Rough sets in floristic description of the inner-city of Warsaw

Abstract

Summary This paper reports an application of rough set theory to floristic description of downtown Warsaw. The original Warsaw - flora dataset consisting of 1181 attributes (presence/absence of vascular plant taxa in 225 grid squares) has been reduced to an 8-attribute subset enabling a minimal discriminant description of the inner-city (Table 1). The distribution of Diplotaxis muralis (Fig. 1) has been recognized as the attribute whose discriminating power is the highest (Table 2). This conforms to an earlier chisquare-based approach. Three taxa out of eight included in the best subset, Diplotaxis muralis, Humulus lupulus and Malva neglecta , had been considered by other authors as good indicator plants for certain urban biotopes. An identification tree extracted from the reduced dataset has been presented (Fig. 2). The tree can be regarded as the shortest (i.e., involving minimum number of species) floristic characteristic of the downtown. Rough sets approach has been found promising and complementary to both expert's informal appraisal and statistical methods. This makes it possible to obtain the minimum knowledge representation and enables the extraction of useful rules from the data. Conceivable drawbacks include local character of the identification rules and their susceptibility to small changes in floristic composition. We conclude that fuzzy rough sets and rough fuzzy sets provide an especially interesting framework for geobotanical analyses. They combine coarseness of rough sets with imprecision of fuzzy approach, which would presumably eliminate the shortcomings encountered.