Conclusion

This study investigated the impact of data assimilation on the representation of daily maximum temperature (Tmax) and heatwaves by comparing ASOS observational data with ERA5 reanalysis data during the summer of 2024 (June–August). In particular, the performance and limitations of ERA5 were evaluated in terms of heatwave frequency, duration, and intensity.

​

The results showed that ERA5 reproduced the overall temperature variability and the broad spatial distribution of heatwaves across the Korean Peninsula reasonably well. The general correlation of Tmax and the spatial patterns of heatwave occurrence were similar to those observed in ASOS data. This indicates that data assimilation–based reanalysis datasets effectively simulate large-scale climate fields and average atmospheric conditions.

​

However, clear limitations were identified in the representation of extreme heat characteristics.

In the general Tmax analysis, a pronounced coastal–inland contrast was found in the spatial distribution of mean bias. ERA5 substantially underestimated temperatures in southwestern coastal and island regions — including Jindo (+3.72°C), Heuksando (+3.21°C), and Goheung (+3.15°C) — where ocean cooling effects within ERA5 grid cells suppressed land surface heating. In contrast, relatively stable and accurate performance was observed over flat inland areas such as Seocheongju, Cheonan, and Sejong, where the bias remained small.

Regarding heatwave characteristics, ERA5 showed the following systematic limitations: it generally underestimated the number of heatwave days, represented long-duration heatwave events as shorter episodes, and simulated lower Tmax values than observations during heatwave periods. These limitations are closely linked to the heatwave definition applied in this study — days with Tmax ≥ 33°C, with events requiring at least two consecutive such days. Because ERA5 systematically underestimates temperatures, days near the 33°C threshold are frequently misclassified as non-heatwave days, which not only reduces the total count of heatwave days but also breaks the continuity of heatwave events, leading to shorter simulated durations. The underestimation tendency further intensified at higher temperature extremes.

Notably, the coastal underestimation pattern identified in the general Tmax analysis was consistently reproduced during heatwave-only periods as well, suggesting that the ocean cooling effect within ERA5 grid cells persistently suppresses peak heat intensity in coastal regions regardless of the analysis context.

​

 In addition, the bias and uncertainty between ERA5 and observations increased during August, when summer heat intensified.  This implies that data assimilation–based reanalysis datasets may have limitations in fully capturing localized and short-term variability associated with extreme heat events.