Quantifying landslide erosion rates over millennial timescales in the Southern Alps and Fiordland, New Zealand

Landslides are a major erosional mechanism in mountain landscapes, play a crucial role in source-to-sink systems by delivering coarse sediment, and constitute a major geohazard. However, quantifying long-term (>102 yrs) landslide sediment fluxes is challenging, because remote sensing offers high-resolution constraints only over the last few years or decades, and landslide scars and deposits are often obliterated in <102 yrs. In the Southern Alps and the Fiordland of New Zealand, current estimates of long-term landslide frequency and erosion rates have previously been derived from mapping landslides over several decades and extrapolating frequency-magnitude relationships to ≥103 yr timescales. However, three main issues may limit the utility of these estimates. First, they were derived using linear landslide area-volume relationships, while recent findings suggest that landslide volumes scale non-linearly with landslide area. Second, they are based on landslide frequencies over the 1940-80s (Southern Alps) and 1960s-2007 (Fiordland). An updated inventory for the Southern Alps extending until 2014 shows that landslide frequency over those particular decades may have been anomalously high. Third, these decadal observations of landslide frequency are limited to the current inter-seismic period and, hence, remain insensitive to variability of landslide frequencies across a full seismic cycle. Here, we address the first two issues by including the most recent landslide observations in the frequency-magnitude relationships and by implementing a new field-calibrated power law area-volume scaling relationship. We address the third issue by extrapolating landslide erosion rates to full seismic cycles using published estimates of seismic and post-seismic versus inter-seismic lake sedimentation rates, and known earthquake recurrence intervals. To estimate landslide recurrence intervals over thousand-year timescales, we avoid the limitations of using recent remote sensing data, and instead propose two new approaches that utilize timescale-appropriate cosmogenic radionuclide measurements.  First, using 17 new in situ 10Be concentrations from recent landslide deposits, we show how these concentrations, combined with drone photogrammetry of landslide scars, can be used to estimate the exposure age of the hillslope before the landslide occurred and, hence, provide information about millennial landslide recurrence intervals. Second, we present preliminary data on paired in situ 14C-10Be concentrations from 9 landslides and 20 catchments, and show how 14C/10Be ratios increase with landslide depth and can be used to track catchment-wide landslide activity. Finally, we examine whether the three updated landslide erosion rate estimates (inter-seismic, full seismic cycle, and 10Be exposure-age-based) are consistent with recently published 10Be catchment-averaged denudation rates, and with longer-term estimates of denudation in the Southern Alps and Fiordland.