23.10.2014 Views

December 2012 Number 1 - Utah Native Plant Society

December 2012 Number 1 - Utah Native Plant Society

December 2012 Number 1 - Utah Native Plant Society

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Utah</strong> <strong>Native</strong> <strong>Plant</strong> <strong>Society</strong><br />

The habitat suitability maps also varied between the<br />

models (Figure 2). The full model (Figure 2a) and the<br />

pruned model (Figure 2b) clearly show a smaller range<br />

than the topo model (Figure 2c). The area of suitable<br />

habitat for the full model was 919,973 acres, the pruned<br />

model 1,386,261 acres and the topo model 47,110,640<br />

acres.<br />

The spatial comparison of models found 571,307<br />

acres of suitable habitat agreement between the models<br />

(Figure 2d). Disagreement analysis of the three models<br />

shows 47,081,482 acres (Figure 2e) and suggests that<br />

the topo model generated the most disagreement between<br />

models.<br />

DISCUSSION<br />

Alternative Suitability Models<br />

The use of alternative models can offer a better understanding<br />

of how environmental variables play a role<br />

in a species distribution. The comparison of the three<br />

models may provide a deeper knowledge of speciesenvironment<br />

relationships and lead to a more rigorous<br />

assessment of potential distributions. The alternative<br />

models may also identify needs in data to further understand<br />

species-environment relationships.<br />

The three models used in this analysis all performed<br />

better than random. The full and pruned models performed<br />

the best (highest AUC scores) while the topo<br />

model over predicted suitable habitat. The jackknife<br />

analysis provided a way to understand variable importance<br />

for the use in the pruned model. This may help<br />

eliminate redundant variables and an overly large<br />

model.<br />

Most pruned models showed a larger area of suitable<br />

habitat when compared with the full models. This suggests<br />

two possible interpretations: 1) the full model may<br />

be over fit, and not predicting all potential suitable habitat,<br />

or 2) the pruned model may be over predicting. Another<br />

interesting aspect of using alternative models was<br />

the use of only topographic variables in the topo model.<br />

Although this model over predicted suitable habitat for<br />

Mesa Verde cactus, other species models show a closer<br />

fit. This model may provide a way to analyze a species<br />

distribution when only elevation data are available, as<br />

Parolo and others (2008) found with Arnica montana.<br />

Spatial Comparison<br />

The spatial comparison of habitat suitability map<br />

output from the three models provides quantification of<br />

uncertainty in the model predictions of suitable habitat.<br />

Levels of uncertainty have important management implications.<br />

Areas with high model agreement that a species<br />

is present but without known occurrences of that<br />

species are target areas for field surveys. Areas of disagreement<br />

may provide insight into variables that con-<br />

tribute to uncertainty that could be better resolved<br />

through additional field work.<br />

Final Habitat Maps and Use<br />

The goal of this project was to identify potential suitable<br />

habitat for species by using alternative models<br />

evaluated with AUC scores and a spatial comparison.<br />

When comparing the three models, the AUC scores<br />

range from good to excellent for Mesa Verde cactus.<br />

The spatial comparison identified the full and pruned<br />

models as having similar predicted areas, while the area<br />

predicted by the topo model was larger. This technique<br />

provides a more rigorous analysis of the potential distribution<br />

of suitable habitat.<br />

The methods described above provide a more<br />

straightforward ecological interpretation of how the environment<br />

affects a species’ distribution. This modeling<br />

technique can also provide a better understanding of<br />

endemic plant distributions. These models can be used<br />

for future field investigations to find new populations<br />

and to identify relationships between climate, geology<br />

and topography with endemics.<br />

ACKNOWLEDGEMENTS<br />

This project was supported by the Gloria Barron Wilderness<br />

<strong>Society</strong> Scholarship and NSF grant 0753163.<br />

Comments from Dr. Deana Pennington and Daniela<br />

Roth greatly improved the manuscript and many thanks<br />

to those who have shared their data and time to improve<br />

this project.<br />

LITERATURE CITED<br />

Araujo, M. and A. Guisan. 2006. Five (or so) challenges<br />

for species distribution modeling. Journal of Biogeography<br />

33: 1677-1688.<br />

Elith, J., C. Graham, R. Anderson, M. Dudik, S. Ferrier,<br />

A. Guisan, R. Hijmans, F. Huettmann, J. Leathwick,<br />

A. Lehmann, J. Lucia, L. Lohmann, B. Loiselle,<br />

G. Manion, C. Moritz, M. Nakamura, Y. Nakazawa, J.<br />

McC Overton, A.T. Peterson, S. Phillips, K. Richardson,<br />

R. Scachetti-Pereira, R. Schapire, J. Soberon, S. Williams,<br />

M. Wisz, and N. Zimmermann, 2006. Novel<br />

methods improve prediction of species’ distributions<br />

from occurrence data. Ecography 29: 129-151.<br />

Hanley, J.A. and B.J. McNeil. 1982. The meaning<br />

and use of the area under a Receiver Operating Characteristic<br />

(ROC) curve. Radiology 29: 773-785.<br />

Heil, K.D. 1984. Mesa Verde cactus (Sclerocactus<br />

mesae-verdae) Recovery plan. U.S. Fish and Wildlife<br />

Service, Region 2, Albuquerque, New Mexico.<br />

Kartesz, J. and A. Farstad. 1999. Multi-scale analysis<br />

of endemism of vascular plant species. In: Terrestrial<br />

ecoregions of North America: A conservation assessment.<br />

Eds. Ricketts, T.H., Dinerstein, E., Olson, D.M.<br />

and C. Loucks. Island Press, Washington, D.C.<br />

74

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!