Fig 1.
Example of the regular (A) and the adaptive (B) grid construction methods applied to the UK farm population. Panel (C) shows a schematic representation of the grid-based infection transmission process. From the infectious farm i (red star) in infectious cell a, spread of infection is initially evaluated at the level of entire susceptible grid cells (b1, b2, b3) and only occasionally evaluated for the individual susceptible nodes (green).
Table 1.
Methods and epidemic model considered in the study and how they relate to each other.
Table 2.
Infection kernel parameter values used in disease simulations.
Table 3.
Parameters for infection kernels used in the comparison to FSR method.
Fig 2.
Speed-up of estimated optimal grid configuration with the conditional subsample algorithm compared to the pairwise algorithm.
Comparison of simulations using the CS algorithm with as cell-size threshold, with the pairwise algorithm for different outbreak stages. The y-axis indicate how many times faster (based on run time) the simulations using the CS algorithm were compared to the pairwise simulations.
Fig 3.
Results from outbreak simulations with the conditional subsample transmission algorithm.
Average run time in seconds for each tested grid cell size up to and including the outbreak stages 10 and 10000 (* indicates estimated optimal grid cell size ). The 5th and 95th percentiles are indicated by the ranges (main panels). Each combination of landscape and grid configuration using the CS algorithm, as well as simulations with the pairwise algorithm for comparison was simulated with 500 replicates. The landscapes were (panels A-F): random uniform, random moderate clustering, random high clustering, USA, Sweden, UK. The regular grid construction method was used for the uniform random landscape and the adaptive grid construction method was used for the other landscapes. The black line indicates a unitless relative expected efficacy of the different grid sizes as indicated by the grid optimum estimation method.
Fig 4.
Transmission algorithm performance for estimated optimal grid size .
Run time in seconds for simulations reaching 10, 100, 1000 and 10000 cumulative infected nodes. The grids were constructed using the adaptive gridding method using the estimated optimal grid size for each of the six different landscapes.
Table 4.
Comparison between predicted optimal grid size, , and best actual grid configuration apart from
.
Fig 5.
Comparison of transmission algorithms to the FSR method for outbreaks of up to 10000 infected nodes.
Outliers represent the results from replicates where the epidemic died out early and consequently took considerably less time than larger outbreaks.
Fig 6.
Comparison of transmission algorithms to the FSR method for simulations of outbreaks that were allowed to run their full course.