Challenges of Nowcasting


George A. Isaac1, Monika Bailey1, Faisal Boudala1, Stewart Cober1,
Robert Crawford1, Ivan Heckman1, Laura Huang1, Paul Joe1,
Jocelyn Mailhot2, Jason Milbrandt2, and Janti Reid1

1Cloud Physics and Severe Weather Research Section, Environment Canada
2Numerical Weather Prediction Research Section, Environment Canada

Certain challenges related to Nowcasting have been identified through participation in two projects: the Canadian Airport Nowcasting (CAN-Now) project and the Science of Nowcasting Olympic Weather for Vancouver 2010 (SNOW-V10). These projects were developed for the aviation and sport communities who had very specific requirements. Nowcasting techniques/forecasts must be tuned to their identified users and this talk will emphasize the challenges discovered during CAN-Now and SNOW-V10. These projects have shown the need for high time resolution (every minute) observations and similar temporal scales for NWP data. An NWP model used for Nowcasting needs to be initialized using a high resolution analysis. Selection of model points to represent stations has been a problem for both coarse resolution model (e.g. 15km) and high resolution models (e.g. 1km). NWP model spin up issues can create difficulties. Getting observations that represent specific sites, especially in mountainous terrain is also a problem. Prediction of non standard variables like ceiling, visibility, runway visible range, precipitation type, wind gusts, requires the development and testing of new algorithms. New techniques were required which blend observations and models into nowcast systems and examples will be given in the talk. Verification of nowcasts based on NWP model data alone and the new nowcast systems have yielded some interesting results. For example, it is difficult to forecast relative humidity and this leads to problems related to forecasting visibility. Wind direction forecasts do not verify well. In general, nowcast techniques do show skill when compared to persistence. However, average statistical scores do not always give a good idea of the skill in forecasting high impact events and a further set of performance metrics should be developed and perhaps tuned to specific users.