Rain forest invasion of eucalypt-dominated woodland savanna, Iron Range, north-eastern Australia: II. Rates of landscape change

Aim To explore rates of rain forest expansion and associated ecological correlates in Eucalyptus-dominated woodland savanna vegetation in north-eastern Australia, over the period 1943-91. Location Iron Range National Park and environs, north-east Queensland, Australia. This remote region supports probably the largest extent of lowland (< 300 m) rain forest extant in Australia. Rainfall (c. 1700 mm p.a.) occurs mostly between November and June, with some rain typically occurring even in the driest months July-October. Methods Interpretation of change in lowland rain forest vegetation cover was undertaken for a 140 km2area comprising complex vegetation, geology and physiography using available air photos (1943, 1970 and 1991). A GIS database was assembled comprising rain forest extent for the three time periods, geology, elevation, slope, aspect, proximity to streams and roads. Using standard GIS procedures, a sample of 6996 10 x 10 m cells (0.5% of study area) was selected randomly and attributed for vegetation structure (rain forest and non-rain forest), and landscape features. Associations of rain forest expansion with landscape features were examined with logistic regression using the subset of cells that had changed from other vegetation types to rain forest, and remained rain forest over the assessment period, and comparing them with cells that showed no change from their original, non-rain forest condition. Results Rain forest in the air photo study area increased from 45 km2in 1943 to 78.1 km2by 1970, and to 82.6 km2by 1991. Rainfall (and atmospheric CO2concentration) was markedly lower in the first assessment period (1943-70). Modelled rates of rain forest invasion differed predominantly with respect to substrate type, occurring faster on substrates possessing better moisture retention properties, and across all elevation classes. Greatest expansion, at least in the first assessment period, occurred on the most inherently infertile substrates. Expansion was little constrained by slope, aspect and proximity to streams and roads. On schist substrates, probability of invasion remained high (> 60%) over distances up to 1500 m from mature rain forest margins; on less favourable substrates (diorite, granites), probability of expansion was negligible at sites more than 400 m from mature margins. Main conclusions (i) Rain forest expansion was associated primarily with release from burning pressure from c. the 1920s, following major disruption of customary A