Spatial distribution of thermodynamic conditions of severe storms in southwestern Europe

- We developed a complementary tool for predicting hailstorms in the MEV.
- Five stability indexes are used in a logistic equation to detect hail occurrence.
- This equation is used to define hail spatial distributions via cluster analysis.
- Hail probability radar data are used to verify the MEV hail distribution.

The Mid-Ebro Valley (MEV) is an area of the northeast Iberian Peninsula with a large number of hailstorms throughout the year. Therefore, new forecasting tools are required to improve the spatiotemporal detection of such storms. Using a database of 100 days with severe storms (SSs) over a 13-year period between 2001 and 2013, we obtained vertical profiles predicted by the WRF mesoscale model. A total of 31 indexes describing conditions of humidity, stability, helicity or precipitable water were obtained from the profiles and input to a binary logistic regression model. The regression model was applied using the forward stepwise method, which indicated that the stability indexes were the most accurate for SS in the study area. Of the 31 indexes, 5 were selected: Showalter index (SI), wind speed at 500 hPa (SPD500), dew point temperature at 850 hPa (Td850), relative helicity between 0 and 3 km (SREH3km), and wet bulb zero height (WBZ). Combination of these indexes in a logistic equation gives the probability of SS risk/no risk in the study area. Results of the logistic equation show a Probability of Detection (POD) of 0.94 and a False Alarm Ratio (FAR) of 0.22. The second part of the article describes regionalization of the study area by SS spatial distribution according to the logistic equation. Thus, using multivariate techniques, we used principal component analysis (PCA) in T-mode and posterior cluster analysis, getting four clusters according to the spatial distribution of SS thermodynamic behavior and the distribution of storms observed via radar data. Clusters 1 and 2 showed probabilities of hail occurrence that were lower than Clusters 3 and 4, mainly affecting the MEV and the eastern end of the study area. Likewise, the predicted hail area was more extensive in those last two clusters. These results provide a new tool that complements those previously developed for this study area, toward improving SS prediction and pinpointing these storms in space and in time.