Tree populations in the wooded savannas of northern Australia lack a well developed seed bank, but instead rely on a persistent ground layer of suppressed juvenile plants <1 m of indeterminate ages. The feral Asian water buffalo has been implicated as a factor in inhibiting the movement of juvenile trees to sapling stage. In a 6-year field study in Kakadu National Park, 656 juvenile plants were monitored for growth and survival, and in exploring the effects of buffalo, clipping surrounding vegetation and ambient fire. Buffalo were removed from one half of the sites across an environmental gradient in a landscape-scale experiment. Juvenile trees grew more slowly but fewer became dormant where buffalo were absent. Experimental clipping of surrounding vegetation produced significantly increased height at the buffalo-absent sites, but not at the buffalo-present sites, demonstrating that the positive effect of buffalo on growth was indirect, by reducing the competitive regime for juvenile trees. Within 3 years of removal of buffalo, ground-level green biomass increased 2–10 times and litter 1.3–2 times over a topographic gradient. Mortality of juvenile trees was not significantly different between buffalo-present and buffalo-absent sites when plots remained unburnt. By contrast, at buffalo-absent sites, three times more juveniles died after late dry-season fires than at unburnt sites, but there was no difference among buffalo-present sites. Overall, the total juvenile tree bank decreased by 7% where buffalo grazed and by 18% where buffalo had been removed. The study demonstrated a major mechanism(s) responsible for recorded changes in vegetation patterns of these savannas, whereby buffalo initiate a cascade of effects by changing ground-level biomass, which change competitive relationships and fuel loads, which then have an impact on tree growth and demography. The results are discussed with respect to dynamics of the juvenile tree bank and implications for long-term sustainability of these wooded savannas.

No abstract

The distributions of the 567 plant species considered to be endemic to the Northern Territory, Australia, were collated from a distributional database comprising about 600 000 records. Endemic species comprise a non-random taxonomic subset of all plants known from the Northern Territory. Because of substantial geographic disparity in collecting effort, we analysed geographic patterning of these endemic species by using both (1) actual records only and (2) interpolated ranges (minimum convex polygons). The geographic distribution of the number of Northern Territory endemic plant species was well predicted by a measure of topographic complexity and climate (particularly rainfall). The observed distributional patterning of endemic species was also influenced by survey effort, but this latter influence was substantially reduced by the use of minimum convex polygons. Both analyses revealed that there was a clear aggregation of endemic species in the 32 000 km2 of the sandstone plateau of western Arnhem Land. This ‘hotspot’ has been previously recognised in coarser-scale assessments of national and international centres of plant biodiversity. Our analysis concluded that 172 species are restricted to this plateau, and that the plateau comprised at least 90% of the distribution of a further 25 species. More broadly, 438 plant species are endemic to the northern part of the Northern Territory (the 316 000 km2 north of 16°S), a level of endemism that may match that of Cape York Peninsula and surpasses that of the Kimberley. The core area for Northern Territory endemic plants, the plateau of western Arnhem Land, is currently threatened, particularly by unfavourable fire regimes.

Litterfall from a Melaleuca forest was investigated as part of chemical cycling studies on the Magela Creek floodplain in tropical, northern Australia. The forest contained two species of tree, Melaleuca cajaputi and Melaleuca viridiflora, with a combined average density of 294 trees ha -l. The M. viridiflora trees had diameter breast height measurements ranging from 11.8 to 62.0cm, median class 25.1-30.0cm and a mean value of 29.2_+ 1.0cm, compared to 13.0 to 66.3 cm, 30.1-35.0cm and 33.5 +_ 1.0cm for M. cajaputi trees. A regression model between tree height, diameter breast height and fresh weight was determined and used to calculate average tree weights of 775_ 1.6kg for M. viridiflora and 1009_+ 1.6kg for M. cajaputi, and a total above-ground fresh weight of 263_+ 0.3t ha -1. The weight of litter recorded each month on the ground beneath the tree canopy ranged from 582-+ 103 to 2176_ 376 g m -2 with a monthly mean value of 1105 -+ 51 g m -2. The coefficient of variation of 52% on this mean indicates the large spatial and temporal variability in litter distribution over the study site. This variability was greatly affected by the pattern of water flow and litter transport during the Wet season. Litterfall from the trees was evaluated using two techniques -nets and trays. The results from these techniques were not significantly different with annual litterfall collected in the nets being 705 _+ 25 g m -2 and in the trays 716-+ 49 g m -2. The maximum monthly amount of litterfall, 108_+ 55 g m -E, occurred during the Dry season months of June-July. Leaf material comprised 70% of the total annual weight of litter, 480-+ 29 g m -2 in the nets and 495 _+ 21 g m -2 in the trays. The tree density and weight of litter suggest that the Melaleuca forests are highly productive and contribute a large amount of material to the detrital/debris turnover cycle on the floodplain.