Covid_Simulation_Model

Background

This model is developed as part of the COVID-19 RAMP response. It simulates the spread of Covid-19 through a community. In the first instance we envisage a community of a similar scale to the City of Glasgow. The intention is that it stochastically simulates the movements of people in near real time (with a 1 hour resolution) and their interactions and the spread through these mixing populations. We can subsequently simulate the implementation of controls, such as lockdowns, in these populations.

Building and Running

Java SDK

The Covid_Simulation_Model targets Java 8, please ensure you have a Java SDK installed.

Gradle

To compile, run or test this model, Gradle must be installed. This manages the libraries utilised and simplifies the build process.

For MacOS:

We recommend using homebrew.

To install homebrew:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install.sh)"

To install Gradle:

brew install gradle

For Debian, Ubuntu, Mint:

sudo apt install gradle

For RedHat, Centos, Fedora:

yum install gradle

For Windows:

Follow instructions here.

Build/Test Guide

To compile and run the tests:

gradle build

Running a Simulation

The built java project takes three positional commandline arguments. The first is a json file defining the disease and population parameters, the second is a json file defining parameters of a particular run, and the third is the simulation number for HPC array runs (which should be 0 for single runs). Example json files are provided in the parameters folder. Information about the example parameters, including their provenance, is available here.

To run the project:

gradle run  --args "parameters/example_population_params.json parameters/example_model_params.json 0"

Output Directory

The model results will be written to a new directory which is created within the directory specified by the outputDirectory parameter (except in the special case where outputDirectory is set to "", and results are written to the current directory).

RNG Seed

If rngSeed is not included in the input parameters, then a random seed will be generated (and its value logged).

Network generation

It is possible to run the model in an alternative mode, where the virus is not simulated, and instead a network of contacts is generated. To enable this mode set the networkOutputDir parameter. (See the 'network_gen' example json file).

Sample outputs

In the RCode folder, you can find example output data from a simulation of the Lothian Health Board, together with associated graphs (and the R script which was used to generate the graphs).

Configuring Lockdowns

Information on configuration lockdowns may be found in here.

Running in Parallel

This simulations are stochastic (based on rngSeed) and should be run multiple times to establish confidence levels in the results. Information on running multiple simulations on HPC setups (specifically DiRAC) can be found in Running on DiRAC.