Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
Snow plays a crucial role in processes regulating ecosystems, the climate, and human development. Mountain snowpack in particular has great relevance for downstream communities. Knowledge about the distribution and properties of the snowpack thus help in planning how to live with this dynamic resource. Snow models employed to simulate mountain snowpack cover a range of physical complexity, but all of them are incredibly dependent on accurate input data. This input data must resolve relevant atmospheric processes at the scales at which snow models are run, often down to 100 m or higher. Processes such as preferential deposition of precipitation, where near-surface flow features and microphysical processes enhance precipitation at the ridge scale, should thus be accounted for. One of the best methods for providing snow models with input data is dynamic downscaling, where meteorological input data is calculated at the resolution of snow models using numerical weather prediction models. This technique has many advantages over statistical downscaling, with the large caveat that is it computationally unfeasible to perform over large areas or time scales. In this thesis, we introduce an intermediate-complexity atmospheric model, HICAR, capable of running at the resolution of most snowpack models. The HICAR model makes use of techniques developed in the field of pollutant transport to efficiently solve for a 3D wind field at the hectometer scale. The technique allows for direct modification of the wind field, enabling parameterizations of steady-state eddy-like structures and thermally driven slope flows. Validation of HICAR's flow fields against non-hydrostatic atmospheric models, as well as observations, demonstrate HICAR's ability to resolve flow features relevant to snowpack modeling. These improvements to the model flow field, in combination with improvements to the model's physics, result in accurate simulation of near-surface atmospheric variables. 2m air temperature, radiative inputs, and precipitation outputs of the model are evaluated against observations from automated weather stations and grided precipitation products. A process-level view of precipitation at the 50m scale is presented using a state-of-the-art microphysics scheme. Results from this evaluation reveal HICAR's ability to simulate preferential deposition of snow, and alter the understanding of the process to include the interaction of near-surface flow features with microphysical process. Finally, the relevancy of the model for snowpack modeling is addressed. HICAR is coupled with an intermediate complexity snow model, FSM2trans. This coupled model, HICARsnow, is shown to resolve patterns of snow accumulation and ablation throughout the snow season. The ability of HICAR to simulate preferential deposition is shown to improve the distribution of snow depth in complex terrain relative to snow model runs using statistical downscaling of precipitation. Feedbacks from blowing snow sublimation on humidity also limit the rate of blowing snow sublimation over the winter season for the two-way coupled snow model compared to the snow model run with statistical downscaling. The arc of this thesis shows that intermediate-complexity atmospheric modeling at the hectometer scale is possible, that it is capable of resolving atmospheric variables relevant to land surface models, and that this translates to better process representation within snow models.
, ,
Michael Lehning, Tobias Jonas, Dylan Stewart Reynolds