Longterm Historic Timeseries
Longterm Historic Timeseries
MeteoGroup is able to reconstruct 40 years of complete historic weather data of any site for which measurements of at least 3 to 4 years are available. This is done by using the available station observations and derive a statistical relationship to the large-scale weather patterns. For the past 40 years gridded weather data through re-analysis projects are available which represent the large-scale weather patterns. Together with the statistical relations they are used to calculate weather "observations" for the same site where the measurements where taken. For the training process 3 years of data is needed. The 4th year of independend data can be used to evaluate the quality of the statistical relations which have been developed. This means that the resulting dataset is delivered with a well-known data quality. Tests for some locations have resulted in an average absolute error for hourly temperature of roughly 0.6 K and no bias. Maximum, minimum and average values even show a much better quality.
By the way, the same methods are also used for location *forecasts*, however the statistical relations are not applied to past data but to forecasted gridded weather data.