Short communicationWhy not WhyWhere: The need for more complex models of simpler environmental spaces
Section snippets
Complexity of ecological landscapes
Ecological landscapes can be divided at the outset into those that are physical characteristics of environments (“scenopoetic variables”) versus those that are shaped and modified by biotic factors (Hutchinson, 1978)—both Stockwell and I refer to the former, and reserve the latter from consideration, at least for the present. Stockwell, however, puts considerable weight on the fact that WhyWhere surveys hundreds of data layers that summarize ecological variation across landscapes, and as such
Simplicity of ecological niches
On the other side of the coin, considerably more serious is WhyWhere's assumption that the ecological niches of species can be summarized in very few, “typically two,” environmental dimensions. This viewpoint flies in the face of decades of ecological research—indeed, dating back to the very beginning of the field, researchers have appreciated a more complex, multivariate nature of factors limiting species’ geographic and ecological distributions (Grinnell, 1917, Grinnell, 1924, Hutchinson, 1957
Other problems
As a means of exploring WhyWhere further, I attempted to apply it to several data sets that I know well. The results of these explorations were not satisfying, and led me to appreciate several additional problems with the program—Fig. 1 presents two representative examples of WhyWhere output when challenged with reconstructing a species’ geographic distribution based on ample species’ occurrence data. I believe that the most fundamental problems are the conceptual points listed above. However,
Conclusion
Given the challenge of sorting through the impressive diversity of geospatial data now available for ecological niche modeling, Stockwell should be congratulated on his development of a data-mining-based approach to this challenge. The idea of integrating data mining and modeling is a useful addition to the literature. What is more, WhyWhere's remote data access features seem also to be a useful development, although I was not able to download or upload data sets as Stockwell had promised.
Acknowledgements
Many thanks to David Stockwell for discussion of these ideas, to Jorge Soberón for a careful read of the manuscript, and to two anonymous readers for insightful comments.
References (28)
- et al.
Current approaches to modelling the environmental niche of eucalypts: implications for management of forest biodiversity
Forest Ecol. Manag.
(1996) - et al.
Predictive habitat distribution models in ecology
Ecol. Model.
(2000) - et al.
GRASP: generalized regression analysis and spatial prediction
Ecol. Model.
(2002) - et al.
Comparing discriminant analysis, neural networks, and logistic regression for predicting species distributions: a case study with a Himalayan river bird
Ecol. Model.
(1999) - et al.
SPECIES: a spatial evaluation of climate impact on the envelope of species
Ecol. Model.
(2002) - et al.
Maximum entropy modeling of species geographic distributions
Ecol. Model.
(2006) Improving ecological niche models by data mining large environmental datasets for surrogate models
Ecol. Model.
(2006)- et al.
Measurement of the realized qualitative niche: environmental niches of five Eucalyptus species
Ecol. Monogr.
(1990) - et al.
DOMAIN: a flexible modeling procedure for mapping potential distributions of animals and plants
Biodiversity Conservation
(1993) - et al.
Predictions and their validation: rare plants in the Central Highlands, Victoria
Novel methods improve prediction of species’ distributions from occurrence data
Ecography
Field tests of theories concerning distributional control
Am. Nat.
Geography and Evolution
Ecology
Error Propagation in Environmental Modelling with GIS
Cited by (32)
Biogeographical, ecological, and phylogenetic analyses clarifying the evolutionary history of Calibrachoa in South American grasslands
2019, Molecular Phylogenetics and EvolutionCitation Excerpt :To avoid highly correlated variables in our dataset, we calculated the pairwise Pearson correlation coefficients, including the 19 bioclimatic layers, in ENMTools 1.4 (Warren et al., 2010). We identified variable pairs with R > 0.75 and discarded those that presented the lowest importance to obtain the best model in a preliminary run (Peterson, 2007). Based on this run result, the eight retained variables per lineage were: clade A (BIO2 = mean diurnal range; BIO3 = Isothermality; BIO4 = temperature seasonality; BIO5 = max temperature of warmest month; BIO8 = mean temperature of wettest quarter; BIO14 = precipitation of driest month; BIO15 = precipitation seasonality; and BIO17 = precipitation of driest quarter); clade B (BIO3, BIO4, BIO5, BIO8, BIO9 = mean temperature of driest quarter; BIO10 = mean temperature of warmest quarter; BIO14; and BIO15); clade C (BIO3; BIO4; BIO5; BIO8; BIO14; BIO15; BIO17; and BIO19 = precipitation of coldest quarter); and clade D (BIO3; BIO4; BIO5; BIO8; BIO11 = mean temperature of coldest quarter; BIO14; BIO15; and BIO17).
Maximum entropy niche-based modelling of seasonal changes in little bustard (Tetrax tetrax) distribution
2008, Ecological ModellingChanges in floral shape: insights into the evolution of wild Nicotiana (Solanaceae)
2022, Botanical Journal of the Linnean Society