GPS total water vapour – potential for future operational observations
John Nash
Met Office, Upper Air Technology Centre,
Beaufort Park, Easthampstead, Wokingham, Berks RG40 3DN
Phone 44-1344-855649, Fax. 44-1344-855897: e-mail
Introduction
The purpose of this paper is to outline the potential for improving knowledge of the distribution of water vapour in the lower troposphere provided by the measurements that can be derived from ground-based measurements of navigation signals from the Global Positioning System (GPS) satellites. This type of measurements would appear to fit well into a future design of the Global Observing system since it is able to fill gaps in observations over the land where satellite based measurements of water vapour prove difficult in the lower troposphere.
As a new system, the potential costs of running operations need to be assessed so that sensible choices about the scales of deployment can be made.
The principles of operation of the system and the challenges to be met in establishing regional networks will be discussed.
Examples of current measurement skills will be illustrated with observations in Europe from the COST 716 real time demonstration network.
GPS sensing technique
The time delay of a GPS signal from transmission to reception at the ground is influenced by the state of the ionosphere, surface pressure, vertical temperature structure and the total water vapour in the atmosphere. The influence of the ionosphere is usually cancelled out by observing the differences in phase delay between the L1 and L2 GPS frequencies. Then, if the position of the satellite is known very accurately, possible errors in the satellite clocks are identified and the height of the ground based sensor is identified accurately, then the total delay along the slant path to the satellite caused by the atmosphere can be computed. The GPS satellites change position relatively slowly with respect to the ground sensor and signals from a given satellite can be received for many hours at a time. It is usually possible to receive reliable signals from between 8 and 12 satellites distributed over a variety of elevations and azimuth from the sensor. In order to solve for the height of the sensor and the total zenith delay it is necessary to observe from several satellites along different slant paths (to reduce height errors to acceptable values). A network of sensors with some ground stations distributed along baselines larger than about 500 km from the centre of the network is also required.
Once the zenith delays are estimated along a variety of slant paths from a given site, the values are mapped into the vertical to yield a total zenith delay. This can then be provided to the potential user to assimilate into a numerical weather prediction model. The total zenith delay can be converted into total water vapour if the surface pressure and temperature are known. It is not clear whether the numerical model background temperature fields are sufficiently accurate to provide a significant advantage in assimilating total zenith delay or whether it is better to use total water vapour estimates derived independently of the model fields.
A considerable research effort is attempting to derive additional information about the distribution of water vapour in the vertical utilising the individual zenith delays along slant paths, particularly at low elevations. This is most likely to work if many GPS ground–based sensors are operated in a network with very small network spacing (5km or less). Whether the results obtained can justify the expense of all the sensors is yet to be established, and I will not consider this area any further in this presentation.
Network demonstrations
The GPS method of measuring water vapour was discussed in Bevis et.al [1], and results from the first proof of concept experiment GPS/Storm were first reported in 1995, by Rocken et al [2]. Subsequently the NOAA-GPS PW network was established by NOAA’s Forecast System Laboratory in the USA. Japan has undertaken extensive research and development using the JMA GPS network, established originally for seismological observations, Naito et al [3] and also see the Report of N. Mannoji, CIMO Rapporteur on this technique for CIMO-XIII. Within Europe various research projects have contributed to the EC supported COST 716 project “ Exploitation of Ground –based GPS for Climate and Numerical Weather Prediction Applications”. National research and development of the technique continues in many other countries.
In late 2000 I was invited to chair the Working Group of COST 716 tasked with establishing options for future operational networks in Europe. This paper will show the results that have been achieved in Europe and review some of the problems associated with establishing operational networks.
Real time observations
As noted above, it has been possible to demonstrate the information content of the GPS water vapour measurements since 1995, but this was usually on observations processed a significant time after the time of observation when the estimates of the precise orbit positions of the satellites became available. It is only within the last few years that the IGS (International GPS Service for Geodynamics) has begun to issue rapid orbital position estimates of sufficient quality that the total water vapour estimates can be achieved with time delays of less than 2 hours after observation. In addition techniques for orbit estimation have now developed to the stage that individual GPS processing centres can also attempt to generate their own rapid orbit estimates.
A second hindrance to real time network operation has been that in many countries where GPS sensors have been deployed for geodetic purposes, real time communications to the sites were poor. Thus, the move towards real time operations requires some investment in communications infrastructure. It is not always clear at the moment whether some GPS networks are currently running in real time or where the networks are being evaluated with observing system experiments based on processing after the event.
Characteristics of total water vapour
The random errors associated with real time total water vapour measurements over the UK are about 1 mm of water vapour and systematic bias is of similar magnitude. It is expected that these errors will not differ widely with geographic location. Fig.1 indicates the distribution of water vapour in the vertical for a saturated pseudoadiabatic atmosphere, and the dependence of total water vapour on temperature in saturated conditions. This dependence on temperature is quite strong, and leads to marked differences in total water vapour across frontal systems, even when there are no significant dry layers present.
When temperatures are cold, total water vapour is less than 10mm even for saturation and a 1-mm error corresponds to an equivalent relative humidity error of greater than 10 per cent. However, in tropical conditions with total water vapour higher than 60 mm in many conditions, then 1mm random error offers a random error in equivalent relative humidity of around 2 per cent, as good as can be achieved by the best radiosonde system. Thus, this is a system, which could be very useful for water vapour measurements in equatorial regions given that the sensor site and communications can be sustained.
Real-time Observations
The COST 716 Demonstration Network has built up within two years to a situation where more than 100 sites are regularly supplying data for processing in real time from more than 15 different countries. Most of the resultant total zenith delays are being supplied to a Data Assimilation centre within 2 hours of completing the GPS measurements. All these processing centres are from the geodetic community and for the future some form of partnership will have to be organised if operations are to continue. Fig.2 shows examples of GPS signals received at Herstmonceux in southern England and processed by several centres within Europe. Total zenith delay and integrated water vapour are plotted as a function of time for two days, with the last data added within 2 hours of the nominal time. Plots such as these can be obtained each day from the relevant COST 716 web page maintained by KNMI, ( During this period the HIRLAM numerical weather prediction model fields were similar to the measurements at Herstmonceux.
Fig. 3 contains time series of measurements from 4 sites in early July 2002, when there were large discrepancies between the model fields and actual measurements over central France for most of the night of 8 July. The situation at 12 UTC on this day over the UK, France, Belgium and the Netherlands is summarised in this plot of measurements from real time GPS sensors. Note the very strong gradient in total water vapour over the UK associated with lower troposphere temperature change across a frontal surface. This system gave very heavy rainfall over parts of the UK that day. In France the low values of total water vapour rose rapidly later in the day as a line of thunderstorms moved eastwards from the coast. The lowest values of total water vapour occurred just in front of the thunderstorms and the high values once the thunderstorms were overhead. These data were taken straight from the web page and were not subjected to any further editing. Similarly data from the majority of the COST 716 stations were used to derive the estimate of the total water vapour field shown in Fig.5 for 06 UTC in the morning on 19 July 2002. On this day, the total water vapour remained lower than 20 mm over France and Spain and much of Germany and Switzerland. High values of total water vapour, 35 to 40 mm were found aligned north – south from Sweden down to Croatia, associated with a trough, with a smaller pool of higher values associated with a trough over the Adriatic Sea. Later in the day thunderstorms developed close to the regions of high total water vapour identified.
Network development
Within Europe there are known to be large numbers of GPS sensors that are not connected up in real time to the current demonstration network. For instance, within the UK there are more than 40 sites used for lighthouse navigation, tide gauges and sea level measurements, maintenance of the national grid reference for surveys and university experiments associated with communications infrastructure. To install 40 GPS sites for operational use would cost far too much for the Met Office in the current economic climate. However, if some form of partnership agreement can be made with these other institutions then it should be possible to build a high resolution-observing network for water vapour for the future. Fig.6 shows the distribution of water vapour around a severe thunderstorm that moved northwards over the UK, with the water vapour produced using observations from a combination of Met office and other institution sites. The air to the north of the storm was saturated so that the gradients in water vapour in this area represent temperature gradients in the horizontal in the lower troposphere.
Analyses such as these have led Met Office forecasters to request that data be delivered with a time lag of less than 1 hour if possible with as high a resolution in the horizontal that can be achieved without large scale investment. Authorisation to develop the basis of an UK GPS observing network has been given.
Summary
Development of GPS water vapour sensing has progressed to the stage where measurements of suitable quality can be delivered in a timely fashion to users.
Effective exploitation of the GPS technique requires working arrangements to be established with other agencies that deploy suitable GPS sensors, e g. for geodesy, seismology, or land survey purposes.
Collaboration in processing observations needs to be established, whether the processing centre is hosted by a Met Service for a given region, or by a national geodetic centre
Acknowledgements: Real time data for the UK has been made possible by the processing centre of the Observatory of Pecny, Czech Republic. Internally within the UK, IESSG, Nottingham University has made available and processed data from sensors within the UK for the detailed post processing studies.
References:
[1] Bevis. M., Businger, S., Herring, T., A, Rocken, C., Anthes, A., and Ware, R..1992 Remote Sensing of the atmospheric water vapour using the Global Positioning System, J. Geophys. Res.,97, 15,787-15801
[2] Rocken, C., T. Van Hove, J. Johnson, F. Solheim, R. Ware, M. Bevis, S. Chiswell, S. Businger, 1995 GPS/Storm-GPS Sensing of Atmospheric Water Vapour for Meteorology, J. Atmos & Ocean . Tech, 12 468-478
[3] Naito, I., Y. Hatanaka, R. Ichikawa, S. Shimada, T. Yabuki, H. Tsuji, and T. Tanaka, Global positioning System project to improve Japanese Weather, Earthquake Predictions, EOS, Transactions, Am. Geophys. Union, 79,301.308.311, June 1998
.
Wet bulb potential temperature[ºC] / 0 / 6 / 12 / 14 / 16 / 18 / 22 / 260 to 1 km / 4 / 6 / 10 / 12 / 13 / 14 / 19 / 23
1 to 2 km / 2 / 4 / 6 / 6 / 8 / 9 / 12 / 16
2 to 3 km / 1 / 2 / 4 / 5 / 6 / 7 / 11 / 14
3 to 4 km / 0.5 / 1 / 2 / 3 / 4 / 5 / 7 / 10
4 to 5 km / 0.5 / 1 / 1.5 / 2 / 3 / 3 / 5 / 8
5 to 6 km / 0.5 / 1 / 1. / 1.5 / 2.5 / 3 / 6
6 to 7 km / 0.5 / 0.5 / 0.5 / 1.5 / 2 / 5
7 to 8 km / 0.5 / 1 / 3
Total water vapour [mm] / 8 / 15 / 25 / 30 / 36 / 43 / 62 / 88
Fig.1 Distribution in the vertical of contributions to total water vapour for a saturated pseudoadiabatic atmosphere
Fig.2 Total zenith delay and associated integrated water vapour measured at Herstmonceux, south-east England, as displayed on the COST 716 web page in real time for two days in February 2002, data processed by three different processing centres and compared to HIRLAM model fields (grey line labelled NWP).
Fig.3 Near real time GPS IWV plots from COST 716 for 7 to 9 July 2002 from Herstmonceux and Camborne in southern UK and two sites in central France see Fig.4
Fig.4 Water vapour around the UK, COST 716 onlyFig.5. Water vapour contours derived from COST716
Fig. 6 Water vapour distributions superimposed by hand on weather radar observations of a severe storm progressing northwards into southern England. Grid for radar displays has spacing of about 23 km. Air to the north and west of the storm is much colder than air to the east. Relatively strong winds were feeding into the storm from the south east, with light winds at low levels to the north and west of the storm.