Projecting Daily Extreme Heat Events in the Iberian Peninsula using Statistical Downscaling with 700, 500 and 300 Geopotential Fields

Global warming is evident in the extreme events (XE) of daily maximum temperature (Tx) in the Iberian Peninsula, but this behaviour is not fully explained by the mean evolution of temperature, Castillo-Mateo et al (2025). In this context, it is clear that projections of XE risk are required for future climates, and they must be obtained using methods specifically designed for extremes.

This work proposes a new statistical tool to obtain daily and local-scale projections for the occurrence of XE, defined as days with Tx above a pre-established threshold. First, the tool relies on a geostatistical model that links the occurrence of XE at each point of the study region with atmospheric covariates at different geopotential levels, taken from grid points in a surrounding area. Second, a selection of AR6 GCM trajectories is performed using criteria that account for (1) the reproduction of the daily frequency and persistence of weather types over the region, and (2) the reproduction of the empirical distribution of ERA atmospheric variables at the daily scale. Third, the projected values of the atmospheric covariates are used as inputs for the statistical model, allowing estimations of daily characteristics at both local and regional scales.

Model estimation is carried out using daily Tx data for 1960--2024 from 36 Spanish stations (European Climate Assessment& Dataset), for June-August. XE is defined by the 95th percentile of Tx for 1991--2020. Covariates consist of geopotential variables at 12 p.m. for pressure levels of 500 and 700 hPa, on a 1º x1º grid over the area 45º--35º N and 10ºW--5ºE, obtained from the ERA5 reanalysis. The statistical models achieve high goodness-of-fit, with AUC values above 0.8 for validation conditions at most stations.

Trajectories from 36 AR6 GCMs are analysed to select those that meet the criteria, and only six trajectories remain. Finally, projections for 2031--2060 are obtained for the Iberian Peninsula under different scenarios.

References

Castillo-Mateo, J., Gelfand, A.E., Gracia-Tabuenca, Z., Asín, J., Cebrián, A. C. (2025). Spatio-temporal modeling for record-breaking temperature events in Spain. Journal of the American Statistical Association, 120, 645-657.