MultiModel Forecasts

Temperature and relative humidity forecasts on Italian weather stations, based on the post-processing of different weather models. The Multimodel Superensemble technique is performed by calculating the weight after a performance evaluation during a period called “learning”. An approach used to provide a more accurate prediction is the development of post-processing schemes, through which a statistical link is sought between the quantities predicted by the models and the meteorological parameters on the ground. The Multimodel Superensemble is a self-regressive linear fit that, using the observed data, determines the average error of the various models over a period called “learning”. It is then assumed that this error is also valid for the next forecast period. In this way, it is possible to obtain a certain prediction by simply making an average of the various available models weighed with this systematic error, different for each one. The weights are recalculated every day, for each station and for each forecast deadline. The Multimodel Superensemble is calculated daily on Italian SYNOP stations and stations in the various regions and is able to provide an output even in case of lack of one or more models, so it represents a very solid tool for the continuous and guaranteed emission of correct station point forecasts with this statistical tool.