Adjoint Methods for Guiding Adaptive Mesh Refinement in Wave Propagation Problems
For researchers using Clawpack/GeoClaw for wave propagation where only a target solution area is needed, this method reduces computational cost while preserving accuracy.
The authors integrated adjoint methods into adaptive mesh refinement (AMR) for wave propagation problems, achieving significant computational time reduction while maintaining solution accuracy. In acoustics and tsunami examples, the adjoint-guided AMR refined only regions influencing a target area, reducing computational cost.
One difficulty in developing numerical methods for hyperbolic systems of conservation laws is the fact that solutions often contain regions where much higher resolution is required than elsewhere in the domain, particularly since the solution may contain discontinuities or other localized features. The Clawpack software deals with this issue by using block-structured adaptive mesh refinement to selectively refine around propagating waves. For problems where only a target area of the total solution is of interest, a method that allows identifying and refining the grid only in regions that influence this target area would significantly reduce the computational cost of finding a solution. In this work, we show that solving the time-dependent adjoint equation and using a suitable inner product with the forward solution allows more precise refinement of the relevant waves. We present acoustics examples in one and two dimensions and a tsunami propagation example. To perform these simulations, the use of the adjoint method has been integrated into the adaptive mesh refinement strategy of the open source Clawpack and GeoClaw software. We also present results that show that the accuracy of the solution is maintained and the computational time required is significantly reduced through the integration of the adjoint method into AMR.