Discrimination of air masses near the extratropical tropopause by multivariate analyses from MOZAIC data

Multivariate analyses (MA) are applied to discriminate air masses near the extra-tropical tropopause in a north Atlantic cut-off low (COL) case, using the Measurement of OZone and water vapour by five Airbus In-service airCraft data. Different MA, such as hierarchical clustering methods (HCM) and K-means clustering and the linear and quadratic discriminant analyses are tested and compared. The resulting methodology is based on the successive use of the principal component analysis, HCM with average linkage and K-means clustering, and finally the linear discriminant analysis. It is found that these tools result in a much better discrimination between air masses than the classical selection criteria. From a study based on data selected during seven isobaric flights (230 hPa) crossing the same COL in June 1995, the MA method allows to obtain the physical and chemical structure of the COL and surrounding air masses, and the identification of air mass directly linked to the stratosphere–troposphere exchanges (STE). This air mass corresponds to a mixed zone between stratosphere and troposphere near the jet stream, with intermediate values and weak gradients of ozone, potential temperature and humidity (the tropopause is poorly defined). These STE air masses could not have been found objectively with classical discrimination criteria. A possible hypothesis to explain these STE is the dynamics associated with strong convection beneath and around the southern side of the COL.