Improving niche projections of plant species under climate change: Silene acaulis on the British Isles as a case study
Empirical works to assist in choosing climatically relevant variables in the attempt to predict climate change impacts on plant species are limited. Further uncertainties arise in choice of an appropriate niche model. In this study we devised and tested a sharp methodological framework, based on stringent variable ranking and filtering and flexible model selection, to minimize uncertainty in both niche modelling and successive projection of plant species distributions. We used our approach to develop an accurate, parsimonious model of Silene acaulis (L.) presence/absence on the British Isles and to project its presence/absence under climate change. The approach suggests the importance of (a) defining a reduced set of climate variables, actually relevant to species presence/absence, from an extensive list of climate predictors, and (b) considering climate extremes instead of, or together with, climate averages in projections of plant species presence/absence under future climate scenarios. Our methodological approach reduced the number of relevant climate predictors by 95.23% (from 84 to only 4), while simultaneously achieving high cross-validated accuracy (97.84%) confirming enhanced model performance. Projections produced under different climate scenarios suggest that S. acaulis will likely face climate-driven fast decline in suitable areas on the British Isles, and that upward and northward shifts to occupy new climatically suitable areas are improbable in the future. Our results also imply that conservation measures for S. acaulis based upon assisted colonization are unlikely to succeed on the British Isles due to the absence of climatically suitable habitat, so different conservation actions (seed banks and/or botanical gardens) are needed.
KeywordsBritish Isles Climate-driven niche modelling Climate extremes Model selection Parsimonious modelling Silene acaulis Variable ranking Variable selection
We thank two anonymous reviewers for their helpful comments that improved this manuscript.
AF and JMA conceived the study. AF performed the GIS and modelling work. AF and JMA wrote the manuscript. MHSAA and JD commented on the paper.
Compliance with ethical standards
Conflict of interest
The authors declare no competing financial interests.
- Aho KA (2013) Foundational and applied statistics for biologists using R. CRC, Boca RatonGoogle Scholar
- Alatalo J, Molau U (1995) Effect of altitude on the sex ratio in populations of Silene acaulis (Caryophyllaceae). Nord J Bot 15:251–256. https://doi.org/10.1111/j.1756-1051.1995.tb00150.x CrossRefGoogle Scholar
- Barry RG, Chorley RJ (2009) Atmosphere, Weather and Climate. Ninth edition. Routledge Editor, NYGoogle Scholar
- Burnham KP, Anderson D (2002) Model selection and multi-model inference, 2nd edn. Springer, New YorkGoogle Scholar
- Cortes C, Vapnik V (1995) Support-vector networks. Mach Learn 20:273–297Google Scholar
- Cox DR (1958) The regression analysis of binary sequences. J R Stat Soc Ser B Methodol 20:215–242Google Scholar
- Freund Y, Schapire RE (1995) A decision-theoretic generalization of on-line learning and an application to boosting. In: European conference on computational learning theory. Springer, pp 23–37Google Scholar
- Guyon I, Elisseeff A (2003) An introduction to variable and feature selection. J Mach Learn Res 3:1157–1182Google Scholar
- Huntley B, Bartlein PJ, Prentice IC (1989) Climatic control of the distribution and abundance of beech (Fagus L.) in Europe and North America. J Biogeogr 551–560Google Scholar
- IPCC (2013) Climate change 2013: the physical science basis. Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change. Cambridge University Press, CambridgeGoogle Scholar
- Kira K, Rendell LA (1992) A practical approach to feature selection. In: Proceedings of the ninth international workshop on machine learning, pp 249–256Google Scholar
- Körner C, De Moraes J (1979) Water potential and diffusion resistance in alpine cushion plants on clear summer days. Oecol Plant 14:109–120Google Scholar
- Lawton JH, May RM, 1995. Extinction rates. Oxford University Press, OxfordGoogle Scholar
- Liu H, Setiono R (1995) Chi2: Feature selection and discretization of numeric attributes. In: Tools with artificial intelligence, 1995. Proceedings, seventh international conference on IEEE, pp 388–391Google Scholar
- Quinlan JR (1986) Induction of decision trees. Mach Learn 1:81–106Google Scholar
- Root TL, MacMynowski DP, Mastrandrea MD, Schneider SH (2005) Human-modified temperatures induce species changes: joint attribution. Proc. Natl. Acad. Sci. U. S. A. 102, 7465–7469Google Scholar
- Rumelhart DE, Hinton GE, Williams RJ, 1985. Learning internal representations by error propagation. DTIC DocumentGoogle Scholar
- Tabachnick B, Fidell L (1996) Using multivariate statistics. Harper, Colling’s, New YorkGoogle Scholar