Modelling landscape leakiness and sediment yields from savanna hillslopes: the critical role of vegetation configuration J.A. Ludwig a , R. Bartley b and A.C. Liedloff c a Tropical Savannas Management CRC and CSIRO Sustainable Ecosystems, PO Box 780, Atherton, Queensland 4883, Australia, john.ludwig@csiro.au b CSIRO Land and Water, PO Box 780, Atherton, Queensland 4883, Australia c Tropical Savannas Management CRC and CSIRO Sustainable Ecosystems, PMB 44 Winnellie, Northern Territory 0822, Australia Keywords: Eco-hydrology; erosion; runoff; soil surface condition; vegetation patchiness EXTENDED ABSTRACT The potential of a savanna hillslope to retain, not ‘leak’, vital soil sediments, has been modelled using a simple landscape leakiness index. This index is very sensitive to the configuration of vegetation cover on the hillslope. We asked the question: does the sensitivity of this index match reality? Runoff and sediment yield data were collected from flumes at the bottom of two sites on a savanna hillslope located near Charters Towers, Queensland, Australia. The two sites were within 200 m of each other and have the same general soils, vegetation and topography – the key difference was the spatial configuration of their vegetation cover. One site had a relatively high and uniform grass cover over the entire hillslope (34% in 2005). The other site had an even higher average grass cover (47% in 2005), but it had a patch of bare soil low on the hillslope. Over three wet- seasons, from 2003-2005, the uniformly grassy site lost almost no sediment (≤0.06 t/ha/y) whereas the site with a bare lower slope lost from 2.0 to 3.1 t/ha/y. The modelled landscape leakiness index, known as the cover-based directional leakiness index, CDLI, indicated that the potential for the uniformly grassy site to leak sediment was low (CDLI = 0.21). The index for the site with a bare patch toward the bottom of its slope was much higher (CDLI = 0.71), indicating a more leaky hillslope. This finding confirms the logic of having CDLI sensitive to the spatial configuration of cover. However, this sensitivity, and the generality, of CDLI remains to be tested for other kinds of patch configurations and vegetation types. We know that CDLI is limited to landscapes where runoff flows are by surface sheeting, not channelling. Thus, CDLI is only likely to apply to relatively gentle and uniform landscapes. We are developing a new landscape leakiness index (LI) that combines remotely sensed vegetation cover and digital elevation data. This new index should be applicable to open landscapes with rough terrain, such as those found in arid and semiarid rangelands around the globe. 361