1 Analysis of digitized aerial photographs taken in 1941 and 1994, using image processing and geographical information system technology, enabled the quantification of change in the coverage of forest and grassland patches that occur within a Eucalyptus savanna matrix in a subcoastal region of the Australian monsoon tropics. The 3058 ha study area was orientated along a low escarpment that separated a sandstone plateau from lowlands that comprised 58% and 42% of the area, respectively. 2 In the 53-year period, humans modified less than 1% of the study area, primarily for road building, and primarily in savanna areas. More than 85% of the study area at both sample times was covered by savanna. However, over the same period, the forest coverage increased from 5.03% to 9.91% of the study area and coverage of grassland decreased from 6.70%, to 2.47%. The aerial photography also showed that tree density in the savanna had increased. although this was not assessed quantitatively. 3 There was an increase in the number of forest patches from 116 to 142. The number of grass an patches decreased (particularly those > 1 ha) from 87 to 59, although the size class distribution of forest and grassland patches was statistically similar for both sample times. 4 A 50-m GIS buffer was used to distinguish creek-lines environments from surrounding catchments. Using this criterion, 14% of the study area was classified as plateau creek-lines and 9% lowland creek-lines. Although the expansion of forest and loss of grassland varied significantly amongst catchment and creek-lines on the plateau and lowlands, the 1941 rank order of coverage of each vegetation type was maintained in these four landscape categories in 1994. In both years the greatest extent of forest and grassland occurred on the lowland catchments, despite their accounting for only one-third of the study area. 5 Transition matrices for vegetation change among the four landscape categories demonstrated that, unlike the other vegetation types, grasslands, particularly on the plateau, had a low probability of remaining unchanged during the study period. 6 The cause(s) of the overall increase of woody biomass across the topographic and edaphic gradient remains unclear but may be related to a period of increased rainfall since the 1970s, as well as to the cessation of Aboriginal landscape burning at the beginning of the study period.