Assessing the effects of fire on vegetation in tropical savannas

Because of its high frequency and generally low intensity, fire in tropical savannas appears to be a different phenomenon from that in other biomes. A recent study of fire in savanna at Munmarlary in northern Australia, analysed by detrended correspondence analysis (DCA) concluded that different fire regimens resulted in negligible changes in the vegetation, a conclusion crucial for fire management in the region. Here, we describe the short‐term impact of an unusually intense fire in an area of tropical open forest. Tree and shrub mortality of 14.3% was recorded within 6 months of the fire, and the composition of the vegetation was changed because of differences between species in mortality rates, which ranged from 4 to 90%. Analyses of variance (ANOVA) of DCA co‐ordinates were unable to detect any change, however. DCA seems inappropriate for analysing vegetation changes after savanna fires, because the floristic changes, compared with those in temperate fire‐prone ecosystems, are subtle and multidirectional. Further, it is shown that rather large plot sizes (2–4 ha) are likely to be required to detect fire treatment differences even as great as about 20% of the mean, given the variability of savanna vegetation, and replicates that are likely to be limited in number. A possible solution is to measure the change over time in permanent plots, rather than attempting to detect treatment differences by sampling on a single occasion.