Optimizing Meteorological Station Placement for High-Resolution Field Reconstruction in Mountainous Terrain

Reconstructing high-resolution geophysical fields from sparse observations is a central challenge for environmental sensing and model evaluation in complex terrain. While high-resolution climate models provide detailed insights, they are computationally expensive and difficult to validate in remote mountainous regions. This work adapts a data-driven sparse sensor placement framework [1] to identify optimized meteorological station locations for an arbitrary number of sensors in complex terrain.

Applied to a mountainous region in northern Norway, our approach can help hydrologists, glaciologists, and climate scientists determine where to place sensors to obtain independent streams of data, supporting a comprehensive representation of variables such as wind speed, humidity, or snow depth. We generalize the original framework by introducing a spatial weighting formulation, allowing users to prioritize specific sub-regions or account for physical constraints such as inaccessible terrain. In addition, prevailing wind patterns are incorporated into the selection criteria, guiding sensor placement toward configurations that capture the most frequent and impactful flow regimes. An orthogonal component approach is further introduced to integrate existing stations, ensuring that newly deployed sensors capture complementary information rather than redundant data. Ongoing work explores the use of the same framework to reconstruct missing or partially degraded measurements when stations are temporarily unavailable, using information from the remaining network.

A key advantage of the framework is its transparency. In contrast to many data-driven or machine-learning-based downscaling approaches, the reconstruction relies on explicit linear algebra operations, providing a traceable link from point observations to a domain-wide target field. For operational safety applications such as monitoring airport winds or avalanche hazards, this offers a computationally efficient and flexible alternative when high-resolution simulations are unavailable.

[1] Xihaier Luo, Ahsan Kareem, and Shinjae Yoo. “Optimal sensor placement for reconstructing wind pressure field around buildings using compressed sensing”. In: Journal of Building Engineering 75 (2023), p. 106855. issn: 2352-7102.