Constraining first-order controls on volcanic repose time by examining cumulative distribution functions

Currently, we lack conceptual models for the processes that govern the time interval between volcanic eruptions, or "repose time", at a specific volcano. A number of studies examine repose time for individual eruptions, including work on short-term probabilistic hazard assessment relying on calculations of repose times. However, little work has focused on longer-term patterns of repose time for a volcano as a whole, as well as the underlying geologic mechanisms controlling these patterns. Here we examine the distribution of eruptions at representative volcanoes within the Cascades Arc to better understand the root causes of their repose times. The available geochronological datasets for these volcanoes are used to investigate the differences in the shape of the cumulative density functions (CDF) of dated eruptions through time for each volcano. Mt. Adams, Mt. Lassen, and Mt. Rainer maintain a broadly constant frequency of eruptions through time ("stationary" behavior) between 600 ka and present, whereas Mt. St. Helens, Middle Sister, and South Sister show an increase in frequency of eruptions through time with a transition around 100 ka ("non-stationary" behavior). Although these results may be biased by higher sampling of younger eruptions and volcanos and/or gaps in the geochronology data, they suggest different processes govern these volcanoes unrelated to geographic location. To assess the minimum number of dated eruptions required to accurately reproduce the shape of the CDF's (i.e., % of "missing eruptions" that are feasible to still understand the eruptive distribution), we conduct Monte Carlo simulations to create synthetic CDF's for a given volcano and explore adding a greater weight to the selection of the younger and larger eruptions. Preliminary results suggest at least 6-27% (avg. 15%) of individual eruptions are required to accurately reproduce the known CDF's for a given volcano. Additional statistical treatment (e.g., "boot-strapping") of the datasets will be applied to further evaluate the biases mentioned above. These results suggest there may be two populations (stationary and non-stationary) of volcanic behavior amongst the modern Cascades volcanoes, which may be important in constraining the drivers of repose time and working toward an increased ability to forecast future eruptions.