One of the purposes of ManUniCast is to help our environmental science students understand the extent that high-resolution (4-km and less horizontal grid spacing) numerical weather prediction (NWP, in meteorology lingo) can predict the details of the weather. What features can be sufficiently resolved in such models? What features are just unpredictable?
For example, compare yesterday's and today's forecasts of today's radar imagery over the UK.
YESTERDAY'S 43-h FORECAST
TODAY'S 19-h FORECAST
VERIFYING RADAR IMAGERY FROM THE MET OFFICE
At first glance, the two forecasts are rather close, showing the showers over the UK and a band of steady precipitation over southern Ireland.
Take a closer look at the observed showers northwest of Scotland—they are mostly isolated cells called open-cellular convection. These showers tend to become more linear as they come across land, perhaps as the low-level wind shear increases due to the increased friction over the land. The 43-h forecast captured the idea of these cells, although certainly not as abundant as in reality. That may be due to the ability of model with 4-km horizontal grid spacing to adequately resolve individual cells.
In contrast, the 19-h forecast shows a greater tendency for the cells to form lines over the water, not what the observed radar is showing at all. This suggests that the wind shear is stronger in this model run than in the later model run, resulting in greater organization.
Indeed, that is exactly what happened. Compare the soundings and see the stronger winds and wind shear in the lowest few hundred millbar in the 19-h forecast compared to the 43-h forecast.
This case was a nice illustration of the sensitivity of the model simulation to small changes in the environmental conditions, illustrating that while we may be able to make good forecasts on the large scale, small-scale forecasts are much more difficult.