Spatial Analysis of Plague in California - Recent Advances and Issues in Environmental Science - page 125

Warmest Month, and Temperature Annual Range) and four precipitation variables
(Precipitation Seasonality, Precipitation of Wettest Quarter, Precipitation of Driest
Quarter, and Precipitation of Warmest Quarter). The log response charts for the two
most important variables used in models of plague in California ground squirrels (Pre-
cipitation in the Wet Quarter and Maximum Temperature of the Warmest Month) re-
fl ect a quadratic response to increasing temperatures and precipitation (Table 3).

Table 3. Maxent final models.

Model AUC

(All rodents)

AUC
( S. beecheyi )

Test omission
rate

Variables % Contribution Response p

Biol6
Biol8
Bio15
Bio5
Bio7

47.3
17.6
15.2
13.2
6.7

-
+
+
Quadratic
+

A 0.876 0.948 0.115 < 0.0001

Biol6
Bio5
Bio17
Bio15
Bio7

-
Quadratic
Quadratic
+
+

48.7
18 14.1 11.9 7.4

B 0.872 0.946 0.115 < 0.0001

Biol6
Bio5
Bio7
Bio18

Quadratic
Quadratic
+
+

C 0.842 0.914 0.192 56.5 23.8 10 9.7 < 0.0001

Biol6
Bio5
Bio18

Quadratic
Quadratic
Quadratic

D 0.835 0.926 0.154 69.5 16.5 13.9 < 0.0001

Test points were used to evaluate omission and 10,000 background points were used to evaluate commission. The re-
ported test omission rate is for equal sensitivity and specifi city. The percent contribution of each variable to the models
refl ects the increase in regularized gain when added to the contribution of the corresponding variable. P-values are
for t-test results for the comparison between positive coyote points and negative coyote points (compared by Maxent
prediction).

Models of plague activity in all rodent species (AUC of 0.835-0.88) and in
California ground squirrels (AUC of 0.913-0.948) based on recent climate condi-
tions accurately identifi ed case locations. All models predicted the highest plague
activity in the Sierra Nevada and along the southern coast under recent climate
conditions (Figure 4 and 5). Models using environmental variables based on squir-
rel data performed well at predicting plague presence in coyotes. All four Maxent
models predicted signifi cantly higher values for pixels that overlapped with positive
coyote specimens (Table 3).
Under future emissions scenarios, our models indicated that climate conditions
will drive (a) an overall decrease in the probability of plague in the state, (b) a subtle
shift to higher elevations as well as (c) a subtle shift to higher latitudes. Future climate
conditions will support increased plague activity in the northern Sierra and central/
north coast counties. However, plague risk associated with climate conditions may
decrease in the southern Sierras and southern inland counties (Figure 6).