Reducing uncertainty in building vulnerability and impact assessment for tephra hazards

Abstract

Tephra hazards from large explosive eruptions can damage and destroy thousands of buildings at once, presenting significant challenges to post-eruption recovery efforts. These impacts can be better anticipated and reduced using insights from pre-eruption building impact assessments. However, our ability to make accurate and informative impact assessments is currently limited by a relatively poor understanding of how buildings perform when exposed to tephra hazards of varying severities. This thesis addresses several key issues with the aim of reducing uncertainty in future tephra hazard impact assessments. To increase the pool of empirical data available for forecasting likely impacts, a new method is developed to conduct remote tephra fall building damage and vulnerability assessments. Using this method, tephra fall building vulnerability models are developed that are the first specifically for buildings common throughout Indonesia. These are also the first models to consider damage less than severe roof collapse, showing that partial roof collapses can occur at tephra loads < 50 kg m-2 (~2.5 – 8 cm thick deposits, depending on density). To make more efficient use of limited data, statistical vulnerability assessment techniques commonly applied in research on other natural hazards were applied to volcanic hazards for the first time. These techniques enabled transparent development of vulnerability models and were used to quantitatively assess, for the first time, how the interaction of multiple types of tephra hazards can influence building vulnerability. Specifically, it was found that compared to when there is no tephra deposit, the presence of a 5 cm thick layer of tephra on reinforced concrete or clay tiled roofs can triple the kinetic energy required for volcanic projectiles to penetrate through roofs. Post-depositional absorption of rainfall into tephra deposits is a major source of uncertainty in tephra fall building damage assessment with previous theoretical work highlighting that load increases > 100% are possible. To better constrain the likely upper limit of loading increases, rainfall simulation experiments were conducted across a wide range of different grainsize distributions. Experiments found that loading increases >30% could not be achieved, even under two hours of intense rainfall, suggesting 100% increases - whilst theoretically possible - are unlikely to occur. By accounting for variations in tephra hazard, building exposure, building vulnerability and rain fall conditions, I conducted a damage estimation sensitivity study at Kelud volcano, Indonesia. The high sensitivity (up to 45 %) to individual variations and the compounding effect of simultaneous variations (up to 85 %) highlighted the importance of propagating uncertainty from all components of the modelling process into impact assessments. To reduce model sensitivity and increase confidence in building damage assessments it is important that future vulnerability assessments more accurately quantify the hazard intensities that buildings experience following future damaging eruptions. A critical gap that future research should address is that no pre-eruption building damage assessments have ever been validated against post-eruption damage observed at the same volcano. Following the methods outlined in this thesis, a higher number of high-quality tephra hazard building damage assessments can be carried out and eventually improved via calibration against observed post-eruption damage.