Tropospheric delay is one of the main sources of measurement error in global navigation satellite systems.It is usually compensated by using an empirical correction model.In this paper,temporal and spatial variations of the global zenith tropospheric delay(ZTD) are further analyzed by ZTD time series from global International GNSS Service stations and annual ZTDs derived from global National Centers for Environmental Prediction reanalysis data,respectively.A new ZTD correction model,named IGGtrop,is developed based on the characteristics of ZTD.Experimental results show that this new 3D-grid-based model that accommodates longitudinal as well as latitudinal variations of ZTD performs better than latitude-only based models(such as UNB3,EGNOS,and UNB3m).The global average bias and RMS for IGGtrop are about-0.8 cm and 4.0 cm,respectively.Bias values for UNB3,EGNOS,and UNB3m are 2.0,2.0,and 0.7 cm,respectively,and respective RMS values 5.4,5.4,and 5.0 cm.IGGtrop shows much more consistent prediction errors for different areas than EGNOS and UNB3m,In China,the performance of IGGtrop(bias values from-2.0 to 0.4 cm and RMS from 2.1 to 6.4 cm) is clearly superior to those of EGNOS and UNB3m.It is also demonstrated that IGGtrop biases vary little with height,and its RMS values tend to decrease with increasing height.In addition,IGGtrop generally estimates ZTD with greater accuracy than EGNOS and UNB3m in the Southern Hemisphere.