*4.2. Experiment Design*

An experiment scenario was designed for the city of Rawalpindi in Pakistan, we took 65 union councils of Rawalpindi district and their total area is 122,982,339.2 m2. We used NetLogo environment for simulations in this region as shown in Figure 6.

In the simulations, we divided agents into two categories; demands and resources. Resource agents are fixed according to the actual location of the resource center while demand agents are created randomly on 186 building nodes according to simulation. Furthermore, travelling distance was estimated using GIS road information instead of Euclidian distance. Each scenario was simulated with and without randomizing traffic delay of maximum 50% of travelling time. These scenarios were simulated 150 times for each experiment according to given factors in Table 3. ANOVA test results are presented in Table 4.

**Figure 6.** RA in the geographic information system (GIS) Space NetLogo implementation.

**Table 4.** ANOVA test summary.


#### *4.3. Result Analysis*

Simulation results of each algorithm were analyzed in R studio with several experiments that are given in Table 3. We performed analysis on the following data to evaluate simulation:


The two algorithms, FCFS and HSL were analyzed for significantly different behavior. As shown in Figure 7a, the HSL behaves slightly better than FCFS. The unallocated demands in HSL are slightly less than that of FCFS allocation. The reason behind this performance is selective allocation of HSL. In HSL, the demand with high severity is allocated at first thus a certain number of demands get served early. Secondly, the effects of duration of demand i.e., short or long jobs on allocation algorithm are observed. It is observed (Figure 7b) that HSL is more sensitive to demand duration as compared to FCFS. As the simulation depicts non-preemptive allocation, the low priority waiting jobs with longer duration stay in HSL wait queues and eventually receive the required resource, and this improves the percentage of allocated jobs. Here each job size is simulated 150 times for numbers of demand sites 5, 10, 15, 20, 25, 30, and 35 respectively. Their requested quantity at each site lies between 1 ≤ random ≤ 10.

**Figure 7.** Effect of number of demand sites on total unallocated demands. (**a**). Percentage of allocated demands w.r.t duration of commitment demands (**b**). Effect of traffic on % of unallocated demands (first come first serve—FCFS) (**c**). Effect of traffic on % of unallocated demands (high severity level—HSL) (**d**).

The effects of traffic delay on resource allocation are presented in Figure 7c,d. A random delay of maximum 50% of the travelling time was introduced for both FCFS and HSL. The increase in unallocated demands is observed due to traffic delay as these delays increase the duration of the commitment of the resources.

The average wait time will be less if we increase the number of resource quantities that fulfils the demand quantities within its time frame. Average wait time also depends on the RA cycle; if it releases resources as soon as possible before the request of other demands, it will, in turn, also affect overall average wait time.

Increase in the number of demands affects the overall resource allocation. If number of demands is less than or equal to number of resources, then its average time is almost near to zero. For all experimental data shown in Table 3, we increased the demand node while resources remained constant with same quantity between 1 ≤ random ≤ 10 for both algorithms. From all experiments, we analyzed that when we have more demands as compared to the available resources then more demands stay unallocated.

#### *4.4. Discussion of Results*

Our work focuses on unallocated demands, as these are a number of deficient resources required in cases of emergencies and disaster. It is seen that even with equal number of demands and resource sites, there is a chance that up to 10% of demands will not be entertained. Pakistan has population to health facility ratio of 1:11,413. Our model shows that in the case of disaster, the resulting resource to demands

ratio is of 1:7. All allocation schemes result in up to 90% deficiency. It is seen that in Pakistan only 4% of patients [20] arriving to the hospital use the emergency services like ambulances. It is due to chronic lack of resources that has resulted in mistrust in the system, thus, most of the emergencies are handled by self-help basis with no medical expert on site. Our study is the first step towards establishing effects of several factors that result in poor disaster management. In our work, we showed the policies of managing or allocating resources does affect the resource management. However, that difference is insignificant as compared to the deficiencies in the system. In future work, we will highlight the areas that are at severe risk due to lack of facilities and distance to resources centers.
