SUJET :

Hi,

Are your readings from a meteotable or meteogram? From your description it appears so.

Cyprus depression rainfall is mostly a cumulus event. Lines of individual cumulus clouds dumping their content in a scattered fashion. It may be pouring in one street and nothing in the next block and vice-versa. It is very much a statistical game. There are many gaps between the showers
.
The high resolution models try their best to predict the scattering of the showers but this is also in the end quite random for a pinpoint such as used by a meteotable as the model "gaps" in the showers are very unlikely to match the shower "gaps" in reality. Because of this rain needs to be visualised and read as an area event using the maps with rainfall contours. These average out the pinpoints of rainfall and give a much more accurate picture.

In fact in the recent rain event the openWRF model and the IMS COSMO model showed very similar area maps and especially the accumulation maps (24 or 48hrs) were very close in results.

BTW, in a lower resolution model such as the global GFS, ICON Arpege and EC models the model averages the rainfall for the larger cell size and a pinpoint reading is often more realistic than one preformed on a high resolution model.
Since yesterday openWRF is using a different cumulus scheme and also a different micro-physics scheme (for the non-cumulus rain. This does some smoothing of the output so there are less so called "gaps" predicted in the scattered showers. This was done as it was observed that the cumulus scheme had a consistent error in the position of back slope reduction in rainfall. After changing this the new scheme caused the micro-physics scheme to have an optimistic snowfall prediction so this was also modified. We will wait for the next depression system to see the results.

David

Thank you very much for the full explanation and for your efforts.

Did openWRF become using a different cumulus scheme and a different micro-physics scheme or are they still under checking?

Hi,
As of 28 Dec. all openWRF production has been set to use Multi-scale Kain-Fritsch cumulus scheme and WSM Single-Moment 6-class micro-physics scheme.
These appear to be providing good detection of convection and satisfactory prediction of precipitation.
We do place primary importance on convection detection as the forecasts are focused on use by the sailing community for whom it is more important to be aware of the potential of squall lines then of the exact amounts of precipitation (which all models have problems with).