2014 GSA Annual Meeting in Vancouver, British Columbia (19–22 October 2014)

Paper No. 248-5
Presentation Time: 2:10 PM

LANDSLIDES THAT LIQUEFY:  IMPLICATIONS OF THE 2014 OSO DISASTER


IVERSON, Richard M. and GEORGE, David L., Cascades Volcano Observatory, U.S. Geological Survey, 1300 SE Cardinal Ct. #100, Vancouver, WA 98683

Some landslides move slowly or intermittently downslope, whereas others accelerate catastrophically and run out long distances across flat or gently sloping terrain. The extraordinary runout of a debris avalanche / flow (DAF) that caused more than 40 fatalities at Oso, Washington on 22 March 2014 can be gauged by empirical indices and also examined mechanistically. The simplest and best-known index compares the total elevation change of the landslide path, H, and the total horizontal distance covered by the path, L. The Oso DAF had a volume of about 8 million cubic meters, H = 180 m, L = 1700 m, and H/L = 0.106, ranking it as one of the most mobile landslides in history. By contrast, several previous episodes of landsliding at the Oso site (most recently the 2006 “Hazel landslide”) had moderate H/L values (> 0.3). Diverse field evidence indicates that the long runout of the 2014 Oso DAF can be attributed to landslide liquefaction and/or liquefaction of the overridden Stillaguamish River floodplain. Both sediment bodies were unusually wet in March 2014 owing to months of exceptionally heavy antecedent precipitation.

Our mechanistic examination of the Oso DAF behavior holds a key implication for evaluation of landslide hazards. We simulated the landslide’s dynamics using our recently developed model (“D-Claw”), which combines physical conservation laws with principles of critical-state soil mechanics, granular mechanics and fluid mechanics. Our results show that the runout of the 2014 Oso DAF can be explained if the wet landslide material had a loose initial state that made it subject to contractive shearing and consequent liquefaction during slope failure. Alternative D-Claw simulations show that the landslide would have been far less mobile if its initial porosity and water content had been only a few percent smaller. In the contractive case, the simulated landslide crosses the 1-km-wide Stillaguamish River floodplain in about 60 s, but in the alternative case it moves only a small fraction of this distance before stopping – similar to the behavior of the 2006 Hazel landslide. This bifurcation of landslide behavior, contingent on nuanced differences in initial conditions, demonstrates that landslide hazard evaluation differs greatly from landslide hazard recognition.