Regional short-term forecast of total column ozone

An algorithm for predicting the regional distribution of total column ozone (Ω) is described. It is based on locally representative regressions between Ω, and temperature and geopotential at different standard levels of the preceding and/or the day considered. The use of model output parameters instead of the locally measured quantities in the regressions is carried out by an additional regression. The quality of Ω prediction is assessed against a persistence forecast in a January case study characterized by predominantly high interdiurnal changes of Ω. The regional Ω distribution above Germany is classified into objectively determined patterns which are described in some detail for the patterns of the case study. For the smaller region, Germany, there is only one regression superior to the persistence forecast. This regression uses temperature: information at four different pressure levels. The reduction of variance is about 50%. For the larger region considered (40°W–40°E, 30–70°N) all regressions offered give about the same amount of improvement of about 30% against a persistence forecast. Possible improvements of the algorithm are proposed.