Close Window

Virginia Polytechnic Institute and State University: "The use of GPS precipitable water measurements to supplement WSR-88D rainfall estimates for flash flood forecasting in southwest Virginia"

Final Report


The overall objective of this study is to find relationships between the GPS Integrated Precipitable Water (IPW) data at Blacksburg VA (BLKW) and KFCX WSR-88D radar data in order to better predict significant convective rain events capable of producing flash flooding. Specific objectives include the following:

  1. Compare the maximum GPS IPW with maximum “cell-based” VIL within 100 km of Blacksburg.

  2. Establish a correlation between the GPS IPW trends and radar-observed trends in convective activity within 100 km of Blacksburg. Convective trends will consider maximum values of reflectivity, rainfall rate and VIL.

  3. Establish a correlation between GPS IPW trends and gage-measured rainfall within 100 km of Blacksburg.

  4. Make direct comparisons between the GPS IPW, GOES sounder PW, and radiosonde-derived PW.

  5. Subjectively compare GPS IPW with the RUC (Rapid Update Cycle) model short term hourly forecasts.

  6. Through realtime operational evaluation, determine if forecasters perceive a benefit to having access to these data as it applies to the short term heavy rain and flash flood forecast problem.


Direct comparisons between the GPS IPW and radiosonde-measured PW indicated a very close agreement, and while comparison with GOES-PW showed the GOES data to have a small wet-bias, overall trends were very similar. While GOES PW data is not available during periods of cloud cover, and the grid point closest to the Blacksburg lat/lon can change every hour, GPS IPW is available every 30 minutes and in near-realtime (30 minute latency) regardless of cloud cover. Comparisons with radar-observed trends showed close correlation, but with peaks in GPS IPW offering little if any lead time. This corroborated subjective evaluation by the forecasters who observed little difference in peaks in IPW trends from peaks in convective activity. Therefore, the focus shifted to correlating decipherable trends in the IPW time series with automated gage measurements of significant rainfall rates in an attempt to identify signals that would provide some lead time. It was also found that wintertime IPW almost replicates the same pattern at different nearby receiver locations as synoptic scale storm systems dominate the signals as they propagate across portions of the network. In contrast, smaller scale fluctuations in the summertime IPW trend showed little correlation between the two sites, suggesting these subtle variations are likely due to mesoscale or even storm scale phenomena, and are likely important to the behavior of convection.

A more detailed trend analysis and comparison with climatological thresholds (the climatology was based on historical radiosonde data) revealed the potential to achieve 1-2 hours or more of lead time before heavy rain was observed at gages within 100 km. This lead time was based on IPW values exceeding and remaining above a particular threshold for a certain period of time. Figure 1 illustrates one example where the IPW crosses the 75th percentile based on that month’s climatology, and remains above for at least four hours. A running average was also calculated to smooth out extremes (or in case only data point fell below the threshold). This time constraint results in the “lead time” being less than the “alert time” (when the threshold is first met) but minimizes the false alarm rate.

The results are promising, and may warrant experimentation of an automated algorithm and more thorough analysis of verification rather than just for a select list of events. Certainly the high temporal resolution data shows promise as an important tool for increasing operational forecasters situational awareness of impending heavy convective rain events.

Figure 1. Plot showing GPS IPW and gage-observed rainfall for June 4-6, 2001. Lower red line is 75th percentile above the
mean (or the mean plus 25%) for the month of June, and the upper red line is the mean plus or two standard deviations
(or the 95th percentile).


There has been a strong interaction between the NWS officials and Virginia Tech faculty and students. The NWS officers are frequent speakers for seminars organized by the various departments at Virginia Tech. Students from Virginia Tech have held part-time positions at the NWS office. There have been numerous tours to the NWS facilities by student groups and classes. There has been a good amount of data transfer from the NWS to the various departments. The NWS personnel are considered an excellent resource for their technical knowledge and are constantly engaged in technical discussions related to both problem formulation and its solution.


Loganathan, G.V., Keighton, S., Gillen, M., Eisenberger, T., Gorugantula, S., Kibler, D.F., and Gutman, S., , “GPS Enhanced Radar Precipitation Estimates for Real-time Applications”, World Water and Environmental Resources Congress: Bridging the Gap, ASCE, Edited by D. Phelps and G. Sehlke, Orlando, FL, May 20-24, 2001.

Keighton, S., Gillen, M., Loganathan, G.V., Gorugantula, S., and Eisenberger, T., “P5.14 The use of GPS integrated precipitable water measurements to supplement WSR-88D parameters in determining the potential for flash flood producing rainfall”, Ninth Conference on Mesoscale Processes, AMS, Ft. Lauderdale, FL, July 30 – August 2, 2001.

Loganathan, G.V., Eisenberger, T., Gorugantula, S., Kibler, D.F., Keighton, S., and Gillen, M., “Assessing Flash Flood Potential with the Aid of GPS Integrated Water Vapor Estimates”, Virginia Water Research Symposium 2001, Charlottesville, Virginia, November 14-16, 2001.

Loganathan, G.V., Eisenberger, T., Gorugantula, S., Kibler, D.F., Keighton, S., and Gillen, M., “Use of GPS Integrated Water Vapor Estimates in Short-term Prediction of Rainfall”, Proceedings of the ASCE Environmental and Water Resources Institute (EWRI) Annual Conference, May 19-22, 2002, Roanoke, Virginia.

S. Gorugantula, " The Use of GPS Integrated Water Vapor Estimates in Short-term Rainfall Prediction", MS Thesis in Progress, Department of Civil and Environmental Engineering, Virginia Tech, Blacksburg, VA, 2002.

An informal presentation on data availability and preliminary findings was presented by S. Keighton at the COMET Symposium on Heavy Precipitation and Flash Flooding, October 2001.

The S. Gorugantula M.S. thesis was summarized and presented to forecasters at WFO Blacksburg during a seasonal Heavy Precipitation and Flash Flooding Workshop, May 9, 2002.


4.1 (Academic Partner) The best single most benefit is the access to the sophisticated instrumentation and the countless hours of fruitful discussion with the NWS officers in sharing and implementing the research ideas. The Virginia Tech researchers will write proposals using the knowledge that they have gained from this research with the NWS officers.

4.2 (Forecaster Partner) Most significantly, forecasters gained an understanding of a new observing system and began incorporating the data into their mesoscale forecast process. The results of the research were presented to operational staff at an on-station workshop. While no specific evidence has been gathered to prove the research has helped improve forecasts (there have been very limited events since the research has been completed), forecasters have been accessing the data plots available from the internet, and comparing the trends with the monthly thresholds, and likely increasing their situational awareness which would typically lead to faster warning decisions. More efficient local displays need to be developed to facilitate operational use of the research results, and this will be possible as soon as local data can be made available through AWIPS. The added experience of producing a paper and making a presentation at an AMS Conference was another important benefit, as was continuing to foster our collaborative relationship with the faculty and students at VA Tech.

One problem encountered had to do with efficiently analyzing all the radar data for the series of events we had identified. The students did not have enough time to do a thorough job with this, which was one of the reasons the focus was placed on comparisons with rain gage data; it was easy to place archived digital data from the automated gages into spreadsheets for analysis. As a result, the objectives related to comparisons with radar data were not completely met, as only limited data sets were used. We also discovered that PW date from RUC hourly forecasts was not an output field and so this comparison could not be used. We could have used 3-hr forecasts from archived Eta model grids but we ran out of time to collect and analyze these data. We were most interested in the accuracy of very short range hourly forecasts from the RUC.