Accuracy evaluation and analysis of the HGPT2 model in Europe

Abstract Accurate acquisition of meteorological parameters is fundamental for improving the performance of Global Navigation Satellite System (GNSS) applications, especially in water vapor retrieval and precise point positioning. As an effective and reliable method, parametric modeling supports robust global-scale estimation of essential variables, including surface temperature (Ts), weighted mean temperature (Tm), and surface pressure (P). Among existing models, the second-generation Hourly Global Pressure and Temperature (HGPT2) model has attracted considerable attention due to its capability to provide meteorological parameters with high spatiotemporal resolution. However, a systematic assessment of its accuracy and applicability across the European region has been lacking. To address this gap, this study presents the first comprehensive evaluation of HGPT2 in Europe, utilizing quality-controlled data from 61 radiosonde stations spanning 2019 to 2021, with outlier removal performed using the Interquartile Range (IQR) method. Results show that for Ts, the bias, root mean square error (RMSE), and mean absolute percentage error (MAPE) are 0.03 K, 4.00 K, and 1.13 %, respectively; for Tm, the corresponding values are −0.09 K, 4.56 K, and 1.34 %; and for P, they are 0.22 hPa, 8.43 hPa, and 0.68 %. Further analysis indicates clear geographic heterogeneity in model accuracy, as well as periodic variations across interannual and seasonal scales. Overall, the HGPT2 model demonstrates favorable accuracy and stability in the European region, establishing it as a reliable reference for practical applications; however, given the spatiotemporal distribution characteristics of its accuracy, there remains considerable potential for further optimization of the model in future studies.