Emmy Scott

Emmy’s research examines complexity, assumption validity, and sensitivity to input data and tuneable internal parameters when computational models are applied to future volcanic hazard scenarios.

Hazard assessment and forecasts in New Zealand is increasingly reliant on the output of computational models that simulate natural phenomena. These models simplify, to varying degrees, the highly complex dynamics that drive these phenomena to produce hazards to varying intensity and spatial footprint (eg ashfall dispersal and thickness, or tsunami inundation and depths). To simulate future scenarios, internal model parameters are typically calibrated to replicate past observations. The underlying assumption is that if a model can be tuned to reproduce past events, it may have the potential to forecast future hazard behaviour. However, this approach often lacks any explicit measure of forecast accuracy and provides an incomplete picture of model appropriateness. 

Emmy’s work is underpinned by three main research areas:

1. Model ontology, complexity, and selection in volcanic hazard forecasting – Emmy explored how differences in model structure and assumptions influence volcanic hazard forecasts. To support this, Emmy developed an ontology and associated frameworks that describe how volcanic hazard models are constructed, how they vary in complexity, and how they can be selected and applied appropriately for different forecasting contexts.
2. Global sensitivity analysis (GSA) – as hazard models become more complex, the number of required input parameters increases, introducing more sources of uncertainty. Although volcanic ash transport and dispersal models (VATDMs) have been validated against past eruptions, they have not yet been rigorously evaluated for forecasting purposes. Emmy applied GSA to two widely used VATDMs, Tephra2 and Fall3D, to identify which inputs, and combinations of inputs, most influence model output. 
3. Real life application using the 17 June 1996 Mount Ruapehu, Aotearoa New Zealand, 28 February 2021 Mount Etna, Italy, and the 3 March 2015 Villarrica, Chile Eruptions – Emmy tested the effects of input parameter variation, both under the condition of no prior knowledge and partial prior knowledge, on simulated tephra thickness using Tephra2. She then tested the forecast accuracy of these simulations, based on deposit mass and impact-based thresholds. 

Emmy’s research offers a transparent approach to assessing forecasting accuracy and provides critical insights into the appropriate use and robustness of volcanic hazard models for forecasting. By identifying key uncertainties and knowledge gaps, it supports more robust and reliable forecasts – ultimately reducing the uncertainty around the scale and likelihood of impacts future natural disasters may have on our communities.