针对对流层延迟影响全球卫星导航系统（GNSS）水汽反演精度的问题，借助河南省连续运行参考站综合服务系统观测数据（HNCORS）对GPT系列模型（GPT模型、GPT2模型、GPT2w模型和GPT3模型）在河南地区的适用性进行系统评估，实验结果表明：GPT3模型获取的气温、气压、水汽压及反演PWV结果精度最高，GPT2模型和GPT2w模型获取结果精度相当，GPT模型获取结果精度最差；与GPT模型相比，GPT3模型获取的气温、气压和水汽压结果 BIAS值精度分别提升了43.05%、39.50%、48.92%,RMS值精度分别提升了48.28%、33.01%、43.50%;GPT模型、GPT2模型、GPT2w模型和GPT3模型基于HNCORS获取的PWV结果平均RMS值分别为3.96、3.17、3.04、2.35 mm,GPT3模型获取PWV结果总体精度优于3 mm，因此在河南地区进行GNSS水汽反演推荐使用GPT3模型。

In response to the impact of tropospheric delay on the accuracy of water vapor inversion in the global navigation satellite system(GNSS), we used observation data from the Henan continuously operating reference stations(HNCORS) to systematically evaluate the applicability of GPT series models(GPT model, GPT2 model, GPT2w model, and GPT3 model) in the Henan region. The experimental results show that the GPT3 model has the highest accuracy in obtaining temperature, pressure, water vapor pressure, and inverted PWV results, the GPT2 model and GPT2w model have the same accuracy in obtaining results, while the GPT model has the worst accuracy in obtaining results. Compared with the GPT model, the accuracy of BIAS values for temperature, air pressure and water vapor pressure obtained by the GPT3 model has been improved by 43.05%, 39.50% and 48.92%, respectively. The accuracy of RMS values has been improved by 48.28%, 33.01% and 43.50%, respectively. The average RMS values of PWV results obtained from the GPT model, GPT2 model, GPT2w model and GPT3 model based on HNCORS are 3.96 mm, 3.17 mm, 3.04 mm and 2.35 mm, respectively. The overall accuracy of PWV results obtained from the GPT3 model is better than 3 mm. Therefore, it is recommended to use the GPT3 model for GNSS water vapor inversion in Henan region.