Abstract
Many procedures exist for identifying sets of sites that collectively represent regional biodiversity. Whereas the mechanics and suitability of these procedures have received considerable attention, little effort has been directed towards assessing and quantifying the effects of varying data inputs on their outcomes. In the present paper, we use sensitivity analysis to evaluate the impacts of varying degrees of (i) survey intensity, (ii) survey extent and (iii) taxonomic diversity on iterative reserve selection procedures. A comprehensive distribution database of the mammalian fauna from the Transvaal region of South Africa is systematically perturbed before implementation of a site selection algorithm. The resulting networks of sites are then compared to quantitatively assess the impact of database variations on algorithm performance. Systematic data deletions result in increased network variability (identity of selected sites), decreased numbers of frequently selected sites, decreased spatial congruence among successive runs and a rapid increase in the number of additional sites required to represent all species present in the region. These effects become particularly evident once data sets are reduced to below 20% of the original data. Consequently, a mixed survey strategy that balances survey effort with survey extent and maximizes taxonomic knowledge is more likely to ensure appropriate planning outcomes.