empirical investigation of why species–area relationships overestimate species losses

It is generally assumed that, when natural habitat is converted to human‐dominated land cover, such habitat is lost to its native species. Most literature assumes that species richness should vary as a function of remaining natural area, following the well‐known species–area relationship (i.e., classic SAR). However, classic SARs have consistently overestimated species losses resulting from conversion of natural forested land cover to human‐dominated landscapes. Moreover, richness is sometimes a peaked function of remaining natural habitat. Recent studies propose modified SAR models based on species' utilization of multiple habitat types, yet none fully explain a peaked species–area relationship. Here, we evaluate the responses of total avian richness, forest bird richness, and open‐habitat bird richness to remaining natural land cover within 991 quadrats, each 100 km², across southern Ontario, Canada. Total bird species richness peaks at roughly 50% natural land cover. Richness of forest birds varies as a classic power function of forested area. In contrast, richness of birds that prefer open habitats does not increase monotonically with either natural‐ or human‐dominated land cover. Richness of open‐habitat species can be predicted when we partition human‐dominated land cover into an “available human‐dominated” component and “lost” habitat. Distinguishing three land‐cover types (natural, available human‐dominated, and lost) can thus permit accurate predictions of species richness in landscapes with differing levels of natural habitat conversion.