Swamp forests dominated by Picea abies contribute strongly to the total plant species richness in boreal forests. The variation in species composition and environmental conditions in swamp forests and the relationships of swamp forests to mires and upland forest, have, however, remained insufficiently known. From a preliminary survey of 57 spruce swamp forests, eleven localities were selected to represent the variation in the study area with respect to size and nutrient status, and altitude and position in the landscape. Physiographic and hydrotopographic descriptions were made. A total of 150 1-m2 plots were distributed on the localities by a restricted random procedure. In each plot, the abundance of all vascular plants, bryophytes and macrolichens was recorded as frequency in 16 subplots, 53 variables (topographic and geographic, tree influence, water-table, water chemical and physical, and soil chemical and physical) were measured. The gradient ( coenocline) structure of vegetation was found by parallel DCA and LNMDS ordinations, interpreted ecologically by analysis of correlations and geostatistical patterns. The two main coenoclines were the same for the full species composition and for vascular plant and cryptogams considered separately. The first gradient was related to soil acidity and nutrient concentrations: plots segregated into relatively poor (and intermediate) and richer swamp forests. Nitrogen availability is considered a decisive factor for species’ responses to this gradient. The second gradient was related to depth to the water table and mesotopographic relief of the swamp-forest surface, varying mostly on fine (0.75-1.5 m) scales. Vascular plants segregate along this gradient due to a trade-off between tolerances to waterlogging and drought, bryophyte are influenced by a complex set of factors. Two minor vegetation gradients were also found; one related to microtopography (extending from flat, lawn-like areas dominated by large bryophytes to more strongly sloping sites dominated by small mosses and hepatics; ‘pocket species’) and one weakly related to the annual water-table amplitude. Relatively strong coenoclines were found that separated entire swamp forests but were uncorrelated with measured variables. These occurred because all swamp forests, notably the richer, had a strong element of uniqueness in species composition, probably because species are recruited from a large species pool during thousands of years. Swamp forest is proposed as a broad term for all peatlands with trees, including mire margin, from which it is essentially indistinguishable. Similarities with, and differences from, open mire and forest on mineral soil are discussed. Some unique features of swamp forests are pointed out. A classification of swamp forests into eight site-types by division of the two main gradients is proposed. Descriptions are provided for the six site-types encountered in the study area. All intact richer swamp forests and a representative selection of poor swamp forests should be protected if maintenance of the biological diversity of (coniferous) forests in general, and swamp forests in particular, is aimed at.
