Predictive vegetation mapping based on soil and topographical data: A case study from Saare County, Estonia

Summary in objectives of this study were to estimate the predictability of vegetation site types (VSTs) on the basis of soil parameters, and to prepare the existing data layers for the compilation of the digital vegetation cover model for the islands of Saaremaa and Muhu (Saare County, Estonia). We used various digitised materials (vector format basic map 1:10 000; digitised soil maps, 1:10 000; vector-format mesoscale landscape units (mesochore) subtypes map, 1:50 000; vector-format CORINE Land Cover map, 1:100 000; the classification of Estonian VSTs and type groups (TGs)), and field observation results from 1999 to 2004. On the basis of the values of the combined-digital data fields (soil types+vegetation of basic polygons of the basic map+special restrictions arising from mesochore type or CORINE land use type) defined using theoretical and field observation data, queries were made to ascertain the distribution of natural VSTs. Only 3.16% of the classification area was rejected due to contradictory query criteria. The resulting synthesis map was verified using three different data layers. The results of the verification indicated a need for additional ordination of the system “soil–vegetation” in the light of Saaremaa Islandʹs varied geomorphology, and also highlighted obvious systematic errors in the databases used for verification. esult of verification, 25% of suitable soil distribution areas were, on average, covered by so-called accurate (well correlated) site types (STs) (in some cases by site TGs); accurate and similar, probably due to their mode of use or some change in abiotic conditions, STs and site TGs accounted for around 82% of the soil distribution areas. Query-based maps can be easily upgraded if the accuracy of the data layers is improved and they can be used in the fields of forestry, nature protection and planning.