Contemporary fire regimes and elephant impacts in Baikiaea-dominated woodlands of the Chobe region, northern Botswana

Study area

Broadly defined, Miombo and related woodlands, including those dominated by Baikiaea plurijuga (Zambezi teak) are included within southern Africa’s Miombo Ecoregion (Timberlake and Chidumayo 2011; Fig. 1a), occurring on typically infertile, acidic, freely draining sandy soils of the ancient African Plateau, under markedly southern hemisphere summer seasonal (November–April) rainfall conditions ranging generally between 600 and 1600 mm MAR. Woodlands dominated by Baikiaea plurijuga extend over 2000 km² under low rainfall conditions in the southern portion of the Miombo ecoregion, with large tracts in Angola, Zambia, Namibia, Zimbabwe, and Botswana (Werger and Coetzee 1978; White 1983; Timberlake et al. 2010; Dziba et al. 2020). Although Baikiaea woodlands are characterised by the dominance of B. plurijuga, they are observed to integrate with other structural types including typical Miombo taxa (e.g. Russell-Smith et al. 2024). Various dominant leguminous taxa (e.g. slow growing Baikiaea, which is characteristic Miombo woodland taxa as well as Brachystegia spp., Julbernardia spp., Pterocarpus spp.), are recognised as being vulnerable to repeated severe fires, leading to shifts in species composition, conversion of wooded forests to open-savannas, and disruption to critical habitats (e.g. Frost 1996; Holdo 2007; Ribeiro et al. 2020b). Mean annual (2001–2019) savanna fire extent across this vast geography ranges between ~20% and 30% under northern higher rainfall conditions, to between 7% and 10% in the lower rainfall south (Fig. 1b); the great majority of fire extent occurs later in the 6–7 month Southern Hemisphere dry season under relatively severe fire-weather conditions (Russell-Smith et al. 2024).

Fig. 1.

(a) Distribution of Miombo sub-formations where the dominant structure is classified as dry deciduous forest and wooded grasslands (Timberlake and Chidumayo 2011) and (b) southern African fire frequency context, 2001–2022, based on a MODIS burned area product (Russell-Smith et al. 2024). The black square outlines the location of the Chobe study area.

The Chobe District is recognised globally for its rich biodiversity and high conservation value, centrally located within the multi-national Kavango-Zambezi (KAZA) conservation area (shared between Botswana, Angola, Namibia, Zambia, Zimbabwe). The KAZA area supports a thriving conservation-based economy with an estimated population of 1.5 million inhabitants (Cumming D.H. 2008, unpubl. data; Stoldt et al. 2020) and over 250,000 elephants, 600 bird species, 128 reptile species, 50 amphibian species, and diverse invertebrate species (Bussière and Potgieter 2023). The Chobe region supports the largest population of African elephants (Loxodonta africana) in the world, hosting ∼131,000 resident elephants while serving as a key migration and dispersal corridor (D. H. Cumming 2008, unpubl. data; Government of Botswana 2021). While Chobe region supports about 80% of Botswana Community-based Natural Resource Management (CBNRM) income (Chevallier 2016), KAZA has recorded the highest rate of forest cover decline since 1990 (FAO 2020)

The Chobe region lies on a flat to gently undulating landscape in the north-eastern portion of Botswana (17°–19°S, 24°–26°E). The climate is semi-arid to sub-humid, with MAR ~650 mm and >90% falling from October to the end of April; with most rain typically recorded in December and January (Skarpe et al. 2014a). This annual cycle results in seasonal accumulation of biomass followed by significant potential for regular and uncontrolled large annual fires.

The Chobe District includes portions of the Okavango and Zambezi basins (Skarpe et al. 2014a). The physical landscape is a result of paleo-environmental alluvial and aeolian regimes influenced by tectonic events (Shaw and Thomas 1993; Skarpe and Ringrose 2014) giving rise to a matrix of vegetation structures that serve both as a resource for wildlife and anthropogenic socio-economic activities. Deep aeolian and alluvial deposits form sands and silts that occupy most of the district with more fertile clay soils formed from lacustrine deposits (Barnes 2001; Smith et al. 2005). The relatively fertile clay soils formed from exposure of Stormberg basalts in the eastern part of the district are suitable and exploited for arable agriculture (Ecosurv 2018) and for pasture grasslands (Fig. 2). Low-lying seasonal drainage pans with finer particles of clay and organic material, usually colonised by elephant-truncated mopane (Colophospermum mopane) shrubs, comprise a general feature of these relatively fertile landforms.

Fig. 2.

Map of Chobe District (21,000 km²) showing the project area outline, land use, firebreaks, major highway (A33), roads and tracks. Major firebreaks (graded to 20 m) primarily constitute land use boundaries.

Chobe vegetation closely follows soil formations and varies from Zambezian broad-leaved dry deciduous forests and shrubby woodland Miombo formations on deep nutrient-poor sands dominated by B. plurijuga (Zambezi teak), Pterocarpus angolensis (Africa teak), Combretum spp. (bushwillow), Schinziophyton rautanenii (mongongo), with mopane woodlands and grasslands confined to relatively fertile seasonally waterlogged clays (Skarpe et al. 2014a). Fragments of riparian vegetation, which in the 1960s used to be continuous (Child 1968; Mosugelo et al. 2002), are found along the northern fringes of the District on the Chobe-Zambezi and Linyanti river systems. The south-western portion of the district features a mosaic of mopane, acacia-dominated and grassland systems (Skarpe et al. 2014a; Fox et al. 2017; Ecosurv 2018)

Land use plays a key role in the management of the Chobe landscape. Forest Reserves are a significant land use type designated regardless of vegetation structure or characteristics, both on state and tribal lands (Government of Botswana 2011). Six Forest Reserves were established from the late 1960s to protect significant tracts of woodlands containing several high value timber species (including B. plurijuga and P. angolensis) with large tracts of oil-producing S. rautanenii forests. Logging took place from 1935 but after determining that harvesting practices were non-sustainable (Wegge 1992; Kgathi and Sekhwela 2012), timber extraction activities ceased in 1994.

The Chobe District is a home to ~28,500 inhabitants (Statistics Botswana 2022) with settlements on the outer fringes of the District. Kasane and Kazungula townships are more urbanised and form a contiguous northern urban area with a combined population concentration of 62% of the district total. The majority of settlements in the District have populations of less than 1000 people (Ecosurv 2018), rendering average district density of 1.4 persons km⁻² (Statistics Botswana 2022).

Burned area mapping

Available annual fire mapping data for the period 1991–2023 imply an apparent decline in burnt area (BA) over the study period, with the great majority of detected fire scars occurring in the LDS period (Fig. 3). During 2001–2023 when both sufficient Landsat and MODIS imagery were available, mean BA was 329 km² in the EDS and 3707 km² in the LDS, representing 8% and 92% of the study area, respectively. Over the 33-year study period, the spatial patterning of BA occurred mostly in the eastern sector of the study area (Fig. 4a), including in both EDS and LDS periods during 2001–2023 assessed using available MODIS imagery (Fig. 4b, c).

Fig. 3.

Burnt Area (BA) mapping for the 21,000-km² study area derived from available imagery sources for the period 1991–2023 (note: insufficient imagery for years 1995,1999, and 2000). BA mapping for EDS (pre-August) and LDS periods included where sufficient with seasonal imagery available (note: insufficient Landsat coverage 1992–1998). Trend line represents apparent decline in annual BA extent over the study period.

Fig. 4.

(a) Fire Frequency Index (FFI) based on burnt area (BA) over a 33-year (1991–2023) period combining available MODIS and Landsat BA products, FFI in (b) EDS, and (c) LDS periods based on BA mapping derived from MODIS imagery 2001–2023. (d) Landcover vegetation structural types, (e) soils distribution (Department of Surveys and Mapping, Botswana), and (f) dry season average elephant distribution (1994, 1999, 2003, and 2012. Note: different sample scales; Department of Wildlife and National Parks, Botswana). Ecozones 1–10 on all maps are proposed management units adapted from the Chobe District Land use Plan (van der Sluis et al. 2017). See Table 1 for Ecozone names.

Landcover vegetation types

Overall, landcover vegetation structural type mapping indicated that the study area comprised 7.6% forest, 26.6% woodland, 49.3% open woodland, 3% shrubland, 6% grassland, 1.3% bare, and 6.3% mixed mopane and grassland. Based on extensive validation data (see Materials and methods), mapping accuracy overall was determined to be 85%, with reliability >90% for open woodland, grassland, and forest types; and 78% for woodland (overlapping both forest and open woodland; Table 1). Forest vegetation was observed to occur in the north of the study area, woodland predominantly in the west, and open woodland and grassland in the east (Table 2; Fig. 4d). Baikiaea forest, associated Miombo woodland and open woodland types, all occurred on deep sandy Arenosols, with grassland and accompanying mopane types on clay-dominated vertisols (Fig. 4e).

Over the 33-year study period, annual fire extent was greatest in grassland and in mixed grassland/mopane (mean FFI = 10 and 11, respectively, which is equivalent to one fire every 3 years), followed by open woodland and shrubland (mean FFI = 5 and 6, respectively, which is equivalent to one fire every 5 and 6 years). Woodland and forest showed the lowest annual fire extent (FFI = 3 each, which is equivalent to one fire every 10 years).

Land use

We ranked the ecozones by fire prevalence from 1 (highest fire prevalence) to 10 (lowest fire prevalence). Fire prevalence was associated generally across communal and public lands, consistently in the LDS period. Rather, most fire (FFI >12) occurred annually across land use types in the central and eastern sectors of the study area, and least (FFI <3) in the extreme north and western sector public lands (Table 1; Fig. 4a).

Elephant distribution

The mean dry season elephant density distribution derived from 4 years of available records is in Fig. 4f. Highest dry season densities are observed in northern portions of the study area (particularly in proximity to perennial water resources associated with the Chobe River wetlands), and are associated generally with forest, woodland, and open woodland vegetation types (Fig. 4d), and highly variable fire occurrence conditions (Fig. 4a).

Historical land use context

Historical records of landscape-scale fire patterning in the study area are scant but presumably involved the widespread application of fire especially by San communities for hunting and gathering, and controlling fire spread through detailed knowledge of ecological requirements and seasonal conditions (Archibald et al. 2012). The 1870s saw the arrival of pastoral settlers who burned virtually every year in the LDS to promote growth of palatable grass species at the expense of woody biomass (Child 1968; Skarpe et al. 2014a). This practice continued through the turn of the century and was followed by commercial timber harvesting from 1935, virtually throughout north-eastern sectors in areas with harvestable timber material (Child 1968; Kgathi and Sekhwela 2012). In 1938, EDS burning was introduced and adopted as the official fire policy of the Forestry Department, confronting the prevailing LDS burning practices of local pastoralists and the concerns of wildlife managers (Child 1968).

Formal moves towards establishment of Chobe National Park in 1967 commenced in the early 1960s, and regional Forest Reserves from the late 1960s. These developments coincided with the relocation of original inhabitants from parts of the Chobe District interior and government enforcement of fire exclusion, suppressing cultural fire practices. In the mid-1980s, Botswana introduced its National Fire Management Policy, which emphasised fire exclusion particularly in Protected Areas like National Parks and Forest Reserves. In the 1990s, these fire exclusion and suppression policies were augmented by implementation of a widespread system of linear graded firebreaks of width ranging from 10 m to 20 m depending on assumed protection priority, including within Protected Areas (Fig. 2). The most recent iteration of Botswana’s fire management legislated framework, the Forest Act of 2007, although allowing for the undertaking of control burns under Permit conditions, essentially reaffirmed the fire exclusion policy.

The trajectory of fire exclusion practice in Botswana follows many other former colonial and subsequent neo-colonial governments in restricting or banning use of fire, including by Indigenous peoples dependent on long-standing traditional fire management practices with claimed ecological and forestry management considerations (Pricope and Binford 2012; Roodt 2015). As reported by other informed observers, such policy and restricted management interventions (e.g. fire-breaks) have manifestly failed to quell the occurrence of extensive destructive LDS wildfires in Botswana (Dube 2013; Cassidy et al. 2022; Dube 2024). Our assessment supports this contention.

The role of elephants

By the turn of the 20th century, elephant populations in the Chobe region had been decimated by hunting for ivory, and only started recovering rapidly from the 1960s (Skarpe et al. 2014a). Evidence suggests that elephant populations pose relatively insignificant agents of habitat change in the more extensive Baikiaea woodlands on deep infertile sands (Childes and Walker 1987; Ben-Shahar 1996a; du Toit et al. 2014). For example, despite large dry season elephant populations fringing the Chobe River (Fig. 4f) with ongoing major impacts on woody vegetation on adjacent fertile floodplain substrates (Mosugelo et al. 2002; du Toit et al. 2014; Skarpe et al. 2014b), elephant populations have little impeded post-logging regeneration of dense Baikiaea forests on sandy facies immediately to their south (Fig. 4d). This can be attributed to observations that B. plurijuga, and various associated common tree taxa (e.g. Erythrophleum africanum and Ochna pulchra), possess foliage unpalatable to elephants and therefore incur limited structural damage (Ben-Shahar 1996b; Ben-Shahar and Macdonald 2002; Holdo 2006, 2007). Conversely, many other regionally dominant taxa, notably including legumes (e.g. Acacia spp., Brachystegia spp., Julbernardia spp., C. mopane), common mid-storey trees (e.g. Combretum spp., Terminalia sericea) and shrubs (Baphia massaiensis and Bauhinia petersiana) are highly palatable to elephants (Ben-Shahar 1996a; Ben-Shahar and Macdonald 2002; Holdo 2007; Skarpe et al. 2014a). In consequence high elephant population pressures and their associated browsing, debarking, pushing-over, uprooting, trampling impacts can result in rapid (infra-decadal) marked vegetation structural and population demographic change (Beuchner and Dawkins 1961; Laws 1970; Anderson and Walker 1974; Guy 1981; Ben-Shahar 1996a; Mapaure and Campbell 2002; Mosugelo et al. 2002; Skarpe et al. 2014a). We can anticipate that under conditions of continuing high and increasing elephant population density, conversion of the remaining woodland to more open-canopied vegetation will continue to occur through destructive impacts especially on palatable taxa, resulting in more fire-prone fuels and likely enhanced habitat opportunities for herbivores and their hunters. We can also anticipate that populations of non-palatable taxa (including Baikiaea) will be competitively advantaged and under conditions of enhanced fire management, regenerating patches will likely continue to re-develop and expand from previously logged, fire prone sites.

Chobe fire regime

Recent studies have outlined contemporary burning patterns in the broader Chobe region, including the predominance of LDS intense fires (Russell-Smith et al. 2021) and identification of local fire density ‘hotspots’ (Fox et al. 2017; Cassidy et al. 2022). While noting the significant effects of frequent anthropogenic LDS fires on net conversion of woodlands (−16%) to shrublands and grasslands in the 21,000 km² Chobe District in the period 1990–2013, Fox et al. (2017) also observed that annual fire occurrence and associated land cover change impacts were strongly mediated by antecedent rainfall conditions. Regional climate challenges in the decades ahead are projected to involve significant deterioration of fire-weather conditions associated particularly with reduced wet season precipitation and soil moisture, rising temperatures and associated evaporation (Trisos et al. 2022) further exacerbating the intensity of fires.

For the most part, fires impacting the study area originate on the eastern side of the Chobe District, ignited along the main north-south road corridor (Fig. 2), and driven westward by prevailing strong dry season south-easterly winds (Deloitte and Touche 1992). In the absence of concerted fire mitigation activities (e.g. EDS prescribed fires; back-burning from fire barriers), the inability of graded fire-break infrastructure to contain this westerly spread of LDS fires is in Fig. 4(c) by significant continuity in FFI across the boundary between eastern Land use Units 1, 2 (Kakulwane Plains, Kasane Forest Extension) and western Chobe National Park Units 5, 6 (Maku Plains, Nogatsaa Valley). The east-west highway corridor separating the northern Chobe Waterfront sector of Chobe National Park from Protected Areas to the south (Fig. 4a), conversely serves as an effective barrier to fire-spread emanating from the south-east.

The current high prevalence of fire also reflects the extended impacts of burning under cattle pastoral management (which had mostly ceased by the 1950s due to tsetse fly), and especially logging regimes from the 1930s to 1990s (Child 1968). Logging of valuable hardwoods (B. plurijuga - Zambezi Teak; P. angolensis - African Teak; and G. coleosperma - African rosewood) was sufficiently intense to reduce woodlands on deep sands in the north-east, extending up to 30-km south from the Chobe River, to ‘scrubby savanna’ with few big trees (Child 1968; Skarpe et al. 2014a). By the 1990s, under diminishing logging pressures, much affected vegetation was reverting to woodland, dominated by patches of deep-rooted Baikiaea regenerating from coppice (Ben-Shahar 1993). The legacy of these pastoral and logging interventions is reflected in the restricted distribution of extant woodland vegetation in Kasane Forest Extension and Maku Plains (Land use Units 2 and 5) (Fig. 4d), and resultant open-canopied fuel conditions conducive for carrying extensive wildfires.

By contrast, the lack of burning particularly in grassy savanna habitats in south-western Protected Area Land use Units (Chobe Forest Reserve; Nogatsaa Valley) does not address the varied habitat and grazing management requirements of diverse herbivore assemblages (e.g. McNaughton 1985; Klop and Prins 2008), significant potential for woody plant encroachment (e.g. O’connor et al. 2014), and significant risk of very extensive wildfires.