Abstract We report erosion rates determined from in situ produced cosmogenic 10 Be across a spectrum of Australian climatic zones, from the soil-mantled SE Australian escarpment through semi-arid bedrock ranges of southern and central Australia, to soil-mantled ridges at a monsoonal tropical site near the Arnhem escarpment. Climate has a major effect on the balance between erosion and transport and also on erosion rate: the highest rates, averaging 35 m Ma −1 , were from soil-mantled, transport-limited spurs in the humid temperate region around the base of the SE escarpment; the lowest, averaging about 1.5 m Ma −1 , were from the steep, weathering-limited, rocky slopes of Kings Canyon and Mt Sonder in semi-arid central Australia. Between these extremes, other factors come into play including rock-type, slope, and recruitment of vegetation. We measured intermediate average erosion rates from rocky slopes in the semi-arid Flinders and MacDonnell ranges, and from soil-mantled sites at both semi-arid Tyler Pass in central Australia and the tropical monsoonal site. At soil-mantled sites in both the SE and tropical north, soil production generally declines exponentially with increasing soil thickness, although at the tropical site this relationship does not persist under thin soil thicknesses and the relationship here is ‘humped’. Results from Tyler Pass show uniform soil thicknesses and soil production rates of about 6.5 m Ma −1 , supporting a longstanding hypothesis that equilibrium, soil-mantled hillslopes erode in concert with stream incision and form convex-up spurs of constant curvature. Moreover, weathering-limited slopes and spurs also occur in the same region: the average erosion rate for rocky sandstone spurs at Glen Helen is 7 m Ma −1 , similar to the Tyler Pass soil-mantled slopes, whereas the average rate for high, quartzite spurs at Mount Sonder is 1.8 m Ma −1 . The extremely low rates measured across bedrock-dominated landscapes suggest that the ridge–valley topography observed today is likely to have been shaped as long ago as the Late Miocene. These rates and processes quantified across different, undisturbed landscapes provide critical data for landscape evolution models.

Numerical models are used to help constrain empirical parameterizations of soil production and transport mechanisms on soil-mantled hillslopes. The neighbourhood algorithm is used to invert soil thickness versus surface curvature data to provide not only more rigorous estimates of model parameter values but determine which of the model parameters are constrained by cosmogenic exposure ages ( Heimsath et al., 2000 ). We show that linear and depth-dependent creep constants can be constrained by simple geomorphometric measurements, such as the distribution of soil thickness on the landform and its relationship to surface curvature. We also show that this unique data set cannot be used on its own to constrain the parameterization of overland flow, another transport mechanism that is thought to play an important role on soil-mantled hillslopes, or to determine if a soil distribution has reached local steady state. We also demonstrate that, to explain the data, soil production must be a function of soil thickness. These conclusions have important implications for our understanding of landscape evolution on medium to long time scales.