Fig 1.
Surveyed sites in Alta Floresta, southern Brazilian Amazon: 19 forest patches (in red and highlighted by a buffer of 1,000 m-radius) and three continuous forest sites (CF1, CF2 and CF3).
The inset map shows the location of the study area (black circle) in relation to the ‘deforestation arc’ (pink-coloured) within the Legal Brazilian Amazonia (delimited in green). Surveyed sites are numbered according to S1 Table. The enlarged forest patch (inset) illustrates the sampling design applied to a small patch (< 2 ha): (a) an array of four pitfall-traps; (b) followed by five live-trap stations, each of which including one Sherman and one wire-mesh trap deployed on the forest floor. Distances between traps (stations) are indicated in the figure. See main text for a detailed description of the sampling design.
Table 1.
Description of landscape, patch and habitat quality variables quantified to examine properties of small mammal assemblages in Alta Floresta, southern Brazilian Amazon.
Fig 2.
Individual species forest-dependency index (FD) as obtained from the ratio between each species abundance within forest patches and that in neighbouring open-habitat matrix areas (log10 x + 0.01).
Bars are coloured according to each species classification in terms of forest-dependency: forest-dependent (no individuals were recorded using the matrix; FD > 1.82); matrix-tolerant (at least one individual was recorded using the matrix or species abundance within forest patches was particularly high; 0 < FD < 1.82); and open-area (more individuals were recorded using the matrix than using forest fragments; FD < 0). The threshold FD = 1.82 corresponds to the maximum value of FD obtained for a species recorded at least once in the matrix. FD was obtained using data from live traps only. Prior to analysis, species abundance was standardized according to sampling effort. Error bars correspond to the FD standard error (see main text for details).
Fig 3.
Relationships between species (a) richness and (b) abundance (log10 x) and forest area (log10); (c) Principal Coordinates Analysis (PCoA) ordination plot; (d) relationship between species composition (PCoA1) and (e) community-averaged forest-dependency index (FD) and forest area (log10 x).
In (a–d), points are colour-coded according to community-average FD values. Lines are the model adjusted for the stronger relationships (P ≤ 0.05), and shaded areas represent the 95% confidence region. Grey dots in (e) represent outlier data not included in model fits (patches 9, 12 and 17). Explanation power (R2) is indicated for each relationship.
Fig 4.
Estimates of averaged models and their 95% confident intervals for predictors of (a) species richness, (b) standardized species abundance (log10 x), (c) species composition (denoted by PCoA axis 1), and (d) community-averaged forest-dependency index.
Table 2.
GLMMs explaining overall species abundance (best model) and incidence (average model).
Fig 5.
Community-averaged forest-dependency (FD) of small mammal assemblages across the fragmented landscape of Alta Floresta, southern Brazilian Amazon, for 1985, 1990, 1995, 2000, 2005, 2010 and 2015.
Forest patches are colour-coded according to the equation: community-average FD = 0.35 log10 forest area– 0.04 log10 forest area2. Forest area explained 61% of community-averaged FD. For information on species FD values see Fig 2, and on the methods used to estimate community-averaged FD, see main text.