Our goal was to test, whether a classification by means of multivariate analysis could be used to revise the traditional vegetation classification of Cajanderian poor Picea abies mires of Finland, perhaps by including climatic variation alongside the ecological variation. We used agglomerative clustering (cluster analysis) for the classification and non-metric multidimensional scaling (NMDS) for interpreting gradients. A four-cluster solution provided a reasonable hierarchic classification with a poorer Rubus chamaemorus–Sphagnum angustifolium main group (including two site types) and a less poor Sphagnum girgensohnii main group (including two site types). The NMDS revealed a climatic south–north gradient (axis 1), an ecological poor–rich (fertility) gradient (axis 2) and a forest–mire gradient (axis 3). Distribution differences among the site types (i.e. climatic result) was a new feature for the Finnish Picea abies mires. The revised vegetation types can help in defining threatened habitat types at the national level as well as can be used in forest planning.