Print PageThe problem of uncertainty in species distribution model (SDM) has been heavily discussed. However, quantitative evaluations of uncertainty of species range shifts are seldom done, probably due to lack of accurate measurements of range shifts. We adopted 2 new indices to show how unreliable a species distribution can be in projecting species ranges to future climates based on example species. Two range shift indices, range increment index (I) and range overlapping index (O) were defined based on fuzzy set theory, and then were applied to 8 example species (belonging to Pinaceae family in China). ArcInfo Maro-language (AML) program was developed to calculate the indices. Maxent model was use to project species ranges to future climates in the end of this century(the period of 2081-2100). Five GCM model predictions on future climates based on 3 green house gas emission scenarios(A1B, A2, B1) , and five bioclimatic variable sets were chosen as model running settings for each species (resulting in 5×3×5×8 runs). Consensus (or ensemble) modeling was also adopted for each scenarios and species to serve as references. ANOVA analysis of the I and O indices was conducted to separate variance components among GCM models differences, variables selecting differences, and green house gas emission differences.
Results/Conclusions
To I index (I > 0 means range expansion, otherwise contraction, with 1.0 meaning 1.0 time larger than original range size, and -1.0 meaning lost all original range), the range of index values can change from -0.85 to 1.86 among the 25 single models (5 GCMs * 5 variable sets) when the consensus index value is -0.09. For the eight species and three scenarios, all ranges of I value change from minus to positive, no matter the consensus model values are minus or positive. On average, the I variation cause by differences among scenarios, predictors, and GCMs accounts for 1.0, 72.6, and 13.5 percent of total variation respectively. To O index (O = 0.0 means no overlapping, and O = 1.0 means completely overlapping), the range of index for single models can change from 0.05 to 0.58, when the value of consensus model is 0.48. On average, the O variation cause by differences among scenarios, predictors, and GCMs accounts for 7.6, 38.4, and 27.5 percent of total variation respectively. Although single SDM model can generally make meaningful predictions on the spatial patterns of species distribution and directions of range shift, it is highly unreliable in making predictions on changes of species range size and overlapping. Consensus modeling is recommended even if it may not guarantee correctness by itself. The fuzzy set defined indices may serve as valuable tools in similar researches.
The problem of uncertainty in species distribution model (SDM) has been heavily discussed. However, quantitative evaluations of uncertainty of species range shifts are seldom done, probably due to lack of accurate measurements of range shifts. We adopted 2 new indices to show how unreliable a species distribution can be in projecting species ranges to future climates based on example species. Two range shift indices, range increment index (I) and range overlapping index (O) were defined based on fuzzy set theory, and then were applied to 8 example species (belonging to Pinaceae family in China). ArcInfo Maro-language (AML) program was developed to calculate the indices. Maxent model was use to project species ranges to future climates in the end of this century(the period of 2081-2100). Five GCM model predictions on future climates based on 3 green house gas emission scenarios(A1B, A2, B1) , and five bioclimatic variable sets were chosen as model running settings for each species (resulting in 5×3×5×8 runs). Consensus (or ensemble) modeling was also adopted for each scenarios and species to serve as references. ANOVA analysis of the I and O indices was conducted to separate variance components among GCM models differences, variables selecting differences, and green house gas emission differences.
