C. Landry, R. Parent, J.-F. Deschênes, A. Giguère, G. Hardy and R. Verret
Division du développement CMC
The SCRIBE Weather Forecast Product Expert System is now capable of
ingesting the latest observations and nowcasting model data to update in
real time the Scribe weather elements. This sub-system has been developed
to minimize the necessary manual adjustments done by the forecaster to
merge the current weather conditions with the forecast.
Version 1.0 of the Scribe Nowcasting is currently using surface
observations, North American radar mosaic data and lighting data from the
Lighting Detection Network. These observations are used to feed short term
forecast models. A statistical model called "PubTools" uses the surface
observations to forecast the probabilities of occurrences of weather
elements. The observed radar reflectivities are projected during the next
6 hours with a vector motion calculated from observed imageries 20 minutes
apart. Finally, an algorithm has been developed at CMC to predict the
probabilities of thunderstorm occurrences based on the forecast position
of the lightning clusters. All these observed and forecast data are
processed by a rules base system to determine the best sequence of weather
elements representing the current observation and short term tendencies.
The first 6 to 9 hours of the Scribe weather elements will thus be
influenced by the nowcasting data. Depending on the weather conditions and
on how well the model handles these conditions, significant changes can be
done to the regular Scribe weather elements. To assess whether these
changes contribute to improve the first hours of the forecast or not,
objective verifications was perforemed. For some weather elements
preliminary results show that in the first 6 hours of the forecast the
Probablily of Detection has increased and the related False Alarm Ratio
has decreased. Other verification scores also indicate an improvement of
the short term forecast performance.