A startling story doing the rounds in UK papers tells of a £97-million supercomputer that can predict the weather a year in advance. Amazing, you might well think. Read on and, perhaps predictably, the claims start to be revised downwards.It turns out that a new technique, called "hindcasting" and which has been used to measure the North Atlantic Oscillation, can really be used only to forecast the UK winter. Still, pretty impressive. After all, it's not like they have much of a summer.But what is really interesting - and no one seems to have come up with a way in which to predict this - is that at exactly the same time, a very different story is doing the rounds in US papers, bemoaning the national weather service's ability to predict anything.Apparently the Hurricane Center in Miami, using data from a reconnaissance aircraft, had forecast only a slight strengthening of category 1 Matthew in the following 12 to 14 hours.The next day, of course, little Matthew exploded into a category 5 hurricane and killed hundreds of people in Haiti and quite a few in the US too.Now, despite the entirely contrary nature of these two stories, this is not a sympathetic argument for #sciencemustfall.Rather it is about why weather matters and why we need to be better able to predict it.With the advent of stiffer water restrictions in the Western Cape, the whole of our country is now even deeper in the grip of drought.We need better meteorological methods to help us anticipate and manage the attendant crisis. Today the SA Weather Service predicts a 60% chance of rain in Gauteng and a 0% chance of rain in the Western Cape.But they have been mistaken so often, it is foolhardy to put any store in such forecasts.All we can do is hope that they are more than right in the first instance and utterly wrong in the second.And, in the meantime, if it is such a difficult science, perhaps a prayer or even a little dance might not go amiss. #rainmustfall...