If conservation efforts are to be successful, it is critical that we understand the relationship between habitat loss and the probability of population extinction. Available evidence suggests a threshold amount of habitat loss at which the probability of population extinction increases from near-zero to near-one following a small additional loss of habitat. The main factors thought to determine this extinction threshold are reproductive rate of the organism, rate of emigration of the organism from habitat, habitat pattern in the landscape (fragmentation), and matrix quality (survival rate of the organism in non-habitat areas). Effective conservation measures require knowledge not only of what factors determine the threshold, but also the relative magnitudes of their effects. Therefore, the purpose of this study was to determine the relative effects of these four factors on the extinction threshold, using a spatially-explicit simulation model. Reproductive rate had the largest potential effect on the extinction threshold. The rate of emigration from habitat also had a very strong predicted effect on the extinction threshold; the higher the rate of emigration, the more habitat was needed for persistence. Matrix quality had a moderate effect, and habitat pattern had a very small predicted effect. The simulations predicted that under certain circumstances up to 58% less habitat is required for population persistence, if a matrix of very low quality is converted to a matrix of very high quality. Matrix quality can be improved by maintaining heterogeneous landscapes with a diversity of vegetated features, and by reducing factors causing direct mortality of dispersers such as road traffic and pesticides. In summary, the simulations suggest that (i) the first priority for conservation should be habitat preservation and restoration, (ii) information on movement rates of organisms is critical for predicting extinction thresholds and (iii) conservation strategies should consider the quality of the whole landscape, including the matrix. (C) 2001 Elsevier Science Ltd. All rights reserved.
