African buffalo (Syncerus caffer) have a large distributional range across the savannas of Africa. Their habitat selection and foraging ecology have been relatively well studied (Landman & Kerley 2001; Macandza, Owen-Smith & Cross 2004; Sinclair 1979; Taylor 1985). Because of their large body size and large groups, buffalo are able to fend off predators (Sinclair, Mduma & Brashares 2003), thereby enabling them to forage in wooded vegetation with relatively low visibility and high predation risk. The large absolute food demands of buffalo, which is a function of their large body size (Illius & Gordon 1987; Wilmshurst, Fryxell & Bergman 2000), combined with their inability to efficiently crop short grass (use of tongue to increase bite size cannot work on short grass) constrain them to foraging in vegetation with sufficient height and biomass of forage (Codron et al. 2008; Illius & Gordon 1987). Buffalo generally avoid heavily grazed regions of short grassland (Bhola et al. 2012; Jarman & Sinclair 1979; Traill & Bigalke 2007), preferring woodlands dominated by tufted and leafy perennial grasses during the wet season and often relying on more productive riverine vegetation during the dry season (Bell 1970; Bennitt, Bonyongo & Harris 2014; Fynn, Chase & Roder 2014; Macandza, Owen-Smith & Cain 2012; Sinclair 1979).

While drylands often support less productive but higher quality forage than wetlands over the wet season (Fynn et al. 2014; Taylor 1985), their moisture-limited position in the landscape results in forage drying out and declining in forage quality and quantity over the dry season (Ellis & Swift 1988; Owen-Smith 2008; Taylor 1985). Thus, wetlands, where permanent water is easily accessible and perennial grasses continue to produce green forage over the dry season, would be expected to be favoured by many herbivores at this time with drylands being favoured over the wet season (Bell 1970; Fynn et al. 2014; Maddock 1979; Taylor 1985).

In contrast to more mobile species such as zebra (Equus burchellii) and wildebeest (Connochaetes taurinus), which often migrate between distinct wet- and dry season ranges (Bartlam-Brooks, Bonyongo & Harris 2011; Maddock 1979), most buffalo studies have found strong overlap between wet- and dry season ranges (Macandza et al. 2012; Ryan 2006; Ryan, Knechtel & Getz 2006; Venter & Watson 2008). Many studies, however, were conducted in small reserves (Landman & Kerley 2001; Ryan et al. 2006; Tshabalala, Dube & Lent 2009) where opportunity for large seasonal movements are limited or in larger reserves where functional heterogeneity of resources may be well developed at the landscape (catena) scale (Bell 1970; Macandza et al. 2012; Perrin & Brereton-Stiles 1999; Sinclair 1979), thereby reducing the need for large seasonal movements (Hopcraft, Olff & Sinclair 2010). However, in large, relatively unfragmented ecosystems and where landscape scale functional heterogeneity may be poorly developed relative to regional-scale heterogeneity (Fynn et al. 2014), larger buffalo movements may be expected (Naidoo et al. 2012; Skarpe et al. 2004).

The Savuti–Mababe–Linyanti ecosystem (SMLE) is part of the > 80 000 km² northern conservation area of Botswana, one of the largest relatively unfragmented wildlife regions in Africa (Fynn & Bonyongo 2011). Despite several studies on buffalo in the SMLE (Fynn et al. 2014; Patterson 1972), detailed seasonal movement and habitat selection patterns of buffalo in the region have not been fully identified and established, especially seeing that no detailed habitat map has been available until this year (2016). Considering their non-specialised mouth anatomy and their large body size–mediated demands for absolute food intake, we expected buffalo in the SMLE to favour vegetation where tufted, leafy grasses provided sufficient height and biomass of grass to enable food intake requirements to be satisfied (Illius & Gordon 1987; Wilmshurst et al. 2000) (Hypothesis 1 – H1).

The strongly developed regional-scale distribution (rather than landscape scale) of functional heterogeneity of wet- and dry season vegetation types in the SMLE (Fynn et al. 2014) is likely to promote extension of buffalo home ranges beyond the landscape scale (Hopcraft et al. 2010). Consequently, we expected seasonal movement patterns of buffalo in the SMLE to match the scale at which functional seasonal vegetation types are distributed in the ecosystem, which is strongly regional (Fynn et al. 2014) (Hypothesis 2 – H2). The objectives of this study were to (1) to examine seasonal movements and habitat selection of buffalo in the SMLE of northern Botswana and (2) to link seasonal movements to the quality and quantity (grass greenness, height and biomass) of vegetation in favoured seasonal regions of the landscape.

All three buffalo had much greater activities in sandveld and mopane woodlands during the wet season and various wetland systems during the dry season (Figure 3). BH1 and BH2 had greater separation of their wet and dry season concentration areas than BH3 (quantified by percentage overlap of their wet and dry season 95% local convex hull isopleths – Table 1; also see Figure 4 vs. Figure 5). BH1 and BH2 demonstrated distinct wet and dry season concentration areas, returning to these respective seasonal locations in successive years (Table 1; Figure 6).

For second-order habitat selection, the buffalo selected mopane and sandveld woodlands over the wet season, whereas riverine and wetland habitats were selected over the dry season (Table 2); for third-order habitat selection, the buffalo selected mopane woodlands over the wet season and acacia grasslands, whereas riverine and wetland habitats were selected over the dry season (Table 3). For more details of the plant species composition and classification of these plant communities and for details of the habitat map development, see a parallel paper in this special issue (Sianga & Fynn 2017). Over both dry seasons, BH1 concentration areas were in wetland and adjacent woodlands where it focused on the Linyanti Swamp in increasing proportions (and regularly across the international border with Namibia) as the dry season progressed (Figure 4). The concentrations of BH2 were in the Selinda Spillway and Kwando River wetland (floodplains) and adjacent woodlands in the 2012 and 2013 dry seasons (Figures 4 and 6). Interestingly, BH2 concentrated its activities in the Tsum Tsum wetlands (floodplains) of the Okavango Delta in the late wet season of April 2013 (which it did not do in 2012) before moving to the Selinda Spillway later in the dry season (Figure 6). BH3 had the wetland habitats of the Savuti Channel and Savuti Marsh at the core of its range, which it focused on during the dry season, and utilised adjacent woodlands on the peripheries of the home range during the wet season (Figure 5).

During the early and late dry season of 2012, wetland habitats (floodplain not sampled in the early dry season) had higher grass greenness, height and biomass than mopane, sandveld and riverine habitats, respectively (p < 0.05; Figure 7), while mopane, sandveld and riverine habitats were not significantly different in grass greenness, height and biomass, respectively (p > 0.05; Figure 7).

A veterinarian registered with the government of Botswana conducted all darting operations in principle with the research permit (EWT 8/36/4 XVII [31]) and the supplementary collaring permit (EWT 8/36/4 XVII [41]) issued by the Ministry of Environment, Wildlife and Tourism and the Department of Wildlife and National Parks (Gaborone, Botswana). Darting operations or removing collars from cows were done from a vehicle and a helicopter, respectively. Collars were removed successfully after two successive years.

Buffalo shifted their seasonal concentration areas between wetlands and woodlands (Figures 3 and 4) as an adaptive strategy to seasonal forage dynamics and water availability over the annual cycle, with BH1 and BH2 having a migratory strategy with little overlap between wet and dry season concentration areas. However, the distance moved from perennial water sources over the wet season varied significantly across the buffalo, with BH3 being more sedentary. Bennitt et al. (2014) and Taolo (2003) reported similar results, noting that buffalo selected contrasting seasonal vegetation types, using woodlands far from permanent water during the rainy season and seasonally flooded vegetation close to permanent water during the early and late flood seasons. As with our findings, several other studies in the region have also found a mix of strategies (migratory vs. sedentary) between buffalo (Bennitt et al. 2014; Bennitt, Bonyongo & Harris 2015; Naidoo et al. 2012, 2014). Cornélis et al. (2011) also noted similar findings where buffalo were observed to range between woodlands and riverine habitats during the wet and dry season, respectively.

A key factor that may influence buffalo concentration patterns during the wet season appears to be the availability of higher quality, soft-leaved grasses (such as Digitaria eriantha) in back country woodlands than in wetlands (floodplains), which in contrast support tougher-leaved grasses adapted to shading and litter accumulation under the more productive conditions of wetlands (Fynn & Bonyongo 2011), and where greater biomass dilutes nutrient concentrations (Jarrell & Beverly 1981), while greater silica contents of wetland versus dryland grasses further reduce digestibility (Mosimane 2015). One of the most favoured grasses by buffalo, D. eriantha, peaks in abundance beyond 10 km from the permanent water of the Okavango Delta and Linyanti Swamps, whereas shorter grazing lawn–forming grasses such as Urochloa trichopus peaked in abundance with 5 km of these extensive wetland systems (Sianga et al. 2017). The medium to tall, leafy high-quality D. eriantha would be expected to be optimal in height and digestibility for maximising protein and energy intake rates in buffalo, which explains why buffalo focused their wet season foraging beyond 10 km from wetland systems (Figure 4). By contrast, wildebeest and impala, which favour shorter grasses, focused within 5 km of these wetland systems (Fynn et al. 2014). This suggests that short and tall grass grazers in the SMLE have distinct spatial niche separation during the wet season driven by distance to wetland systems. Similar observations have been made in the Maasai-Mara, where buffalo favour the taller grass areas within the reserve and wildebeest and gazelles the shorter grass areas outside the reserve (Bhola et al. 2012). Thus, a high conception rate of buffalo during the wet season is probably because of the availability of high-quality green forage of optimal height and biomass (Ryan, Knechtel & Getz 2007). Thus, concentration during the wet season by the three buffalo in vegetation types (sandveld and mopane) where tufted, soft-leaved grasses are more abundant supports our first hypothesis (H1). Drying up of ephemeral pans over the dry season forced buffalo to move away from favoured wet season ranges towards permanent water sources (Figures 3 and 4), which included floodplains and swamps (wetlands) and adjacent woodlands within the study area. It is this forced movement away from favoured wet season ranges towards permanent water that likely contributes to reduced grazing pressure in favoured wet season ranges and, therefore, higher abundances of favoured grasses such as D. eriantha far from permanent water. This indicates that artificial water provision in back country woodlands far from permanent water may be an unwise management option that could reduce niche diversity in the system leading to declines in taller grass grazers such as buffalo, as well as roan and sable antelope.

These floodplain and swamp (wetlands) systems provided buffalo not only with reliable drinking water during the dry season but also with forage significantly greener than that found in woodlands, as observed in other studies in the region (Bennitt et al. 2014; Fynn et al. 2014; Taolo 2003). More productive wetland areas are likely to provide a critical reserve and critical buffer resources (Owen-Smith 2002) for the late dry season, or elsewhere referred to as key resources (Illius & O’Connor 2000), especially during drought years when more preferred resources have been depleted (Owen-Smith 2002). The ability of wetlands to provide taller green forage with higher energy and protein levels relative to drylands during the late dry season has been shown to greatly elevate buffalo population productivity (Taylor 1985). The taller grass of wetlands is likely to be especially critical for buffalo, which are highly vulnerable to competition for forage (Bhola et al. 2012), owing to their tongue sweep foraging strategy, which cannot deal with short grass.

Thus, our findings of increasing use of swamp, marsh and floodplains (wetlands) as the dry season progressed by our three buffalo together with similar observations in other studies (Prins & Beekman 1989; Taylor 1985; Tinley 1977; Vesey-FitzGerald 1960; Western 1973), demonstrate the importance of wetland systems as dry season key-resource habitat for buffalo, as is the case for cattle and other wild herbivores across Africa (Fynn et al. 2015).

Our collared buffalo migrated at an intermediate scale from wetlands to woodlands rather than from wetlands to the high-quality Mababe Depression grasslands, favoured by a migratory zebra population over the wet season (Sianga 2014) and, thus, our second hypothesis (H2) that buffalo seasonal movements would match the regional-scale distribution of functional seasonal vegetation types was not supported. Though this finding does not support our H2, it does not rule out the possibility of buffalo migrating into the depression during the wet season because not all buffalo herds in the ecosystem were fitted with collars to determine their movements. Also, aerial surveys conducted by Fynn et al. (2014) did not observe any buffalo in the Mababe Depression at that time, suggesting that buffalo probably do not favour the Mababe Depression during the wet season, unlike the zebra population (Sianga 2014). However, the buffalo appear to show adaptive variation in seasonal presence to changing environmental conditions in different years. For instance, in 2011, BH1 and BH2 avoided floodplains and swamps (wetlands) in their usual dry season concentration areas of the Linyanti Swamps and Selinda Spillway (2012 and 2013) because of the exceptionally high floods of 2011, being forced to use the more elevated floodplains of the Savuti Channel. This shift in location of different dry season concentration areas emphasises the importance of having a large spatial scale available for movement (Fynn et al. 2014), which enables adaptation to changing conditions between years. Finally, the finding of cross-border movements by BH1, together with those of Naidoo et al. (2014) and Patterson (1972), demonstrates the importance of Transfrontier Conservation Areas such as this one (Kavango Zambezi Transfrontier Conservation Area, KAZA), which transcends the borders of Angola, Namibia, Botswana, Zimbabwe and Zambia allowing adaptive foraging to a variety of different functional seasonal resources.