Observing three dimensional water vapour using a surface network of GPS receivers

Abstract. Atmospheric water vapour is highly variable both in space and time. In an operational sense, only radiosonde provide vertical information on water vapour. Radiosondes are generally launched two to four times per day at synoptic times and sample primarily synoptic scales. For nowcasting purposes these observations are very valuable but obviously lose their importance with elapsing time. Water vapour observations from a surface network of Global Positioning System (GPS) receivers can fill this information gap. In this paper, a GPS network is used to observe integral water vapour quantities along the line of sight, so-called Slant Water Vapour (SWV). Using a variational technique (3DVAR) a three-dimensional water vapour field is reconstructed and its performance is investigated by assimilating SWV observations deduced from a simulated atmosphere (so-called nature run). The forecasts from a high resolution limited area model (HIRLAM) embedded in the synthetic atmosphere of the nature run is compared to the separate GPS-3DVAR estimates. This experiment showed that assimilation of SWV resulted in a smaller bias and standard deviation than the HIRLAM forecast with the nature run. Besides simulated data, real SWV observations are used to assess impact. Two experiments were conducted; one with a HIRLAM six hour forecast as a background field (updated every six hours) and one with persistence as background (updated every hour). The first experiment showed a reduction of the bias between radiosonde observations compared to HIRLAM forecast. The second experiment, which has no information inherited from HIRLAM, showed to have smaller biases with independent radiosonde observations than the HIRLAM analysis. The used network, however was too sparse to detect water vapour inversions correctly.