Evaluating the Accuracy of MERRA-2 Reanalysis Data Against In-Situ Observations Under Varying Weather Conditions

Abstract

Accurate estimation of surface atmospheric parameters is essential for climate and environmental studies, especially in regions with limited ground-based observations. This study addressed a key gap by evaluating Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) reanalysis data against in-situ measurements of solar radiation, temperature, and wind speed under varying weather conditions. In-situ data were collected in Osogbo, Osun State, Nigeria, using a Professional Weather Station and a Solar Meter (SM205), while MERRA-2 data were obtained from the NASA GIOVANNI platform, covering January 1 to March 31, 2025. Data were synchronized to hourly resolution and categorized by weather type. Statistical analysis applied linear, quadratic, and logarithmic regression models, with performance evaluated using Root Mean Square Error (RMSE) and Coefficient of Determination (R²). Results showed strong agreement between MERRA-2 and in-situ temperature across most conditions, with the quadratic model performing best. Under sunny conditions, R² reached 0.994 and 0.948, with RMSEs of 0.27 °C and 0.62 °C. Overcast days also showed good reliability (R² = 0.860 and 0.767), though accuracy declined during rainfall (R² = 0.377; RMSE = 2.61 °C). For solar irradiance, performance varied by condition: the quadratic model performed best on sunny Day 1 (R² = 0.759), while the logarithmic model gave the lowest RMSE on Day 2 (114.78 W/m²). Overcast and rainy Day 2 favored the quadratic model (R² = 0.925 and 0.881). Wind speed showed poor agreement across all conditions, with best R² = 0.289 and RMSE up to 8.97 m/s.