一种新的多因子约束下的NWP反演ZTD残差改正模型

对流层延迟是GNSS定位的主要误差源之一，利用NWP模型的气象数据积分反演ZTD是当前研究热点.然而，采用两大气象预报中心(ECMWF和NCEP)的再分析资料反演ZTD的残差一般在±60 mm之间浮动，预报资料反演的ZTD的精度更差，都不能直接用于精密定位.一般是先将此反演的ZTD作为初值，设定先验方差，将残差作为未知参数求解.NWP反演的ZTD的精度，将直接影响对流层和模糊度参数在滤波过程中收敛速度.前人的研究表明，NWP反演ZTD的残差大小与测站所在纬度相关，利用纬度与残差的相关函数可提高NWP反演ZTD的精度，但效果并不明显.针对以上问题，比较ECMWF和NCEP再分析资料反演ZTD的精度，然后分析精度较高的ECMWF资料反演的ZTD的残差随温度、湿度、纬度、季节等因子变化的规律，并结合基于最小绝对残差法的多项式拟合方法拟合残差，提出一种新的多因子约束下的NWP反演ZTD的残差改正模型，从而提高NWP反演ZTD的精度.为验证模型的性能，以133个IGS站高精度ZTD为参考，拟合2015年ECMWF反演ZTD的残差，构建残差改正模型，并利用此模型改正2016年ECMWF反演的ZTD.实验结果表明：在纬度高于15°的地区，NWP反演的ZTD的平均残差和均方根误差比使用模型前分别减小了86.9%和36.3%.另外，对于较低纬度地区，此残差改正模型的效果不明显.

A new residual correction model for ZTD inversion of NWP under multi-factor constraints

Tropospheric delay is one of the main error source in GNSS positioning. Using the meteorological data of NWP model to invert ZTD is the current research hotspot. However， the ZTD residuals inverted by the reanalysis data of the two meteorological forecast centers (ECMWF and NCEP) are generally floating between ±60 mm， and the precision of ZTD inverted by Forecast data would be worse. As a result， both of them are not able to be used directly for precise positioning. The accuracy of ZTD inverted by NWP will directly affect the convergence speed of the troposphere and ambiguity parameters in the filtering process. Previous research shows that the residual size of the ZTD inverted by the NWP is related to the latitude of the station， and the correlation function between latitude and ZTD residual would improve the precision of the ZTD inverted by the NWP， but the effect is not obvious. To solve this issue， researchers usually take this ZTD as an initial value and set a priori variance， and take the residual as a unknown parameter. Immutable and inaccurate prior variance will directly lead to the troposphere and ambiguity parameters of user station converge slowly， even divergence in the filtering process. In view of the above problems， this paper compares the accuracy of ZTD inverted by ECMWF and NCEP reanalysis data， and then analyzes the change law of the residual of ZTD inverted by ECMWF data with temperature， humidity， latitude， season and other factors. On this basis， a new residual correction model for NWP ZTD inversion under multi-factor constraints is proposed in this paper to improve the accuracy of ZTD by NWP inversion. This model fit the ZTD residual using the polynomial fitting based on the least absolute residuals method. It will provide more accurate priori information for NWP ZTD. To verify the performance of this model， the high precision ZTD of 133 IGS stations are taken as reference， and the residual ZTD inverted by 2015 ECMWF are fitted to establish residual correction model. Then this model is applied to correct the ZTD inverted by 2016 ECMWF. Experimental results show that， in higher latitude areas (above 15 °)， the yearly residual and RMS error of the ZTD inverted by ECMWF is reduced separately by 86.9% and 36.3% than that before using the dynamic stochastic correction model. In addition， in the lower latitude areas， this residual correction model is not obvious.