Avalanche forecasting remains a cognitively demanding task, requiring experts to integrate heterogeneous data sources under time pressure. While advances in numer ical modelling and monitoring networks have expanded the available information, tools that effectively combine and visualise these data sets for operational use are still limited. This thesis develops and evaluates a prototype dashboard designed to support avalanche forecasters through integrated, interactive visualisation of snowpack, mete orological, and avalanche data. The system brings together interpolated grid fields, point-based measurements, and clustering results from an ensemble HDBSCAN* approach, presented within a multi-tab layout (Cluster, Point, and Grid views). The design follows established geovisualisation principles, emphasising spatio-temporal context, intuitive filtering, and consistency across representations. To assess the dashboard’s effectiveness, I conducted a structured user study including performance tasks, ease-of-use ratings, and a System Usability Scale (SUS) survey. Results show that grid data were interpreted most intuitively, while point data offered detailed accuracy at the cost of higher cognitive load. Cluster-based representations provided a useful overview but were limited by the modest quality of the underlying ensemble HDBSCAN* clustering (Adjusted Rand Index = 0.28). Despite these challenges, participants rated the dashboard as straightforward and coherent, and reported that its interactive design supported efficient pattern recognition. The main contribution of this work lies in demonstrating how multiple avalanche relevant data sources can be integrated into a single, user-centred dashboard concept. Beyond its immediate application, the study highlights both the opportunities and limitations of applying ensemble clustering in this context, and establishes a benchmark for future research. The approach and design principles are transferable to other natural-hazard domains where experts must balance complex, uncertain data under time pressure.
Building similarity graph...
Analyzing shared references across papers
Loading...
Nils Besson
Building similarity graph...
Analyzing shared references across papers
Loading...
Nils Besson (Tue,) studied this question.
www.synapsesocial.com/papers/69d894ce6c1944d70ce05bff — DOI: https://doi.org/10.5167/uzh-433557