligators and saltwater crocodiles suggest that individuals measuring > 1.8 m can inflict serious injuries, and individuals measuring ≥ 2.4 m carry out fatal attacks (Caldicott et al., Reference Caldicott, Croser, Manolis, Webb and Britton2005; Fukuda et al., Reference Fukuda, Manolis, Saalfeld and Zuur2015).
Overall, most victims were swimming, bathing or fishing, but disaggregating data on activity of victim when attacked by age and gender reveals distinct profiles (Fig. 5). Thomas (Reference Thomas2006) and Wallace et al. (Reference Wallace, Leslie and Coulson2011) found similar results in the Okavango Swamps (Botswana) and lower Zambezi (Zambia), respectively. Our data show that until the 1980s most victims were performing domestic chores or crossing water when attacked, but since then these activities have been superseded by swimming and fishing.
For high risk areas there are a number of mostly low-cost actions that can be taken. Local authorities could facilitate safe water crossings, and safe access to water for swimming (particularly near rural schools) or domestic needs, including alternatives such as water tanks, piped water and protective enclosures.
Provincial conservation authorities and district municipalities could create, equip and train teams to capture and remove problem crocodiles. Where such teams already exist, it would be helpful to make them known to the public. If departmental resources are limited, a system of licensing private individuals (as in the USA) could be trialled (Dutton et al., Reference Dutton, Waller, Carbonneau, Hord, Stiegler and Woodward2014; King & Elsey, Reference King and Elsey2014). Some commercial crocodile farmers already provide this service on an ad hoc basis. Removing crocodiles requires the creation of clear protocols for disposing of captured crocodiles.
Educating children should be a priority, particularly in identified high-risk areas. Outreach activities could be supported with existing materials (Pooley, Reference Pooley2015b; Pooley, Reference Pooley2017) that provide information on crocodile biology and behaviour, their ecological and conservation importance, as well as advice on avoiding and responding to attacks.
Provincial conservation authorities should appoint knowledgeable spokespersons to brief the public in the event of a crocodile attack (or alleged attack). The accuracy of reporting would be improved by keeping detailed records of attacks, and building better communication between police, coroners and conservation authorities to ensure accurate information on causes of death are reported. In South Africa, where crocodiles are farmed but not ranched (i.e. captive bred but not sourced from the wild) and there is no link between farming and the country's wild populations of crocodiles, and in a region where taboos against the eating of crocodiles have recently been overturned (Viljoen, Reference Viljoen2014; Zulu, Reference Zulu2015), tolerance for wild crocodiles should not be taken for granted.
SP thanks Vince Egan, Tal Fineberg, Mick Reilly and Freek Venter for information on management policies. We are grateful for constructive comments from Martin Fisher and two anonymous referees. Fig. 1 was created by Martin Fisher. SP is funded by the Lambert Bequest at Birkbeck University of London.
Collection and collation of attack data and historical information, writing: SP; contribution of insights on crocodile populations and management: XC, HB; statistical analyses and figures: GP.
Conflicts of interest
This research abided by the Oryx guidelines on ethical standards.