In South Africa, where savanna elephant (Loxodonta africana) populations and large trees have co-existed on fenced-off landscapes such as within the Kruger National Park (KNP), the impact of elephants on large trees is viewed by some as unsustainable (Asner et al. 2016; Edge et al. 2017). State officials and conservation managers are continuously involved in discussions concerned with managing the impacts of elephants in an attempt to protect large trees and maintain the natural system in an ‘ideal’ state, usually based on colonial historical records (Kerley et al. 2008) or driven by tourist perceptions or expectations of an aesthetic landscape (Edge et al. 2017). These records, however, coincide with the near-extermination of elephants in South Africa from over 100 000 individuals, as a consequence of recreational and subsistence hunting (Whyte 2001). Furthermore, an outbreak of rinderpest in the 19th century resulted in a herbivore population crash, decreasing the numbers of smaller browsers that would usually feed on tree seedlings (Skarpe et al. 2004).

In the post-culling era, elephant numbers have steadily increased (Ferreira, Greaver & Simms 2017), although their distribution ranges have decreased as a result of fencing off protected areas (Whyte 2001). South Africa is therefore dealing with the complexities of managing increasing elephant numbers in restricted ranges or landscapes with homogenous resource distribution, where an ‘ideal’ environmental state is the conservation goal to benefit tourism. The present management strategy of the South African National Parks is focused on how elephants use their resources in relation to their distribution rather than absolute elephant numbers (Ferreira et al. 2017; SANParks 2012). The interplay between elephants and large trees is complex to the extent that no single species or factor can be thought of as solely responsible for what is essentially a multifaceted ecological process (Greyling 2004).

The aims of this review were to evaluate management methods to protect trees from elephant effects, in both small and large protected areas. We thereby firstly discuss the importance of large trees and how elephants interact with them, followed by a critical evaluation of the past management practices to reduce elephant numbers for protection of large trees. Lastly, we discuss the available management options to address the concern for the loss of large trees within a framework that highlights the various factors that potentially cause concern and the complexities of addressing a single mechanism within a heterogeneous landscape.

Large trees (> 5 m in height) have a particular aesthetic significance to both the public and conservation managers (Shannon et al. 2008), being considered as important landscape features where they occur (Edge et al. 2017; Owen-Smith et al. 2006). The addition or reduction of large trees across a threshold is thus an indication of changing ecological states (Dublin, Sinclair & McGlade 1990). Importantly, a change in ecological state is not necessarily negative, as environments of a degraded nature may be recovering from past management practices (Young 2000).

Large trees have important ecosystem functions and play a significant role in the cycling of nutrients (Ludwig, De Kroon & Prins 2008), reduction of evapotranspiration and conductance for species which occur below the crowns of trees (Belsky 1994), as well as providing a forage source for fauna (Kerley & Landman 2006). Other than providing food and shelter, large trees provide nesting sites for both vultures and raptors (Vogel et al. 2014).

Elephants impact trees in a variety of means, ranging from bark-stripping and branch breakage, towards more destructive impacts such as uprooting and stem snapping (Greyling 2004). The type and intensity of elephant impacts are determined by a variety of factors, including tree species (Shannon et al. 2008), tree height (Cook et al. 2017) and elephant sexual dimorphism (Greyling 2004). Elephants have particular forage preferences for species, including the marula (Sclerocarya birrea), knobthorn (Senegalia nigrescens) and red bushwillow (Combretum apiculatum) trees (Shannon et al. 2008). Tree height can exacerbate elephant impact. For example, marula trees between 5 m and 11 m in height are more vulnerable to uprooting and stem snapping when compared with trees above 11 m (Cook et al. 2017). Elephant bulls, being larger bodied, have heavier impacts on food species compared to cows (Greyling 2004).

More recent studies focused on the positive ecological/biodiversity consequences of elephant disturbance regimes (Guldemond, Purdon & Van Aarde 2017). The ability of elephants to modify and alter woody structure, however, can increase habitat complexity by modifying the woody structure (Sianga et al. 2017). The feeding habits of elephants can lead to ‘hedging effects’ on certain tree species (Lombard et al. 2001; Styles & Skinner 2000), which increase trees’ vulnerability to both insect invasions and fire damage due to a reduction in tree height, volume or exposed inner tissues (Jacobs & Biggs 2002). The impact of elephants above a certain threshold can lead to loss of particular tree species within a landscape, as well as a decrease in the landscape’s structural diversity (Asner et al. 2016; Eckhardt, Van Wilgen & Biggs 2000; McCleery et al. 2018).

Some studies have found that elephants can have a negative influence on biodiversity (Kerley & Landman 2006; McCleery et al. 2018) and woody canopy cover within protected areas (Asner et al. 2016). Extensive elephant feeding habits can displace black rhinoceros (Kerley & Landman 2014), which leads to increased predation of meso-herbivores owing to the opening of habitat (Tambling et al. 2013) and decreased foraging opportunities for micro-herbivores (Hrabar & Du Toit 2014). McCleery et al. (2018) have found that the combination of fires and elephant impact can lead to the removal of large trees in southern African savannas, decreasing the diversity of birds, bats and terrestrial small mammals.

Elephants play an important role in the propagation of large trees, in terms of both dispersal and germination of tree seeds. Mature bulls can transport seeds to a maximum distance of 65 km away from their source (Bunney, Bond & Henley 2017). Concurrently, the acid treatment in their digestive system promotes the germination of seeds (Cochrane 2003), while the deposition of seeds in parcels of organic mulch further promotes germination (Cochrane 2003). These processes are promoted by the inefficiency of elephants’ digestive systems and their ability to produce 14–20 droppings per day with each defecation weighing around 11 kg, thereby enabling them to comfortably produce at least 150 kg of wet dung a day (Owen-Smith 1988).

Elephants also modify the landscape as ecosystem engineers, major tree pruners, effective composting agents and seed dispersers, thereby increasing biological diversity on a micro-scale (Valeix et al. 2011). Depending on the level of impact, this process has been found to increase the nutrient quality of impacted plants, escalate the overall biodiversity of the landscape and promote a mosaic of elephant impact-tolerant plant species closer to water points (Gaylard 2015; Kohi et al. 2011).

Hence, landscapes used by elephants will have altered vegetation structure, which leads to a higher diversity of ants, reptiles and frogs through the creation of micro-habitats (Nasseri, McBrayer & Schulte 2010; Palmer et al. 2008). The pruning activities of elephants can stimulate plant growth and also stimulate shoot production in height levels accessible for other browsers, thereby promoting foraging opportunities for other herbivores (Mograbi et al. 2017). Thus, further research is required on the management methods available for reducing elephant impact on the environment, and how biodiversity responds to changes in levels of elephant impact over time.

In the past, the biodiversity objectives of the KNP were realised through the reduction in elephant numbers by culling operations (Whyte 2001). However, elephants represent only one facet of a multifaceted ecological process when it comes to survival rates of large trees (Figure 1). Elephant density alone does not explain demographics of large trees (Guldemond et al. 2017). This is in part due to many factors involved in the survival and recruitment rates of large trees (Helm, Scott & Witkowski 2011), as well as the spatial distribution of elephants (Sianga et al. 2017).

Elephant encounter rate and residence time with large trees are influenced by their density, as well as their spatial distribution (O’Connor, Goodman & Clegg 2007). The spatial distribution of elephants in turn depends on a variety of factors. These include both forage (Young, Ferreira & Van Aarde 2009) and water availability (De Knegt et al. 2011), rainfall (Birkett et al. 2012), the thermal environment (Kinahan, Pimm & Van Aarde 2007), fire (Woolley et al. 2008), roads when used as footpaths by elephants or because of increased forage productivity on their verges (Smit & Asner 2012), fences (Vanak, Thaker & Slotow 2010), terrain ruggedness (Wall, Douglas-Hamilton & Vollrath 2006) and the social and safety benefits of the areas they occupy (De Knegt et al. 2011).

Elephants, fire, soil and elevation have been found to be the main determinants of tree extirpation, but the rates of extirpation are also dependent on the plant species being affected (Asner et al. 2016; O’Connor et al. 2007). Species such as the mopane tree (Colophospermum mopane) are able to re-coppice post-elephant impact (Styles & Skinner 2000), while the weak coppicing abilities of knobthorn trees leave them more vulnerable to elephant impact (MacGregor & O’Connor 2004). Despite current increased atmospheric carbon dioxide levels, which promote the growth of woody vegetation, there is little bush encroachment where there are elephants and high-intensity fires (Stevens et al. 2016). Overall, large tree survival rates are thus influenced by the heterogeneity of the landscape in terms of elephant use and the interplay with differences in fire intensity.

Although elephant management interventions should not be considered in isolation from other drivers influencing large tree population dynamics, a number of options are still applicable if the management objective is to protect large trees. The mitigation strategies presented here focus on the ecological consequences without considering ethical concerns and can be divided into the following three types:

Lethal methods for population reduction: Past approaches have aimed at directly reducing elephant numbers with particular reference to culling and translocation. However, controlling an elephant population at an arbitrarily placed number prevents density-dependence feedbacks from playing a natural role in reducing the population’s growth rate (Owen-Smith 2005; Robson & Van Aarde 2018). Culling of a population which is experiencing an exponential growth phase will habitually result in an increase in the birth rates of the population, as the availability of food per head is increased (Caughley 1983). Furthermore, inter-regional movements of elephants have been observed into areas where elephants have been previously culled, nullifying the intentional plan of protecting the vegetation through culling (Van Aarde et al. 2005). Definition of zones where culling operations are meant to take place have also been found to mismatch the actual spatial and temporal movements of elephants (Delsink et al. 2013).

Elephant numbers can also be affected by poaching (Wittemyer et al. 2014), disease outbreak (Grobler et al. 1995) and drought (Loveridge et al. 2006), but these population control measures remain undesirable as they are often unexpected and notoriously difficult to manage. Poaching, in particular, has led to the demise of one-third of the African continental elephant population in the space of 7 years across primarily Central and East Africa (Chase et al. 2016; Wittemyer et al. 2014). Recently, poaching has also increased in the southern African states with the KNP experiencing the highest poaching incidents in decades (Lindsay et al. 2017).

Legal hunting has been suggested as a means to control elephant numbers; however, hunting is a highly selective activity, as bulls of particular age categories and with sought-after physical traits are targeted (Stalmans, Attwell & Estes 2002). For these reasons, hunting has not been listed as a population control method as it could result in undesirable skewed sex ratios and age structures within populations (Milner, Nilsen & Andreassen 2007).

Non-lethal methods for population reduction: Non-lethal methods used to control elephant numbers include the use of contraceptives (Delsink et al. 2007) and translocation (Grobler et al. 2008). The success of contraception as a management tool to control population numbers has been shown in a number of reserves in South Africa, including the Greater Makalali Private Game Reserve and Tembe Elephant Park (Bertschinger et al. 2018).

Translocation, however, is dependent on new available space for elephants, which is a limiting factor in South Africa (Grobler et al. 2008). Depending on the scale of the operation, translocation, as a means of reducing elephant numbers, may also have the same effect on elephant growth rates as culling had in the past (Caughley 1983). Both these non-lethal methods are currently of prime importance in smaller reserves with limited options for expansion or dispersal of animals and high tourism investment.

Overall, elephant impact is regulated by resource abundance and thereby, this impact can be spatially and temporarily altered by modifying how accessible resources are to elephants, thus promoting heterogeneity (Sianga et al. 2017). Modified elephant impact, both spatially and temporally, leads to greater levels of biodiversity on a regional scale (Gaylard 2005; Sianga et al. 2017).

Surface water availability is an important driving factor behind elephant distributions (Gaylard, Owen-Smith & Redfern 2003). A scarcity of water can result in refugia localities for particular plant species (Eckhardt et al. 2000), as elephant movement ranges are naturally heterogeneous in accordance with surface water distribution (Gaylard et al. 2003). Elephant impact is thus homogenised in areas containing a high density of water points, as movement ranges are spatially and temporally reduced (Gaylard 2015). Smit, Grant and Whyte (2007) found that bulls in the KNP are more likely to occur further from natural rivers in comparison to breeding herds, making use of the availability of artificial waterholes to expand their spatial range. The KNP has used the closure and spatial distribution of artificial waterholes to manipulate elephant movement ranges, as well as their impact on vegetation (Ferreira et al. 2017; Macfadyen et al. 2019; O’Connor et al. 2007; Shannon et al. 2008). Intense levels of impact occur around the surface water points, while impact decreases as distance to surface water increases (Gaylard 2005). Sianga et al. (2017) found that large tree populations were abundant > 10 km from surface water in the Okavango Delta and Linyanti Swamps. Furthermore, the eventual scarcity of resources around a limited supply of water could lead to a situation where an elephant population that is double its size but increasing at half its rate would result in the same surplus of individuals (Owen-Smith 2005; Robson & Van Aarde 2018).

Waterhole closure in the KNP has already reduced the annual population growth rate from 6.5% to 4.2% over a 12-year period (Ferreira et al. 2017). It has, however, been suggested that a lag phase may occur between the closure of waterholes and a reduction in elephant impact on trees in the KNP (SANParks 2012). Furthermore, this method may not be viable in small protected areas where it is not spatially realistic to create gradients of elephant impact. Importantly though, reducing waterholes can also reduce the residency of water-dependent browsers of seedlings, such as impala (Skarpe et al. 2004). Reducing environmental pressures on the seedlings of large trees will aid in the recruitment of mature individuals.

Other methods used to manipulate elephants’ distribution include the use of barriers or hard boundaries. Fences are the most commonly used barriers, but a variety of other barrier types can also be implemented (SANParks 2012).

In open systems, elephants can adapt their behaviour when exposed to temporal and spatial variability of resources (Owen-Smith 2005). The colonisation of new areas by less risk-adverse bulls (Whyte 2001) could be of great value in alleviating impact on vegetation within current home ranges, as bulls exert greater levels of impact on the vegetation in comparison to cows and calves (Greyling 2004). This dispersal can be temporally delayed, as Druce et al. (2008) found that older elephant bulls in South Africa’s Phinda Private Game Reserve took 1 month to move into two neighbouring reserves following the dropping of fences, while younger bulls and breeding herds took 5 to 8 months to follow. Dispersal of elephants can be a result of increased densities of elephants and this process could be disrupted by pre-emptive culling, which does not allow source areas to reach the required densities, which would encourage dispersal. Dispersal movements can be further encouraged by ensuring that important movement corridors, identified through the movements of collared individuals, are kept open and protected (Douglas-Hamilton, Krink & Vollrath 2005).

Elephants may also avoid areas if deemed unsafe. Fear landscapes can influence elephant distribution patterns as elephants avoid areas where threats to their safety are perceived as spatially predictable, but the timing and type of threat remains unpredictable (Cromsigt et al. 2013). Indeed, the density of elephants immediately decreased in zones where culling had commenced in KNP, before elephants immigrated back to these zones within the following years (Van Aarde, Whyte & Pimm 1999). Conversely, safety benefits can thus be used as one method to attract elephants to particular areas. Hence, environmental manipulation could be applied and monitored in an adaptive management approach to encourage elephants into particular areas where their population growth rates would be limited by natural processes (Robson & Van Aarde 2018).

Tree sanctuaries can be formed from the direct protection of the resource (i.e. large trees). In addition to preserving the aesthetic importance of landscape features such as large trees, mature specimens could serve as important seed banks for future recruitment programmes (Western & Maitumo 2004).

Wire-netting tree trunks have been found to increase the survival rate of large trees as the technique essentially prevents bark-stripping by elephants. However, treated trees do remain susceptible to branch breakage, uprooting or main stem snapping, albeit at lower frequencies (Derham, Henley & Schulte 2016).

The use of African honeybees (Apis mellifera subsp. scutellata) has proved to be a highly effective, albeit costly, mitigation method for protecting individual trees from elephant impact (Cook et al. 2018). Potentially, the costs can be offset by the production of honey and the additional pollination services obtained from active beehives.

Other methods used to protect large trees involve the packing of rocks and pyramids around the base of a tree to a distance of up to 5 m from the stem (SANParks 2012). The efficacy of this technique has not been quantified in the scientific literature, although anecdotal evidence indicates that it could be effective if laid out correctly.

Importantly, methods that directly protect the individual tree will have a small spatial effect, but no lag time with regard to their effectiveness. These methods are applicable in both small and large protected areas where individual trees are in need of protection.

Lastly, the artificial propagation of seedlings of woody species favoured by elephants in exclusion experiments can be viewed as another method to increase the density of food plants while reducing encounter rates with elephants (Hofmeyr 2005). This method, while not yet tried in South Africa, provides an alternative option to lowering elephant densities to achieve the same effect. Experimental exclosures in the Mapungubwe National Park have demonstrated the potential of artificial propagation as a means of assisting big tree regeneration (Scholtz 2007). Artificial propagation has further been used to help diminish human–elephant conflict in Thailand (Van de Water & Matteson 2018).

Each of the proposed mitigation strategies have certain advantages and disadvantages associated with them (Table 1), many of which also need to be evaluated from an ethics perspective (broadly listed from most severe to least severe ethical concern in Figure 1 and Table 1), as increased interference and disruption of intact social systems can occur when moving from strategy three to strategy one. However, elephants are continually exposed to a wide range of stresses across the landscape, and their response to such stressors can be used to evaluate the ethics and effectiveness of proposed mitigation strategies. Understanding how elephants respond to human-induced fear (Douglas-Hamilton et al. 2005) and resource manipulation (Purdon & Van Aarde 2017) will be important when mitigation strategies are implemented.

In the Great Limpopo Transfrontier Park (GLTP, 37 572 km²), all three abovementioned mitigation strategies are being carried out across the system. The density of elephant poaching is the highest in the Limpopo National Park (Lunstrum 2014), which would affect elephant densities and create sink areas driven by fear (Van Aarde & Jackson 2007). These landscapes of fear influence both the density and spatial distribution of elephants concurrently, albeit undesirably because poaching is known to disrupt the social structure and demographics of the population and is hard to control (Jones et al. 2018). Future research is required to evaluate the effect that poaching in the Limpopo National Park and the slow increase within the KNP will have on the higher elephant population density and vegetation composition found within the KNP and neighbouring reserves. Key to elephant management within the KNP would be to increase the safety benefits within the neighbouring Limpopo National Park to which elephants will naturally respond in keeping with seasons and as this reserve only has a seasonal and natural spread of available surface water.

Historically, the KNP reduced elephant numbers through culling and although these practices are no longer implemented partly because they were unsuccessful in achieving the desired outcome – for example the protection of large trees – the KNP is experiencing some of the highest poaching records in its history (Lindsay et al. 2017). However, as the KNP still has an expanding elephant population (Ferreira et al. 2017), management is preferentially focusing on the second mitigation strategy by primarily controlling elephant distribution through artificial waterhole closure and have already closed two-thirds of the 365 artificial waterholes and 50 earth dams since it began its water stabilisation programme in the early 1930s (Purdon & Van Aarde 2017; SANParks 2012).

The Associated Private Nature Reserves (APNR) to the west of the KNP have a saturated water landscape and also a high density of landowners and lodges (Peel 2009). The landscape use implemented in this self-funded protected area may thus not lend itself to the aforementioned mitigation strategies. The APNR have, however, successfully implemented various mitigation strategies aimed at directly protecting the resource (large trees) such as wire-netting (Derham et al. 2016), African honeybees (Cook et al. 2018) and rocks and pyramids (Henley & Cook 2018). In smaller reserves (< 1000 km²), elephant range size is often a function of the size of the reserve (Roux 2006).

The manipulation or closure of waterholes may have a limited effect on reducing elephant impact on large trees in the APNR due to the immense number of waterholes distributed across private properties. As previously discussed, translocation may be a temporary option, although limited by the number of reserves that can support elephant populations (Grobler et al. 2008). Smaller reserves should focus on methods that directly protect large trees from elephant impact (Table 1) and investigate the potential of contraception for managing their elephant numbers (Table 1).

Botanical reserves that exclude elephants from particular floral communities within smaller reserves can also ensure the survival of large tree species and their seed banks (Lombard et al. 2001). Efforts should also be focused on large tree regeneration and recruitment, by considering factors such as seed predation (Helm et al. 2011), seedling herbivory (Skarpe et al. 2004) and fires (Smit et al. 2016), which are known to affect large tree survival, even in the absence of elephants (Helm & Witkowski 2012).

Conservation managers are faced with the difficulties of fulfilling their mandate of protecting biodiversity in human-dominated landscapes and what biodiversity should be protected where objectives may be in conflict. Can large trees and elephants coexist and what strategies should managers implement to optimise biodiversity goals? These strategies should reflect on (1) implementing limited or no interventions when ecological processes are playing out (Biggs et al. 2008), (2) restoring ecological processes and opportunities if the landscape is termed ‘degraded’ (Wassenaar, Ferreira & Van Aarde 2007), and (3) mimicking desired ecological processes if restoration is not possible (SANParks 2012). Hence, an adaptive management plan needs to consider and continually evaluate whether the proposed mitigation strategy will lead to the desired effect with:

The three possible mitigation strategies available to potentially protect large trees can each be evaluated given the size of the reserve and historical perspectives. Protected areas such as the KNP, which forms part of a large open system, have opted for environmental manipulation as the management strategy. Although environmental manipulation may be viable for a reserve of this size, it may not prove appropriate in smaller protected areas. Here, a combination of directly protecting the resource (large trees) from elephant impact, in combination with translocations or contraceptive programmes, may prove more appropriate.