Assessing the accuracy of kernel-fitting and projections of spatial risk at key early timepoints in outbreaks

Abstract Rapid decision-making during exotic animal disease outbreaks necessitates the early characterisation of transmission dynamics; however, in practice, a strategic balance between accuracy and speed is required. This study presents a simplified and practical approach for field response, leveraging initial kernel estimates from spatial data to predict outbreak risks without the need for high-dimensional data or complex models. Using the 2001 United Kingdom foot-and-mouth disease outbreak as a case study, we investigated the temporal stability of kernel estimates and evaluated their utility for short-term spatial risk stratification during the early phases of an outbreak. Focusing on the five most affected regions, we estimated the optimal kernels on a weekly basis during the first month and monthly thereafter, validating their stability against full-outbreak datasets. Furthermore, we evaluated whether infection pressure derived from weekly updated kernels captured short-term (7-day) spatial patterns of disease spread. Our analysis revealed that the power law type 2 kernel was consistently selected as the best fitting kernel across all study regions. Furthermore, the results demonstrated that reliable kernel estimation became feasible within high-incidence clusters as early as seven days post-notification, whereas low-incidence clusters required a higher cumulative number of infected premises or a longer observation period to achieve stable estimates. Infection pressure derived from the kernels demonstrated strong discriminatory ability for identifying high-risk areas across multiple evaluation approaches, including quintile-based categorisation, ROC analysis, and Cox proportional hazards models. These findings indicate that even with minimal initial data and without complex parameters, it is possible to generate practically valid biosecurity guidelines. The proposed real-time kernel approach offers a practical tool for supporting early-stage surveillance and control decisions by enabling rapid identification of relatively high-risk areas and facilitating efficient allocation of limited biosecurity resources under uncertainty during the early phases of an animal disease outbreak.