Ecosystem‐scale effects of megafauna in African savannas

Natural protected areas are critically important in the effort to prevent large‐scale megafaunal extinctions caused by hunting and habitat degradation. Yet such protection can lead to rapid increases in megafauna populations. Understanding ecosystem‐scale responses of vegetation to changing megafaunal populations, such as the case of the African elephant Loxodonta africana in savannas, requires large‐scale, high‐resolution monitoring over time. From 2008 to 2014, we repeatedly surveyed the fate of more than 10.4 million woody plant canopies throughout the Kruger National Park, South Africa using airborne Light Detection and Ranging (LiDAR), to determine the relative importance of multiple environmental, biotic and management factors affecting treefall rates and patterns. We report a mean biennial treefall rate of 8 trees or 12% ha^?1, but with heterogeneous patterns of loss in both space and time. Throughout Kruger, the influence of elephant density on treefall was matched only by spatial variation in soils and elevation, and all three factors co‐dominated park‐wide treefall patterns. Elephant density was up to two times more influential than fire frequency in determining treefall rates, and this pattern was most pronounced for taller trees (> 2 m in height). Our results suggest that confining megafauna populations to protected areas, or reintroducing them into former or new habitat, can greatly alter the structure and functioning of the host ecosystem. Conservation strategies will need to accommodate and manage these massive ecological changes in the effort to save megafauna from extinction, without compromising system functionality.     

Ecological impact of large herbivores on the woody vegetation at selected watering points on the eastern basaltic soils in the Kruger National Park

A lack of quantitative data on the impact of large herbivores on woody vegetation around watering points in the Kruger National Park gave rise to an in depth study. This paper reports on trends found at two artificial watering points on the basaltic soils in the Satara Land System. As a result of the almost complete lack of woody individuals in the immediate vicinity of the watering points, this area could be described as a 'sacrifice area'. Shrub density increased with distance from the watering point, with the impact of large herbivores on shrub density extending up to 2.8 km. The practice of providing artificial water in troughs at windmills in the Kruger National Park causes the formation of a multiple 'piosphere' effect, with the existence of a zone of high utilization in woody vegetation extending far beyond the sacrifice area.     

Responses of woody vegetation to exclusion of large herbivores in semi‐arid savannas

The Nkuhlu large‐scale long‐term exclusion experiment in Kruger National Park was designed to study the long‐term effects of large herbivores on vegetation. One treatment excludes elephants, another excludes all herbivores larger than hares and another one comprises an open, control area. Vegetation monitoring was implemented in 2002 when a baseline survey was conducted prior to exclusion. Monitoring was repeated 5 years after exclusion. Data from the surveys were analysed to establish how structure and composition of woody vegetation had changed 5 years after herbivore exclusion. The analysis showed that neither plant assemblage nor mean vegetation height had changed significantly since exclusion. However, both species richness and density of woody plants increased 5 years after exclusion of all large herbivores, but not after the exclusion of elephants alone. One already common species, Dichrostachys cinerea, became more common after excluding all large herbivores compared with either no exclusion or elephant exclusion, possibly leading to competitive suppression of other species. Species other than D. cinerea tended to either increase or decrease in density, but the changes were insufficient to induce significant shifts in the overall assemblage of woody plants. The results indicate that after 5 years of exclusion, the combined assemblage of large herbivores, and not elephants alone, could induce changes in species richness and abundances of woody plants, but the effect was so far insufficient to induce measureable shifts in the assemblages of woody plants. It is possible that assemblages will change with time and increasing elephant numbers may amplify future changes.     

Flood‐deposited wood creates regeneration niches for riparian vegetation on a semi‐arid South African river

Question : The formation of large woody debris (LWD) piles during floods has significant impacts on riparian succession through pioneering plants often establishing in association with wood. We assess the importance of LWD for seed regeneration of riparian plants after a century‐scale flood disturbance in a semi‐arid environment. Location : The Sabie River within Kruger National Park in the semi‐arid northeast of South Africa. Methods : Our approach was to quantify the riparian soil seed bank, to record the frequency of establishment of riparian plants in woody debris piles, and to conduct experimental out‐plantings of common riparian trees in plots with and without LWD. Results : We found the abundance and diversity of seedlings were higher in soils taken from wood piles than from open reference areas, and most seedlings were herbaceous species. Surveys indicated that numbers of seedlings recorded within woody debris were significantly greater than in open reference areas or within established vegetation. Seedling establishment in various cover‐types also varied for different riparian tree species. Experimental out‐planting of seedlings of two riparian tree species (Philenoptera violacea and Combretum erythrophyllum) revealed that, after 433 days, planted seedlings survived only in woody debris piles. Conclusion : LWD formed after a large flood creates heterogeneous patches that may influence post‐disturbance regeneration of riparian vegetation by providing a variety of environmental niches for seedlings establishment. We suspect that higher seedling survival in LWD is due to increased moisture (particularly in the dry season) and nutrients, and protection from seasonal flooding and herbivory.     

Patterns of alien plant distribution at multiple spatial scales in a large national park: implications for ecology, management and monitoring

Aim Spatial scale is critical for understanding and managing biological invasions. In providing direction to managing alien plant invasions, much emphasis is placed on collecting spatially explicit data. However, insufficient thought is often given to how the data are to be used, frequently resulting in the incompatibility of the data for different uses. This paper explores the role of spatial scale in interpreting, managing and monitoring alien plant invasions in a large protected area. Location Kruger National Park, South Africa. Methods Using 27,000 spatially‐explicit records of invasive alien plants for the Kruger National Park (> 20,000 km²) we assessed alien plant species richness per cell at nine different scales of resolution. Results When assessing the patterns of alien plants at the various scales of resolution, almost identical results are obtained when working at scales of quarter‐degree grids and quaternary watersheds (the fourth level category in South Africa's river basin classification system). Likewise, insights gained from working at resolutions of 0.1–0.5 km and 1–5 km are similar. At a scale of 0.1 × 0.1 km cells, only 0.4% of the Kruger National Park is invaded, whereas > 90% of the park is invaded when mapped at the quarter‐degree cell resolution. Main conclusions Selecting the appropriate scale of resolution is crucial when evaluating the distribution and abundance of alien plant invasions, understanding ecological processes, and operationalizing management applications and monitoring strategies. Quarter‐degree grids and quaternary watersheds are most useful at a regional or national scale. Grid cells of 1 to 25 km² are generally useful for establishing priorities for and planning management interventions. Fine‐scale data are useful for informing management in areas which are small in extent; they also provide the detail appropriate for assessing patterns and rates of invasion.     

The influence of fire frequency on the structure and botanical composition of savanna ecosystems

Savannas cover 60% of the land surface in Southern Africa, with fires and herbivory playing a key role in their ecology. The Limpopo National Park (LNP) is a 10,000 km² conservation area in southern Mozambique and key to protecting savannas in the region. Fire is an important factor in LNP's landscapes, but little is known about its role in the park's ecology. In this study, we explored the interaction between fire frequency (FF), landscape type, and vegetation. To assess the FF, we analyzed ten years of the Moderate resolution Imaging Spectroradiometer (MODIS) burned area product (2003–2013). A stratified random sampling approach was used to assess biodiversity across three dominant landscapes (Nwambia Sandveld‐NS, Lebombo North‐LN, and Shrubveld Mopane on Calcrete‐C) and two FF levels (low—twice or less; and high—3 times or more, during 10 years). Six ha were sampled in each stratum, except for the LN versus high FF in which low accessibility allowed only 3 ha sampling. FF was higher in NS and LN landscapes, where 25% and 34% of the area, respectively, burned more than three times in 10 years. The landscape type was the main determinant of grass composition and biomass. However, in the sandy NS biomass was higher under high FF. The three landscapes supported three different tree/shrub communities, but FF resulted in compositional variations in NS and LN. Fire frequency had no marked influence on woody structural parameters (height, density, and phytomass). We concluded that the savannas in LNP are mainly driven by landscape type (geology), but FF may impose specific modifications. We recommend a fire laissez‐faire management system for most of the park and a long‐term monitoring system of vegetation to address vegetation changes related to fire. Fire management should be coordinated with the neighboring Kruger National Park, given its long history of fire management. Synthesis: This study revealed that grass and tree/shrub density, biomass, and composition in LNP are determined by the landscape type, but FF determines some important modifications. We conclude that at the current levels FF is not dramatically affecting the savanna ecosystem in the LNP (Figure 1). However, an increase in FF may drive key ecosystem changes in grass biomass and tree/shrub species composition, height, phytomass, and density.     

Do feral buffalo (Bubalus bubalis) explain the increase of woody cover in savannas of Kakadu National Park,Australia?

Aim To study changes in woody vegetation in both floodplains and eucalypt savanna over a 40‐year period using multi‐temporal spatial analysis of variation in density of a large introduced herbivore, the Asian water buffalo (Bubalus bubalis). Feral buffalo built up to high densities in the study area until c. 1985, after which a control programme almost eliminated the animals. From 1990, low densities of managed buffalo were maintained inside an enclosure. We compared trends in woody vegetation when buffalo were high‐density feral, low‐density managed or absent. Location The study area was located in and around a 116‐km² buffalo enclosure inside Kakadu National Park, in monsoonal northern Australia. Methods We analysed sequences of digitized and geo‐rectified aerial photographs, acquired in 1964, 1975, 1984, 1991 and 2004, to chart changes in woody cover on the floodplain and in the savanna. On the floodplain we assessed whether trees were present at these times at 14,568 points, and buffalo density was estimated from the density of animal tracks. In the savanna we estimated woody cover at pre‐selected sites. Generalized linear modelling was used to analyse changes in woody vegetation, using elevation and presence of woody vegetation in neighbouring points on the floodplain, and buffalo regime and initial woody cover in the savanna. Results Changes in animal track density reflected park‐wide historical estimates of buffalo numbers. Tree cover increased in both floodplain and savanna, but this was only weakly related to buffalo density. The best predictor of whether a floodplain cell converted from treeless to woody, or the converse, was the woodiness of neighbouring vegetation. There was slightly less thickening with high buffalo densities. In savanna, low densities of managed buffalo were weakly associated with increases in tree cover relative to either high densities of feral buffalo or no buffalo. Main conclusions Our study indicates that buffalo are not a major driver of floodplain and eucalypt savanna dynamics. Rather, the observed increase in woody cover in both savanna and flood plains concords with regional trends and may be related to increased atmospheric CO₂, increasing rainfall and changing fire regimes during the study period.     

Relationship between small-scale structural variability and simulated vegetation productivity across a regional moisture gradient in southern Africa

The observed variability in vegetation structure within landscapes was used as the basis for model estimates of the range of annual productivity of landscape patches at four sites along a moisture gradient in southern Africa ranging from 879 to 365 mm mean annual rainfall. Principal components of patch‐scale variability in leaf area, woody biomass and vertical leaf profiles were derived from intensive characterization of the small‐scale spatial structure of woody vegetation at each site. For each site, the mean and extremes of the principal component distribution parameterized an ecophysiology model of vegetation productivity. Vegetation was most heterogeneous at intermediate locations along the rainfall gradient. Variability in vegetation structure led to a range of annual productivity within one site (600 mm) that accounted for 68% of the total range in mean productivity across all sites. Patch‐scale estimates of tree productivity were found to be primarily correlated to annual rainfall (r²=0.66, P=0.001) and not woody leaf area (r²=0.01, P=0.75), while grass productivity was found to be related to values of woody leaf area (r²=0.77, P<0.001) and not annual rainfall (r²=0.11, P=0.29). This result indicates that life‐form interactions have a significant role in controlling vegetation productivity across the rainfall gradient. The findings of this study emphasize the importance of considering heterogeneity rather than mean structure when modeling productivity, particularly when considering dynamic vegetation structure, where differences between landscape patches may not be well represented in the mean structure.     

Thresholds of potential concern as benchmarks in the management of African savannahs

In the Kruger National Park (KNP), South Africa, ecosystem managers use a series of monitoring endpoints, known as thresholds of potential concern (TPCs), to define the upper and the lower levels of accepted variation in ecosystems. For woody vegetation, the current TPC suggests that woody cover should not drop by more than 80% of its 'highest ever' value. In this paper, we explore the utility of palaeoecological data in informing TPCs. We use calibrated fossil pollen data to explore variability in vegetation at two sites over the past 5000 years, to provide a long-term record of changes in woody vegetation cover and a context for interpreting more recent vegetation change. The fossil pollen data are calibrated using studies of modern pollen and vegetation from KNP; arboreal pollen percentage was simulated using pollen-landscape modelling software for savannah landscapes of varying woody vegetation cover, and the relationship between vegetation and pollen data was quantified using nonlinear regression. This quadratic equation was then applied to fossil pollen data in order to estimate woody vegetation cover from arboreal pollen percentages. Our results suggest that the TPCs have not been exceeded during the period represented in the pollen record, because estimated woody vegetation cover has remained above 20% of its highest ever value. By comparing the fossil pollen data with TPCs, our study demonstrates how palaeoecological data can be presented in a form that is directly relevant to management objectives.     

Elephants limit aboveground carbon gains in African savannas

Understanding the drivers of vegetation carbon dynamics is essential for climate change mitigation and effective policy formulation. However, most efforts focus on abiotic drivers of plant biomass change, with little consideration for functional roles performed by animals, particularly at landscape scales. We combined repeat airborne Light Detection and Ranging with measurements of elephant densities, abiotic factors, and exclusion experiments to determine the relative importance of drivers of change in aboveground woody vegetation carbon stocks in Kruger National Park, South Africa. Despite a growing elephant population, aboveground carbon density (ACD) increased across most of the landscape over the 6‐year study period, but at fine scales, bull elephant density was the most important factor determining carbon stock change, with ACD losses recorded only where bull densities exceeded 0.5 bulls/km². Effects of bull elephants were, however, spatially restricted and landscape dependent, being especially pronounced along rivers, at mid‐elevations, and on steeper slopes. In contrast, elephant herds and abiotic drivers had a comparatively small influence on the direction or magnitude of carbon stock change. Our findings demonstrate that animals can have a substantive influence on regional‐scale carbon dynamics and warrant consideration in carbon cycling models and policy formulation aimed at carbon management and climate change mitigation.     

Linking riparian vegetation types and fluvial geomorphology along the Sabie River within the Kruger National Park, South Africa

Vegetation types were studied in relation to the fluvial geomorphology along the mixed bedrock‐alluvial Sabie River within the Kruger National Park, Mpumalanga, South Africa. Six vegetation types were identified using TWINSPAN analysis, namely: Phragmites mauritianus , Phyllanthus reticulatus , Breonadia salicina , Combretum erythrophyllum , Diospyros mespiliformis and Spirostachys africana vegetation types. Spirostachys africana and Diospyros mespiliformis vegetation types were found to occur predominantly on the stable, infrequently flooded macro‐channel banks, while the remaining four vegetation types were found almost exclusively along the more geomorphically and hydrologically dynamic macro‐channel floor. The degree of bedrock or alluvial influence was identified as being an integral factor in the distribution of the four macro‐channel floor vegetation types at both the morphological unit and the channel type scale. The geomorphological continuum from the bedrock influenced bedrock anastomosing channel types, to mixed anastomosing and pool‐rapid channel types, to the fully alluvial braided channel types, is reflected in the change in species composition from Breonadia salicina vegetation type, to Phyllanthus reticulatus and Phragmites mauritianus vegetation types, to Combretum erythrophyllum vegetation types, respectively. Given the vegetation/fluvial geomorphology links established, changes in vegetation composition are proposed in response to scenarios of geomorphological change as a result of progressive sedimentation.     

6. GENERAL NOTES

Stutterheim, C. J. 1982. Breeding biology of the Redbilled Oxpecker in the Kruger National Park. Ostrich 53:99-90.

The nest of the Redbilled Oxpecker Buphagus erythrorhynchus in the Kruger National Park is a natural hole in a tree where no excavation is required. No evidence of a territorial system WBS observed and only the nesting tree is defended. Mammal hair, dung, grass and rootlets are used for nesting material. The average clutch size was 2.8 eggs with a mean incubation period of 12,6 days. The average nestling period was 30 days. The Redbilled Oxpecker can raise three broods in a season of 176 days such as in the 1973/74 breeding season. The activity area of one breeding group was 7,0 km². The breeding unit consists of two to five birds with helpers of both sexes. All the birds in a group help to select a nest site, build the nest and feed the young. Only one male and one female participate in incubation. Post-hatching development was studied in 13 chicks.     

How widespread is woody plant encroachment in temperate Australia? Changes in woody vegetation cover in lowland woodland and coastal ecosystems in Victoria from 1989 to 2005

Aim Encroachment or densification by woody plants affects natural ecosystems around the world. Many studies have reported encroachment in temperate Australia, particularly in coastal ecosystems and grassy woodlands. However, the degree to which published studies reflect broad-scale changes is unknown because most studies intentionally sampled areas with conspicuous densification. We aimed to estimate changes in woody vegetation cover within lowland grassy woodland and coastal ecosystems in Victoria from 1989 to 2005 to determine whether published reports of recent encroachment are representative of broad-scale ecosystem changes. Location All lowland grassy woodland and coastal ecosystems (c. 6.11 × 10⁵ ha) in Victoria, Australia. Four major ecosystems were analysed: Plains woodlands, Herb-rich woodlands, Riverine woodlands and Coastal vegetation. Methods Changes in woody vegetation cover from 1989 to 2005 were assessed based on state-wide vegetation maps and Landsat analyses of woody vegetation cover conducted by the Australian Greenhouse Office’s National Carbon Accounting System. The results show changes in woody cover within mapped patches of native vegetation, rather than changes in the extent of woody vegetation resulting from clearing and revegetation. Results When pooled across all ecosystems, woody vegetation increased by 18,730 ha from 1989 to 2005. Woody cover within Riverine woodlands and within Plains woodlands each increased by >7000 ha. At the patch scale, the mean percentage cover of woody vegetation in each polygon increased by >5% in all four ecosystems: Riverine woodlands (+9.2% on average), Herb-rich woodlands (+7.6%), Plains woodlands (+6.7%) and Coastal vegetation (+5.9%). Regression models relating degree of encroachment to geographic and climatic variables were extremely weak (r² ≤ 0.026), indicating that most variation occurred at local scales rather than across broad geographic gradients. Main conclusions At the scale of observation, woody vegetation cover increased in all lowland woodland and coastal ecosystems over the 16-year period. Thus, published examples of encroachment in selected coastal and woodland patches do appear to reflect widespread increases in woody vegetation cover in these ecosystems. This densification appears to be associated with changes in land management rather than with post-fire vegetation recovery and is likely to be ongoing and long-lasting, with substantial implications for biodiversity conservation and ecosystem services.     

Dual phylogenetic origins of Nigerian lions (Panthera leo)

Lion fecal DNA extracts from four individuals each from Yankari Game Reserve and Kainji‐Lake National Park (central northeast and west Nigeria, respectively) were Sanger‐sequenced for the mitochondrial cytochrome b gene. The sequences were aligned against 61 lion reference sequences from other parts of Africa and India. The sequence data were analyzed further for the construction of phylogenetic trees using the maximum‐likelihood approach to depict phylogenetic patterns of distribution among sequences. Our results show that Nigerian lions grouped together with lions from West and Central Africa. At the smaller geographical scale, lions from Kainji‐Lake National Park in western Nigeria grouped with lions from Benin (located west of Nigeria), whereas lions from Yankari Game Reserve in central northeastern Nigeria grouped with the lion populations in Cameroon (located east of Nigeria). The finding that the two remaining lion populations in Nigeria have different phylogenetic origins is an important aspect to consider in future decisions regarding management and conservation of rapidly shrinking lion populations in West Africa.     

Use of vegetation classification and plant indicators to assess grazing abandonment in Estonian coastal wetlands

Question: What are the effects of grazing abandonment on the vegetation composition of Estonian coastal wetlands? Location: Vormsi Island and Silma Nature Reserve in western Estonia, Europe. Methods: Local knowledge and field reconnaissance were used to identify current and historical management levels of wetland sites within the west Estonian study area. Nine study sites, with varying management histories, were selected comprising an area of 287 ha. A total of 198 quadrats were taken from 43 distinct vegetation patches in five of the sites. TWINSPAN analysis was used to identify community type, and a phytosociological key was constructed for character taxa. This vegetation classification was then applied within a GIS‐based context to classify all the study sites, using a ground survey technique and 1:2000 scale air photos. Results: We identified 11 different brackish coastal wetland community types. Indicator species were defined with community characteristics for the seven main vegetation types readily recognisable in the field. Coastal wet grasslands were most extensive in grazed sites, or sites that had been more intensively grazed, while abandoned sites were largely composed of Phragmites australis stands, tall grassland, and scrub. Site variations based on vegetation composition were significantly correlated with past grazing intensity. Plant community types showed significant edaphic differences, with particularly low soil moisture and high conductivity and pH for open pioneer patches compared to other vegetation types. Conclusion: Abandonment of traditionally grazed coastal grasslands threatens their characteristic biodiversity. This study found that grazing abandonment reduced the extent of coastal wetland grasslands of particular conservation value. Nevertheless, plant species of conservation interest were found across the sequence of community types described. The study shows that grazing is an important factor influencing coastal wetland plant communities but suggests that vegetation distribution is affected by environmental variables, such as topography.     

Random Forest characterization of upland vegetation and management burning from aerial imagery

Aim The upland moorlands of Great Britain form distinctive landscapes of international conservation importance, comprising mosaics of heathland, acid grassland, blanket bog and bracken. Much of this landscape is managed by rotational burning to create gamebird habitat and there is concern over whether this is driving long‐term changes in upland vegetation communities. However, the inaccessibility and scale of uplands means that monitoring changes in vegetation and burning practices is difficult. We aim to overcome this problem by developing methods to classify aerial imagery into high‐resolution maps of dominant vegetation cover, including the distribution of burns on managed grouse moors. Location  Peak District National Park, England, UK. Methods Colour and infrared aerial photographs were classified into seven dominant land‐cover classes using the Random Forest ensemble machine learning algorithm. In addition, heather (Calluna vulgaris) was further differentiated into growth phases, including sites that were newly burnt. We then analysed the distributions of the vegetation classes and managed burning using detrended correspondence analysis. Results Classification accuracy was c. 95% and produced a 5‐m resolution map for 514 km² of moorland. Cover classes were highly aggregated and strong nonlinear effects of elevation and slope and weaker effects of aspect and bedrock type were evident in structuring moorland vegetation communities. The classification revealed the spatial distribution of managed burning and suggested that relatively steep areas may be disproportionately burnt. Main conclusions Random Forest classification of aerial imagery is an efficient method for producing high‐resolution maps of upland vegetation. These may be used to monitor long‐term changes in vegetation and management burning and infer species–environment relationships and can therefore provide an important tool for effective conservation at the landscape scale.     

Upland plant community classification in Elk Island national park,Alberta, Canada,using disturbance history and physical site factors

Our objective was to classify upland sites sampled from an area in and around Elk Island National Park, Alberta, Canada, into community types and interpret them in terms of variation in environmental factors, including fire and wild ungulate herbivory. The vegetation from 36 sites was classified into 6 community types using TWINSPAN. These types could be effectively interpreted and explained with canonical correspondence analysis ordination procedures. Based on a forward regression, topography, ungulate use and prescribed burning related closely to the first two canonical axes, accounting for 47.2 and 21.1% of the species-environment relation among sites, respectively. Unique and meaningful combinations of environmental variables influenced community-type understory composition and structure, as well as the characteristics of the tree overstory. This information provided the basis for a preliminary state and transition model of vegetation dynamics for these rangelands, which could be used to assist Park managers in manipulating plant communities within the landscape using ungulate removal and prescribed burning programs.     

Stream order controls geomorphic heterogeneity and plant distribution in a savanna landscape

We posed the question: does viewing a savanna as a network of streams linked to a matrix of terrestrial hillslopes provide a useful framework to research and understand plant distribution in these landscapes? Our study area, the Phugwane River network, lies in the semi‐arid savanna of Kruger National Park, South Africa. We examined changes in hillslope geomorphology from first‐, third‐ and fifth‐order hillslopes with regression equations. The distribution of geomorphic boundaries was enumerated by moving window analysis and the relationship between geomorphology and plant distribution was explored through ordination. First‐order hillslopes had a simple geomorphology, fewer geomorphic boundaries and a relatively homogeneous plant assemblage. By contrast, fifth‐order hillslopes were more complex in geomorphology, with more boundaries and a relatively heterogeneous vegetation pattern. Stream order classification of a savanna drainage network resulted in landscape units distinguishable by geomorphology, geomorphic boundaries and vegetation pattern. Therefore, the drainage network is a useful template to expose and organize the complexity in savanna landscapes into easily managed and researched units. This perspective should inform a shift from single‐scale phytosociological views of homogeneous vegetation units towards multi‐scale conceptualizations of savannas as water dependent ecosystems.     

The irreversible cattle-driven transformation of a seasonally flooded Australian savanna

Aim Anecdotal historical and photographic evidence suggests that woody vegetation is increasing dramatically in some northern Australian savanna habitats. Vegetation change in savannas has important implications for pastoral land‐use, conservation management, and landscape‐scale carbon storage, and informs theoretical debates about ecosystem function. This study seeks to determine the nature, extent and cause(s) of woody vegetation change in a seasonally flooded alluvial savanna habitat. Location The study area is located within the seasonally inundated alluvial zone of the tidal portion of the Victoria River, Northern Territory, Australia. The study area has been grazed by domestic stock since c. 1900, prior to which the area was inhabited and more likely regularly burnt by Aboriginal people for thousands of years. Methods Digital georeferenced aerial photographic coverages were used to examine and quantify woody vegetation change between 1948 and 1993. Transect surveys of woody and herbaceous vegetation were carried out to ground‐truth air‐photo results and determine the nature and causes of observed vegetation changes. Results There has been a dramatic increase in woody vegetation cover throughout the study area. Vegetation change patterns are roughly uniform across the full range of edaphic habitat variation and are unrelated to the depositional age of fluvial sediments. Two woody species, Eucalyptus microtheca and Excoecaria parvifolia, are predominantly responsible for observed increases. Demographic analyses reveal that woody invasions have been episodic and indicate that in most locations peak woody species establishment occurred in the mid‐1970s. Grasses are almost absent in a majority of habitats within the study area. Instead, large areas are covered by scalded soil, dense invasive weed populations, and unpalatable forbs and sedges. What grasses do occur are predominantly of very low value for grazing. The condition of the herbaceous layer renders most of the study area almost completely non‐flammable; what fires do burn are small and of low intensity. Main conclusions Multiple working hypotheses explaining observed patterns of woody vegetation increase were considered and rejected in turn. The only hypothesis consistent with the evidence is as follows: (1) observed changes are a direct consequence of extreme overgrazing by cattle, most likely when stocking rates peaked in the mid‐1970s; (2) prolonged heavy grazing effected the complete transformation of much of the herbaceous vegetation to a new state that is not flammable; and (3) in the absence of regular fire mortality, woody vegetation increased rapidly. The relatively treeless system that existed in 1948 was apparently stable and resilient to moderate grazing levels, and perhaps also to episodic heavy grazing events. However, grazing intensity in excess of a sustainable threshold has forced a transition that is irreversible in the foreseeable future. Stable‐state transitions such as this one inform debates at the heart of ecological theory, such as the nature of stability, resilience, equilibrium and carrying capacity in dynamic savanna ecosystems.