Statistical analysis of triggered landslides: implications for earthquake and weathering controls.
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
We first aim to review the external perturbations which can lead to landslide failure. We review the perturbations and associated processes in five sections: i) increase of slope angle, ii) increase of load applied on the slope, iii) rise of groundwater level and pore pressure, iv) frost weathering processes and v) earthquake loading. Second, we analyse the New Zealand landslide catalogue, integrating all landslides recorded in New Zealand between 1996 and 2004. We analyse the New Zealand landslide series in time and rate and find a strong correlation in landslide occurrences. The time correlation found between landslide occurrences for events occurring more than 10 days apart is not found to be driven by the earthquake-landslide nor the landslide-landslide interactions. We suggest the climate-landslide interactions drive, non-linearly and beyond the empirically reported daily correlation, most of New Zealand landslide dynamics. Third, we compare the landslide dynamics in time, space and rate of New Zealand, Yosemite cliffs (California, USA), Grenoble cliffs (Is`ere, French Alps), Val d’Arly cliffs (Haute-Savoie, French Alps), Australia and Wollongong (New South Wales, Australia). Landslides are resolved as correlated to each other in time for all catalogues. The New Zealand, Yosemite, Australia and Wollongong landslide daily rates are well fitted by a power law for rates between 1 and 1000 events per day, suggesting that the same mechanism(s) are driving both the large landslide daily crises and the single events. The joint analysis of the six catalogues permits to derive parameters that allow to sort the relative landslide dynamics in each of the six areas. From the most re-active landslide area (New Zealand) to the less re-active area (Grenoble), the global trend of the different parameters are: i) decreasing departure from randomness; ii) decreasing maximum daily rates and area over which the trigger operates; iii) decreasing landslide triggering for landslides occurring one day apart; iv) decreasing global interaction to earthquake, rainfall and temperature. Fourth, we compare earthquake aftershock space distributions with landslide space distributions triggered by the Chi-Chi MW7.6 earthquake (Taiwan), by the MW7.6 Kashmir earthquake (Pakistan), by the MW7.2 Fiordland earthquake (New Zealand), by the MW6.6 Northridge earthquake (California) and by the MW5.6 Rotoehu earthquake (New Zealand). We resolve the seismic aftershock and landslide normalised number of events to display roughly similar patterns with distances, when comparing the landslide distributions to their aftershock distribution counterparts. When comparing the five landslide - aftershock distribution pairs for a given mainshock, we do not resolve a clear common pattern. Then we compare landslide and aftershock distance distributions to ground motion observations (Peak Ground Acceleration, Peak Ground Velocity and Peak Ground Displacement) and we find no linear scaling of the number of landslides or aftershocks with none of the ground motion variables. We suggest that landslides and aftershocks are driven by the same mechanisms and shed light on the Peak Ground Displacement and associated static stress changes on landslide triggering.