Covid-19 spread: Reproduction of data and prediction using a SIR model on Euclidean network

TL;DR

This paper models Covid-19 spread using an SIR model on an Euclidean network, fitting real data and predicting epidemic duration with high accuracy, emphasizing the importance of spatial dependence.

Contribution

It introduces a spatially-aware SIR model on an Euclidean network that accurately fits Covid-19 data and predicts epidemic resolution times.

Findings

The model fits China's Covid-19 data with high accuracy.

The spatial dependence follows an approximate power law with exponent ~1.85.

The model can predict when the epidemic will be over locally.

Abstract

We study the datafor the cumulative as well as daily number of cases in the Covid-19 outbreak in China. The cumulative data can be fit to an empirical form obtained from a Susceptible-Infected-Removed (SIR) model studied on an Euclidean network previously. Plotting the number of cases against the distance from the epicenter for both China and Italy, we find an approximate power law variation with an exponent $\sim 1.85$ showing strongly that the spatial dependence plays a key role, a factor included in the model. We report here that the SIR model on the Eucledean network can reproduce with a high accuracy the data for China for given parameter values, and can also predict when the epidemic, at least locally, can be expected to be over.