Garbage In Garbage OutAdded : Monday at 13:45 When we build the numerical models here at Metcheck, one of the most important layers which we add is what we call the OBS layer. Basically, we produce the forecast each hour which is fantastic (in our own words), but a forecast is no good if it's wrong, so every 30 minutes we grab sky and temperature observations from around the world and then compare them directly to what the model forecast.
Not just that, but we compare several models with the OBS data to see which is the most accurate and then that has a higher priority in the next model run.
If we step back one stage, then the observational data which goes into the numerical models is equally important. This comes from various sources like weather observations from manned stations, automated ones, satellite, aircraft, weather balloons and also RADAR data.
All this gets crunched up into BUFR data and then goes into the model starting conditions. Great eh? But take a look at the radar for today :-
You can see some very light and patchy showers around, but no doubt there are some areas seeing light drizzle which the radar misses. Some models compare it to the infra red satellites which have precipitation probability built in and these images look like this :-
But, neither is perfect. Even so, these images are then broken up into gridded chunks which go into the starting data and as a result, the model comes back with something like this :-
It's a sort of "blurry" image of where the radars/satellites and observations showed rain and then the model builds its own view of it.
The problem is, radars around the world aren't perfect. They show up anaprop sometimes which is usually cleaned up, but other times they miss light precipitation, especially where the verification process compares it to their own models which suggest no rain and therefore the radar removes it.
A great example of the limitation of this was yesterday across the UK where light drizzle on and off was missed by the radars which in turn fed back into the models which in turn affected the low cloud cover across parts today.
Numerical weather models are great, but they aren't perfect. There will always be some bits of garbage which get thrown in at the start, it's for the meteorologist to figure out when such situations might occur and then view the subsequent model output with a slightly different view, as well as wearing gloves obviously.
METEOROLOGIST : MARSH |