Extended Abstract (88K)7.4
Observations and Nowcasting in SCRIBE
Claude Landry, MSC, Dorval, QC, Canada; and M. Ouellet, R. Parent, J. F. Deschênes, and R. Verret
The SCRIBE Weather Forecast Product Expert System is capable of generating automatically or interactively any type of weather products for a region or a specific locality. However, the data that feeds the system is produced only from numerical weather prediction (NWP) models or statistical models. Therefore, the SCRIBE product generator is totally unaware of recent weather events, and this limitation is particularly acute for weather products that are generated long after the model run. This "blind" effect would generally result in forecasts that are not as up-to-date in their first 18 to 24 hours, were it not for the adjustments made by the operational forecasters. One of the key impetus for the work presented in this paper was to minimize these necessary manual adjustments.
The Observation and Nowcasting SCRIBE sub-system actually under development will merge the SCRIBE forecast events with the latest local observations and very short range forecast data. Different algorithms are used to generate nowcast data: multiple discriminant analysis is used to generate probabilistic forecasts over the next 6- to 12-hours at a one hour time resolution of several weather elements reported in the hourly observations; vector motions are calculated and used to displace into the future radar echoes at a one hour time resolution over the next six hours based North-American radar composite imagery and NWP 700 hPa forecast winds; vector motions are also calculated and used to project into the future areas with lightning strikes at a one hour time resolution. A first prototype will be presented showing how surface observations, radar data and lightning data are processed to update the SCRIBE weather element concepts and how the different nowcasting algorithms are used to project observation data in the short term. Although it is too early to make objective verification, it is expected that this approach will allow to reduce by up to 50 % the time spent by forecasters to quality control the initial parts of the SCRIBE forecasts. Furthermore, the added value to the SCRIBE forecasts in the initial period is expected to be of the order of 5-10%.
Session 7, European and Other International Applications (RO0M 613/614)
Tuesday, 13 January 2004, 11:00 AM-12:00 PM, Room 613/614
Previous paper