Simple rules to minimise exposure to coseismic landslide hazard

Abstract

Abstract. Landslides constitute a hazard to life and infrastructure and their risk is mitigated primarily by reducing exposure. This requires information on landslide hazard on a scale that can enable informed decisions. Such information is often unavailable to, or not easily interpreted by, those who might need it most (e.g. householders, local governments and non-governmental organisations). To address this shortcoming, we develop simple rules to minimise exposure to coseismic landslide hazard that are understandable, communicable and memorable, and that require no prior knowledge, skills or equipment to apply. We examine rules based on two common metrics of landslide hazard, (1) local slope and (2) upslope contributing area as a proxy for hillslope location relative to rivers or ridge crests. In addition, we introduce and test two new metrics: the maximum angle to the skyline and the hazard area, defined as the upslope area with slope >40∘ from which landslide debris can reach a location without passing over a slope of <10∘. We then test the skill with which each metric can identify landslide hazard – defined as the probability of being hit by a landslide – using inventories of landslides triggered by six earthquakes that occurred between 1993 and 2015. We find that the maximum skyline angle and hazard area provide the most skilful predictions, and these results form the basis for two simple rules: “minimise your maximum angle to the skyline” and “avoid steep (>10∘) channels with many steep (>40∘) areas that are upslope”. Because local slope alone is also a skilful predictor of landslide hazard, we can formulate a third rule as “minimise the angle of the slope under your feet, especially on steep hillsides, but not at the expense of increasing skyline angle or hazard area”. In contrast, the upslope contributing area has a weaker and more complex relationship to hazard than the other predictors. Our simple rules complement but do not replace detailed site-specific investigation: they can be used for initial estimations of landslide hazard or to guide decision-making in the absence of any other information.

Insights from the topographic characteristics of a large global catalog of rainfall-induced landslide event inventories

“…It has become common practice to map areas affected by landslide-triggering earthquakes to build a spatially complete picture of landslide impacts [Tanyas et al 2017], and the inventories that are generated have been used to produce hazard maps [Jibson et al 2000, Harp et al 2011, susceptibility models [Garcia-Rodriguez et al 2008, Xu et al 2012, guidelines for hazard zonation [Milledge et al 2019] and global alerting systems [Nowicki Jessee et al 2018]. Landslide-event inventories are also required to explore the landscape response to tectonic and climatic forcings [e.g., Malamud et al 2004, Korup et al 2012, Marc et al, 2016.…”

“…For the assessment of each parameter by itself, we calculate the relative ratios of the distributions for each variable for the topography and the landslide populations. This follows the example of other recent studies [Marc et al 2018, Milledge et al 2019]. For both the landslide parameter distribution and the parameter distribution for the topography, we divide the values into bins, normalizing by the total size of the distribution.…”

“…CC BY 4.0 License. the landslide and topography data by the median value for the parameter in question prior to splitting the data into bins [Marc et al 2018, Milledge et al 2019. For example, in the case of slope, we calculate the median slope value for each inventory and divide the each binned interval by the median slope value calculated across the mapped area.…”

Automated landslide detection outperforms manual mapping for several recent large earthquakes

“…Even in the case of susceptibility mapping, the impact of discrepancy between inventories may be less severe than the direct pixel-based comparison suggests. Sensitivity analyses for a range of locations and of landslide hazard and susceptibility models indicate that both the statistics of key causative factors (Milledge et al, 2019) and the resultant hazard maps (Ardizzone et al, 2002) remain remarkably consistent between inventories despite considerable locational mismatch (i.e. <50% overlap between inventories).…”

The Preservation of Climate‐Driven Landslide Dams in Western Oregon

“…To isolate whether dams occur at characteristic drainage areas and valley widths relative to those expected from the SN, we calculated the ratio of probability, PDF R , which is akin to a relative hazard (e.g., Milledge et al, 2019;Rault et al, 2019) (Rault et al, 2019). We then plotted the 90% confidence intervals for these randomly generated PDFs and noted where PDF R is greater than the 90% confidence interval.…”