next up previous
Next: Parallel Results Up: Verification and Performance Previous: Verification and Performance

Comparison of Selected Outputs

Although several outputs were produced and analyzed, three main statistics were chosen for comparison: average daily travel distance of deer, deaths due to weight loss, and year-end population size. Since daily travel distance and weight loss deaths are dependent upon available forage and water levels, these statistics are used to verify the hydrology and vegetation components, as well as the foraging phase of the deer component. Year-end population sizes are used to verify the reproduction phase of the deer component. Statistics are plotted for an initial deer population size of 10,000. Statistics are also plotted for 2,000 and 20,000 deer population sizes and are presented in Appendices A and B respectively. Random initial deer locations and other characteristics were generated separately for the sequential and parallel programs. The comparisons shown are for single simulation runs for each of the serial and parallel models for each initial population size.

Figure 12 illustrates the average daily travel distance per year for an initial deer population of 10,000, with discrepancies ranging from 0% to 8% between the sequential and parallel models. A deer's travel distance is related to the amount of available forage. Larger available forage quantities result in smaller travel distances while smaller available forage quantities result in larger travel distances. The slight increase in travel distance in the parallel model can be explained by the fact that deer waiting on the queue to forage are required to travel further if the grid cells to which they are assigned have no available forage by the time they are allowed to graze. Since the total number of weight loss deaths is large and very similar in both models, a graph showing the log of these numbers is presented in Figure 13. A difference of only 3% was noted between the two models over the entire 23 year simulation. This slight difference can be attributed to the increase in travel distance in the parallel model, since increased travel distance results in increased energy expenditures and therefore less weight gain. The graph of year-end population sizes shown in Figure 14 illustrates a difference of no more than 5% between the two models.

  
Figure 12: Average daily travel distance per year for an initial deer population of 10,000.

  
Figure 13: Weight loss deaths per year for an initial deer population of 10,000.

  
Figure 14: Year-end population sizes for an initial deer population of 10,000.

In order to examine the distribution and abundance of deer across the landscape, yearly distribution maps were created. Figures 15 through 19 illustrate the actual grid positions of every member of the deer population in selected simulation years. These figures represent those years showing the greatest change from previous years. Each dot represents a single 500m grid cell, which may contain more than one deer, thus different colors are used to represent the number of deer located on each grid cell. A legend describing the colors used for the distribution maps is provided in Figure 15 shows the initial random deer distribution in both models on the first day of the simulation. Figure 16 shows an almost exact distribution of deer in the two models on the first day of year 4. Results become less exact in the beginning of year 15, as shown in Figure 17. This difference is due to the random placement of deer on their outer search perimeters since mostly low quality forage is available during the previous simulation year. Year 18 (Figure 18) shows a distribution similar to year 4, but with a larger deer concentration in the upper left portion of the map. Finally, Figure 19 again shows very similar distribution patterns.

 

  


Figure 15 (top row): Deer distribution at the beginning of Year 0 in (a) the sequential model and (b) the parallel model.

Figure 16 (bottom row): Deer distribution at the beginning of Year 4 in (a) the sequential model and (b) the parallel model.

 

  


Figure 17 (top row): Deer distribution at the beginning of Year 15 in (a) the sequential model and (b) the parallel model.

Figure 18 (bottom row): Deer distribution at the beginning of Year 18 in (a) the sequential model and (b) the parallel model.

 

  


Figure 19 (top row): Deer distribution at the beginning of Year 20 in (a) the sequential model and (b) the parallel model.

Figure 20 (bottom): Deer distribution map legend.



next up previous
Next: Parallel Results Up: Verification and Performance Previous: Verification and Performance



Michael W. Berry (berry@cs.utk.edu)
Wed Oct 11 14:53:18 EDT 1995