Performance Evaluation of ERA5 Reanalysis against Radiosonde Observations in Representing Atmospheric Vertical Variability on a Seasonal Scale in the Maritime Continent of Indonesia

Abstract

ERA5 reanalysis is widely applied in meteorological studies, but its performance requires evaluation in regions with complex seasonal dynamics such as the Indonesian Maritime Continent (BMI). This study assesses the ability of ERA5 to represent seasonal atmospheric vertical variability using radiosonde observations. The analysis covers December 2021 to November 2024 at three stations, namely Soekarno-Hatta, Juanda, and I Gusti Ngurah Rai, for temperature, relative humidity, zonal wind, and meridional wind at eight standard pressure levels from 925 to 100 hPa. Evaluation methods include Hovmöller analysis, Mean Absolute Error (MAE), Mean Error (ME), correlation coefficient, scatter plots, and box plots for December–January–February (DJF), March–April–May (MAM), June–July–August (JJA), and September–October–November (SON) seasons. Results indicate that ERA5 represents temperature very well, with MAE values of 0.3 to 1.1°C and correlations exceeding 0.9, although errors increase at 100 hPa. Wind components also show high correlations above 0.9 but exhibit underestimated amplitudes, particularly for zonal winds during DJF. In contrast, relative humidity shows good agreement in the lower layers but significant limitations in the upper layers, with MAE values up to 58 percent, correlations below 0.6, and overestimation above 300 hPa. These findings confirm ERA5’s reliability for temperature and wind analyses, while humidity remains a key source of uncertainty in tropical maritime regions.