The effects of the experimental fire treatments (relative to unburnt) at Kapalga can be broadly characterized as follows:

• Late - increased tree mortality (potential reduction in tree density by up to 20%); significant but surprisingly limited effect on biodiversity
• Early - little effect on tree mortality, but significant effects on biodiversity
• Progressive - few unique results

Two recurring results emerge from Kapalga. First, although our experimental treatments encompassed the most extreme fire regimes possible, many plants and animals showed little or no response at all. Much of the savanna biota is therefore highly resilient to fire, and their dynamics are driven primarily by other factors. Notable exceptions to this are riparian vegetation and associated stream biota. Second, in many cases where fire had a marked effect, the contrast was between burnt versus unburnt, rather than between high versus low intensity. This indicates that fire frequency is a particularly important factor, more so than previously recognised.

There are two clear management implications from these results. First, the overall resilience of the savanna biota in relation to fire gives managers considerable flexibility. The savanna biota is fundamentally driven by other factors (such as patterns of rainfall), so it is not a case of one 'mistake' with fire causing long-term or possibly irreversible damage. This is unlike the situation for particularly fire-sensitive habitats, including riparian vegetation.

Despite such general resilience, fire still has important effects on the savanna biota, and needs to be managed accordingly. The second management implication from Kapalga is that fire frequency needs to be given more consideration than has previously been the case. This is a particularly relevant issue in the Top End, where most of the country is burnt every one or two years. Our Kapalga results indicate that plant recruitment and biodiversity can only be optimised if substantial areas of savanna are maintained with a fire frequency of once every 3-5 years. However, given the high incidence of unauthorised burning, it is not clear how fire can be best managed to achieve this goal. CSIRO is currently addressing this question.

Kapalga has provided important information for conservation managers, but many questions still remain. There will always be more to learn about the ecological effects of fire in northern Australia, which means that conservation managers can never be totally confident that they have got it right. It is therefore critically important that effective monitoring strategies are implemented, such that management is continually refined by new information in an adaptive management framework.

In collaboration with partners through the NT Bushfires Council and the proposed Tropical Savannas Management CRC, CSIRO is embarking on a new phase of fire research in northern Australia. The priority areas of this research are: 1. Further field and modelling work on fire and tree dynamics; 2. Management options for maintaining substantial areas with relatively low fire frequency; 3. The use and biodiversity impacts of wet season burning; and 4. The extent to which results from Kapalga are more generally applicable.

Finally, a number of philosophical issues relating to fire management for nature conservation need more thought. For example, it is not always clear what are 'good' or 'bad' ecological outcomes. Moreover, management needs to approached from a regional scale, such that the 'generally' best fire regime is not necessarily best at all.