The ZTD consists of the hydrostatic delay (ZHD) and wet delay (ZWD). The ZHD can be well calculated from surface meteorological data, ranging from 1.5 to 2.6 m, which accounts for 90% ZTD. It derives from the relationship with hydrostatic equilibrium approximation for the atmosphere. Under hydrostatic equilibrium, the change in pressure with height is related to total density at the height h above the mean sea level by
where ρ(h)and g(h) are the density and gravity at the height h. It can be further deduced approximately as
where k is constant (2.28 mm/hPa) and p0 is the pressure at height h0 [Davis et al., 1985]. It shows that the ZHD is proportional to the atmospheric pressure at the site. The ZWD is highly variable due possibly to varying climate, relating to the temperature and water vapor. Unfortunately, only fewer IGS sites have meteorological instruments which can directly obtain the real ZWD or PWV. If one calculated the ZWD or PWV using the meteorological data from the European Centre for Medium-Range Weather Forecasts (ECMWF), it has some differences relative to the real observation results with meteorological instrument data [Hagemann et al., 2003]. Therefore we here analyze the variation and relationship between atmospheric parameters at GPS stations equipped with the meteorological instruments. For example, the GPS station Wettzell (WETT), Germany has equipped the meteorological instruments. The positions of humidity and temperature sensors are the same as the GPS antenna, and the pressure sensor is 10.5 m below the GPS antenna. The data frequencies of relative humidity, temperature and pressure are all 15 min, and their accuracies are 1.5% (relative to height), 0.3°C and 0.1 mbar, respectively. Figure 4 shows an example of the time series of various atmospheric parameters at the site WETT from the year 2002 to 2006. From top to bottom panels, it sequentially denotes zenith tropospheric delay (ZTD), zenith hydrostatic delay (ZHD), zenith wet delay (ZWD), surface temperature, pressure, and relative humidity. We examine how much of that correlation between surface measurements. It has been noted that the ZHD is highly proportional to the atmospheric pressure at the site and relatively stable, while the ZWD is positively correlated with the temperature and also correlated with the relative humidity. This is due to the combined effects of increasing evaporation and a strong increase in the water vapor saturation pressure. The correlation coefficient between ZWD and surface temperature is 0.81, and the remaining is maybe correlated to the water vapor. The correlation coefficient between ZTD and ZWD is about 0.95, reflecting a good correlation between ZTD and ZWD variations. In addition, we further analyze and compare these parameters at other GPS stations with meteorological data (MATE (Italy), ONSA (Sweden), NYAL (Norway), TSKB (Japan), WES2 (USA), ALGO (USA), FAIR (Alaska, USA), KOKE (Hawaii, USA), and HART (Australia)) and it has shown almost the same correlations with the WETT station. Therefore it has been indicated that the seasonal variations of ZTD are due primarily to the wet component (ZWD), even though the wet delay is only 10% of the total delay (ZTD). In addition, the ZHD is proportional to the atmospheric pressure (equation (5)), while the pressure is mainly related to height (seeing the following equation (6)), and therefore the ZHD is almost constant, again showing that the seasonal variations of ZTD are due primarily to the ZWD.
 The mean ZTD values at all GPS sites are shown in Figure 5 as a color map. It has been noted that lower ZTD values are found at the areas of the Tibet (Asia), Andes Mountain (South America), Northeast Pacific and higher latitudes (Antarctica and Arctic), and the higher ZTD values are concentrated at the areas of middle-low latitudes (also see Figure 6). Figure 7 shows the distribution of ZTD at all IGS sites with the altitude (above the global mean sea level). It has been clearly seen that the ZTD values decrease with increasing altitude. This is due to the atmospheric pressure variations with the height increase. Atmospheric pressure is the above any area in the Earth's atmosphere caused by the weight of air. Air masses are affected by the general atmospheric pressure within the mass, creating areas of high pressure (anticyclones) and low pressure (depressions). Low pressure areas have less atmospheric mass above their locations, whereas high pressure areas have more atmospheric mass above their locations. As elevation increases, there are exponentially fewer and fewer air. Therefore atmospheric pressure decreases with increasing altitude at a decreasing rate. The following relationship is a first-order approximation to the height (http://en.wikipedia.org/wiki/Air_pressure):
where P is the pressure in Pascals and h is the height in millimeters. On the basis of equation (5), ZHD can be expressed as 2.28 * 10(5−h/15.5). As the ZHD accounts for 90% of ZTD, we can further deduce the approximate ZTD at all GPS sites as an empirical formula:
where the units of ZTD and h are in millimeters, respectively. Comparing GPS-derived ZTD with the empirical formula estimations (Figure 7), it has shown a good consistency.