This is regarding the precip type radar and in particular I am adressing this to Retron who seems to know something about this. The precip type radar I have raintoday.co.uk is basically the same as the one the bbc sometimes uses. Retron claims that the raw data does not contain precip type information; I genuinely believe him. But that leaves some questions unanswered, I know for a fact that raintoday does not just overlay model data because:

1) The highest resolution models still only update every hour. I can believe you may interpolate to half an hour or even 15 minutes. But every 5 minutes is a stretch (yes the precip type part of the radar does change along with the intensity every 5 minutes).

2) It cannot be based on observations as the resolution is as high as the intensity and it occurs over the sea aswell.

3) I have compared it directly to the NMM and I can rule out an interpolation based on that, the precip type pixels are not identical.

4) We can rule out lower resolution models given the high res of the precip type colouring.

5) Euro4 is not high resolution enough, and although AROME might be the resolution is the same in scotland and England. This only leaves the ultra high res UKV model. But the UKV would be ludicrously expensive for a relatively cheap service of raintoday, plus there is still the interpolation issue.  

Any thoughts. Another idea I have is that they do use model data such as the NMM (or possibly the UKV) but then use a complicated formula that takes into account intensity and 'corrects' based on that. I will say the radar is very accurate in my experience but I would like to know if I am just purchasing model data that I can acsess for free.

Radar data doesn't offer precip type. Precip type is generated using model data and / or current obs. On TWO current obs are used but there have been probs this winter because of the changes to the way the MetO cache data  on their servers. 

I don't know about TWO (and Raintoday), but NW uses NMM for their precipitation type overlay - you get to see large chunks at a time change from "sleet" to "rain", for example, whereas we all know in reality that it won't change all at once over a square area!

Raintoday's radar is indeed very well-regarded, although it too is based on a model (quite possible NMM too, as it was freely available a few years ago when they implemented it). I suspect as Q says there is some clever compensation going on in their algorithm, although for obvious reasons they're very unlikely to say exactly what they do!

Yes NW makes it pretty obvious that it is a model overlay; raintoday is much cleverer if that is indeed what it does.