Figure 8. The effect of SOAs on epidemic size – source of infection.
Each data point is the number of infected holdings per seed, generated from 400 stochastic simulations of FMD epidemics starting at that time of year, each with five seeds. The model keeps track of how each holding becomes infected, which can be via one of three routes: 1) Movements; 2) Local spread; and 3) Intra-SOA spread. The grey line shows the total number of infected holdings per seed for the normal Control model, with no SOAs or linked holdings. The grey line is therefore composed of Movement and Local spread infection events only. The red line shows the number of Movement and Local spread infection events (per seed) for the SOA model, which increases the size of epidemics throughout the year. However, the green line shows all (Movements+Local+intra-SOA) infection events (per seed) for the SOA model. As can be seen, the biggest contributor to the increase in epidemic size as a result of including SOAs is intra-SOA spread itself– other premises within the SOA becoming infected. Epidemic sizes are measured in terms of number of infected holdings – not number of infected animals.
doi:10.1371/journal.pone.0035089.g008
Another measure of risk is the geographical spread of the epidemic. In the model, GB is subdivided into a grid consisting of squares of 100 km2, where each square contains the relevant holdings as defined by their easting/northing coordinates. The number of grid squares with at least one infected holding at the end of the simulation is a proxy for the extent of geographical spread. The imposition of distance limits to constrain intra-SOA spread has little effect on geographical spread (Figure 7), indicating that long range jumps infecting new parts of the country are rare, and that the increased geographical spread is predominately local with neighbouring grid squares becoming infected. Therefore, under the worst-case scenario (one infected – all infected) of intra-SOA spread, although an increase in logistical (veterinary, slaughter teams etc.) resources may likely be required to handle any increase in the number of infected premises, these resources would not be required in a much larger geographical area as a result of intra-SOA spread.
Modelling – CTS Links
Although CTS Links form dyads, numerous holdings are involved in more than one link. Therefore, CTS Links can be amalgamated together to form CTS Chains, which can be imported into the model and treated as sets of holdings equivalent to SOAs. The inclusion of CTS Chains alone (without SOAs) into the model does substantially increase epidemic size (Figure 9), although CTS Chains have less of an effect on epidemic size than SOAs. Furthermore, CTS Chains substantially increase the geographical spread of the epidemic (Figure 9), resulting in smaller but more geographically dispersed epidemics when compared to SOAs. As in the SOA model, we then applied distance limits to intra-CTS Chain spread of 50 km, 16 km, and 8 km (Figure 9). In contrast to SOAs, the distance limits are important, with the 50 km limit having a large impact on epidemic size, and the 16 km and 8 km limits reduce epidemic sizes still further (Figure 9). However, increased epidemic sizes are still observed even with severe distance limits. The geographical spread of epidemics when CTS Chains are incorporated into the model was also examined. The distance limits again have more of an effect on geographical spread when compared to SOAs (Figure 9), with the 8 km limit reducing the geographical spread to close to that observed in the control model with no linked holdings at all. This again highlights the fact that holdings linked via CTS Links are more geographically dispersed than those linked via SOAs.