Texas Precipitable Water Vapor

From GPS

Project Outline

• Texas Precipitable Water Vapor From GPS - This Page
  • Project Outline
  • Problem Statement
  • List of Sites
  • Solution Methodology
  • Results
    • April 5, 1997 - The Dry Day
      • Contour Plot of Estimated PWV
      • Contour Plot of NCEP PWV
      • Contour Plot of PWV Differences
      • National Weather Service Reports
      • Analysis
    • April 11, 1997 - The Wet Day
      • Contour Plot of Estimated PWV
      • Contour Plot of NCEP PWV
      • Contour Plot of PWV Differences
      • PWV Time Series for West Texas
      • PWV Time Series for Central Texas
      • PWV Time Series for East Texas
      • National Weather Service Reports
      • Analysis
        
  • For Further Information

Problem Statement

If you have ever lived in Texas, you probably heard the words, "If you don't like the weather in Texas, wait 5 minutes." The weather in Texas is highly variable and unpredictable. The people of Texas experience droughts one year and unending thunderstorms the next (1996, 1997 for example). An effort needs to be made to understand the weather over Texas and to formulate weather forecasts that Texans can use to anticipate weather related problems.

In order to forecast weather, certain conditions are required, one of which is water vapor content of the atmosphere. Previously, quantifying the water vapor content of the atmosphere has required radiosondes or water vapor radiometers, which are both impractical, expensive, and fails to provide a dense distribution of data. The Global Positioning System (GPS) provides a method that solves all those problems, either from ground based or space based GPS receivers. This project demonstrates the procedure for quantifying precipitable water vapor (PWV) from ground based GPS receivers.

I processed data for April 5 and April 11, 1997, for sites in Texas, Oklahoma, and New Mexico. I picked April 5th because it was a relatively dry day, and April 11th because it was more humid (due to data limitations at the time of processing, I was unable to find a good stormy day). Because this is essentially a feasibility demonstration, surface meteorological data was not available. I obtained atmospheric data from the National Centers for Environmental Prediction (NCEP) and used it as if it were surface data recorded at the site.

List of Sites

The sites were part of the CORS network, a network of continuously operating reference sites that provides data over the web along with apriori site coordinates. Note that, while the CORS network claim continually operating sites, several of the Texas sites run by the Texas Department of Transportation only record data between 13:00 and 24:00 hours.

New Mexico

Pietown (PIE1)
White Sands (WSMN)

Oklahoma

Haskell (HKLO)
Lamont (LMNO)
Purcell (PRCO)
Vici (VCIO)

Texas

Amarillo (AML5)
Aransas Pass (ARP3)
Arlington (ARL5)
Austin (AUS5)
Beaumont (BEA5)
Corpus Christi (CORC)
El Paso (PASO)
Galveston (GAL1)
Houston (HOUS)
Lake Houston (LKHU)
Lubbock (LUBB)
McDonald Observatory (MDO1)
Odessa (ODS5)
San Antonio (ANTO)

Solution Methodology

• Data is downloaded

GPS RINEX observations from JPL
JPL FLINN analysis GPS orbits
NCEP Atmospheric Data

• Process GPS observations for 24 hours using JPL's GIPSY Software (Gregorius, 1996)
  • Estimate site coordinates
  • Zenith delay is allowed to adjust stochastically to absorb errors

• GPS observations are reprocessed in 1 hour increments
  • Site coordinates held fixed to values estimated in previous step
  • A zenith delay correction (ZDC) is estimated stochastically as a random walk every 10 minutes

• Total zenith delay (TZD) is calculated from GIPSY output and input.
  • Apriori Hydrostatic (Dry) Delay (AHD) is calculated by GIPSY from the station heights

ADD = 2.29951*e^(-0.000116*Height)

• Apriori Wet Delay (AWD) is arbitrarily set to 0.1 m.
• TZD = AHD + AWD + ZDC

• Atmospheric pressure and temperature from NCEP atmospheric model is interpolated to desired location on the surface and at the desired time
  • Data is in 1 x 1 degree grid
  • Distinct pressure levels (as opposed to height levels)
  • Every 6 hours

• Zenith Dry Delay (ZHD) is calculated from NCEP surface pressure using Elgered's relation (Elgered et al., 1991)

• Zenith Wet Delay (ZWD) is calculated

ZWD = TZD - ZHD

• Why is dry delay added in and then subtracted out? Because GIPSY's apriori model for dry delay is independent of atmospheric pressure. Errors in the dry delay model will appear in the estimated ZDC.

• The mean temperature is calculated from the surface temperature using an empirical linear relationship

• ZWD is converted to integrated water vapor (IWV)

• PWV is IWV scaled by the inverse of the density of water

WOW!!! That was dry! Show me the WATER!!!

April 5, 1997 was a dry day... Click Here to See Results

April 11, 1997 was not so dry... Click Here to See Results

For Further Information

• Water Vapor From GPS
  • Water Vapor from GPS Receivers
  • Water Vapor from Ground Based GPS Receivers
  • NOAA Environmental Research Laboratories
  • Forecast Systems Laboratory
  • UNAVCO's Real Time Water Vapor
  • Slant Path Water Vapor Measurement Using GPS

• Water Vapor
  • AGU Web Site: Water Vapor in the Climate System. A Special Report.

• GPS Related Sites
  • IGS's Home Page
  • JPL's Home Page
  • The Institute of Navigation
  • GPS Overview
  • University of Texas Department of Geography GPS Site

• Atmospheric Models and Weather Related Sites
  • An Introduction to Atmospheric and Oceanographic Datasets
  • Meteorological Data Guide for Geodesists
  • IWIN (Interactive Weather Information Network)
  • Surface Maps From Texas A&M
  • PSC Weather Center
  • Weather Information From Other Servers
  • Allentown Weather Center - FTP Sites
  • warn1.htm

• References

Acknowledgements: I'd like to thank Tatyana Pekker and Gerard Kruizinga for their assistance and experience in using the NCEP atmospheric data. I'd like to thank Katherine Quinn, a graduate student from MIT, who's web pages were instrumental in my obtaining the atmospheric data. I'd also like to acknowledge that the GMT script for creating the contour plots was obtained from UNAVCO; additional GMT help came from Philip Curell. I also appreciate the hours of work that Kevin Bowman put into getting GIPSY up and running on our workstations. And finally, I'd like to thank Dr. R. Steven Nerem for his assistance and advice regarding this project.

This page created for the Remote Sensing course at the University of Texas at Austin. Any questions or comments regarding this page should be e-mailed to Michael Gabor (mgabor@csr.utexas.edu).

Last Modified 05 May 1997