Title:	LivEpiMod
DNr:	SNIC 2020/5-35
Project Type:	SNIC Medium Compute
Principal Investigator:	Tom Lindström <tom.lindstrom@liu.se>
Affiliation:	Linköpings universitet
Duration:	2020-02-01 – 2021-02-01
Classification:	10699
Keywords:

Abstract

This project for Livestock Epidemiological Modeling (LivEpiMod) focus on Transboundary animal diseases (TADs), which are a major threat to agricultural system potentially affecting food security and economy. Many tools used to understand potential disease spread and the effect of response actions on that spread inherently require an underlying model. Our models, the Animal Movement Model (AMM) and the Disease Outbreak Simulation (DOS), provide a rigorous, quantitative predictions about cattle shipment and the spread, size, duration and spatial risk of an FMD, or other animal disease, outbreak at the national level for planning and response purposes. Our team has significant expertise and previous successes with livestock shipment and TADs spread modeling. AMM is a Bayesian hierarchical model and uses sampled data on cattle shipments in a MCMC algorithm to predict unobserved cattle shipments in space and time at the national level. This technique is an efficient method yet with heavy run time costs when processing large amount of data. DOS is a simulation model using a gridding algorithm to reduce run time costs yet a large outbreak do have significant run time costs and to fully use the model and test scenarios one have to make a huge amount of replicates including initially infected farms as well as realization for each and test of different control strategies. AMM parallelization will be implemented as shared-memory programming while DOS parallelization will be set as of replicated runs. Here, we will also refine AMM prediction to include the size and type of premises shipping and receiving in addition to the shipment size. These refinements will facilitate more detailed risk assessment, application to a wider range of diseases, and will be an appropriate input for a wider range of disease models. We propose to demonstrate the utility of AMM by using its predictions to determine the geographic areas that ultimately feed into slaughter surveillance and to recommend allocation of slaughter surveillance for improved geographic coverage. This information can inform surveillance plans to maximize early or first detection as well as improving surveillance for endemic diseases. DOS is a fully premises-based TAD model with two transmission routes: long range due to shipments informed by AMM and local spread using a spatially implicit, density-dependent kernel parameterized from published foot and mouth disease outbreak data. We use sensitivity analysis to address likely variation from published parameters. We extend DOS to take inputs from AMM and further to incorporate within-premises dynamics in order to test their importance at the national scale.