Electronic Supplement Part A: Generalized Description of Hog Lake Stratigraphy, Hog Lake Radiocarbon Ages and OxCal Chronologic Models
Site Stratigraphy
The trenches within Hog Lake exposed finely bedded silt and clay with a few, thicker layers of sand(Figure SA1). The character of the sediments varies throughout the depression as a function of proximity to their source. Typically, units get finer towards the depo-center and coarsen towards their source. Unit descriptions provided here are taken from descriptions on the north wall of trench T2 (T2N, cut 3), with units above 400 described from the downthrown (NE) side of the fault, and units 400 to 750 described from the upthrown (SW) side of the fault.
Organic accumulations separate many of the strata, with the amount of organics varying across different areas of the depression. In some cases, a dense, cm-thick peat-like accumulation of organics probably represents burial of an organic mat during a large storm event. Such a layer can be seen to thin, disappear, or merge into an organically-enriched, darkened soil (A horizon) towards the pond margins. We interpret the organic layers as periods of non-deposition and soil formation, or in the case of the wettest/lowest areas of the depression, organic accumulation. The sediment layers, in contrast, are interpreted to represent punctuated storm events when sediment is transported from the watershed to the basin. The uppermost strata are capped by a thick A horizon soil that indicates that the site is substantially dryer now than for most of its past 4000-year history.
Some of the organic units merge laterally with strongly oxidized, bright orange sediments that we interpret as surface burn layers. Some of these burn layers are present in all exposures and probably represent complete burning of the site. For other such strata, the burning was apparently limited to the dry sections of the depression, as oxidized layers west of the fault, where the surface is relatively uplifted and water-free during the dry season, transitioned to unburned organic-rich peat-like strata in the deeper parts of the depression east of the fault that were presumably wet at the time of the burn.
The units are numbered from top to bottom, with increasing numbers downward with increasing age, as we excavated deeper trenches during the latter part of the investigation (see full-resolution logs in Part C of this electronic supplement). We primarily numbered the major units that could be traced from trench to trench, and we excavated fault-parallel trenches for stratigraphic correlation so that all common strata share a common unit designation. Not every stratum has been designated a unit number. In some areas of the depression, units splayed into multiple discrete stratums, and this was particularly the case towards the primary depo-center near trench T2 (Figure 5, main paper). As we initially named the units in our first trench, T1, this required subdividing the units into multiple sub-units and designating them with a small-case letter, such as 149a, 149b, and 149c. We did this where the designation of discrete strata were important in the interpretation of faulting events. In other cases, there are more discrete units defined for some trench logs, although we did not assign every recognizable stratum with a unit name. We also did not name some units where the correlation was uncertain or speculative.
OxCal Chronologic Models
In order to develop our chronology of dated earthquakes, we used OxCal version 4.2.3 (Bronk-Ramsey, 2013) to calibrate the radiocarbon ages to calendar ages, and well as to employ stratigraphic ordering constraints and develop probability density functions of event ages in stratigraphic context. Out of the 111 samples dated, 14 samples were initially excluded from the initial OxCal model as being obviously too old, either relative to other dates within the same unit, or relative to dates in the surrounding units. These excluded samples are highlighted in red on Table A-1.
We built two models in OxCal in order to examine the effects of how the selection of radiocarbon ages affected the event ages. Table A-1 shows which samples were used in each model in the columns labeled “Included in Model 1” and “Included in Model 2”. Model 1 (Plate A-1) was our attempt to use as many of the radiocarbon ages as possible, and consisted of 97 radiocarbon ages. As Plate A-1 shows, in general the calibrated ages show good stratigraphic ordering, with very few age reversals (older dates above younger dates). In order to aid us in our selection for alternative OxCal models, we also plotted the OxCal-calculatedIndividual Agreement Indices (IAI), shown after each sample name. According to the OxCal user manual, the IAIs are useful for identifying which samples do not agree with the model, and typically the IAI should be above A=60 (Bronk-Ramsey, 2013). Effectively, it is a number that shows how much the priors are reweighted, and low values indicate samples with calibrated ages that are possibly inconsistent with ages stratigraphically above and below, or have been significantly reweighted into low probability areas of the calibrated age range. As Plate A-1 shows, most radiocarbon ages in Model 1 have IAIs greater than 60, although there are several radiocarbon ages that dip below this, and some that have IAI = 0. For example, the two dates within Unit 250 are not only inconsistent with eachother, but are also inconsistent with ages above and below, and may be affecting the IAI for other samples, such as date in unit 235. Another measure of the overall model is provided by the Model Agreement Index (Amodel) calculated by OxCal, which is a measure of how likely the model is as a whole, given the data. In general, OxCal recommends the Amodel be greater than 60. In the case of Model 1, Amodel = 0, suggesting that even a limited number of radiocarbon ages with low IAIs are dragging down the model as a whole.
Using the IAIs as a guide, we constructed Model 2 (Plate A-2), which consisted of 85 radiocarbon ages. Samples with low IAI’s in Model 1 were considered for removal, especially those samples that were inconsistent with ages stratigraphically above and below, as well as radiocarbon ages that were inconsistent with ages within the same unit. In general, our metric for “inconsistent” was that the calibrated age posterior range was older than other nearby samples, and our assumption was that these samples had an inherited age before they were incorporated into the deposit. Care was exercised in which samples were removed, because some samples with low IAIs were only low because of the influence of other low IAI samples. However, plotting the prior and posterior distributions makes it fairly easy to inspect the model visually and predict the effects of removing or retaining different radiocarbon dates. In Model 2, only 7 radiocarbon ages have IAI< 60, and the overall Amodel = 45. In comparing Model 1 to Model 2, it appears that the omission of problematic dates only changes a few individual event ages by a few decades (Figure 11 in main text), suggesting that the subtraction of a few individual radiocarbon ages has a small effect on the overall modeled event chronology. Although the Amodel score for Model 2 is below the OxCal-recommended value of Amodel = 60, the removal of the 7 low-scoring samples would have very little effect on the overall model other than the Amodelscore, as the surrounding high-scoring radiocarbon ages are the controlling dates in the sequence. Based on the relatively small differences between the models, we conclude that the overall chronology is well-resolved and would be stable, even if other low scoring samples were removed from the models.