Machine-Learning Discovery of Signature Atmospheric Flow Patterns in Complex Terrain

It is known that the influence of complex terrain can drive local circulations or regional scale flow, but our understanding of flow patterns of atmospheric species over and within complex terrain such as mountainous or urban areas, and the impact of climate change on those flow patterns is limited. Capturing small-scale turbulence and at various horizontal resolutions, especially at resolutions relevant to processes at both regional and urban scale, is critical to understanding how meteorological processes interact among natural and urban structures. Using Self Organizing Maps, we examine the impact of local terrain on atmospheric transport and dispersion under different synoptic-scale and local-scale forcings using output from simulations performed by the Weather Research and Forecasting (WRF) model, the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) dispersion model, and observations from a ground-based network. We summarize these flow patterns into those most likely to obtain in a specific area, and record diurnal, monthly and seasonal flow differences with an eye towards predictability.