Our research concerns the meteorological causes on wind speed extremes by Granger causality. We investigate 16 hourly meteorological parameters from the ERA5 database [1] and identified 62 extreme events based on a wind speed threshold in our dataset. We explore the causal effects by Granger causal inference, namely by the heterogeneous graphical Granger model (HMML) [2]. In a large set of events we find that wind speed has a non-Gaussian distribution and a better fit can be achieved with distributions from the exponential family. Contrary to the classical Granger model, proposed for causal inference among Gaussian processes, HMML is able to detect causal relations in short multivariate time series with distributions from the exponential family, including inverse Gaussian and gamma distributions. Furthermore it uses an optimization procedure based on the minimum message length principle. HMML has shown to be superior in precision over baseline causal methods in previous research [2]. We evaluate the discovered causal connections in extreme events and compare the results to a control group, which is an equal sized set of events without high wind speeds. The variables specific humidity, wind speed, and temperature have most frequently been identified to be causal to wind extremes. Furthermore specific humidity, wind speed and divergence are statistically more frequent causes to wind speed in the extreme group compared to the control group. [1] Hersbach, H, Bell, B, Berrisford, P, et al. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 2020; 146: 1999– 2049. https://doi.org/10.1002/qj.3803 [2] Hlaváčková-Schindler, K., Plant, C. (2020) Heterogeneous graphical Granger causality by minimum message length, Entropy, 22(1400). pp. 1-21 ISSN 1099-4300 MDPI (2020). [3] A. Fuchs, HMML, version 1.1.1. [Online]. Available: https://git01lab.cs.univie.ac.at/a1106307/hmml.

Top