Estimating the Potential Origin of African Swine Fever on the Korean Peninsula: Backcasting North to South Transmission

African swine fever (ASF) is a transboundary viral disease causing severe impacts on national animal health systems in wild and domestic suids. Since its official report in North Korea in 2019, ASF has posed a persistent threat to livestock production, public health, and ecological safety on the Korean Peninsula. Wild boars are recognized as a key reservoir and vector facilitating long-distance spread of ASF, particularly across national borders. However, in North Korea, critical information on outbreak locations, wild boar population density, and transmission pathways remains inaccessible, making risk assessment and preparedness extremely challenging.
   
This study aims to estimate the potential origin and spatial and temporal spread of ASF in North Korea, despite severe data limitations, by following the method from Ko and Cho et al. (2023), applying an agent-based modeling (ABM) and machine-learning framework. In the model, we simulate the wild boar migration and ASF virus transmission. Wild boar sounders were represented as agents whose movement, social structure, reproduction, and contact behaviors were parameterized using ecological and physiological information from the literature-based database. ASF transmission was simulated through local contacts among agents in a spatially explicit landscape, and infection trajectories were tracked over time to estimate transmission pathways and the timing of potential arrival at the Demilitarized Zone (DMZ).
   
Two introduction scenarios were examined based on proximity to reported outbreaks in northeastern China and prior epidemiological evidence: Usi County in Jagang Province, the only officially reported outbreak site in North Korea (Scenario 01), and Hoeryong City in North Hamgyong Province, where suspected early mortality events were reported (Scenario 02). Repeated simulations were conducted for each scenario to identify dominant spread patterns and temporal dynamics.
   
While Scenario 01 successfully reproduces the large-scale southward diffusion pattern toward the DMZ, and Scenario 02 remains constrained mainly by topography, the model fails to capture the short elapsed time from the emergence of ASF in North Korea to its arrival at the DMZ in 2019. This temporal mismatch indicates that, although wild boar-driven spatial spread is plausibly represented, additional mechanisms—such as human-mediated long-distance transmission, earlier widespread circulation before official reporting, or multiple introductions including trade-related pathways—are likely required to explain the observed dynamics.
   
Overall, this study demonstrates how agent-based modeling can be used to reconstruct plausible disease spread scenarios in data-scarce regions and provides insights for prioritizing transboundary surveillance and control strategies along the Korean DMZ.