2.1. Comparison of Heatwave Frequency
Differences in Heatwave Days between ERA5 and ASOS
ASOS observations showed high heatwave frequencies mainly over southern inland and southeastern regions, with some stations experiencing more than 40 heatwave days. ERA5 reanalysis data reproduced a broadly similar spatial pattern; however, the overall number of heatwave days was generally lower. The largest discrepancies appeared over southern inland and eastern coastal regions, where ERA5 simulated substantially fewer heatwave days than the observations, with some stations showing differences greater than 20 days. In addition, the discrepancy tended to increase at stations with higher observed heatwave frequencies.
These differences are mainly related to the grid-averaged characteristics of ERA5 reanalysis data. Through spatial averaging, ERA5 tends to smooth localized extreme temperature events, which may weaken urban heat island effects, basin-related warming, and other local-scale heat extremes. In coastal regions, the inclusion of ocean areas within ERA5 grid cells likely contributed to lower simulated heatwave frequencies due to the moderating effect of the ocean. On the other hand, a few inland stations showed slightly higher heatwave frequencies in ERA5 than in ASOS observations. This may be related to representativeness differences between point observations and grid-averaged ERA5 values, although further analysis would be required to identify the exact causes.
Overall, ERA5 reproduced the general spatial pattern and regional tendency of heatwaves reasonably well, but it tended to underestimate the occurrence of intense extreme heat events compared to station observations.
Fig. Spatial Distribution of Heatwave Days
The nationwide average number of heatwave days was approximately 19.9 days in ASOS observations, while ERA5 showed only about 8.1 days. This indicates that ERA5 significantly underestimated heatwave frequency compared to observations.
This underestimation is related to the spatial averaging characteristics of ERA5. Near the heatwave threshold of 33°C, even a small temperature difference can determine whether a day is classified as a heatwave event.
Fig. Comparison of National Mean Heatwave Days
The comparison of Tmax between ASOS and ERA5 showed a clear positive correlation, indicating that ERA5 reproduced the overall temperature variability reasonably well. However, ERA5 tended to show lower temperatures in extreme hot conditions.
2.2. Comparison of Heatwave Duration
Differences in Heatwave Duration between ERA5 and ASOS
In ASOS observations, the mean heatwave duration was around 3 days, with some stations exceeding 5 days. In contrast, ERA5 showed shorter durations of about 1–2 days on average.
Similarly, the maximum heatwave duration exceeded 20 consecutive days at some ASOS stations, whereas ERA5 mostly showed durations shorter than 10 days.
This result suggests that heatwave continuity in ERA5 was frequently interrupted when temperatures fell slightly below the 33°C threshold.
Fig. Distribution of Mean and Maximum Heatwave Duration
ASOS observations showed prolonged heatwaves mainly over southern inland and southeastern regions. ERA5 reproduced similar spatial patterns but generally showed shorter durations overall.
In particular, coastal regions exhibited weaker long-duration heatwaves in ERA5, likely due to the cooling influence of nearby ocean areas included within the grid cells.
→ Therefore, although ERA5 captures the broad spatial characteristics of heatwaves, it has limitations in reproducing long-lasting extreme heat events observed in reality.
Fig. Spatial Distribution of Maximum Heatwave Duration
2.3. Comparative Analysis of Heatwave Intensity
Comparison of Heatwave Intensity (Tmax) between ASOS and ERA5
The preprocessing procedure and datasets used in this analysis were identical to those applied in the previous analyses.
This analysis was conducted only for days classified as heatwave events in both ASOS and ERA5 datasets.
The purpose of this approach was not simply to compare the occurrence of heatwaves, but to evaluate how well ERA5 reproduced the actual intensity of the same heatwave events.
Heatwave intensity was defined using daily maximum temperature (Tmax). The bias characteristics of the reanalysis data were analyzed through the Tmax difference between ASOS and ERA5 (ASOS − ERA5).
The comparison of Tmax during heatwave days showed a positive correlation between ASOS and ERA5, although the correlation coefficient (R = 0.38) was relatively low. This suggests that ERA5 has limitations in fully reproducing localized extreme temperature variability.
The mean bias was approximately +0.77°C and the RMSE was about 1.23°C. Most data points were distributed below the 1:1 reference line, indicating that ERA5 generally underestimated actual heatwave intensity.
In particular, the difference between ASOS and ERA5 became larger when observed Tmax exceeded 33°C, showing that the underestimation tendency of ERA5 intensified during extreme heatwave conditions.
→ In summary, ERA5 reproduces the occurrence of heatwaves reasonably well, but tends to conservatively represent the peak intensity of extreme heat events.
Fig. Comparison of Tmax Scatter Plot during Heatwave Events
The monthly distribution of Tmax bias from June to August showed that the underestimation tendency of ERA5 became stronger as summer progressed.
August exhibited not only the largest median bias but also the widest Box and Whisker ranges. This indicates that the uncertainty and discrepancy between observations and reanalysis data increase during periods of more severe heatwaves.
In addition, positive outliers were clearly observed in all months, indicating cases where ERA5 significantly underestimated actual heatwave intensity.
→ As heatwave intensity increases, the ability of ERA5 to reproduce extreme high temperatures tends to weaken.
Fig. Monthly Tmax Bias (Box Plot)
The spatial analysis conducted only for heatwave days also revealed a clear underestimation tendency of ERA5 in coastal regions.
Large Tmax differences between ASOS and ERA5 were observed in southeastern coastal cities such as Busan, Ulsan, and Changwon, showing patterns similar to those found in the general summer temperature analysis. This indicates that ERA5 tends to underestimate extreme heat intensity in coastal areas even during heatwave periods.
This phenomenon is interpreted as the result of ocean cooling effects within ERA5 grid cells, which suppress the strong surface heating occurring over land. In contrast, inland regions showed relatively small biases and more stable reproduction performance.
→ Therefore, when using ERA5 for heatwave intensity analysis, additional bias correction is particularly necessary for coastal regions.
