Testing a novel cave-based proxy for palaeo-earthquake shaking on the Alpine Fault, Aotearoa/New Zealand.

Speleoseismology aims to reconstruct palaeoseismic records by dating pre- and post-damage speleothem calcite. A common approach is to infer palaeo-earthquakes from evidence of coinciding damage features (e.g., rockfall and broken speleothems) at multiple locations, which can be challenging in regions of high tectonic strain where short recurrence intervals of large earthquakes require dating of an impractically large number of damage features. Alternative approaches concerned with dating successive growth changes in individual speleothems (e.g., axis changes and growth hiatuses) are better suited to high-seismicity settings, as closely spaced events are more readily resolved. However, the origins of these growth changes can be ambiguous.

This study tested a novel geochemical proxy for quantifying ground shaking that is amenable to high-resolution speleothem studies, and potentially more diagnostic of earthquake damage. We evaluated the hypothesis that past large earthquakes temporarily elevate Mg/Ca in cave drip waters via incongruent carbonate dissolution following host rock fracturing (ICD_C), leading to corresponding Mg enrichments in speleothem calcite. To do this, we examined a well-dated Holocene stalagmite (GT1) from a cave near the Alpine Fault, which is Aotearoa/New Zealand’s longest (>500 km) active onshore fault and a major source of seismic hazard. The locality is 4 km from the Alpine Fault’s northern section, which typically ruptures every 414–470 yr in a major (M_W >7) to great (M_W >8) earthquake, resulting in shaking intensities of MMI >VIII at the study site (MMI: Modified Mercalli Intensity).

We present a record of Mg/Ca variability in GT1 since ~5 ka, obtained by laser ablation inductively coupled plasma mass spectrometry along the stalagmite growth axis, and constrained temporally by >40 U–Th ages. Preliminary data show high baseline Mg concentrations in GT1 that cannot be explained solely by other mechanisms of drip water Mg/Ca enrichment (i.e., prior calcite precipitation), suggesting an ongoing contribution of Mg to drip waters by ICD_C. Anomalous Mg peaks are therefore interpreted as high-intensity shaking events that temporarily elevated drip water Mg/Ca above baseline values. Post-2.5 ka Mg peaks are generally more subtle (30–50% enrichment) than pre-2.5 ka peaks (40–100%). Magnesium peaks are also strongly associated with brown-stained laminae inferred to reflect soil-derived organics. We propose that the high-Mg/high-organics horizons represent large earthquakes that both fractured the host rock and enhanced the mobilisation of organics from overlying soil.

We compared the GT1 record with a proximal and independent 1.4-kyr record of well-dated seismically triggered lacustrine turbidites. Given the subtle nature of Mg peaks in this interval, we consider those associated with physical growth changes (i.e., growth onset/cessation and/or axis change) as more likely to represent earthquakes. Of nine Mg peaks identified, five are associated with major physical growth changes. Of the four largest (MMI >VIII) shaking events in the lake turbidite record, which correspond to northern Alpine Fault surface-rupturing earthquakes, three overlap in age with a GT1 Mg peak and physical growth change. Further, two of the three historic earthquakes that generated MMI ≥VII shaking at the study site also overlap in age with a Mg peak.