An experimental and numerical study of a landslide triggered by an extreme rainfall event in northern Australia

In February/March 2007, an extreme rainfall event occurred in the Jabiru region of the Northern Territory of Australia. Rainfall of 784 mm fell in a 72-h period. This rainfall event resulted in 49 separate landslides occurring in the adjacent, but remote and inaccessible region of Arnhem Land. The landslides were extensively mapped and characterised. A common feature of the landslides was their relatively surficial nature. This paper reports on laboratory and field tests to characterise the material properties of the slide material and the underlying, more competent material. One particular, large and relatively accessible landslide was chosen for detailed investigation. The experimental data are used to carry out seepage and slope stability analyses, taking account of changes in the degree of saturation (and thus the negative pore water pressure or suction) in the slope material during the rainfall event in question. Using a parametric study in which various material parameters were varied around the measured mean values, it is shown that the failure of this particular slope could have been predicted using relatively straightforward seepage and limit equilibrium slope stability analyses, coupled with the relevant rainfall data, as long as the contribution of matric suction to the engineering characteristics of the slope material was accounted for. The work also highlights the importance of in situ conditions at the time a particular rainfall event (particularly an extreme event such as that considered in this paper) occurs. If the slope has a relatively high degree of saturation, manifested as a low initial in situ suction, it is more susceptible to rainfall triggering a slope failure. Although this observation is not novel, the investigation described in this paper confirms the importance of ambient in situ conditions and provides an indication of how the likelihood of landslide occurrence at this particular site may in the future be quantified, i.e. by focussing on antecedent rainfall history.