Abstract
This paper presents a high-order numerical framework for simulating nonlinear coupled partial differential equations (PDEs) describing bacterial spread and infection. By integrating fourth-order compact finite difference (CFD) schemes with an Alternating Direction Implicit (ADI) approach, the authors achieve significant computational efficiency and point-wise accuracy. A rigorous maximum-norm error analysis is provided, proving the method's stability and convergence. Numerical experiments demonstrate that the ADI compact scheme reduces computation time from days to minutes compared to standard compact methods while maintaining high resolution.
Introduction
Modeling the dynamics of diseases like cholera requires solving complex nonlinear PDEs that track bacterial concentrations and human population densities. Accurate simulations are critical for capturing sharp gradients and ensuring local outbreak detection. Traditional numerical methods often struggle with either computational cost or lack of point-wise error guarantees. This study addresses these challenges by developing a linearized, fourth-order compact ADI method that provides "maximum-norm" error estimates, ensuring reliability at every grid point.
Methodology
The core of the proposed approach lies in several mathematical innovations designed to enhance both accuracy and speed. A pivotal step is the introduction of auxiliary variables $p$ and $q$. This transformation, shown in the system below, linearizes and decouples the original nonlinear equations, facilitating a more straightforward numerical analysis.
The spatial discretization utilizes Compact Finite Difference (CFD) schemes. Unlike standard finite difference methods that require wide stencils for high-order accuracy, CFD achieves fourth-order precision using a minimal set of neighboring points. To handle the 2D spatial domain efficiently, the Alternating Direction Implicit (ADI) method is employed, decomposing the 2D problem into a series of 1D problems. Furthermore, the authors utilize a temporal-spatial error splitting technique to prove unconditional stability, meaning the simulation remains stable regardless of the time step size relative to the grid spacing.
Results
The method was validated through synthetic numerical experiments. Convergence tests confirmed the theoretical predictions of fourth-order spatial and second-order temporal accuracy.
The computational gains were particularly striking. In a $160 \times 160$ grid simulation, a standard compact scheme required over 7 days to complete. In contrast, the ADI-enhanced version reached the same level of accuracy in just 3 minutes. For larger $280 \times 280$ grids, the ADI method finished in 36 minutes, demonstrating its scalability for high-resolution modeling.
The simulations also successfully captured biological phenomena such as bistability. Depending on the initial bacterial load, the system correctly predicted either disease extinction or the transition to a stable endemic state, even under noisy initial conditions.
Discussion and Conclusion
The development of maximum-norm error estimates ($L^\infty$) represents a significant mathematical contribution, as it guarantees accuracy at every individual grid point—a necessity for biological monitoring where local spikes are critical. The decoupling strategy via auxiliary variables proves to be a powerful tool for simplifying complex nonlinear systems without sacrificing precision.
While the current model assumes a constant susceptible population and uses a square domain, future work will focus on extending these methods to complex 3D geometries and more dynamic biological assumptions. This research provides a robust foundation for high-performance epidemic modeling, bridging the gap between rigorous mathematical analysis and practical computational efficiency.
Summary updated