Memory-based incremental parameter updating of a generic stochastic plant epidemic model

In plant-disease surveillance, timely and accurate estimation of transmission parameters is critical for informed decision-making. Here, I present a sequential Monte Carlo method that incrementally updates key parameters of a stochastic compartmental epidemic model as new incidence data are collected daily. My approach couples a Gillespie stochastic simulation algorithm for disease epidemiology with a memory-based particle-resampling scheme that allows real-time inference of primary and secondary infection rates even when observed infection counts are low. Using a synthetic outbreak to mimic typical field epidemics, I show that posterior means for transmission parameters converge to within an average of[~] 10% of true values by Days 10-15 post-introduction. Concurrently, ensemble-based short-term forecasts achieve R2[~] 0.8 by Day 2 and exceed R2{approx} 0.93 by Day 30. Computational costs remain modest; each daily update completes in under 4 seconds on standard hardware, highlighting the feasibility of integrating this method into automated surveillance platforms. While validation against synthetic data shows strong performance, I discuss potential challenges in real-world applications, including real data, model misspecification, latent infection dynamics, and spatial heterogeneity. This sequential-Bayesian approach provides a scalable, uncertainty-aware solution for real-time parameter estimation and forecasting in stochastic plant-epidemic systems, laying the groundwork for adaptive management of crop-disease outbreaks.