Earthquake Hazards in Hawaii: Challenges in Assessing Earthquake Probabilities in a Volcanic Environment

The complicated volcanic and tectonic processes in Hawaii present challenges for estimating the probabilities of earthquakes using methods developed for standard probabilistic seismic hazard analysis (PSHA), which usually considers only tectonic earthquakes with stationary rate. These processes can range from volcanic eruptive and intrusive activity near the summit calderas and rift zones to gravity-driven deformation on the flanks, and they vary over space and time. Likewise, the characteristics of the seismicity in Hawaii, which ranges from earthquake swarms to M≥7 ruptures, can also vary over space and time, making it difficult to determine a long-term rate and Gutenberg-Richter b-values that are appropriate to use for long-term (e.g., 50-year) seismic hazard assessments.

These challenges in characterizing the seismicity of Hawaii have been encountered as the USGS updates its seismic hazard model for Hawaii. Here we present some of the approaches that have been used to develop the earthquake rate model for the updated seismic hazard map. In particular, we focus on declustering, which is common practice in standard PSHA. We examine the effects that different declustering methods, including Reasenberg (JGR, 1985), the nearest-neighbor method of Zaliapin et al. (PRL, 2008), and an ETAS-based stochastic declustering method (Llenos and Michael, 2019), have on the estimated earthquake rate as well as the Gutenberg-Richter b-value in space and time. The estimated earthquake rate can vary widely depending on what declustering method is used, and so the final hazard model will have some sensitivity to the choices that are made in declustering. Moreover, one of the main purposes of declustering is to obtain a Poissonian rate of independent earthquakes, but it fails to do so in Hawaii because of the variety of volcanic and tectonic processes that can cause earthquake rates to vary over time. Therefore, the final hazard model will also be sensitive to how a long-term seismicity rate in Hawaii is defined.