On a similar topic so I posted in this thread. Does anyone know how much historical data is input and/or retained in each new forecasting run? Also any idea on how often the "crunching" program is updated/changed to reflect errors between what was predicted and what actually happened? Finally, is it valid to compare like for like when runs have been generated using different versions of the algorithms?
Just curious really - have been fascinated by the fact that computers models are used to predict atmospheric conditions for years but never realised that there was so much info available on the subject until I found this site! If you read the BBC website about how they do the monthly outlook it really only gives the impression that they use historical data and hence a LR forecast based on the probability or likelyhood of what has gone before happening again based on the current prevailing conditions, time of year etc etc.
Great site - keep up the good work
Wysi
