Elephants are allogenic ecosystem engineers (Jones, Lawton & Shachak 1994) in that they substantially modify the environment through their unique foraging and feeding habits (Laws 1970). In areas where elephants are confined by fences and human settlements, their population density can increase locally, leading to extensive modification of habitat that can potentially have negative consequences on ecosystem processes and other organisms (Dublin & Hoare 2004). For example, Vogel et al. (2014) proposed that elephants could affect vulture and raptor nesting sites by allowing opportunities for insect and fungal attack after bark stripping, and therefore reducing tree survival. Guldemond and Van Aarde (2008) list a number of studies where high densities of elephants were found to negatively impact woody vegetation, but caution that the findings are site-specific and dependent upon environmental conditions. Within the Kruger National Park (KNP), managing elephants in relation to vegetation changes has been a longstanding topic of concern (Ferreira et al. 2011; Gaylard & Ferreira 2011; Young, Ferreira & Van Aarde 2009). Large trees are of particular concern, because elephants are one of only a few biotic forces that can directly and rapidly modify this key feature of the savanna landscape (Asner et al. 2015; Laws 1970; Morrison, Holdo & Anderson 2015). Large trees play an important role in the biogeochemical cycles of the savanna as well as indirectly affect the distribution of numerous sympatric species that use the trees for refuge, shade, nesting, food, and other activities (Bernhard-Reversat 1982; Nasseri, McBrayer & Schulte 2010; Vogel et al. 2014).

Elephants can affect trees in a variety of ways, including bark stripping, breaking branches, breaking the main stem, or uprooting the entire tree (Boundja & Midgley 2010; Henley 2013). Bark stripping and branch breaking expose trees to insect attack and greater damage from fire, both of which may result in tree mortality (Helm et al. 2011; Vogel et al. 2014). Trees are particularly vulnerable to a type of bark stripping known as ringbarking, where the bark is removed around the entire circumference of the tree (Helm et al. 2011; Ihwagi et al. 2009). Tree species differ in their vulnerability to impact because elephants selectively browse for particular species or because of certain characteristics that could predispose a tree to extirpation (Boundja & Midgley 2010; O’Connor, Goodman & Clegg 2007). Boundja and Midgley (2010), for example, found that elephants preferred larger trees and therefore targeted larger, less abundant species in Hluhluwe-Imfolozi Park, South Africa. Elephants are also known to favour trees with certain chemical composition, choosing to browse trees with high levels of leaf nutrients such as calcium, magnesium, potassium, and protein (Holdo 2003). Although extensive branch breaking and felling can alter the form of a tree, species that readily coppice after such events can survive and continue to grow if their roots remain intact (Eckhardt, Van Wilgen & Biggs 2000; Henley 2013; Ihwagi et al. 2009). However, the continuing loss of habitat for elephants because of expanding human populations and/or fencing compresses their range and confines them to certain areas for unnaturally long periods of time. This often results in continuous and persistent impact that alone or in combination with other factors such as fire or insect attack leads to increased risk of tree mortality (Boundja & Midgley 2010; Mapaure & Moe 2009; Van Aarde, Jackson & Ferreira 2006).

Many conservation management plans focus on reducing elephant numbers in order to reduce impact, often assuming a direct linear relationship between the two (Van Aarde et al. 2006). However, some research has suggested this is not the case and that other factors may be involved (Young et al. 2009). Environmental manipulation, translocation, contraceptives, and culling have all been suggested and used at various times in South Africa’s history to reduce elephant densities (Du Toit, Rogers & Biggs 2003; Van Schalkwyk 2008). Alternatively, management could focus on protecting the trees themselves. Wrapping wire netting around the trunk of a tree is one such technique that was first employed in Kenya (Gordon 2003). If wire netting decreases elephant impact in other areas and over the long term, it could serve as a cost-effective and ecologically valuable solution to continuous elephant activity in confined areas. The aim would not be to protect all trees of a certain species and thereby potentially increase the impact on other, unprotected species. Instead, the technique can be used to protect individual trees or areas of specific species to ensure there will always be enough mature specimens to populate the surrounding landscape with their seeds.

The objective of this study was to assess the effectiveness of wire netting in reducing elephant impact to three relatively abundant, iconic, larger (> 5 m) tree species, namely Sclerocarya birrea, Acacia nigrescens, and Lannea schweinfurthii. Wire netting is designed to reduce bark stripping, but we examined all forms of impact by elephants to determine if there was a relationship between wire netting and elephant impact in general.

Elephant impact was common in this study with 2381 out of 2668 (89%) trees being impacted by elephants in some way or the other (Table 2). A total of 74 unprotected trees (5.8%) and 213 protected trees (15%) did not have any elephant impact (Table 2). In total, 25% of unprotected trees were bark stripped by elephants, while only 1.7% of wire-net-protected trees were bark stripped by elephants (Table 2). DBH was not significant in determining the likelihood of bark stripping (P = 0.08); however, wire netting (P < 0.001) and species (P = 0.03) were significant (Table 3). Acacia nigrescens had significantly more bark stripping than L. schweinfurthii (G = 36, df = 1, P < 0.001), which had significantly more bark stripping than S. birrea (G = 7.2, df = 1, P < 0.01; Figure 3a). There was no significant interaction between species and whether a tree was protected by wire netting, meaning that regardless of species differences, wire netting was effective at reducing the number of trees that were bark stripped (P = 0.75; Table 3).

In addition, the severity of bark stripping was reduced on trees protected by wire netting (G = 55, df = 8, P < 0.001) (Figure 4). The number of trees in all impact score categories was reduced with wire netting, and no tree wrapped with wire experienced impact in the highest two categories, 9 and 10 (Figure 4). Only three trees (0.2%) protected with wire netting were bark stripped more than 50% of the circumference of their trunk, compared with 85 (6.0%) without wire netting. In addition, no tree with wire netting was ring barked (i.e. class 10; 100% bark stripped), compared with 23 unprotected trees.

The presence or absence of wire netting was not a significant factor for explaining the occurrence of elephant impact other than bark stripping (branch breaking P = 0.49; felling P = 0.51; Table 3; Figure 3b). Branch breaking was recorded for 69% of unprotected trees and 64% of protected trees (Table 2). DBH had a significant effect on the occurrence of branch breaking (P = 0.001; Table 3), with smaller trees more likely to have their branches broken by elephants (t = 3.80, df = 1788, P < 0.001). In addition, species was an important factor influencing branch breaking (P < 0.001; Table 3). Acacia nigrescens had significantly fewer branches broken compared with S. birrea (G = 346, df = 1, P < 0.001) and L. schweinfurthii (G = 208, df = 1, P < 0.001) (Figure 3b). There was no significant difference in branch breaking between S. birrea and L. schweinfurthii. DBH was a significant factor in influencing whether a tree had been felled (P < 0.001; Table 3). Felled trees were significantly smaller in DBH than trees that remained standing (t = 11.3, df = 954, P < 0.01). Although 15% of trees without wire netting were felled and only 7% with wire netting suffered from this impact, once DBH was accounted for, species and treatment were not significant factors (Table 3).

Although wire netting was not successful at significantly reducing the proportion of trees that had their branches broken or were felled by elephants, it did have a significant influence on the distribution of the level of impact (G = 73.8, df = 9, P < 0.001; Figure 5). Trees that were protected with wire netting were more likely to experience lower levels of elephant impact than unprotected trees. Sixty two per cent of the time the highest impact recorded was branch breaking. Therefore, it appears that wire netting reduced the degree of branch breaking in trees that were impacted in this way by elephants.

Wire netting was effective at reducing the relative frequency and severity of bark stripping by elephants in this study. Since high levels of bark stripping are known to affect tree survival (Helm et al. 2011; Ihwagi et al. 2009), wire netting could prevent mortality by decreasing both the number of trees that are bark stripped by elephants and the occurrence of severe impact. Although wire netting did not significantly decrease the frequency of branch breaking or felling, it did reduce the occurrence of high levels of all types of impact. Trees that are likely to experience severe elephant impact, specifically ringbarking, main stem breaking, or uprooting are susceptible to mortality (O’Connor et al. 2007) and therefore wire netting could help these trees survive in areas with high elephant densities.

Along with the absence or presence of wire netting, tree species and size were important factors in explaining damage for some measures of impact. DBH did not significantly affect the frequency of bark stripping; however, this could be because most of the tagged trees were of a similar size. If we included a wider range of tree sizes in the study, then it is possible that some sizes of trees would benefit more from wire netting. For example, Smallie and O’Connor (2000) found that trees taller than 4 m were selected by elephants for bark stripping. Further research would be needed to see if smaller trees would also benefit from wire netting and how the wire netting of such trees could accommodate girth expansion over time. Acacia nigrescens had high levels of bark stripping suggesting it might be preferred by elephants for this type of impact. Physical properties of the bark itself may account for the differences in frequencies of bark stripping between the species (O’Connor et al. 2007). In addition, elephants are known to bark strip larger trees and this result might also be because A. nigrescens were significantly larger than S. birrea and L. schweinfurthii in this study (Boundja & Midgley 2010; Smallie & O’Connor 2000). Since wire netting was effective regardless of species, it could benefit all three of the species assessed in this study. However, it might be particularly useful in reducing elephant bark stripping in preferred species such as A. nigrescens.

In contrast to other methods of reducing elephant impact such as culling and altering water-surface availability that focus on reducing elephant numbers or altering spatial distribution patterns, wire netting instead attempts to reduce impact by protecting the trees themselves. Some research has used elephant exclosures to study the relative impact of elephants on trees (Western & Maitumo 2004). However, as a management tool these exclosures also exclude other large herbivores such as giraffe from accessing the trees (Goheen & Palmer 2010; Western & Maitumo 2004). In contrast, wire netting maintains the functionality of the trees for other organisms in the ecosystem by allowing them to access the trees for refuge, shade, food, and other activities. Wire netting is a relatively low maintenance strategy of reducing elephant impact. In the current system, the wire netting did not need to be replaced from the time it was applied in 2004 until the last survey in 2012 and very little maintenance was required. Depending on the scale at which trees are wire-net protected, it could be cost-effective. Electric fencing has been used to keep elephants out of specific areas but it is more expensive and can be problematic to maintain (Kiokoe et al. 2008). In addition, wire netting wrapped around the trunk of a tree is certainly less visible to tourists than electric fencing.

This study aimed to assess the effectiveness of wire netting in reducing elephant impact on trees in the APNR. We found that wire netting reduced the proportion and severity of elephant impact and contributed to the survival of trees. As elephant densities in the APNR and other areas continue to increase, there will need to be strategies to maintain these iconic and valuable large tree species. In such areas, the use of wire netting could serve to alleviate elephant impact.