IntroductionWater vapor is one of the most important components of the Earth's atmosphere. It is the source of precipitation, and its latent heat is a critical ingredient in the dynamics of most major weather events. As a greenhouse gas, water vapor also plays a critical role in the global climate system: it absorbs and radiates energy from the sun and affects the formation of clouds and aerosols and the chemistry of the lower atmosphere. Despite its importance in climate and weather prediction, water vapor has been one of the most poorly measured and least understood components of the Earth's atmosphere. Researchers at FSL and elsewhere are utilizing recent technology to reverse this situation.
The ability to use the Global Positioning System (GPS) to make accurate refractivity measurements under all weather conditions has led to the development of a promising new meteorological observing system for NOAA. The first and most mature application of ground-based GPS meteorology involves the measurement of integrated (total column) precipitable water vapor (IPW) in the atmosphere. The GPS-IPW technique is more advantageous than conventional water vapor observing systems because of its low-cost, high-measurement accuracy, all weather operability, and long-term measurement stability. Further, GPS-IPW requires no external calibration, operates unattended for long periods with high reliability, and is easily maintained. Since GPS-IPW measurements are compatible with satellite data retrievals, they provide an independent method for calibrating and validating global satellite observations.
These positive attributes, however, are accompanied with one major disadvantage: GPS-IPW provides no direct information about the vertical distribution of water vapor in the atmosphere. In an attempt to mitigate this deficiency, researchers at government laboratories and universities around the world are investigating the best ways to use GPS-IPW as a "proxy quantity" for moisture profiles in weather forecasting.
In this article we discuss how IPW is now calculated from GPS signal delays and the potential use of slant-path measurements in numerical weather prediction models. Preliminary results of the effect of GPS-IPW on numerical weather prediction, the demonstration network, data and product availability, and plans for the operational network are also described.
Calculating IPW from GPS Signal Delays
GPS signals are delayed as they pass through the Earth's atmosphere (Figure 1). The signal delay caused by the presence of free electrons in the ionosphere makes the largest contribution to the total atmospheric delay. Because the ionosphere is a dispersive medium, the velocity of the GPS signals is frequency dependent and its impact can be effectively eliminated by using dual frequency receivers.
Figure 1. Signal delays caused by the atmosphere.
Below the ionosphere, in the electrically neutral portion of the atmosphere, refraction (that is, slowing and bending) of the GPS signal is caused by changes in temperature, pressure, and water vapor. Most of this delay occurs in the troposphere, which extends from about 9 km at the poles to about 13 km at the equator. The primarily tropospheric delay consists of a hydrostatic (or dry) component caused by the mass of the atmosphere and a wet component (the wet delay) caused by the dipole moment of the water vapor molecule. The contributions of the wet and dry components of the tropospheric signal delay are in the same proportion as the wet and dry components of the atmosphere.
FSL currently collects GPS observations from a demonstration network of 55 sites (Figure 2) and processes them to produce IPW measurements every 30 minutes using the scheme shown in Figure 3.
Figure 2. A map of the NOAA-FSL Global Positioning System Integrated Precipitable Water (GPS-IPW) Demonstration Network (55 sites) as of October 1999.
Figure 3. The FSL-developed data processing scheme used to produce IPW measurements.
The first step in obtaining IPW from GPS observations is to determine the zenith-scaled delay caused by the neutral atmosphere. This delay is commonly referred to as the zenith tropospheric delay (ZTD), and is calculated from carrier phase and range observations made by networks of GPS receivers. The calculation is made using GPS analysis software such as GAMIT (GPS At MIT), which in addition to the GPS observations, requires improved satellite orbits and parameters describing the orientation of the Earth in space and time. Next, the ZTD is separated into its wet and dry components using additional observations made by collocated surface meteorological sensors. The zenith- scaled hydrostatic delay (ZHD) is caused by the mass of the atmosphere directly above the site and can be estimated with great accuracy from a surface pressure measurement. The wet signal delay (ZWP) is caused by water vapor along the paths of the radio signals to all satellites in view, about 6 to 8 with the current GPS satellite constellation.
ZWP is calculated simply by subtracting the hydrostatic delay from the tropospheric delay. The resulting wet delay can be mapped into IPW with an error of about 5 degrees using a quantity that is proportional to the mean vapor pressure-weighted temperature of the atmosphere (Tm). Tm may be estimated from a climate model, the surface temperature derived from a numerical weather prediction model, or measured directly using remote sensing techniques. FSL is planning to utilize model-derived Tm estimates operationally.
Integrated precipitable water calculated from GPS signal delays is physically identical to integrated measurements or retrievals made by other upper-air observing systems including rawinsondes, ground-based microwave water vapor radiometers, or satellite microwave and infrared instruments including sounders and interferometers. Comparisons of GPS and radiosonde-derived total column water vapor have been carried out continuously since 1996 under all weather conditions at the DOE Atmospheric Radiation Measurement site near Lamont, Oklahoma. Results from 3600 comparisons (to September 1999) indicate a mean difference of 0.08 mm and a standard deviation of 2.1 mm (Figure 4).
Figure 4. Scatterplot of GPS and rawinsonde observations of integrated precipitable water vapor at the ARM CART site near Lamont, Oklahoma, between January 1996 and September 1999.
Slant-Path Signal Delay Measurements
Recent investigations by FSL director A.E. MacDonald and Yuanfu Xie (of the Forecast Research Division) of the potential use of line-of sight estimates of path-integrated water vapor (derived from slant-path GPS signal delay measurements) to retrieve the 3-D moisture field have been very interesting and potentially significant. The experiments involve assimilating simulated slant-path moisture measurements from a wide area network of closely spaced stations into the Quasi-Nonhydrostatic (QNH) model using variational techniques. In recent research, their simulations indicate that it may be possible to recover the three- dimensional structure of the moisture field from a densely spaced network of ground-based GPS receivers making a single line-of- sight, or slant path, measurement of the signal delay to all satellites in view. The configuration of the GPS satellite constellation as seen from Boulder, Colorado, between 1200 and 1300 UTC on 28 September 1999 is shown in Figure 5. A GPS satellite moves across the sky at the rate of about 30 degrees per hour. Although 10 satellites are visible above the horizon in this example, six to eight would be more typical at any one time.
Making a slant-path signal delay measurement with the same accuracy as a zenith-scaled measurement is not trivial. The sources of measurement error that are successfully managed through geodetic modeling of the zenith tropospheric signal delay will have to be dealt with in other ways. Although some of the most important information about the structure of the atmosphere can be obtained from low-angle observations, measurement errors increase significantly along with the negative impact of multipath reflections from nearby obstacles as satellites approach the horizon. One way to mitigate these problems is to utilize advanced GPS receivers and antennas that maximize the ability to track satellites under all conditions and reject multipath reflections. Unfortunately, not all problems can be eliminated through the selection of hardware, and advanced data processing techniques will be needed as well.
Research at Scripps Institution and the University of Hawaii into ways to monitor the accuracy of GPS orbit predictions suggests that these techniques can also be used to reduce systematic errors in slant-path signal delay or refractivity measurements to individual satellites.
Figure 5. Configuration of the GPS satellite constellation as seen from Boulder between 1200 and 1300 UTC 28 September 1999.
Effect of GPS-IPW Data on the Accuracy of Numerical Weather Prediction
Since 1997, parallel runs with and without GPS have been carried out using the research version of the Rapid Update Cycle (RUC-2) model to assess how GPS-IPW data affect the accuracy of numerical weather prediction. Results from the first two years using optimal interpolation techniques have been encouraging despite the fact that the observations came from only a limited number of widely spaced sites. Model runs using data acquired from more sites over a larger area through September 1999 confirm improvements in forecast accuracy, especially under conditions of active weather. Therefore, NOAA meteorologists expect significant improvements in short-term forecasts of clouds, precipitation, and severe weather when high-resolution numerical weather prediction models routinely use data from a nationwide network of GPS-IPW systems in conjunction with data from other observing systems and advanced data assimilation techniques.
The decision to implement ground-based GPS Meteorology (GPS-Met) as a next-generation upper-air observing system will be supported in part by promising assessments such as this one. In anticipation of a favorable decision, network design and implementation options for a national network of ground-based GPS- IPW systems are being evaluated at FSL.
GPS-IPW Demonstration Network
The rapid development of the GPS-IPW Demonstration Network for meteorological remote sensing has been made possible by a fortuitous synergy with the positioning and navigational applications of GPS by the U.S. Coast Guard and U.S. Department of Transportation. As of October 1999, the data acquisition component of the demonstration network consisted of 55 GPS-IPW systems operating in the continental United States and Alaska.
Thirty-four systems are currently installed at NOAA Profiler Network (NPN) sites, seven at sites belonging to other NOAA organizations or institutions affiliated with NOAA, 11 belong to the U.S. Coast Guard Maritime Differential GPS (DGPS) system, and three are at the Department of Transportation Nationwide Differential GPS facilities. Typical sites from each organization are illustrated in Figures 6 – 9.
Figure 6. GPW-IPW installation at the NOAA Profiler Network site at Platteville, Colorado.
Figure 7. GPS-IPW installation at the Scripps Institution of Oceanography at La Jolla, California.
Figure 8. GPS-IPW installation at the U.S. Coast Guard Differential GPS site at Cape Canaveral Air Force Station.
Figure 9. GPS-IPW installation at the DOT National Differential GPS site at Whitney, Nebraska.
In addition to supporting the assessment of GPS as a possible next-generation upper-air observing system, the GPS-IPW Demonstration Network is designed to help NOAA accomplish the following tasks:
All ground-based observing systems in the GPS-IPW Demonstration Network consist of dual-frequency GPS receivers and antennas, and collocated surface meteorological sensors.
These systems are located at sites where shelter, power, and communications are available to operate and collect data from the instruments, and transmit these data in real or near real time to one of two locations. The generalized flow of data and products from the network is illustrated in Figure 10.
Figure 10. Flow of data and products from the GPS-IPW Demonstration Network.
Data and Product Availability
GPS and surface meteorological observations from the GPS-IPW Demonstration Network sites are available to the general public in near real time through the NOAA National Geodetic Survey. Information and raw data may be acquired via the Web.
Processed data, including GPS signal delays and integrated precipitable water vapor, are available shortly after improved NAVSTAR GPS satellite orbits and Earth Orientation Parameters are available from one of the International GPS Service (IGS) tracking stations. This usually occurs within 24 hours of the close of the day, but efforts to accelerate the process and make improved orbits available within 1–3 hours are well underway. IPW and other products may be acquired from the FSL Demonstration Division, GPS-Met Observing Systems Branch.
Plans for Network Expansion and Improvement
Our primary goals in 2000 are to continue to expand the demonstration network, demonstrate distributed data processing using low-cost PCs instead of high-end workstations, and implement real-time data processing.
Expansion of the Demonstration Network – Now that all NOAA Profiler Network sites have been equipped with GPS-IPW systems, expansion of the network will mostly proceed by installing GPS Surface Observing System (GSOS) packages at U.S. Coast Guard Maritime DGPS and Department of Transportation NDGPS sites. Depending on the availability of funds and the status of interagency agreements under review, during the next year we hope to install 21 new systems at DGPS sites (mostly in the Mississippi Valley and Great Lakes regions), 11 NDGPS sites, and one at the Department of Energy ARM facility at Point Barrow, Alaska (Figure 11).
Figure 11. Planned expansion of the Demonstration Network to about 92 sites during 2000.
Implementation of Real-time Data Processing – We define real- time data processing as acquiring and processing GPS and ancillary observations to yield signal delay or IPW calculations within a single numerical weather prediction assimilation cycle. In the case of the Rapid Update Cycle, running operationally at the National Centers for Environmental Prediction, this is approximately 75 minutes.
Real-time data processing techniques are being tested and evaluated in a collaborative effort involving FSL, the Scripps Permanent Orbit Array Center, and the University of Hawaii at Manoa. Techniques involve acquiring data from a subset of the IGS global tracking network and using these observations to produce an improved retrospective orbit with only about 2-hour latency. An orbit prediction that covers the data gap is also made, and it is this short-term prediction that is used to calculate IPW. In theory, the error in a prediction that spans only 2 or 3 hours will be proportionally less than an error made over an interval of 36 – 48 hours.
Real-time quality control techniques are also under evaluation. The most promising involve continuous monitoring of the relative positions of a number of sites, and using these data to infer changes in orbit accuracy for specific satellites in the constellation. When a problem is encountered, the satellites are removed temporarily from the ephemerides until an updated orbit can be produced.
Distributed Data Processing – Recent advances in low-cost PC processor speed and memory will be utilized to perform data processing in a fully distributed environment. During this year we have demonstrated the ability to partition a large network into smaller subnetworks, process each independently in substantially less time with no significant loss of accuracy and precision.
Operational GPS-IPW Network Implementation Strategy – The expansion of the GPS-IPW Demonstration Network to about 200 sites, and the transition from retrospective to real-time data processing will enable us to assess the impact of these data on weather forecast accuracy. Based on the results of these studies, a decision to implement ground-based GPS-IPW as a next-generation upper-air observing system for NOAA is expected. The following plan has been developed to expand the demonstration network to an operational network of about 1000 sites with an average station spacing of somewhat less than 100 km (Figure 12).
Figure 12. Expected configuration of the NOAA/FSL GPS-IPW Demonstration Network by 2005. Sites in Hawaii and the Caribbean Sea are not shown.
Figure 13. An Automated Surface Observing System (ASOS) installation at Cape Hatteras, North Carolina.
Figure 14. GPS receiver placement on top of FSL's new office building, the David Skaggs Research Center.
Figure 15. A map of the AWIPS offices.
[Editor's Note: More information on the topics covered here is available by contacting Seth Gutman, who can provide copies of published articles which include a list of references.]
(Seth Gutman is Chief of the GPS-Met Observing Systems Branch, part of the Demonstration Division headed by Margot Ackley. He can be reached by e-mail. Kirk Holub is a Systems Analyst in the same division.)