Abstract
A simple method of forecasting the mean monthly temperature in Stockholm has been developed. It is based mainly on data from a 96 year long record of local temperature. In addition to persistency other statistical interrelations, dynamical methods and synoptic experience have been used. The result from a 10 year test period shows that the forecasts have caught about one third of the possible amount above guessing. The presented method gives results which are significantly better than the results of pure persistency and of the US monthly forecasts. This systematic method may be used for other places and could be combined with other methods.