对流层天顶延迟模型研究进展及其在中国区域的精度分析

Research progress of zenith tropospheric delay model and its accuracy analysis over China

对流层延迟是空间大地测量技术的主要误差源之一,数据处理中对流层延迟的修正需要借助对流层延迟模型.本文首先从物理原理出发,梳理了对流层天顶延迟模型的研究历程和最新进展.按照时间顺序,对流层延迟模型先后经历了依赖实测气象资料的经典模型、不依赖实测参数的经验模型和以数值模型气象资料为基础的高分辨经验模型三个发展阶段.其次,本文利用中国区域内219个GPS测站2014—2015年两年实测的天顶延迟,对后两类经验模型中国际最新通用的代表模型UNB3m和GPT2w在中国境内的实际精度进行评估.精度评估结果显示:UNB3m模型在中国地区的平均Bias为-0.85 cm,平均RMSE为5.14 cm,其精度不随计算时间分辨率的变化而显著变化;模型参数的空间分辨率对GPT2w模型在中国地区的精度的影响不大,但GPT2w模型精度随计算时间分辨率的提高显著下降,2 h分辨率时GPT2w模型的平均RMSE分别为8.07 cm(1°参数文件)和7.97 cm(5°参数文件),1天分辨率时GPT2w模型的平均RMSE分别为3.49 cm(1°参数文件)和3.59 cm(5°参数文件);受水汽分布的影响,时间上,两个模型在冬季的精度相对最高,在夏季的精度相对最差,空间上,两个模型在高纬度和高海拔地区的精度相对较高.以上分析可为中国区域用户对流层延迟模型的选择提供参考.

Tropospheric delay has been recognized as one of the major error sources for many radio space geodetic techniques. The elimination of it is normally dependent on the tropospheric delay model. Firstly, proceeding from the physical principles, the research history and the latest progress of zenith tropospheric delay model are summarized in this paper. According to the chronological order,the tropospheric delay model has undergone three stages of development： classical model relying on observed meteorological data,empirical model not dependent on measured parameters and high resolution empirical model based on numerical weather model data. Secondly,the actual accuracy of UNB3m and GPT2w,the latest international representative models of the latter two types of empirical models,is evaluated over China using the measured zenith tropospheric delays of 219 GPS stations from 2014 to 2015. The accuracy assessment results are as follows. The average Bias of the UNB3m in China is-0. 85 cm and the average RMSE is 5. 14 cm.The accuracy of UNB3 m does not change significantly with the calculated time resolution. The accuracy of GPT2w in China is not affected by the spatial resolution of the parameters,but it decreases significantly with the improvement of the calculated time resolution.The average RMSE of GPT2w with 2 hours resolution are 8. 07 cm（ 1°grid parameters file） and 7. 97 cm（ 5° grid parameters file）respectively. The average RMSE of GPT2w with 1 day resolution are3. 49 cm（ 1° grid parameters file） and 3. 59 cm（ 5° grid parameters file）. Affected by the distribution of water vapor,the two models achieve the highest accuracy in winter and the lowest accuracy in summer,and they also perform relatively better at higher latitudes and higher altitudes. The above analysis can provide a reference for the selection of zenith tropospheric delay models in China.