search

informatics-lab / precip_rediagnosis

Project to use ML to re-diagnose precipitation fields from ensemble model fields
0 stars 0 forks source link

Incorporate/Implement code for deciding whether dynamic or convective rain #21

Open stevehadd opened 2 years ago

stevehadd commented 2 years ago

Nigel Roberts has some code or an algrithm for deciding whether the weather (and specifically therain) is convective or dynamic. We should include this as a feature in our dataset.

stevehadd commented 2 years ago

Link to releavnt existing code: https://github.com/metoppv/improver/blob/master/improver/utilities/textural.py https://github.com/metoppv/improver/blob/master/improver/cli/convection_ratio.py

Info from Nigel Roberts: That convective ratio is something different to the texture algorithm I had mentioned, although it is useful for the MOGREPS-G problem and may be good enough for now. MOGREPS-G produces "Convective precipitation" and "dynamic precipitation" and the ratio of those tells you whether the model thinks it is convective or dynamic. The issue with that is that MOGREPS-UK doesn't have that distinction - all precipitation is "dynamic". Similarly, the radar only has precipitation and doesn't distinguish between showers or non showers without an algorithm. In the future MOGREPS-G may also not have a distinction when the new convection scheme comes in, although because of the usefulness of having the distinction for downstream products, it may get output anyway. Then there may be the need for an algorithm to discriminate between dynamic and convective precipitation when the model can't.