Towards Improved Uncertainty Quantification and Visualization for Aftershock Forecasts in the Pacific Northwest

Date: 4/24/2019

Time: 06:00 PM

Room: Grand Ballroom

Probabilistic aftershock forecasts based on seismicity models are crucial for earthquake risk reduction. However, forecasts have uncertainty stemming from several sources, which must be rigorously quantified to properly assess seismic risk. We build several models for on-shore seismicity in the Pacific Northwest (PNW) of North America, based on the Epidemic Type Aftershock Sequence (ETAS) method. ETAS is a parametric space-time point process model that describes background seismicity and aftershock clustering. We combine three earthquake catalogs for the PNW from 1970-2001, including data on measurement errors for reported magnitudes and locations for all earthquakes. We discuss the effects of estimating parameters differently across two- and three-dimensional space. Unlike other seismicity models, we propagate the known measurement errors through the model to obtain more realistic uncertainties on estimates of earthquake rate and its parameters. We use a simulation-based approach that isolates the effect of measurement error and model stochasticity on these model outputs. We present results showing spatial dynamics in estimated PNW earthquake rates and ETAS parameters. The effect of measurement error on model outcomes is more nuanced and also influenced by space and time. Implications for aftershock forecasting are discussed. We also introduce early work on the improved visualization of such model uncertainties. In an experiment, we evaluate the effectiveness of three competing visualizations of aftershock forecasts and their uncertainties. We provide the experimental design and first results of this evaluation.

Presenting Author: Max Schneider

Authors