Background Seasonal influenza annually creates a serious burden for people worldwide. The influenza surveillance system in China primarily indicates the intensity of the influenza epidemic trend, but it cannot provide accurate forecasts and early warnings. Meteorological conditions and antigenic variation rarely participated in optimizing the influenza model, which not only limits the simulation and prediction of influenza trends but also the development of reasonable preventive measures. Methods The weekly influenza-like illness percentage (ILI%) data from Baoji (BJ) and Qinhuangdao (QHD) were collected from the Chinese Center for Disease Control and Prevention. Data on meteorological factors were used from the China Meteorological Administration spanning from 1 January 2010 to 31 December 2019. The SIRS model, driven by temperature and relative humidity data, was employed to simulate and predict seasonal influenza over multiple years in two cities of Northern China: Baoji (Shaanxi Province) and Qinhuangdao (Hebei Province). Results Our model simulations indicated that both raw and smoothed meteorological data could be used to capture multi-year seasonal influenza trends. When antigenic variation was not incorporated, simulations based on raw data and those based on smoothed data yielded consistent performance within each city: in Baoji (RMSE_S = 1.59, <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\({R}_{s}^{2}\)</EquationSource> </InlineEquation>=0.45) and Qinhuangdao (RMSE_S = 0.5, <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\({R}_{s}^{2}\)</EquationSource> </InlineEquation>=0.56). Critically, incorporating antigenic variation was associated with a marked improvement in model performance. This improvement was evidenced by two key findings: in Baoji, this inclusion enabled a better representation of the 2016 peak and fitted the model more closely to the observed incidence in the following two seasons; in Qinhuangdao, it led to a more accurate prediction of the 2018 outbreak peak (raw data: RMSE_P: 0.55 < 0.67, <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\({R}_{p}^{2}\)</EquationSource> </InlineEquation>: 0.48 > 0.24; smoothed data: RMSE_P: 0.47 < 0.67, <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\({R}_{p}^{2}\)</EquationSource> </InlineEquation>: 0.62 > 0.24). Conclusions This study demonstrates that integrating antigenic variation data with meteorological drivers is critical for accurate influenza forecasting in temperate China. These findings provide a mechanistic basis for proactive surveillance. By combining real-time meteorological and genomic data, such systems would enhance early warning and help optimize the timing of interventions like targeted vaccination.

Modeling the dynamics of seasonal influenza in response to meteorological conditions and antigenic variation: a simulation study

期刊论文
出版模式：

开放获取
发表日期：

2026年2月27日

查看全文

Yinglin Liang,
Zhaobin Sun,
Wei Hua,
Hsiang-Yu Yuan,
Wenxi Ruan,
Demin Li,
Zining Yuan

摘要

Background

Seasonal influenza annually creates a serious burden for people worldwide. The influenza surveillance system in China primarily indicates the intensity of the influenza epidemic trend, but it cannot provide accurate forecasts and early warnings. Meteorological conditions and antigenic variation rarely participated in optimizing the influenza model, which not only limits the simulation and prediction of influenza trends but also the development of reasonable preventive measures.

Methods

The weekly influenza-like illness percentage (ILI%) data from Baoji (BJ) and Qinhuangdao (QHD) were collected from the Chinese Center for Disease Control and Prevention. Data on meteorological factors were used from the China Meteorological Administration spanning from 1 January 2010 to 31 December 2019. The SIRS model, driven by temperature and relative humidity data, was employed to simulate and predict seasonal influenza over multiple years in two cities of Northern China: Baoji (Shaanxi Province) and Qinhuangdao (Hebei Province).

Results

Our model simulations indicated that both raw and smoothed meteorological data could be used to capture multi-year seasonal influenza trends. When antigenic variation was not incorporated, simulations based on raw data and those based on smoothed data yielded consistent performance within each city: in Baoji (RMSE_S = 1.59, \({R}_{s}^{2}\) =0.45) and Qinhuangdao (RMSE_S = 0.5, \({R}_{s}^{2}\) =0.56). Critically, incorporating antigenic variation was associated with a marked improvement in model performance. This improvement was evidenced by two key findings: in Baoji, this inclusion enabled a better representation of the 2016 peak and fitted the model more closely to the observed incidence in the following two seasons; in Qinhuangdao, it led to a more accurate prediction of the 2018 outbreak peak (raw data: RMSE_P: 0.55 < 0.67, \({R}_{p}^{2}\) : 0.48 > 0.24; smoothed data: RMSE_P: 0.47 < 0.67, \({R}_{p}^{2}\) : 0.62 > 0.24).

Conclusions

This study demonstrates that integrating antigenic variation data with meteorological drivers is critical for accurate influenza forecasting in temperate China. These findings provide a mechanistic basis for proactive surveillance. By combining real-time meteorological and genomic data, such systems would enhance early warning and help optimize the timing of interventions like targeted vaccination.