December 2012 Number 1 - Utah Native Plant Society
December 2012 Number 1 - Utah Native Plant Society
December 2012 Number 1 - Utah Native Plant Society
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Calochortiana <strong>December</strong> <strong>2012</strong> <strong>Number</strong> 1<br />
Table 1. Environmental Variables<br />
BIO 1-Annual Mean Temperature<br />
BIO 2-Mean Diurnal Range<br />
BIO 3-Isothermality<br />
BIO 4-Temperature Seasonality<br />
BIO 5-Maximum Temperature of Warmest Month<br />
BIO 6-Minimum Temperature of Coldest Month<br />
BIO 7-Temperature Annual Range<br />
BIO 8-Mean Temperature of Wettest Quarter<br />
BIO 9-Mean Temperature of Driest Quarter<br />
BIO 10-Mean Temperature of Warmest Quarter<br />
BIO 11-Mean Temperature of Coldest Quarter<br />
BIO 12-Annual Precipitation<br />
BIO 13-Precipitation of Wettest Month<br />
BIO 14-Precipitation of Driest Month<br />
BIO 15-Precipitation Seasonality<br />
BIO 16-Precipitation of Wettest Quarter<br />
BIO 17-Precipitation of Driest Quarter<br />
BIO 18-Precipitation of Warmest Quarter<br />
BIO 19-Precipitation of Coldest Quarter<br />
DEM-Digital Elevation Model 1km<br />
Slope Angle-Derived from DEM<br />
Slope Aspect-Derived from DEM<br />
Geology-Landform Description<br />
Alternative Suitability Models<br />
Three final models were built for each species: 1)<br />
full model, 2) pruned model, and 3) topo model. The<br />
full model contained all variables. The pruned model<br />
was based on findings from a jackknife analysis, used to<br />
evaluate individual variable importance in model development.<br />
The jackknife method evaluates variable predictive<br />
strength by excluding each variable and creating<br />
a tentative model with the remaining variables (Phillips,<br />
Anderson and Schapire 2006). Then tentative models<br />
are created using each variable in isolation (Phillips,<br />
Anderson and Schapire 2006). Tentative models are<br />
then compared to the full model. The pruned model is<br />
then produced with only important predictive variables<br />
found during the jackknife analysis. The goal of the<br />
pruned model was to remove redundant variables and<br />
provide a better fit to the most important environmental<br />
predictors, when compared with the full model. The<br />
topo model used only elevation, slope angle and slope<br />
aspect as variables. There is a known correlation between<br />
elevation and climate (precipitation and temperature).<br />
The topo model was used to see how predictions<br />
of current species distributions based on topography<br />
alone (elevation, slope and aspect) compared to models<br />
with climate data (temperature, precipitation, slope and<br />
aspect). All three models are evaluated for accuracy<br />
with an AUC score.<br />
Spatial Comparison of Model Output<br />
The habitat suitability maps produced by the three<br />
models were then compared spatially to identify places<br />
of predictive agreement between models (consistently<br />
predicted present or absent), and places where predicted<br />
area of suitable habitat were in disagreement. A spatially-explicit<br />
(by pixel) comparison was performed following<br />
the methods proposed by Parolo and others<br />
(2008), which produced two output maps. The first map<br />
identifies areas of maximum agreement between the<br />
models and the second map identifies areas of minimum<br />
agreement between models.<br />
RESULTS<br />
Two hundred and eleven endemic plants were identified<br />
to have 10 or more representative location points<br />
for accurate modeling. The least number of points used<br />
for training data were seven with three test points, and<br />
the largest was 168 training points with 71 testing<br />
points. Model accuracy varied across species with AUC<br />
values ranging from 0.6423 to 1. The jackknife analysis<br />
demonstrated that slope aspect was the least predictive<br />
variable for the most species and Precipitation Seasonality<br />
(BIO15) was the highest predictive variable for the<br />
most species (Table 2).<br />
Example of ENM for Sclerocactus mesae-verdae<br />
One example of a rare endemic plant modeled is the<br />
Mesa Verde cactus (Sclerocactus mesae-verdae). Mesa<br />
Verde cactus is listed as Threatened by the US Fish and<br />
Wildlife Service and a recovery plan was written in<br />
1984 (Heil 1984). The recovery plan suggested that<br />
populations should be monitored to determine their stability<br />
(Ladyman 2004). This cactus is restricted to populations<br />
in the Four Corners region of New Mexico and<br />
Colorado. Mesa Verde cactus occurs in salt-desert scrub<br />
communities, typically in the Fruitland and Mancos<br />
shale formations, but has also been found to grow in the<br />
Menefee Formation overlaying Mancos shale (Roth<br />
2001). It is most frequently found on the tops of hills or<br />
benches and along slopes and at elevation ranging from<br />
4900 to 5500 ft (Roth 2001). Annual precipitation varies<br />
from approximately 8 to 20 cm. Average temperatures<br />
65