Fig 1.
Map of the study area highlighting the biogeographic sub-regions of the Atlantic forest (A), the area sampled in our study and the geographic distribution of the thin-spined porcupine previously known through the study of Oliver and Santos [16] (B). Evidence data reported by these authors are presents in the map. Geographic Projection: Datum WGS 1984.
Fig 2.
Potential geographic distribution of the thin-spined porcupine (Chaetomys subspinosus) in the Atlantic Forest, Brazil, based on the binary ensemble model of three algorithms: Bioclim, Maxent and Generalized Linear Model (GLM).
The binary ensemble model is shown by the overlap of three models (A). The most reliable distribution based on the ensemble model and survey empirical data are shown in the map (B). The confirmed occurrence points used in the modeling are also shown in the map (C), as well as the sites where the target species was reported (D) and not reported (E) by locals during interviews. Geographic Projection: Datum WGS 1984.
Fig 3.
Predicted environmental suitability for the thin-spined porcupine (Chaetomys subspinosus) in the Atlantic Forest, Brazil, based on continuous ensemble values estimated through of three algorithms: Bioclim, Maxent and Generalized Linear Model (GLM).
Ensemble suitability values are the average suitability of rescaled values produced by three algorithms. Suitability categories were defined using ArcGis natural breaks (Jenks) as: very low (0 to 0.140), low (0.140 to 0.359), medium-low (0.359 to 0.558), medium-high (0.558 to 0.715) and high suitability (0.715 to 1). Geographic Projection: Datum WGS 1984.
Fig 4.
Protected areas and remaining Atlantic forest in the potential geographical distribution of the thin-spined porcupine (Chaetomys subspinosus) differentiated according to class of the environmental suitability predicted by modeling procedures, which are: lower (0 to 0.558), medium-high (0.558 to 0.715) and high suitability (0.715 to 1).
In detail, the following regions are shown: (A) Sergipe state; (B) northeast of Bahia; (C) southeast of Bahia and northeast of Minas Gerais; and (D) Espirito Santo state. Protected areas are numbered and can be identified following Table 1.
Table 1.
Federal and state protected areas in the potential geographical distribution of the thin-spined porcupine (Chaetomys subspinosus) and the total of protected forest area in the predicted suitable and high-suitability climatic zone for the species.
Fig 5.
Influence of the percentage of forest cover (A) and mean environmental suitability (B) on the proportion of the interviews in which the thin-spined porcupine (Chaetomys subspinosus) was reported by locals (herein proportion of positive interviews) in different regions of the species’ potential geographical distribution. The relationships between these three variables are shown in the bubble chart (C), where the size of the bubble represents the percentage of positive interviews in each region. Regions (n = 36) were delimited overlapping a hexagon grid of the cell size equal to 5000 km2 on the species’ geographical distributions. Only cells with five or more interviews were considered. The line with a breakpoint represents the curve adjusted for the piecewise relation identified according to the “better” model shown in Table 2.
Fig 6.
Relationship between the percentage of forest cover and the proportion of the interviews in which the non-threatened Bahia porcupine (Coendou insidiosus) was reported by locals (herein named incidence rate) in different regions of the Chaetomys’ potential geographical distribution (A). Relationship between the rate of incidence of Coendou insidiosus and Chaetomys subspinosus (B). Regions (n = 36) were delimited overlapping a hexagon grid of the cell size equal to 5000 km2 on the species’ geographical distributions. Only cells with five or more interviews were considered in the analysis.
Table 2.
Akaike (AIC) based model selection for the proportion of positive records of evidence of the presence of thin-spined porcupines (Chaetomys subspinosus) in landscapes (hexagons = 5000 km2) from Atlantic forest obtained by interviews with locals along the species’ extent of occurrence.
Piecewise and Generalized Linear Models (GLM) used the mean environmental suitability predicted by species distribution modeling and percentage of the forest cover as fixed factors and weighted the response variable by the number of the interviews. We also show the number of degrees of freedom (Df), AIC differences (Δi), and Akaike weights (ωi).