Complex patch geometries maximize species richness at the expense of forest specialists

Habitat loss and fragmentation are the greatest threats to reptiles and amphibians (herpetofauna) around the globe, but especially in the Neotropics where high diversity and ongoing land-use change coincide. Persistence of biodiversity in fragmented systems relies both on characteristics of habitat patches, and on the permeability of the landscape that separates the patches (the ‘matrix’). We sought to understand: (a) how the herpetofauna community differs between forest reserves, patches, corridors, and matrices, and (b) the landscape characteristics that increase suitability of a habitat patch. We conducted herpetofauna surveys in patches, corridors, matrices, and preserves (54 total sites) in a fragmented landscape in Costa Rica for three seasons. We recorded 1663 individuals of 52 species. We found that the herpetofauna community differed, and had lower richness and abundance, in the matrix compared to the other three habitat types. Patches and corridors supported a similar community to the forest preserves, demonstrating the conservation value of small forest remnants. Water body presence was an important predictor of richness and abundance in both patches and matrices. While total richness increased in patches with more edge, this was driven by the response of generalist species, whereas the prime indicator species of forest preserves decreased in patches with complex shapes. The differing response to landscape characteristics between specialist and generalist species demonstrates the importance of considering specific taxa when setting conservation goals, rather than using richness measures alone. Our findings can help guide preservation of forest fragments to optimize biodiversity conservation in mixed forest-pastoral landscapes.

Methods
Our study was centered around Las Cruces Biological Station (LCBS) in southern Costa Rica in Puntarenas Province (8.78°N, 82.96°W). Additional details on selected sites can be found in the published manuscript. 

We completed herpetofauna surveys at 54 sites (20 patches, 20 matrices, 10 corridors, 4 reference forests) in each of three field seasons: March 2019, June/July 2019, and March 2020. We employed three survey techniques to obtain representative samples of the entire herpetofaunal community. Diurnal surveys included a visual encounter survey (VES) and a leaf litter quadrat. Nocturnal surveys included a VES and a 10-min anuran call survey. 

All individuals were identified to species when possible, with a few exceptions. Diasporus vocator and Diasporus diastema were collapsed into Diasporus spp. due to difficulties in discerning the two species both visually and by call surveys. Additionally, the rhodopis-species group contains seven small leaf litter frogs in the genus Craugastor that are notoriously difficult to identify in the field, although most were presumably Craugastor stejnegerianus based on our location.

For our visual encounter surveys, we walked a 100 m transect. For leaf litter quadrat surveys, we selected a 5x5-m area along our transects with leaf litter, and would rake through the leaves and grass to uncover any herpetofauna. For our anuran call surveys, we turned off our headlamps and sat in silence for 10 minutes while recording ambient sounds on a Tascam DR-40 recorder set to a sampling frequency of 44.1 kHz (frog species at our sites have calls that range in dominant frequency from ~450 – 7500 Hz). To identify frog calls, we listened to each of our surveys while viewing them in Audacity 2.4.2, and every time a potential frog call was heard, we went through the spectrograms of all possible species from the region until we found one that matched the frequency range and spectrogram pattern. Calls were treated as presence/absence data because we could not determine whether multiple calls of the same species were a single or multiple individuals. Additional information on our survey techniques is available in the published manuscript.