时间频率学报

为了评估当前常用对流层延迟修正模型的性能，选取SAASTAMOINEN、BLACK、EGNOS及GPT2w 4种模型进行了详细分析。由低到高选择了6个不同纬度的测站，通过比较模型气象数据与实际气象数据来评估修正效果。研究首先使用4种非实测气象参数模型方案(EGNOS、GPT2w、GPT2w+SAASTAMOINEN及GPT2w+BLACK)修正对流层延迟。随后，利用2020年IGS(International GNSS Service)的实际气象数据，对6个测站分别应用SAASTAMOINEN和BLACK模型进行修正，并以2020年IGS天顶对流层延迟(zenith tropospheric delay, ZTD)作为真实值进行对比。根据均方根误差(RMSE)和平均绝对百分比误差(MAPE)两个精度指标，分析评估各模型在选定的6个测站中的修正效果，结果表明：EGNOS模型在中高纬度地区表现出较高的修正精度且计算简便，但在低纬度地区的表现较差；GPT2w模型提供了修正精度，但其建立过程复杂，且上述两种基于非实测数据的模型无法有效反映ZTD的短期快速变化；实测气象数据能够准确反映ZTD的短期变化且精度较高，但数据获取难度相较于模型数据更高且容易缺失。这些发现为对流层延迟修正模型的选择和应用提供了重要参考。

To evaluate the performance of commonly used tropospheric delay correction models, a detailed analysis was conducted on four models: SAASTAMOINEN,BLACK,EGNOS,and GPT2w.Six stations, representing a range of latitudes from low to high, were selected to assess correction efficacy by comparing model-based meteorological data with actual meteorological data.Initially, four correction strategies utilizing unmeasured meteorological parameters were applied(EGNOS,GPT2w, GPT2w+SAASTAMOINEN,and GPT2w+BLACK).Subsequently, actual meteorological data from the 2020 IGS(International GNSS Service) was used to apply the SAASTAMOINEN and BLACK models to the six stations, with the 2020 IGS zenith tropospheric delay(ZTD) serving as the reference true value.Based on two accuracy indicators, namely the root mean square error(RMSE) and the mean absolute percentage error(MAPE),the correction effects of various models at the selected six measurement stations were analyzed and evaluated.The results show that: The EGNOS model demonstrated high accuracy and ease of use at mid-to high-latitudes, but showed lower performance at low latitudes.The GPT2w model achieved the higher correction accuracy, though its complexity is a notable drawback, and the above two models based on non-measured date were insufficient for capturing short-term rapid changes in ZTD.But actual meteorological data accurately reflects short-term ZTD variations and offers high precision, data acquisition is more difficult than model data and is prone to missing.These findings provide valuable insights for selecting and applying tropospheric delay correction models.