Item

Abstract

Glaciated alpine catchments feature high suspended sediment yields, with severe implications, e.g. for downstream water quality and flood hazard. However, the impact of past and future climate on these systems is not well understood, due to the limited availability of records and methods for future projections. We tested the feasibility of estimating sediment export at decadal scales using Quantile Regression Forest (QRF), a non-parametric data analysis method. For estimates of past sediment yields, we used short-term daily records (i.e. time series of a few years) of suspended sediment concentrations (SSC) and long records of the most important hydro-climatic drivers. The QRF models were trained independently for two nested and partially glaciated catchments, Vent and Vernagt, in the Upper Ötztal Alps in Tyrol, Austria. Themodels performed well in estimating daily sediment export, and outperformed traditional sediment rating curves. An analysis of the reconstructed time series indicated that there were significant positive trends in the annual suspended sediment yields at both gauges, with distinct step-like increases around 1981, which correspond to the onset of accelerated glacial melt and respective threshold behaviour. To estimate future sediment yields, we used hydroclimatic scenarios (2020-2100) as input. The method proved to be applicable, so the trends in sediment export of can be estimated for the next two or three decades. Modelling into the future is constricted, e.g. by projected glaciers disappearance around 2070 in some scenarios. Such major changes in the sedimentological functioning of the catchments that will likely not be represented adequately by our models.