Species Distribution Modeling for Conservation Purposes

Abstract

Species distribution models (SDMs) can be useful for different conservation purposes. We discuss the importance of fitting spatial scale and using current records and relevant predictors aiming conservation. We choose jaguar (Panthera onca) as a target species and Brazil and Atlantic Forest biome as study areas. We tested two different extents (continent and biome) and resolutions (~4 Km and ~1 Km) in Maxent with 186 records and 11 predictors (bioclimatic, elevation, land-use and landscape structure). All models presented satisfactory AUC values (>0.70) and low omission errors (<23%). SDMs were scale-sensitive as the use of reduced extent implied in significant gains to model performance generating more constrained and real predictive distribution maps. Continental-scale models performed poorly in predicting potential current jaguar distribution, but they reached the historic distribution. Specificity increased significantly from coarse to finer-scale models due to the reduction of overprediction. The variability of environmental space (E-space) differed for most of climatic variables between continental and biome-scale and the representation of the E-space by predictors differed significantly (t = 2.42, g.l.= 9, P < 0.05). Refining spatial scale, incorporating landscape variables and improving the quality of biological data are essential for improving model prediction for conservation purposes.

Mapping black panthers: Macroecological modeling of melanism in leopards (Panthera pardus)

“…For each of these sets of location records, we used 70% of the points for training and 30% for testing the models, with the data being sampled using the bootstrap routine [48]. For all runs, we used the following parameters and configurations: random seed, convergence threshold of 1E-5, 500 iterations and 10,000 hidden background points [49]. The model performance was assessed by the AUC (Area Under Curve) value for the Receiver Operating Characteristic (ROC) curve based on sensitivity versus specificity of the response between occurrence data and predictors, incorporating a binomial probability as a null model [48–51].…”

New records of spiders (Arachnida, Araneae) from the state of Roraima, northern Brazil

“…This type of data can be used in species distribution models for many purposes, such as conservation (Ferraz et al 2012). These collection data are also crucial because they are a permanent record of a species at a given place and time (Funk & Richardson 2002).…”

Biogeography of polymorphic phenotypes: Mapping and ecological modelling of coat colour variants in an elusive Neotropical cat, the jaguarundi (Puma yagouaroundi)

“…The data were sampled using the bootstrap routine (Pearson, 2007). All runs were configured to incorporate a random seed, a convergence threshold of 1E-5 with 500 iterations and 10 000 hidden background points (Ferraz et al, 2012). Model performance was assessed by the area under curve (AUC) value for the receiver operating characteristic (ROC) curve based on sensitivity (omission rate) versus specificity (fractional predicted area) of the response between occurrence data and predictors, incorporating a binomial probability as a null model (Pearson, 2007;Tôrres et al, 2012;Calabrese et al, 2014).…”