The matrix effect: how agricultural matrices shape forest fragment structure and amphibian composition

Abstract

Aim Habitat loss and fragmentation are considered the main drivers of species population declines and extinctions in the world. The large-scale replacement of natural habitats with human-modified habitats, such as the replacement of forests with agricultural and livestock farming areas, creates a scenario where natural habitat patches are immersed in an inhospitable land use matrix. We sought to evaluate how forest fragments (FFs) are influenced by the different surrounding landscape components, where we assessed how agricultural matrices that differ in structure and seasonality affect the structure of FFs and their amphibian assemblages.Location Atlantic Forest, Brazil.Methods Using diverse methods including satellite imagery, historical land use and field data, we evaluated the effects of representative agricultural matrices on the structure of FFs and their amphibian composition, species richness and abundance.Results Our results suggest that the structure of FFs and amphibian composition, species richness and individual abundance are intimately linked to the type of matrix and its proportion around FFs. The nature of the matrix is more likely to affect species richness, individual abundance and composition of amphibian assemblages than FF size, reducing the species-area relationship in FFs. There is strong support for the hypothesis that certain agricultural matrices facilitate FF exposure to the elements and a decrease in forest integrity within the fragment, which is reflected in a loss of strict forest inhabitants. Those matrices that allow for a greater exposure of FFs also appear to favour their occupation by habitat generalists.Main conclusions Surrounding matrices affect the structure and dynamics of forest remnants and should therefore be considered in management plans of severely fragmented landscapes. The influence of the land matrix should be urgently assessed, as it has effects on species occurrences and distribution. This is especially relevant considering the increasing deforestation rates in biodiversity hotspots world-wide.

Effects of agriculture and topography on tropical amphibian species and communities

“…Moreover, our results show that streams surrounded by agriculture may have comparatively lower amphibian occurrence probabilities, indicating the importance of the surrounding matrix. The area surrounding forest fragments, and how it is used, is a key element that determines whether a fragment can support species (Melo et al 2013, Ferrante et al 2017. Most observed species (80%) in our study are forest specialists (Appendix S4) that are expected to avoid open areas, such as agricultural lands Metzger 2010, Almeida-Gomes andRocha 2014).…”

“…, Ferrante et al. ). Most observed species (80%) in our study are forest specialists (Appendix ) that are expected to avoid open areas, such as agricultural lands (Dixo and Metzger , Almeida‐Gomes and Rocha ).…”

“…, Ferrante et al. ). However, our results show that detection probability is far below perfect for all species and variable across the rainy season (i.e., the amphibian breeding season).…”

“…, Ferrante et al. ). If detection error is not properly incorporated, parameter estimates in models may be biased and results can be misleading (Ruiz‐Gutiérrez and Zipkin , Tingley and Beissinger ).…”

“…It can be difficult to determine the impacts of habitat loss on amphibians because the detection probabilities of amphibian species during sampling events are nearly always imperfect (Mackenzie et al 2002, Zipkin et al 2012, Walls et al 2014. Many studies investigating the effects of landscape characteristics on amphibian populations and communities fail to consider that observational data may be biased due to detection errors in field surveys (Becker et al 2007, Ficetola et al 2009, 2011, Mendenhall et al 2014, Quesnelle et al 2015, Ferrante et al 2017. If detection error is not properly incorporated, parameter estimates in models may be biased and results can be misleading (Ruiz-Guti errez and Zipkin 2011, Tingley andBeissinger 2013).…”