ICEBUFFER: Global glacier buffering of streamflow droughts in a changing climate using deep learning

Global warming exacerbates water scarcity by increasing the frequency of hydrological droughts.
These droughts can be buffered by glaciers that mitigate runoff variability on multiple timescales.
However, as glaciers retreat, the reliability of this unique natural water reserve becomes increasingly
uncertain, potentially increasing drought stress. While the fields of drought modelling and glacier
change modelling have evolved strongly, these two disciplines are often treated independently,
leading to high levels of uncertainty. To address this, the ICE-BUFFER project will integrate recent
advances in Deep Learning (DL) into a unified modelling framework to advance drought prediction in
data-scarce regions. The overall objective is to provide the first global assessment of the buffering
capacity of glaciers for streamflow droughts in a changing climate. To achieve this, an evaluation
framework will be developed to assess the ability of glacio-hydrological models to capture the
hydrological signatures of glacierised catchments. This new framework will be used to evaluate ICESTREAM,
a novel DL-based glacio-hydrological model. The culmination of ICE-BUFFER will be the
global application of ICE-STREAM to unravel the influence of glaciers on the propagation of
meteorological into streamflow droughts. An open platform (Glacier Buffer Simulator) will be
developed to disseminate the results of ICE-BUFFER widely.