The sensitivity of species distribution modeling to scale differences

Species distribution modeling is one of the most effective habitat analysis methods for wildlife conservation. We evaluated the sensitivity of species distribution modeling to different grain sizes and extent sizes from 30 m to 4950 m using maximum entropy (MaxEnt) modeling. The grain size represents a unit for analysis, whereas the extent size defines the scope of the analysis in a way that reflects the environmental data for the area in which the species of interest occurs. We compared the resulting suitability indexes and habitat areas based on two approaches. The first approach increases the extent size for a fixed grain size. The second approach increases the grain size and the extent size by equal amounts. The suitability index based on the first approach (R2 = 0.34) was greater than the suitability index based on the second approach (R2 = 0.89). The first approach was fitted to a logarithmic function with a critical point at approximately 0.5 km, converging to about 0.76. In contrast, the second approach showed a linear decrease to values less than 0.5. The distribution of habitat area found with the second method (R2 = 0.87) was broader than that found with the first method (R2 = 0.63). The relationship between the extent size and the landscape index of the first method can be displayed as a power-law graph with a critical point of 0.5 km. The method of expanding extent size has greater accuracy, although the time that it requires for data processing is long. The results of this study suggest that the maximum grain size should be approximately 1.5 km. If the grain size is greater than 1.5 km, the accuracy of the habitat suitability index decreases below 0.6, and the predicted habitat suitability increases dramatically.