Description
This study investigates the main large-scale drivers of precipitation variability. Using monthly reanalysis data spanning 46 years (1979–2025), it examines the causal pathways driving precipitation over Europe and the North Atlantic during the winter season. To this end, causal inference methods are applied, with the aim of revealing temporal and spatial structures that differ from those identified through correlation-based approaches. The causality algorithm, based on conditional independence, is implemented using both first- and higher-order techniques to assess nonlinearities in the teleconnections between atmospheric variables. The North Atlantic Oscillation (NAO), a statistical artifact summarizing the leading mode of sea level pressure variability in the North Atlantic, accounts for a significant share of precipitation variability, so it is used as a benchmark. The overarching goal is to reduce uncertainty in precipitation prediction by relying solely on physical variables, both by increasing the explained variance in regions where the NAO already shows skill, and by providing additional insight in regions where it does not. Preliminary results show a modest improvement in prediction skill over the NAO baseline, while offering a clearer physical interpretation of the mechanisms involved. The linear vs. nonlinear analysis suggests that teleconnections in this region are predominantly linear, with higher-order nonlinearities either absent or below the detection threshold of the method.
| Field of study | Earth & Climate Physics |
|---|---|
| Supervisor | Jens Hesselbjerg Christensen |