Evaluation of ERA5 in Severe Convective Environments against soundings in Spain

Characterizing pre‑convective environments is fundamental to research on deep moist convection, yet the most direct observations (radiosondes) are spatially sparse and typically available only twice per day. For this reason, reanalysis datasets, such as ERA5, are widely used, although their limitations must be considered. This study evaluates the reliability of ERA5 in representing convective environments, focusing on  proximity soundings of thunderstorms and severe convective weather. The analysis is conducted over Spain, a region in the western Mediterranean.

Multiple convective parameters derived from ERA5 were compared with those obtained from radiosonde observations. The poorest agreement occurs for boundary‑layer‑related and parcel‑based parameters, particularly convective inhibition (CIN). The strongest bias affects extreme CAPE values, which ERA5 consistently fails to capture (missing 70% of events with CAPE > 3000 J kg^-1) due to its inability to represent extreme temperature and moisture conditions in the boundary layer.

In contrast, many parameters show strong consistency between ERA5 and soundings, including kinematic variables (e.g., vertical wind shear, mean wind, Bunkers storm motion) and several thermodynamic indicators (e.g., freezing‑level height, low‑level relative humidity, DCAPE, equivalent potential temperature). Severe‑weather environments as a whole (hail, downbursts, heavy precipitation and tornadoes) exhibit biases similar to those of thunderstorm environments not showing these hazards. However, very‑large‑hail cases are associated with extreme CAPE, and thus are particularly affected by its underestimation.

The systematic omission of extreme CAPE, especially along the Mediterranean coast, may have important implications for future climatological studies and hazard‑assessment applications. The agreement between our findings and earlier studies in Europe and North America suggests that several of the identified biases are likely systematic features of ERA5 rather than region‑specific issues.