Tephra fallout hazard assessment with uncertainty quantification: a case study from Cotopaxi and Guagua Pichincha volcanoes, Ecuador

Abstract

The use of numerical models aimed at producing probabilistic maps is becoming more and more a common practice for tephra fallout hazard assessment. However, it is important to complement such maps with a quantification of the major sources of aleatoric/epistemic uncertainties, to help stakeholders and decisionmakers in taking informed decisions. In this contribution, we present an example of uncertainty quantification applied to a tephra fallout hazard assessment. The study is related to two volcanoes (Cotopaxi and Guagua Pichincha) threatening the capital city of Ecuador, Quito. Uncertainty was quantified with respect to three aspects: 1) the numerical model itself; 2) the probability of occurrences of different eruptive styles; 3) the range of variation of three eruptive input parameters (total fallout mass, eruption duration, average plume height). For point 1), the model used (which couples the plume model PLUMEMoM and the tephra dispersal model HYSPLIT) was tested in reproducing recent eruptions from South American volcanoes. This step allowed quantifying the difference between real (observed) and modelled values of several parameters, including mass loading, from which we derived coefficients of average model overestimation and underestimation. Concerning points 2) and 3), we performed an expert judgement (elicitation) session involving 20 experts of different countries and areas of expertise. This allowed deriving detailed uncertainty ranges that we used to i) sample the eruptive input parameters at each iteration during hazard map production; ii) linearly combine maps of different eruptive magnitude/style according to their relative probability of occurrence. The final products of this study are hazard maps of different formats and hazard curves for 10 sensitive sites in the city of Quito. Each of these maps/curves is presented as a set of three maps/curves (“lower”, “mean” and “upper”) which quantify the major sources of uncertainty.