Slant Water Vapour Retrievals using GNSS
van der Marel, H.¹; De Haan, S.²; De Vries, J.^2
1Delft University of Technology; ²KNMI

As GNSS signals pass through the Earth’s atmosphere they are delayed by water vapour and other molecules in the atmosphere. The standard approach in GNSS software is to estimate the the total delay in the zenith direction (ZTD) as an extra parameter. In this paper the standard method for estimating delays in the zenith direction is extended to estimating delays in the line-of-sight direction, the so-called slant delays. These slant delays in turn can be used in numerical weather prediction models, as they contain important information on water vapour in the line-of-sight direction.

Compared to the more traditional zenith delays (ZTD) and vertically integrated water vapour (IWV), slant delays (STD) have several advantages: Firstly, when slant delays are used in numerical weather forecasting models it is known exactly which part of the atmosphere is sampled. GPS measurements are not homogeneously distributed over the sky, and also the satellite constellation is changing over the period of one day, and thereby the part of the sky that is sampled for the ZTD computation also changes over one day. The slant delays are much more accurate in this respect as we know precisely the elevation and azimuth of the underlying observations. Secondly, when using the slant delays one has information on gradients and non-isotropic irregularities in the atmosphere. Thirdly, slant delays are less sensitive for assumptions about the atmospheric homogeneity and mapping functions. Also, with slant delays it is much safer to use low elevation observations. Finally, there are more slant delays than zenith delays. In principle slant delays can be computed every measurement epoch, but this is not the ideal approach as it produces a huge amount of data. Therefore, the slant delays are smoothed into 5-10 minute intervals.

On the other hand, one may not simply use the slant delays as independent and uncorrelated observations. Instead, the slant delays should be considered as a sort of preprocessed raw observations, whereby all non-meteorological effects such as station positions, ocean loading, tides, carrier phase ambiguities, antenna phase delay calibration, multipath effects and satellite and receiver clocks, have been removed from the GPS data.

Slant delays are more sensitive to unmodelled site dependent multipath effects than the ZTD. Therefore, it is essential that the site dependent multipath effects are removed from the data, which otherwise may wrongly be interpreted as an atmospheric signal. This has been achieved by residual stacking. Very significant site and antenna related effects were found, which after correction, improved results considerably. Also, for different receiver types, different elevation dependent weighting functions were found. The back-substitution into the GNSS processing also led to an improvement in estimates of position and tropospheric (zenith) delays. Furthermore, residual stacking proofed to be a very useful tool in detecting and diagnosing problems in the processing.

In the paper also first results from validation with numerical weather prediction models are shown.