Dry season fires are a feature of the tropical savannas of northern Australia. As part of a landscape‐scale fire experiment, we examined the effects of fire regimes on tree survival in a tropical savanna in Kakadu National Park, northern Australia. The fire regimes were annual early dry season (June) fires, annual late dry season (September) fires, and, no fire (control). Prescriptive, experimental fires were lit annually, between 1990 and 1994, in replicate compartments, each 15–20 km2. In addition to the prescribed fires, however, one of the control compartments, which had been unburnt for seven years, was burnt by an unplanned, high intensity fire (~ 20 000 kW m−1) in September 1994. This provided an opportunity to compare the impacts on the tree stratum of frequent, prescribed burning at various intensities, and a single unplanned fire. In all fire regimes, stem survival was substantially lower than whole‐plant survival, and decreased linearly with increasing fire intensity. Significantly, stem death following the single, high intensity 20 000 kWm−1 fire (75%) was comparable to that of a regime of annual late dry season burning for five years, at an average intensity of c. 8000 kWm−1. In the high intensity unplanned fire, stem survival showed a non‐linear response to stem size, being least in the small (< 10 cm DBH) and large (> 40 cm DBH) size classes, and highest in the intermediate size classes. Stem survival was also species‐dependent, being higher in the dominant Eucalyptus miniata than in the subdominant, broad‐leaf deciduous trees. In the absence of fire for 5–10 years, the structure and composition of the tree stratum of these savannas tends to become more complex than in sites burnt more frequently, especially by high intensity fire. Such a long‐term absence of fire may be a conservation objective for some areas of savanna. However, build‐up of fuel to near maximal levels can occur in 2–4 years without fire. This may predispose the savannas to high‐intensity, late dry season fires. Whatever the fire‐management goal within a given patch of savanna, whether it be the prescribed use of fire on a biennial basis, or the exclusion of fire at a semidecadal scale, careful attention still needs to be given to the consequences of fuel build‐up in fire‐excluded sites.

In a landscape-scale experiment, fires were lit in replicate catchments 15-20 km2 in area, either early in the dry season (June) or late in the dry season (September) between 1990 and 1994. For each fire, Byram-intensity was determined in representative one ha areas of Eucalyptus miniata – E. tetrodonta open-forest, with a ground stratum dominated by annual grasses. Fuel weights were measured by harvest, fuel heat content was assumed to be constant, and the rate of spread was determined using electronic timers. Fuels consisted primarily of grass and leaf litter, and ranged from 1.5 to 13 t ha-1; in most years, average fuel loads were 2-4 t ha-1. Rates of spread were generally in the range of 0.2-0.8 ms-1. The mean intensity of early dry season fires (2100 kW m-1) was significantly less than that of the late dry season fires (7700 kW m-1), primarily because, in the late dry season, there was more leaf litter, fuels were drier, and fire weather was more extreme. Crown fires, a feature of forest fires of high intensity in southeastern Australia, were not observed in the Kapalga fires. Fire intensity was a very good predictor of both leaf-char height and leaf-scorch height for fires between 100 kW m-1 and 10,000 kW m-1, the range in which the majority of experimental fires fell.

Changes in vegetation composition and structure are described for grassland and heathland communities on the Bogong High Plains, in the Victorian Alpine National Park. The data are based on long-term records collected from permanent reference plots over the period 1945 to 1994 from plots established in 1945, 1946 and 1979. In the Pretty Valley grassland plots, established in 1946, cattle grazing has prevented the large-scale regeneration of a number of tall, palatable forbs and short, palatable shrubs, while in the absence of grazing, the cover of these Life forms increased substantially. The amount of bare ground and loose litter was significantly greater on the grazed compared with the ungrazed plot. Between 1979 and 1994, there was little or no identifiable trend in the cover of Vegetation or bare ground at either the Pretty Valley grazed site, or two additional grazed grassland sites established nearby in 1979. The current condition of grazed grassland on the Bogong High Plains is interpreted as stable, yet degraded. Improvement in condition will occur in the absence of grazing. In the Rocky Valley open heathland plots, established in 1945, increases in shrub cover over the study period were due to growth of shrubs following the 1939 bushfires that burnt much of the Bogong High Plains. From 1945-1979 shorter-lived shrubs increased in cover; since 1979, these shrubs have senesced, and are being replaced mainly by grasses. On the grazed plot longer lived, taller shrubs have continued to increase in cover and are not senescing. Between 1979 and 1989, total shrub cover declined on the ungrazed plot, but increased on the grazed plot. There was no evidence that grazing has reduced shrub cover, and therefore potential fire risk, in open heathland. These findings have significant management implications for the Alpine National Park and are consistent with those from other regions in the Australian alps.

In this study, we quantify the density, cover and obstruction width of vegetation patches, the roughness of landscape surfaces, and the diversities of plants and grasshoppers with distance from cattle watering- points. We used distance from water as a surrogate for a gradient in grazing pressure. Fourteen study sites were located in the Victoria River District of northern Australia, seven from a water point on Kidman Springs Station on calcareous red loam soils and seven from a water point at Mount Sanford Station on craclcing-clay black soils. At each study site transect lines were oriented within the landscape to run downslope (i.e. in the direction of flows of run-off). We measured the intercept length and obstruction width of perennial vegetation patches along these lines. Plant diversity was measured in quadrats positioned along each line and grasshopper diversity was determined by species counts on each site. We also surveyed the roughness of the landscape surface along each line. A rough surface will tend to slow run-off. hence increase time for water infiltration and soil-water storage. Surface roughness declined near water, as did the density, cover and obstruction width of perennial vegetation patches. Grasshopper and plant species richness also declined near water. These declines suggest strong linkages between landscape filmtion. biodiversity and inipacts of cattle grazing and trampling. Cattle will always create 'sacrifice zones' around watering-points. However, the area of this impact on function and diversity can be minimised by managing the timing and intensity of paddock use. Key words: grazing gradients. Kidman Springs, landscape function. Mount Sanford, piosphere, surface roughness, tropical grasslands, tropical savannas

The reproductive phenology-the annual and spatial variation in tree fecundity, ovule development and seedfall-of Eucalyptus miniata Cunn. ex Schauer and E. tetrodonta F. Muell. is described at a tropical savanna site in northern Australia from 1992 to 1994. There was substantial inter-annual variability in fecundity of both species at the individual and the population level. The proportion of ovules which survived did not vary significantly between years. Seed production and seedfall varied substantially between years with large seed yields in 1994 and low seed yields in 1993. Eucalyptus miniata and E. tetrodonta are separated in time (by 2-8 weeks) with respect to peak periods for all key phenology events-budding, flowering and seedfall. Both species are non-serotinous with all seed shed within 8 months of ovule initiation. Peak seedfall occurred up to 1 month before the first wet season rains for E. miniata and coincided with these rains for E. tetrodonta. Post-dispersal conditions for germination and establishment are more likely than seed supply to limit seedling establishment of E. miniata and E. tetrodonta.