Driving me to distraction
I know the feeling, David! Regional text forecasts are generated by
human forecasters. The site-specific forecasts are computer-generated
with no human input i.e. they are raw computer output which, in my
opinion, is not sufficiently reliable to be offered as end-user
material. In other words they are not fit for purpose.
--
Norman Lynagh
I can only refer to my past experience with MeteoGroup but the site-specific forecasts were driven partly by human forecaster edits. That's what the lead-forecaster did. They also responded automatically to past performance through regression relationships to past observations, bias checking, and sat and radar imagery of course. Be grateful that the temperatures you see posted on weather forecast graphics are model output statistics (i.e. moderated in the above way) and not raw model output! Another issue is how best to convert hourly weather codes to all-day weather symbols, such as we see on the outlook section of TV weather forecasts and online. It is more difficult than it sounds!
I'm not trying to play-down any discrepancies and how confusing they may be to users. Surely the best thing we can do to ascertain London' sunshine for Saturday is to look at model output showing cloud cover - the explanation probably lies in the fact that cloud cover will be highly variable here on Saturday.
Julian
|