Environmental and allometric drivers of tree growth rates in a north Australian savanna

Frequent fire is a feature of the mesic Australian savannas, yet little is known about its effect on tree growth rates. Growth data are critical for the development of demographic models to understand the consequences of different land management regimes. We used generalized linear mixed effects models and information-theoretic multi-model inference to analyse annual diameter growth of adult trees (¡Ý5 cm diameter at breast height, DBH) and saplings (¡Ý1.5 m height and <5 cm DBH) and height growth of juveniles (<1.5 m height) in response to fire, tree size, stand basal area, annual rainfall, and for adult trees, the presence of the introduced Asian swamp buffalo (a potentially growth-limiting herbivore). The analysis was based on two medium-term datasets from natural eucalypt savanna in Kakadu National Park in the Australian monsoonal tropics. One dataset was based on a 7-year experiment examining the effect of buffalo removal during which three fires occurred, and the second was derived from a 4-year study in which three experimental fire treatments (annual early dry season fire, annual late dry season fire and no fire) were imposed. An unplanned, extremely severe fire burnt some of the previously unburnt trees in the final year of this latter experiment, so we considered it as an additional fire treatment. Despite high variance, there was strong evidence for differential effects of the fire treatments on tree growth, and these effects differed among the three size groups. Late and extreme fires resulted in a high proportion of individuals with negative DBH increments (due to burning or shedding of bark, and also to stems being killed and replaced by small stems), and decreased average DBH growth of saplings (−0.17 and −1.38 cm year−1, respectively¡ªcf. 0.22 cm year−1 for no fire) and adult trees (0.01 and 0.03 cm year−1, respectively¡ªcf. 0.18 cm year−1 for no fire). Early fires decreased DBH growth of saplings (0.11¡ªcf. 0.22 cm year−1 for no fire) but increased growth of adult trees (0.29¡ªcf. 0.18 cm year−1). Height growth of juvenile trees was reduced by early fires (0.01¡ªcf. 0.06 m year−1 for no fire) but was increased by late fires (0.10 m year−1), probably because juveniles are physiologically active early in the dry season, but are effectively dormant in the late dry season. When stand basal area was high there was evidence of growth suppression in adults, saplings and juveniles. Growth of adult trees was lower in years with high rainfall and where buffalo were removed, possibly because of increased competition from ground layer vegetation. Our study casts doubt on the sustainability of the frequent, early dry season fires that have become a management goal in much of northern Australia.