Meteorological applications of a surface network of Global Positioning System receivers

This thesis presents meteorological applications of water vapour observations from a surface network of Global Positioning System (GPS) receivers. GPS signals are delayed by the atmosphere due to
atmospheric refraction and bending. Mapped to the zenith, this delay is called Zenith Total Delay (ZTD). The ZTD can be separated in a hydrostatic and a wet delay. The first can be approximated by the
surface pressure, while the second is related to Integrated Water Vapour (IWV); this relation depends on the surface temperature. Currently, radiosondes are the only operational source of upper air humidity observations, however these observations are sparse in space and time. GPS IWV can fill this gap, albeit that the observable is an integral quantity. A study of the correlation of radiosonde
observations and IWV shows that the change over time of IWV is closely related to the change in specific humidity at 2 km.

The quality of the GPS IWV, assessed by comparison with numerical weather prediction (NWP) model and radiosonde observations, shows accuracies of 14 mm in ZTD and 2 kg/m2 in IWV.

In the processing of GPS signals it is assumed that the atmosphere through which the signals propogate is symmetric and close to climatology. A consequence of this assumption is that when a strong water vapour gradient is present at a GPS site, systematic errors of around -3 mm in ZTD are observed (which corresponds to approximately -0.5 kg/m2 in IWV).

Geostationary satellites can observe upper tropospheric water vapour in cloud free areas. Timeseries analysis of the change in GPS IWV and
change in upper tropospheric water vapour gives a rough estimate of the change in vertical water vapour distribution. A spectral analysis of
timeseries of the residual signal of GPS zenith delay estimates shows that there is a relation between the power of the residual signal and
Convective Available Potential Energy (CAPE) from radiosonde. A correlation of 0.6 is observed which is remarkable because the power of the residual is based on a timeseries of one hour of GPS residuals
while CAPE is derived from atmospheric profile information.

Real-time GPS IWV maps are constructed using a two-dimensional variational technique. These maps are validated against NWP analyses and forecast fields. The statistics show that the IWV maps are of good quality. Two thunderstorm cases show the applicability of these maps for nowcasting.

A comparison study shows that slant delay estimates from GPS contain information with a three-dimensional character. This three-dimensional
character is exploited further by assimilating slant delay observations using a three-dimensional variational (3DVAR) scheme. The used GPS network had a horizontal resolution of approximately 100 km, which impaired the resolution of the analysed water vapour field. Nevertheless, when compared to independent radiosonde observations, the bias of 3DVAR GPS was smaller than the NWP six hour forecasts.