Stochastic alluvial fan and terrace formation triggered by a high-magnitude Holocene landslide in the Klados Gorge, Crete

Abstract

Alluvial fan and terrace formation is traditionally interpreted as a fluvial system response to Quaternary climate oscillations under the backdrop of slow and steady tectonic activity. However, several recent studies challenge this conventional wisdom, showing that such landforms can evolve rapidly as a geomorphic system responds to catastrophic and stochastic events, like large-magnitude mass wasting. Here, we contribute to this topic through a detailed field, geochronological, and numerical modelling investigation of thick (>50 m) alluvial sequences in the Klados catchment in southwestern Crete, Greece. The Klados River catchment lies in a Mediterranean climate, is largely floored by carbonate bedrock, and is characterised by well-preserved alluvial terraces and inset fans at the river mouth that exceed the volumes of alluvial deposits in neighbouring catchments of similar size. Previous studies interpreted the genesis and evolution of these deposits to result from a combination of Pleistocene sea-level variation and the region's long-term tectonic activity. We show that the >20 m thick lower fan unit, previously thought to be late Pleistocene in age, unconformably buries a paleoshoreline uplifted in the first centuries CE, placing the depositional age of this unit firmly in the late Holocene. The depositional timing is supported by seven new radiocarbon dates that indicate middle to late Holocene ages for the entire fan and terrace sequence. Furthermore, we report new evidence of a previously unidentified valley-filling landslide deposit that is locally 100 m above the modern stream elevation, and based on cross-cutting relationships, it predates the alluvial sequence. Observations indicate the highly erodible landslide deposit as the source of the alluvial fill sediment. We identify the likely landslide detachment area as a large rockfall scar at the steepened head of the catchment. A landslide volume of 9.08×107 m3 is estimated based on volume reconstructions of the mapped landslide deposit and the inferred scar location. We utilise landslide runout modelling to validate the hypothesis that a high-magnitude rockfall would pulverise and send material downstream, filling the valley up to ∼100 m. This partial liquefaction is required for the rockfall to form a landslide body of the extent observed in the valley and is consistent with the sedimentological characteristics of the landslide deposit. Based on the new age control and the identification of the landslide deposit, we hypothesise that the rapid post-landslide aggradation and incision cycles of the alluvial deposits are not linked to long-term tectonic uplift or climate variations but rather stochastic events such as mobilisation of sediment in large earthquakes, storm events, or ephemeral blockage in the valley's narrow reaches. The Klados case study represents a model environment for how stochastically driven events can mimic climate-induced sedimentary archives and lead to deposition of thick alluvial sequences within hundreds to thousands of years, and it illustrates the ultrasensitivity of mountainous catchments to external perturbations after catastrophic events.