2.6 Community detection
In the state of SC, epidemiological surveillance and control units are organized at various levels starting at the municipality-level. For that reason, we aggregate the movements made by farms in each municipality, considering the municipality to be a node. This analysis captured spatial connectivity amongst different municipalities via animal movements, including pairs of locations that shared common direct and indirect connections within the network. Community detection was performed using an algorithm based on the calculation of LinkRank, a concept derived from Google’s PageRank (BRIN; PAGE, 1998) that takes into account the incoming links of a given municipality and the incoming links of its neighbors (GRISI-FILHO et al., 2013).
We also calculated modularity as a measure of the strength of community structure, where values of 0.3 – 0.7 generally indicate moderate to strong community structure (GRANOVETTER, 1973).
All analysis and data visualization were performed in R statistical software (R CORE TEAM, 2017) version 2.4.2 using the packages: igraph version 1.1.2 (NEPUSZ, 2006), rgdal version 1.2‐8 (KEITT et al., 2018), circlize version 0.4.1 (GU et al., 2014), EpiContactTrace version 0.12.0 (NÖREMARK; WIDGREN, 2014), tidyverse (Hadley Wickham, 2017), sf (PEBESMA, 2018), and the Illustrator CC5 software (GOLDING, 2017). Shapefiles to produce the maps were obtained from the Instituto Brasileiro de Geografia e Estatística(www.ibge.gov.br).
Results
Data
The movements included in this study compromised 99.9% of the total available records. A small proportion of records were removed because the number of animals was listed as zero animals or >20000, which were considered errors. In this swine trade network, there were 11,606 premises, and 6,0645,853 pigs involved in 320,814 movements during the period of study.
Data trade description
Within SC, the dominant trade flow of animals originated from breeding-farms and CSBF and passed through Nurseries to Finisher farms. Finishers received a significant amount of movements from Nursery, Breeding and CSBF premises and, sent the largest flow of movements to Slaughterhouses.
We illustrated the main movements within and between types of premises in Figure 2; for visualization purposes, atypical movements representing less than 1 % of the total of possible connections were excluded (Figure 2). The full description is available in the supplementary table 1.