Maxent modeling for predicting potential distribution of goitered gazelle in central Iran: the effect of extent and grain size on performance of the model

The spatial scale of environmental layers is an important factor to consider in developing an understanding of ecological processes. This study employed Maxent modeling to investigate the geographic distribution of goitered gazelle, Gazella subgutturosa (Guldenstadt, 1780), in central Iran using uncorrelated variables at a spatial resolution of 250 m. We used spatial downscaling to downscale WorldClim data to 250-m resolution. We evaluated the sensitivity of the model to different grain and extent sizes from 250 m to 3 km. We compared the performance of the model at different scales using suitability indexes (AUC) and predicted habitat areas. Two models performed with AUC values higher than random (AUC(un) = 0.957, AUC(pu) = 0.953). The distribution of potential habitats at 250-m grid size was strongly influenced by bioclimatic data, vegetation type and density, and elevation. There were few spatial divergences between uncorrelated and pruned models. The mean AUC across eight different spatial scales ranged from 0.936 to 0.959. There was a significant negative correlation between grain size and AUC (R-2 = 0.57). An increase in grain size increased the predicted habitat area. The extent size and AUC showed a positive correlation (R-2 = 0.18). Predicted suitability habitat also decreased as extent size increased (R-2 = 0.49). Spatial congruence AUC fluctuated within a small range and the maximum difference occurred between models of 1 x 1 and 2.5 x 2.5 km. These results showed that an increase in extent size is more accurate than an increase in grain size, and the maximum accuracy for predicting distribution of goitered gazelle in Iran was obtained if the grain size and extent size were 750 m.