Precipitation, drifting and blowing snow, avalanches, ice melt, permafrost thaw and debris flows are important mountain processes that depend on atmospheric forcing. All these processes are expected to change significantly with increasing climate change. However, a quantitative assessment of changes in these processes has not yet been achieved. One reason is that strong non-linearities prevent a direct application of climate change scenarios, which are available from climate models. This is even true for downscaled climate change scenarios such as currently generated in the CH2018 project, which will produce daily mean values but not sub- hourly values needed as input for models that describe the processes above. In addition, these processes are most important / dangerous for extreme weather situation, which may not be adequately represented in downscaled climate predictions. We therefore propose specific small-scale downscaling of climate change drivers for these processes. The focus of the work will be on mountain processes and the slope scale, while the work should also find a useful application for the vast polar ice sheet areas.
Generate realistic meteorological forcing at the slope scale and in high temporal resolution for local mass movement processes such as snow transport, avalanches and melt.
The approach will be to use a combination of dynamical and statistical downscaling based on the regional CH- 2018 scenarios and results. The dynamical approach is state of the art and currently the only reliable method to get high-resolution climate and weather forcing for very complex terrain. In a series of past and current dissertations, the WRF and ARPS models have been used to investigate small scale meteorological fields including precipitation in highly complex terrain. This expertise is proposed to be applied to the climatological downscaling required in this project. Specifically, the climate change signal as diagnosed in CH-2018 will be superimposed on regional weather model runs representing critical (and therefore extreme) weather situations for the processes in question. This new technique should then allow to characterize the influence of climate change on the boundary conditions for very local simulations with the WRF weather model. This method should on be suitable to represent extreme situations. The method automatically generates weather input at high temporal resolution, which is required for the process models. Since very high resolution dynamical downscaling is computationally extremely expensive, we will also explore a further method. Meteorological forcing is basically determined by a set of highly nonlinear differential equations, which describe air motion and thermodynamics of the atmosphere. Close to the surface, flows are highly turbulent and computational effort increases dramatically. In the context of this project, it will be explored in how far machine learning or a weather generator can be used to replace dynamical downscaling after a training period. First experience has been gathered with high-resolution wind fields in the CRYOS group at EPFL and with weather generators in diverse groups of WSL and the ETH domain. Both methods are promising and have the potential to achieve a larger volume of slope-downscaled climate change scenarios at affordable computational cost.
We will compare different modelling approaches which are currently used at the WSL to downscale climate variables from global and regional scales to sub-kilometer scales. The focus of this comparison is to test the performance of different downscaling methods in reproducing small-scale processes which play a governing role in local-scale mountain climate. The comparison study comprises two efficient mechanistic approaches (TopoSCALE, CHELSA) and two dynamic meteorological models of different complexity (WRF, ICAR).