Dynamical modeling and analysis of the impact of zonal prevention and control under normalized management on African Swine Fever transmission in China.

Background: China's first outbreak of African swine fever (ASF) occurred in August 2018, rapidly spreading to all 31 provinces within three months, causing massive losses to the pork industry. After emergency measures such as a nationwide ban on swill feeding and enhanced biosecurity halted the initial epidemic wave, management transitioned to routine zone-based control. However, regional disparities in farm density and inconsistent policy implementation still pose challenges to effective prevention. Static-parameter models commonly used in ASF research fail to capture the dynamic nature of viral spread, and quantitative assessments of zone-specific interventions remain limited. Based on China's field experience, this study investigates how regional heterogeneity drives ASF transmission and quantifies the effectiveness of zonal control strategies, providing an evidence-based foundation for optimizing interventions. Methods: Monthly records of ASF-related deaths and culls from August 2018 to December 2019 were collected to construct a three-dimensional non-autonomous transmission model (swine-swill-environment) with time-varying parameters. A fully connected feed-forward neural network (FNN) was trained using PyTorch to approximate these parameters. Optuna was employed to optimize the network architecture, and a composite loss function was defined for joint parameter estimation. The effective reproduction number was computed in real-time to quantify transmission risk. Global sensitivity analysis was performed using Latin-hypercube sampling combined with partial rank correlation coefficients (PRCC) to evaluate the impact of zone-specific policies and single or combined interventions (swill-feeding ban, biosecurity compliance, safe carcass disposal). Results: Model-based scenario simulations indicate that zone-based control significantly suppressed ASF transmission, with effectiveness influenced by implementation timing and regional balance. Relative to no zoning, initiating zones in September 2018 reduced peak infected pig numbers by 26% and total deaths by 63%. When zoning began in October with a balanced (1:1) farm-density ratio, overall mortality decreased by 30% compared with an unbalanced (4:1) ratio, confirming "early plus balanced zoning" as the optimal strategy. Among single interventions evaluated against an October-initiated 2:1 farm-density baseline, doubling biosecurity awareness reduced total deaths by 62%, whereas intensified safe carcass disposal was positively correlated with mortality. A standalone swill-feeding ban was only marginally effective compared to zoning alone. However, combining all three interventions significantly reduced cumulative deaths to close to 480 thousand heads, demonstrating a clear synergistic effect. Conclusion: Swill feeding and pronounced regional heterogeneity are critical drivers of ASF spread. Combined interventions consistently outperform single measures. An early, balanced zoning strategy coupled with heightened producer biosecurity awareness represents the most effective management approach. These findings provide robust evidence to refine ASF prevention policies in China and serve as a technical reference for global epidemic management.