Investigating information content of paraglider data in a meteorological context

Summarise with VIRAT
Authors: Dominik Krah
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Accurate and reliable numerical weather prediction remains challenging in mountainous regions, while specifically in alpine areas of high importance for health and environmental reasons. Weather data grid resolutions insufficient to resolve the typically small scaled atmospheric variability in those geographies constitute such a challenge. To overcome this challenge, the present thesis investigates local upsampling of numerical weather prediction output in alpine regions, specifically aided by data automatically captured by paragliding. As this data is highly available in the relevant areas and paragliding is inherently weather dependent, such investigations are well motivated. In this thesis an exploratory dataset analysis, a novel preprocessing and vectorization approach for the paraglider trajectories and initial experiments are presented. The preprocessing contains B-Spline based resampling of the trajectories and finding main movement directions by means of minimum description length. Vectorization is done by a Bag of Words based approach, quantizing the main movement directions into directional bins by means of HEALPix. Experimental results indicate that deep learning based upsampling of weather prediction data benefits from the paraglider data.

Fields of Study & Subjects

Meteorological Phenomena and Simulations Aerospace and Aviation Technology Environmental Monitoring and Data Management

Publication details

Journal / source
ePLUS
Publication date
1 January 2026
Language
EN
Document type
Thesis