Prediction of flu epidemic activity with dynamical model based on weather forecast

TL;DR

This paper presents a weather-dependent SIRS model that accurately predicts flu activity, enabling short-term epidemic forecasts by linking temperature variations to disease dynamics.

Contribution

It introduces a novel SIRS model with variable reaction rates based on temperature, providing analytical insights into epidemic resonance and improving predictive accuracy.

Findings

SIRS model reproduces real flu activity curves effectively.

Temperature-dependent reaction rates enhance prediction accuracy.

Analytical explanation of epidemic resonance phenomena.

Abstract

The seasonality of respiratory diseases (common cold, influenza, etc.) is a well-known phenomenon studied from ancient times. The development of predictive models is still not only an actual unsolved problem of mathematical epidemiology but also is very important for the safety of public health. Here we show that SIRS (Susceptible-Infected-Recovered-Susceptible) model accurately enough reproduces real curves of flu activity. It contains variable reaction rate, which is a function of mean daily temperature. The proposed alternation of variables represents SIRS equations as the second-order ODE with an outer excitation. It reveals an origin of such predictive efficiency and explains analytically the 1:1 dynamical resonance, which is known as a crucial property of epidemic behavior. Our work opens the perspectives for the development of instant short-time prediction of a normal level of flu activity based on the weather forecast, and allow to estimate a current epidemic level more precisely. The latter fact is based on the explicit difference between the expected weather-based activity and instant anomalies.