@ John Holmes & wysiwyg, many thanks for the informative replies! Yes, with aerodynamics it's non-linear equations that are trimmed down and require massive computing power (Computational Fluid Dynamics aka "CFD"). It's not the most accurate method and gets worse when dealing with turbulent flow, etc. So the weather, being a vastly scaled up version of this, complete with a huge range of thermodynamic factors as well, is clearly not going to be easily modeled with current technology and computing power, which is what got me thinking about a different, more statistical "matching" approach rather than taking fragmented / extrapolated data sets and feeding them into algorithms that essentially consist of non-linear equations that have been dumbed down into manageable linear ones! As things stand, it's low quality initial data fed into necessarily non-comprehensive equations, which is why (it now seems to me) a high level of accuracy beyond a few days is unattainable from the models alone. I may indeed email the Met Office but I think they might be reluctant to let me know too much detail about their modeling as I'm guessing its proprietary information!??