|Abstract||Aim To quantify the regional-scale spatio-temporal relationships among rainfall, vegetation and fire frequency in the Australian wet–dry tropics (AWDT).
Location Northern Australia: Cape York Peninsula, central Arnhem, central Kimberly, Einasleigh Uplands, Gulf Fall Uplands and northern Kimberley.
Methods Monthly ‘fraction of photosynthetic active radiation absorbed by green vegetation’ (fAPAR) was decomposed into monthly evergreen (EG) and monthly raingreen (RG) components using time-series techniques applied to monthly normalized difference vegetation index (NDVI) data from Advanced Very High Resolution Radiometer (AVHRR) imagery. Fire affected areas were independently mapped at the same spatio-temporal resolution from AVHRR imagery. Weather station records were spatially interpolated to create monthly rainfall surfaces. Vegetation structural classes were derived from a digitized map of northern Australian vegetation communities (1 : 1,000,000). Generalized linear models were used to quantify relationships among the fAPAR, EG and RG signals, vegetation structure, rainfall and fire frequency, for the period November 1996–December 2001.
Results The fAPAR and EG signals are positively correlated with annual rainfall and canopy cover, notably: EGclosed forest > EGopen heathland > EGopen forest > EGwoodland > EGopen woodland > EGlow woodland > EGlow open woodland > EGopen grassland. Vegetation height and fAPAR are positively correlated, excluding the special case of open heathland. The RG signal is highest where intermediate annual rainfall and strong seasonality in rainfall coincide, and is associated with vegetation structure as follows: RGopen grassland > RGwoodland > RGopen forest > RGopen heathland > RGlow woodland > RGopen woodland > RGlow open woodland > RGclosed forest. Monthly RG tracks monthly rainfall. Annual proportion of area burnt (PB) is maximal where high RG coincides with low EG (open grassland, several woodland communities). PB is minimal in vegetation where both RG and EG are low (low open woodland); and in vegetation where EG is high (closed forest, open heathland).
Conclusions The RG–EG scheme successfully reflects digitally mapped tree and grass covers in relation to rainfall. RG–EG patterns are strongly associated with fire frequency patterns. PB is maximal in areas of high RG, where high biomass production during the wet season supports abundant fine fuel during the dry season. PB is minimal in areas with high EG, where relatively moist fuel limits fire ignition; and in areas with low EG and RG, where a relative short supply of fuel limits fire spread. |