Results
Our analysis included 76 datasets – 37 invertebrate, 12 mammal, 9 bird, 9 herptile, and 9 plant metacommunities – totaling 4401 species and 1190 patches (Fig. 2). Biodiversity declined with habitat fragmentation (SL > SS) in 19.9% of the SLOSS comparisons, showed no response (SS = SL) in 34.7% of the comparisons, and increased with fragmentation (SS > SL) in 45.2% of the comparisons (Fig. 2). If we consider only cases where one of the three SLOSS outcomes was at least twice as likely to occur as the other two outcomes (i.e., > 66% of the comparisons), we found negative fragmentation effects in 9 metacommunities (11.8%), no effect in 20 metacommunities (26.3% of the sample), and positive fragmentation effects in 28 metacommunities (36.8%). Biodiversity was typically higher in sets of many small patches than in sets of a few large patches in plants (81% vs. < 1%), invertebrates (45.2% vs. 18.4%), birds (44.3% vs. 17.5%), and mammals (36.7% vs. 19.5%). Conversely, biodiversity of amphibians and reptiles was typically higher in sets of a few large patches (21.6% vs. 48.8%).
Taxonomic identity had a significant effect on the SLOSS outcome, explaining 19% of the variation, with variation explained increasing to 27% when including an additive effect of patch size evenness in the model (marginal R2). The probability of observing positive responses to fragmentation in biodiversity increased from herptiles (βamp&rep = 7.95; SD = 3.98; z = 1.99;p = 0.04) to mammals (βmamm = 10.01; SD = 3.79; z = 2.64; p < 0.01 ), birds (βbirds = 10.80; SD = 3.97; z = 2.72; p < 0.01 ), invertebrates, (βinvert = 11.31; SD = 3.64; z = 3.10; p< 0.01 ), and plants (βplants = 16.46; SD = 4.19; z = 3.92; p < 0.01 ).