Species density of foliage-dwelling spiders in field margins: a simple, fuzzy rule-based model

Field margins are grassy strips or banks between arable fields or meadows that are not cultivated or ploughed and contain only single shrubs or trees. They not only support beneficial arthropods, e.g. predators of crop pests, but can also be of potential value to species of concern, making them a critical target for conservation of biodiversity in agricultural landscapes. Most ecological data and most knowledge of ecological relationships are imprecise, uncertain or ambiguous. An overemphasis on the precision of mathematical models, thus, does not necessarily translate into a greater representation of reality. In contrast, fuzzy expert systems permit the representation and processing of 'soft' ecological knowledge in terms of natural language. This is done by sets of IF-THEN rules that relate the variables (expressed in the form of fuzzy sets such as 'rather cold', 'comparably few species', etc.) among each other. Data on 96 field margins in southern Germany and their foliage-dwelling spider assemblages, which had been collected from May to September 1994, were used for fuzzy modelling of the effects of margin width, margin density, frequency of mechanical disturbance and vegetation-architectural complexity on species density of spiders. Habitat factors were ordered according to their hierarchical effectiveness: margin width and disturbance determine habitat persistence, habitat persistence and margin density determine colonisation potential, and colonisation potential and herbaceous-plant cover determine species density. Forty-five rules are necessary to relate these factors. The predictive power of the fuzzy model was surprisingly high. The mean average error between predicted and observed number of species for test data, which was not used for model development, was less than 1.4 species (observed values ranged between 3 and 25). A multiple-regression model showed a mean average error of 3.17 species. We conclude that fuzzy models are suitable for representing the soft knowledge of field-margin/spider-assemblage relationships. The model permits time- and manpower-saving validation on larger scales by qualitative rather than quantitative data collection and by the inclusion of remote sensing data. We believe that fuzzy models could provide a valuable means for formulating measures necessary for conservation of biodiversity in agricultural landscapes. (C) 2000 Elsevier Science B.V. All rights reserved.