Several approaches to interpreting the Cascadia paleoseismic record are used to derive relationships between fault area, slip, and moment and to compare the results with the scaling relationships determined by Somerville et al. (2015) for recent subduction‐zone events. In two models (CA12a and CA12b), taken from Goldfinger et al. (2012), paleoevents are classified into five characteristic areas (CA), with the slip during each event estimated based on the time between the event and either the following or the previous event. In model CA14, taken from Scholz (2014), slip on four characteristic segments is determined from the plate tectonic convergence rate, assuming a constant stress drop. In model CL, introduced in this article, the fault length for paleoevents is defined by the along‐strike length over which the observations have been correlated; width and slip are interpolated from model CA14. CA12a and CA12b show large scatter compared with the global compilation because of large variations in slip for a given area. Models CA14 and CL reproduce the relationship derived for asperities (defined as patches in finite‐fault models with slip >1.5 times the average slip). These models can be reconciled with the total area and average slip from Somerville et al. (2015) by increasing the fault area and decreasing the slip using scaling factors derived from the analysis of recent earthquakes (CLmod1) or by reducing the slip by a factor of ∼8 (CLmod2). CLmod1 implies that the paleoearthquake observations are controlled by high‐slip patches, whereas CLmod2 implies that much of the plate tectonic convergence is accommodated aseismically. A scenario intermediate between CLmod1 and CLmod2 is considered most likely. This study demonstrates the value of using scaling relationships based on modern earthquakes as a tool for evaluating earthquake histories derived from paleoseismic data.

Online Material: A table summarizing the source parameters for the different models presented in this article to enable figure reproduction. Figures that summarize the data points used by Somerville et al. (2015), data points from a new analysis that became available after preparation of the main article (Ye et al., 2016), and results for the turbidite mass model (Goldfinger et al., 2012) for estimating seismic moment.

You do not currently have access to this article.