Fig. 3: Modeling slip to match biomarker measurements.

The forward modeling procedure considers variation in fault half width, slip magnitude, number of earthquakes, and distance of a sample from the fault. Here, we show an example for sample PP952, assuming five earthquakes with slip of 50 m. a A successful model fit for all biomarkers using a = 0.0151 m. The first panel shows the temperature with distance from the fault for different time-steps. Lower panels show the final biomarker fraction reacted with distance from the fault (heavy lines) with blue shaded regions showing uncertainty in biomarker kinetics. Measured range of each biomarker anomaly is shown as boxes (height). The width of the boxes represents the sample width, limited by the constraint that the sample is outside of the slipping zone. Boxes are shaded red for successful matches and gray if unsuccessful. b Model is unsuccessful for a = 0.0173 m due to a mismatch between modeled and measured C₃₇ total fraction reacted. c Plot of distance from the fault vs. half width with hotter colors showing higher probability of a successful model (successful models/total model runs). Gray circles indicate unsuccessful models (P = 0). d Lower panel shows the probability of a successful model at the range of earthquake slips tested. Minimum slip with a nonzero probability is indicated with the red arrow. Range of slip estimates for the Tohoku-Oki earthquake shown by the blue shaded region⁴. Upper panel shows the slip vs. number of earthquakes with probability indicated by color as in c. e For samples with biomarker anomalies we find the minimum slip distance of a repeating event (up to 3.2 km of total displacement) needed to match the biomarker anomalies. These modeling results indicate that any of the faults with heating anomalies could have hosted the 2011 Tohoku-Oki earthquake, and that all observed biomarker anomalies require earthquakes with a minimum of 10 m slip per event. Red symbols indicate features with clear biomarker anomalies. Hollow red symbols represent sample PP948, which has alkenone concentrations below the quantification limit and is not modeled (Supplementary Note 3) and samples PP944 and PP951 for which structural evidence is ambiguous (Supplementary Note 3).