https://orcid.org/0000-0002-7287-0588
-
Views
-
Cite
Cite
Jonathan Koh, Jonathan Koh’s contribution to the Discussion of ‘Inference for extreme spatial temperature events in a changing climate with application to Ireland’ by Healy et al., Journal of the Royal Statistical Society Series C: Applied Statistics, Volume 74, Issue 2, March 2025, Pages 315–316, https://doi.org/10.1093/jrsssc/qlae089
Close - Share Icon Share
Extract
I congratulate the authors on an insightful and thought-provoking work, one that touches on pertinent topics like r-Pareto process inference with missing data and uncertainty quantification.
It is well known that the main drivers of heat extremes are (e.g. see Zeder & Fischer, 2023): (i) global thermodynamics: increases in greenhouse gases lead to global warming; (ii) local land surface feedbacks: soil moisture deficits can exert a positive feedback on heat waves through the modulation of energy fluxes at the surface (e.g. Seneviratne et al., 2010); (iii) regional dynamic conditions due to diabatic and adiabatic warming, or advection: a persistent high-pressure system can lead to prolonged periods of heat by inhibiting cloud formation and precipitation (e.g. Pfahl & Wernli, 2012).
Driver (i) relates to the climate, while the latter two relate more to the weather. The current paper considers the first driver with the use of CO2 and climate model summary statistics as potential covariates. Albeit in a machine learning context but also with r-Pareto processes, the 2024 preprint ‘Using spatial extreme-value theory with machine learning to model and understand spatially compounding weather extremes’ by Koh et al. looked at drivers (ii) and (iii), and found the co-occurrence of high- and low-pressure systems to be relevant predictors for the frequency, magnitude, and spatial dependence of heat extreme events across Western Europe. A possible avenue of research would be to consider all three drivers of heat extremes in the r-Pareto process framework, e.g. to assess how facets of the weather will be impacted by climate change (e.g. Steinfeld et al., 2022). Using climate model simulations for this purpose could prove challenging, due in part to their systematic underestimation of low-frequency variability (e.g. Vautard et al., 2023).
