I, like many others, check the likely weather for the coming days on a website which provides an hour-by-hour forecast of the weather in Reading. There are many such websites, each shows a different forecast because their data comes from different sources or they tweak the forecast for each location in a different way. My personal impression is that the hour-by-hour rainfall forecasts are rarely correct, especially during summer. Let me explain why I think this is the case.
Summer rain in the UK often falls from showers, rather than larger-scale weather systems (typically weather fronts). These showers are quite small, typically ranging from a few kilometres to a few tens of kilometres across, and are much more difficult to predict than the large-scale weather systems largely due to their small size. Until fairly recently, the computer models used to predict the weather were unable explicitly to predict showers on this scale, instead the rain from showers was predicted by a statistical approach called “parameterization”. But with recent advances in computing technology the computer models are now able to predict these small showers explicitly. However, whether the forecasts of showers are skillful is still an open question.
My research aims to assess how useful these new high-resolution models are when it comes to predicting these small showers. We focus on regions of the UK where there is significant terrain (Scottish and Welsh mountains, the Lake District, Pennines and south-west England) because the terrain can be quite an important factor in controlling when and where it rains. For example, the western half of the UK receives much more rainfall than the eastern half because the significant hills and mountains are located near the west coast.
What we find when we assess the forecasts of showers in these high-resolution models is that the terrain can influence the location, timing and intensity of the showers. However, the influence of the terrain can change quite substantially for only a small change in wind speed or direction. Therefore, to predict the location and timing of showers accurately it is very important to predict the large-scale weather pattern accurately. A much more accurate forecast of the large-scale is required to predict the location of the individual showers than to predict other, larger weather features.
My original point was that I felt the location-specific forecasts of summer rain showers are rather inaccurate. In short, by presenting a forecast for a single location in a single, hour-long window the forecast user takes the forecast literally (i.e. rain shower at 4 pm and dry at 5 pm). In reality, though computer forecast models can now explicitly represent showers, they have relatively little skill in predicting the location and timing of individual showers, even in regions where the terrain makes the rain more predictable. My feeling is that the presentation of highly localized, time-specific forecasts is misleading because there is little to no skill at predicting showers at this scale, though there is skill in predicting the existence of showers somewhere in a larger area. Online forecasts would appear more accurate if they replaced the location specific forecasts with forecasts for broader areas. By communicating the forecast in this way, it would be less misleading, wrong less often and much more useful.