Comparative evaluation of NASA, ERA5, and observational data for accuracy and reliability

Abstract

This study evaluates NASA POWER, ERA5, and local in-situ meteorological data at Bal & imath;kesir, Turkey (2019-2022), focusing on temperature, pressure, and wind speed. Station coordinates (39.6472 degrees N, 27.8861 degrees E) were bilinearly interpolated from the four surrounding ERA5 (0.25 degrees x 0.25 degrees) and NASA POWER (0.5 degrees x 0.625 degrees) grid points to obtain point-scale estimates. Descriptive statistics, Pearson correlation, RMSE, and bias were calculated, and differences were tested using Wilcoxon signed-rank tests (all p < 0.001). Results show that both reanalysis products capture broad temperature patterns, with NASA POWER outperforming ERA5 (RMSE 2.56 degrees C vs. 3.33 degrees C; bias -1.06 degrees C vs. -1.01 degrees C). For pressure, NASA POWER again leads (RMSE 7.70 hPa; bias -7.60 hPa) while ERA5 underestimates by -35.23 hPa (RMSE 35.27 hPa). Wind speed exhibits the greatest discrepancy: ERA5's grid-cell mean of 8.03 m/s greatly exceeds the local station's 1.33 m/s, whereas NASA POWER's 4.24 m/s remains closer but still overestimates (RMSE 3.47 m/s; bias 2.91 m/s). The wind-speed gap reflects spatial-scale mismatch and coarse roughness parameterization: ERA5 averages across varied land covers and underestimates sheltering, while local terrain features greatly slow winds. We recommend selecting datasets by application-favoring NASA POWER for temperature and pressure studies-and employing high-resolution mesoscale modelling (e.g., WRF at <= 1 km) to resolve sub-grid terrain effects and develop bias-correction or machine-learning calibration schemes for ERA5 wind assessments.