Habitat conversion is a primary threat to terrestrial biodiversity, and fishing is a primary threat to marine biodiversity; both threats can be ameliorated by establishing reserves, if we know where the biodiversity can be found. However, scientists have catalogued only a fraction of the 5-10 million species on Earth. Furthermore, the distributions and habitat requirements of many species that live in remote, inaccessible or uncharismatic locations and communities are poorly mapped. The current state of the art is to “hope” that if we represent certain well-known biodiversity features in a network of reserves then we will conserve the remainder of biodiversity along the way. Using statistical models and reserve selection software, we evaluate how well this “hopeful conservation” might work, and explore how some ecological knowledge can lead us to modified reserve design principles that perform better. We demonstrate our points using a hypothetical region in which we have no maps of biodiversity, and with two empirically-derived case studies: vegetation communities in Queensland, Australia, and mammals in sub-Saharan Africa.
Results/Conclusions
The standard approach effectively assumes each species equally likely to be anywhere in region, and best conserves common species. If species ranges and reserves are compact, then representation becomes a spatial sampling problem: many small reserves and uniform spacing are most likely to represent a species with a critical reserve size/spacing relative to species range size, and sufficiently small and numerous reserves can guarantee representation. However, a reserve that is too small has no conservation value, creating tradeoffs between big reserves that conserve wide-ranging species and small reserves that can represent localized species. We find that “multiscale” reserve networks can equitably protect species of varying extent. Many reserve planning exercises attempt to minimize socioeconomic costs (acquisition or opportunity costs). We find that, because value is spatial autocorrelated, this systematically biases against “unlucky” species that happen to inhabit high-value areas. Actually achieving representation goals with such a strategy requires a much larger reserve network, and is actually more costly than one that ignores costs. In conclusion, if we include biologically realistic models of the spatio-temporal distribution of conservation features, we can improve reserve design. Lacking biological information, the best approach may be to simply place reserves on a latitude/longitude grid.
