The species-area relationship (SAR) is an essential conservation tool, allowing extinction rates to be predicted based on habitat loss. Naturally formed islands are often used as a baseline when predicting the effects of isolation, age, and matrix type on SARs, but they are anachronistic analogs for remnant habitat “islands” created by land-use change. Here, we contrast forest islands created by damming rivers with forest fragments created by land clearing to examine the effects of the matrix (i.e., land vs. water) on the SAR of birds. We compare 50 SARs based on 1,963 bird species in tropical and subtropical forests, including 1,016 forest remnants. Extinction rates predicted from forest remnants within a terrestrial matrix (N = 680) were 14.0% lower than human-made forest islands in an aquatic matrix (N = 336). Moreover, matrix effects became more pronounced when examining only 1,506 forest-dependent bird species, whose predicted extinction rate declined by 16.5% and richness nearly tripled (from 7 to 18) in small (~1 ha) forest remnants in a terrestrial matrix compared to human-made forest islands. We found no support that variables linked with metapopulation dynamics (regional forest cover), extinction debt (isolation time), and global biogeography (elevation, latitude, and regional species pool) explained the observed variation among SARs. Our analysis provides a fundamental test of one of the theoretical foundations of conservation biology. It demonstrates that forest fragments smaller than 10,000 ha, which represent one of the most threatened and numerically abundant tropical and subtropical forest habitats on left Earth, host substantially more forest-dependent bird species than predicted by the biogeography of naturally formed islands. Our findings highlight the potential role of human-modified habitats in curbing extinctions.
countryside biogeography, extirpation rates, fragmentation, island biogeography, working lands