Data source
Near-surface air temperature data, measured 2 meters above ground level ($T_{2m}$) are first obtained from the ERA5 reanalysis. The dataset provides hourly values at a horizontal resolution of 0.5° in both latitude ($\phi$) and longitude ($\lambda$). As an initial step, daily mean temperatures are computed from the hourly data.
Identifying extremes
For each geographical location and calendar day of the year, the 10th and 90th percentiles of the $T_{2m}$ distribution from 1991 to 2020 are calculated and denoted as $T_{2m}^{10}(\phi,\lambda,day)$ and $T_{2m}^{90}(\phi,\lambda,day)$, respectively. While cold extremes (or cold spells) are defined as when temperatures fall below the 10th percentile ($C(\phi,\lambda,day)=T_{2m}(\phi,\lambda,day)<T_{2m}^{10}(\phi,\lambda,day)$), warm extremes (or heat waves) are defined as when temperatures rise above the 90th percentile ($W(\phi,\lambda,day)=T_{2m}(\phi,\lambda,day)>T_{2m}^{90}(\phi,\lambda,day)$). By definition, cold and warm extremes occur 10% of the time.
Identifying persistent extremes
Persistent extremes are identified when cold ($C(\phi,\lambda,day)$) or warm ($W(\phi,\lambda,day)$) extremes are present for any period of 3 consecutive days or longer.
Grouping persistent extremes into spatiotemporal extreme events
Persistent extremes that are neighbors in the space and time dimensions are grouped together using the connected component labeling method. This yields geographically-coherent areas that are affected by the same extremes.
Statistics
Once spatiotemporal extreme events are identified, several properties can be assessed
- Area: Total area affected over the entire lifetime of the event.
- Duration: Total duration in days from the date of first to last recorded persistent anomalies $C_p$ or $W_p$ associated with an event.
- Standardized anomalies: $T_{2m}$ anomalies standardized for every grid point and calendar day ($\sigma(\phi,\lambda,day)$).
- Event center: location of maximum time-integrated standardized anomalies for all grid points and time-steps included in the event. Identifies persistent large anomalies.
- Countries affected (country names are identified with ArcGIS Online and are provided here only for geographical reference)