| Fig.1 | The Desert tortoise (Gopherus agassizii), a federally threatened species, in the Mojave Desert, California. |
| Fig.2 | Location and elevation maps of Fort Irwin, California, which encompasses approximately 260,000 ha with elevations of 390-1865 m. Since 1979, Fort Irwin has been the Army's National Training Center, which provides large areas of force-on-force military training. |
| Fig.3 | Soil compaction map (a) used to create training intensity map (b) for Fort Irwin, California. No soil compaction or training occurred within the yellow-shaded area. |
| Fig.4 | Initialization maps for desert tortoise population density (a) and vegetative cover (b) used in scenario 1 for Fort Irwin, California. These maps were created using a back-propagation neural network analysis by correlating ground truth data with satellite imagery (see text for more detail). |
| Fig.5 | Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation after running scenario 1 for 250 years (c), the carrying capacity of vegetation (d), and mean index of habitat suitability for desert tortoises after running scenario 1 for 250 years for time 0 (e), and time 250 years (f) at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 1. Scenario 1 used the back-propagation neural network analysis to derive initialization maps for tortoise population density and vegetative cover (see Fig. 4). No new training occurred in the model after time step 0. |
| Fig.6 | Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), and mean index of habitat suitability (d) for desert tortoises after running scenario 2 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 2. Scenario 2 used Figs. 5a and 5c as initialization maps for tortoise population density and vegetative cover, respectively. No new training occurred in the model after time step 0. |
| Fig.7 | Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), and mean index of habitat suitability (d) for desert tortoises after running scenario 2 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 2. Scenario 2 used Figs. 5a and 5c as initialization maps for tortoise population density and vegetative cover, respectively. No new training occurred in the model after time step 0. |
| Fig.8 | Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), mean index of habitat suitability (d) after running scenario 4 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 4. Scenario 4 used Figs. 5a and 5c as initialization maps for tortoise population density and vegetative cover, respectively. Spatial variation of training occurred during scenario 4. |
| Fig.9 | Training intensity map for scenario 5 at Fort Irwin, California. Training was excluded from the yellow-shaded areas. |
| Fig.10 | Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), and mean index of habitat suitability (d) for desert tortoises after running scenario 5 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 5. Scenario 5 used Figs. 5a and 5c as initialization maps for tortoise population density and vegetative cover, respectively. Spatial and temporal variation of training occurred during scenario 5. |
| Fig.11 | Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), and the mean index of habitat suitability (d) after running scenario 6 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 6. Scenario 6 used Fig. 5c as the initialization map for vegetative cover. Each cell across the landscape was initialized with the maximum number of tortoises (415 females/km2) that occurred in scenario 1 at time 250 years. No new training occurred in the model after time step 0. |
Fig. 2. Location and elevation maps of Fort Irwin, California, which encompasses approximately 260,000 ha with elevations of 390-1865 m. Since 1979, Fort Irwin has been the Army's National Training Center, which provides large areas of force-on-force military training.
Fig. 3. Soil compaction map (a) used to create training intensity map (b) for Fort Irwin, California. No soil compaction or training occurred within the yellow-shaded area.
Fig. 4. Initialization maps for desert tortoise population density (a) and vegetative cover (b) used in scenario 1 for Fort Irwin, California. These maps were created using a back-propagation neural network analysis by correlating ground truth data with satellite imagery (see text for more detail).
Fig. 5. Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation after running scenario 1 for 250 years (c), the carrying capacity of vegetation (d), and mean index of habitat suitability for desert tortoises after running scenario 1 for 250 years for time 0 (e), and time 250 years (f) at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 1. Scenario 1 used the back-propagation neural network analysis to derive initialization maps for tortoise population density and vegetative cover (see Fig. 4). No new training occurred in the model after time step 0.
Fig. 6. Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), and mean index of habitat suitability (d) for desert tortoises after running scenario 2 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 2. Scenario 2 used Figs. 5a and 5c as initialization maps for tortoise population density and vegetative cover, respectively. No new training occurred in the model after time step 0.
Fig. 7. Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), and mean index of habitat suitability (d) for desert tortoises after running scenario 2 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 2. Scenario 2 used Figs. 5a and 5c as initialization maps for tortoise population density and vegetative cover, respectively. No new training occurred in the model after time step 0.
Fig. 8. Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), mean index of habitat suitability (d) after running scenario 4 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 4. Scenario 4 used Figs. 5a and 5c as initialization maps for tortoise population density and vegetative cover, respectively. Spatial variation of training occurred during scenario 4.
Fig. 9. Training intensity map for scenario 5 at Fort Irwin, California. Training was excluded from the yellow-shaded areas.
Fig. 10. Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), and mean index of habitat suitability (d) for desert tortoises after running scenario 5 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 5. Scenario 5 used Figs. 5a and 5c as initialization maps for tortoise population density and vegetative cover, respectively. Spatial and temporal variation of training occurred during scenario 5.
Fig. 11. Mean spatial distribution of desert tortoises (a), change in mean number of desert tortoises (b), mean percent aerial cover of vegetation (c), and the mean index of habitat suitability (d) after running scenario 6 for 250 years at Fort Irwin, California. The mean was obtained from 100 simulations of scenario 6. Scenario 6 used Fig. 5c as the initialization map for vegetative cover. Each cell across the landscape was initialized with the maximum number of tortoises (415 females/km2) that occurred in scenario 1 at time 250 years. No new training occurred in the model after time step 0.
