KNOWLEDGE-BASED AND INDUCTIVE MODELING OF ROUGH FESCUE (FESTUCA-ALTAICA, FESTUCA-CAMPESTRIS AND FESTUCA-HALLII) DISTRIBUTION IN ALBERTA, CANADA

The distribution of three rough fescue species (Festuca altaica, F. ca mpestris and F. hallii) in Alberta was modelled using knowledge-based and inductive approaches. The first used differences in temperature re sponses defined in growth cabinet experiments, and simple logical algo rithms operating on monthly mean climate surfaces. The second used a p oint database of surveyed botanical composition to define relationship s between the species location and climatic factors. Agreement between the zones defined by knowledge-based logical modelling, inductive mod elling from points and interpolation of botanical composition was gene rally good. Botanical composition was used as an abundance measure to enhance estimates of conditional probability of presence in inductive modelling. Additional peaks in the distribution of conditional probabi lity of presence with certain climate variables allowed the identifica tion of a sub-zone attributable to F. campestris which was smaller tha n that produced by logical modelling. Modelled zones from both methods agreed with published descriptions of distribution and intergrading b etween species. Choice of method depends on the relative availability of site data versus the amount of knowledge of species behaviour. (C) 1997 Elsevier Science B.V.