Forum 2/2003

In This Issue

Introductory Page

NOAA SOSTM

Mesoscale NWP

Managing Metadata

Object Data System

GPS Forecasting

Data Distribution

Weather Products

Briefs

Recent Publications

Contact the Editor
Nita Fullerton


Design/Layout:
Will von Dauster
John Osborn


Best Viewed With
Internet Explorer

2/2003 Forum - GPS/Forecasting By Seth Gutman, Rudy Pursaud, and Sher Wagoner

Introduction

Since 1994, FSL has been evaluating the scientific and engineering bases of ground-based GPS meteorology (GPS-Met) and assessing its utility for operational weather forecasting, climate monitoring, satellite calibration and validation, and improved differential GPS (DGPS) positioning and navigation at continuously operating reference stations (CORS), as shown in Figure 1.

GPS Concept

Figure 1. Differential GPS positioning concept in which errors are calculated at fixed CORS in the geographic vicinity of a remote GPS site and corrections are broadcast or made available, for example, through the Internet.

(Courtesy USCG Navigation Center, 1999).

The requirement for high accuracy GPS-Met retrievals with lower latency is driven mainly by the trend toward shorter forecast cycles and higher spatial resolution in mesoscale numerical weather prediction (NWP) models, and their use by weather forecasters in subjective forecasting and/or model verification.

GPS and ancillary surface meteorological observations, along with improved satellite orbits, must be available on demand. Data processing hardware and techniques must provide GPS-Met retrievals quickly enough to be assimilated into the current model cycle. Model data assimilation techniques must minimize the errors in estimating the initial state of a numerical forecast that come from spatial and temporal aliasing when interpolating discrete observations into an "analysis increment" field. While more GPS-Met retrievals can minimize horizontal aliasing, they can do little to minimize vertical aliasing that comes from assimilating any vertically integrated quantity — such as satellite radiances, zenith tropospheric signal delays, or GPS integrated precipitable water (GPS-IPW) retrievals — into an NWP model. This is mainly because the forecast background error at a discrete vertical level must be estimated from the difference between observed and forecast integrated quantities. For now, we must rely on improved data assimilation techniques coupled with the more efficient use of complementary observing systems to improve the three-dimensional description of moisture in the atmosphere.

FSL has conducted data denial experiments since 1998 to determine the statistical impact that GPS-IPW retrievals have on 3-hour moisture and precipitation forecasts in the central United States. Results from experiments over the last five years indicate more or less continuous improvement in forecast skill as the GPS-Met network expands. Improvements are observed in relative humidity forecast accuracy at all levels below 500 hPa, and all precipitation levels above "trace." The impact steadily decreases with the length of the forecast; it is usually substantial from 0 – 2 hours, small but consistent at 3 hours, and negligible from 0 – 12 hours. The largest impacts usually occur during active weather, and usually manifest themselves as changes in the locations of boundaries such as fronts and dry lines.

GPS Background

The U.S. Department of Defense developed GPS as a dual-usemilitary and civiliansatellite-based position, navigation, and time transfer system. It provides all-weather, passive, three-dimensional position, velocity, and time information anywhere on Earth. GPS currently provides two levels of service: a restricted precise positioning service (PPS) with a horizontal accuracy of less than 20 meters, and an unrestricted standard positioning service (SPS) with a horizontal accuracy of about 30 meters.

GPS is used by federal government agencies as an integral component of their activities. In addition to all branches of the U.S. military, other agencies include the Department of Transportation (DOT), National Oceanic and Atmospheric Administration (NOAA), U.S. Geological Survey (USGS), U.S. Forest Service (USFS), and the National Aeronautics and Space Administration (NASA). Specific applications include:

  • One of DOT's "10 Most Wanted" initiatives for national transportation safety, the Positive Train Control concept involving the application of digital data communications, automatic positioning systems wayside interface units, and control technologies to manage and control railroad operations.

  • National Intelligent Transportation Systems (ITS) applications, such as navigation and route guidance; automatic vehicle location, public safety services, "Mayday" relief, roadway maintenance, public transit, general railroad operations, land surveying, and automated highway systems.

  • Navigation aids for aviation in all phases of flight.

  • Maritime safety along the coasts of the continental United States, the Great Lakes, Puerto Rico, portions of Alaska and Hawaii, and a greater part of the Mississippi River Basin.

Also, many state and local government agencies across America are now using GPS to improve statewide transportation facilities, and support local and regional planning, surveying and mapping, design, construction, intelligent transportation systems, and emergency (911) response applications. By all accounts, the improvements in engineering accuracy and productivity in the public and private sectors resulting from the use of GPS have made it one of the best investments in the national infrastructure.

Improving GPS Accuracy

In general, these users routinely require levels of GPS position accuracy and signal availability that are beyond the current capabilities of both the PPS and SPS. This is especially true when the activities relate to safety of life or property, as is frequently the case in most federal and a growing number of state services.

Position accuracy limitations are caused by uncertainties in the parameters used to calculate position using the technique called trilateration. These parameters include the satellite orbits, satellite and receive clocks, receiver noise, and multipath interference. Errors are caused by the slowing and bending of the GPS radio signals by the upper and lower atmosphere as they travel from the vacuum of space to a receiver at or near the surface of the Earth. However, the major source of GPS position error now comes from the ionosphere and troposphere, since the discontinuance in 2000 of "Selective Availability," the procedure of denying most nonmilitary users full accuracy of the GPS SPS by "dithering" the satellite clock and degrading the broadcast ephemeris.

Mitigating Techniques

Techniques have been developed to mitigate most of these problems, improve position and time accuracy as well as signal availability especially in urban environments. The most effective techniques are called augmentation and error modeling, or stochastic estimation.

Augmentation – This technique involves adding information to the signals received from the GPS satellites to increase integrity, accuracy, and availability. This is usually accomplished through the DGPS process, in which errors are calculated at fixed CORS in the geographic vicinity of a remote GPS site, and corrections are broadcast or made available through some other medium such as the Internet (Figure 1).

Error Modeling – Predictions of the impact of the atmosphere (ionosphere and/or troposphere) on the GPS signal can either use independent observations and a physical model, or estimates of the expected (average) conditions at a place and time that are usually derived from diurnal behavior or climate data.

Stochastic Estimation – This is a process of solving for the tropospheric delay as a free parameter so it can be eliminated to improve the accuracy of the GPS solution. These techniques form the basis of GPS-Met. Signal delays, caused mostly by moisture variability in the troposphere, are modeled as a nuisance parameter and eliminated in high accuracy (geodetic) positioning. Delay estimates provide extremely accurate measurements of wet refractivity that are transformed into integrated (total column) water vapor in near real time under all weather conditions.

Continuously Operating Reference Stations

Civilian GPS users routinely achieve centimeter-level or better accuracy using carrier phase dual-frequency receivers (to compensate for the ionosphere), differential surveying techniques, and geodetic processing software that permits stochastic estimation of ionospheric and tropospheric delays. The GPS CORS make this possible by providing information to help reduce inherent ambiguities.

Global CORS networks have been established by organizations such as the International GPS Service (http://igscb.jpl.nasa.gov/) to provide high accuracy GPS orbits, data tracking products, and other products for a wide range of international scientific and engineering applications such as geodetic plate motion studies, earthquake deformation, satellite tracking, and GPS meteorology.

A National CORS System is being spearheaded by NOAAs National Geodetic Survey (NGS) in cooperation with government, academic, commercial, and private organizations to collectively sponsor and operate the new system. The strategy is to expand the system so that all points in the conterminous United States will be located within 200 kilometers of an operational site at all times, an increase from the current 87% to 100%.

Local CORS networks are being aggressively deployed by the state DOTs in Alaska, Colorado, Florida, Ohio, Michigan, North Carolina, and Texas (to name a few). This will facilitate more specific DGPS applications, including:

  • Creation of geographic databases for use in Emergency 911 systems.
  • Highway inventories such as milepost markers, right-of-ways, guardrails, and bridges.
  • Emergency response services for police, fire, and rescue.
  • Automatic vehicle location for public transit and other fleets.
  • Snow plow guidance for low-visibility situations.
  • Inventory of railroad crossings and road centerlines.
  • Land-use planning.
  • Tracking hazardous materials from origin to destination.
  • Mapping pavement-condition data, safety data, accident data, and traffic data.

Estimating Signal Delays and IPW

It is possible to use the data from federal and state CORS sites to estimate the tropospheric signal delay at each site with extremely high accuracy because of stringent instrument requirements for high accuracy GPS positioning.

The tropospheric signal delay is caused by the refractivity of the nondispersive or electrically neutral atmosphere, and is associated with temperature, pressure, and water vapor in the lower 9 – 16 km. If the surface pressure is known at the elevation of the GPS antenna with reasonable accuracy (approximately 0.5 hPa), then the total wet and dry refractivity directly above the site can be objectively separated with little error. Mapping the resulting wet signal delay into integrated or total column precipitable water vapor (IPW) is accomplished in a straightforward manner if the mean vapor weighted temperature of the atmosphere is known. The latter is usually estimated using the surface temperature and a simple linear regression to worldwide radiosonde observations. However, it has recently been shown that a somewhat better estimate of temperature of the atmosphere can be derived from a numerical weather model over the conterminous United States. The most accurate IPW retrievals are always made when the GPS antenna and surface meteorological sensors, especially the pressure sensor, are in close proximity. FSL is installing surface meteorological sensors at all DOT Nationwide Differential GPS (NDGPS) sites. This collaborative effort evaluates the impact of NDGPS observations on weather forecast accuracy, especially as it impacts multimodal transportation safety. NOAA surface sensors have approximately the same accuracy as instruments installed at typical Road Weather Information System (RWIS) or mesonet sites established by most state DOTs to monitor and transmit weather surface conditions in near real time. In the following example, the GPS antenna and surface meteorological instruments are separated by about 10 meters horizontally and 1.7 meters vertically.

IHOP Experiments

During the IHOP experiment in spring 2002, comparisons were made of GPS and radiosonde-measured water vapor (Figure 2). Radiosondes were launched at theDepartment of Energy Atmospheric Radiation Measurement (ARM) facility located about 9 kilometers south of the GPS-Met instrument at the Lamont, Oklahoma, NOAA Profiler Network site, LMNO2. During this IHOP experiment, 222 independent GPS and radiosonde water vapor comparisons were made under all weather conditions. The mean difference between the two observing systems is 0.2 millimeters of IPW, the standard deviation is 1.7 millimeters, and the correlation coefficient is better than 0.98. These results are typical of GPS-IPW accuracy at the CORS sites where GPS and accurate surface meteorological sensors are collocated.

Time Series - GPS and Radiosonde - IHOP

Figure 2. Time series plot of GPS and radiosonde water vapor measurements during the IHOP experiment, from 13 May – 25 June 2002.

It is always desirable to have surface meteorological sensors installed at CORS sites, but it is not always practical to do so from an engineering or economic standpoint. With this in mind, the decision must be made as to how far apart to place these observations and still yield useful results for weather forecasting. From an objective forecasting perspective, they should be close enough so that the IPW retrieval error is smaller than the numerical weather prediction analysis error without the GPS. FSL determined the latter to be 3 – 5 millimeters IPW, depending on location.

A series of experiments were performed to determine the dependence of atmospheric pressure prediction accuracy on vertical and horizontal distance, geographic region, and time of year. Pressure data from operational National Weather Service Automated Surface Observing System (ASOS) barometers were compared with data derived from high quality digital barometers at the CORS sites. Comparisons were made at 17 sites during the winter and 13 sites during the spring of 2001. The sites were separated by 3 – 53 kilometers horizontally, and 2 – 200 meters vertically.

We verified that the vertical pressure gradient dominates the equation since it exceeds the horizontal gradient by approximately five orders of magnitude. Even under nonhydrostatic conditions, when the horizontal pressure gradient is significant and local wind flow is high, the impact, while important, tends to be spatially localized and relatively short-lived.

The appropriate way to extrapolate pressure, we concluded, is to continue a nearby observation vertically from a quasi-equapotential surface (for example altimeter setting) to the actual elevation of the GPS antenna, rather than estimate it from a digital terrain model. The primary reason for this is that digital elevation models (DEMs) have finite resolution and only reflect the average terrain in a grid cell. Under conditions of high vertical relief, DEMs can carry substantial errors that will introduce significant errors into the estimation of the zenith hydrostatic delay.

Figure 3 shows a comparison of GPS and radiosonde-measured water vapor at the Gaylord, Michigan, CORS site belonging to the Michigan DOT (NOR2). The distance between the GPS antenna and the ASOS located at the Gaylord, Otsego County Airport (KGLR) is 11.8 meters vertically and 3050.8 meters horizontally.

Time Series - GPS and Radiosonde - Gaylord, Michigan

Figure 3. Time series plot of GPS and radiosonde water vapor measurements at Gaylord, Michigan, between 15 August – 1 October 2002. The pressure sensor and GPS antenna are separated by about 3 kilometers horizontally and 12 meters vertically. The mean difference, standard deviation, and correlation coefficient are 0.2 mm, 1.59 mm, and 0.984, respectively.

Results

One of the most significant discoveries of the ongoing assessment of how GPS affects weather forecast accuracy is that the magnitude of its impact increases commensurate with expansion of the GPS network. The significant growth achieved in 2002 is the result of the near real-time availability of CORS data provided by state DOTs, and the ability to use existing surface meteorological sensors to parse the GPS-derived signal delays into their wet and dry components without having to deploy a dedicated sensor package at each site.

The configuration of the GPS-Met network as of October 2002 (Figure 4) consisted of two types of GPS-Met sites. The "backbone sites" of collocated surface meteorological sensors that are operated and maintained by the U.S. DOT and NOAA, and the "infill sites" which are operated and maintained by other organizations including state and local government agencies, universities, and the private sector. Infill sites may or may not have collocated surface meteorological sensors, but the major distinction is that the infill sites carry no obligation concerning maintenance or data availability.

NOAA GPS-Met Demonstration Network

Figure 4. Configuration of the NOAA GPS-Met Demonstration Network in October 2002. Backbone sites are represented by triangles, and infill sites are denoted with circles.

Summary

These evaluations of the utility of ground-based GPS meteorology (GPS-Met) for operational weather forecasting, climate monitoring, satellite calibration and validation have been promising. As the number of sites in the GPS networks increases, primarily through the addition of CORS sites belonging to state DOTs, we expect to see a continuing trend toward increased positive results, especially regarding precipitation forecasts (Figure 5). This hypothesis will be tested in 2003, since it will be the first winter season with substantial coverage in the Great Lakes region.

Comparison of 24-hr precip. forecast verification statistics

Figure 5. A comparison of 24-hour precipitation forecast verification statistics for 2000 and 2001. In each bar chart, red represents the improvement in forecast skill, blue is worse skill, and gray is no change. The metrics are equitable threat score (EQTS), probability of detection (POD), and bias. In general, improvement is observed at all levels of precipitation above 0.01 inch in 2000, and above 0.1 inch in 2001.


Note: A complete list of references and more information on this and related topics are available at the main FSL Website http://www.fsl.noaa.gov, by clicking on "Publications" and "Research Articles."

Acknowledgements: The authors wish to acknowledge the contributions of individuals from numerous state departments of transportation who not only assisted us in retrieving their data in near real-time, but also provided technical and programmatic assistance in all phases of this effort. In particular, we wish to thank Bob Ramsey of the Alaska Department of Transportation, Frank Kochevar of the Mesa County Colorado Department of Highways, Beverly Sutphin and Scott Harris of the Florida Department of Transportation, Dave Albrecht of the Ohio Department of Transportation, Andrew Semenchuk of the Michigan Department of Transportation, Gary Thompson of the North Carolina Geodetic Survey, and Matt Bryant of the Texas Department of Transportation.

(Seth Gutman is Chief of the GPS-Met Observing Systems Branch in the Demonstration Division, headed by Margot Ackley. He can be reached at Seth.I.Gutman@noaa.gov or 303-497-7031. Rudy Pursaud is associated with the U.S. DOT Federal Highway Administration in Washington, D.C. Sher Wagoner is a Senior Systems Analyst at FSL, and is also affiliated with the Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO.)


FSL Staff FSL in Review FSL Forum (Other Editions) Publications


ln/͋[|!8[:#R&M5C <0=/Vp[{0ǴhHz|&a{gW6U`,