Backend
A Second-Order Semi-Implicit Projection Time-Stepping Scheme for the Shallow-Water Equations on the Sphere Using Compatible Finite Elements
arXiv cs.NA0 views
arXiv:2611.08169v1 Announce Type: new Abstract: We propose a second-order, semi-implicit time-stepping scheme for the shallow-water equations on the sphere discretised with compatible finite elements. The scheme splits the advection operator from the wave operator, applying a DG-upwind SSPRK3 method to the advection part and an $H(\text{div})$-conforming implicit discretisation to the wave part, solved efficiently via hybridisation. The scheme follows a three-step symmetric splitting: an explicit half-step, a full advection step, and an implicit half-step that simultaneously projects the velocity back into the $H(\text{div})$ space. This symmetry guarantees global second-order accuracy and confines the stability restriction to the CFL condition of the advection step alone. The time-step can therefore be extended beyond this limit by subcycling the advection step without additional implicit solves. The method is implemented in Firedrake and validated against a standard suite of spherical shallow-water benchmarks, demonstrating stable results for Courant numbers up to 1 (with two advection subcycles) with second-order finite elements, second-order convergence, and good parallel scaling.
Source:https://arxiv.org/abs/2611.08169PreviousSigmoid Tensor Decomposition for Compact Representations of Binary DataNextUniversal Dimension Formula for Planar $C^r$ Splines at Degree $3r+1$
Related posts
Terminal-Bench: Why Agent Benchmarks Need Continuous Integration to Stay Valid
0dev.to
GitHub's New Git Storage Architecture: How Thousands of Agents Push to One Repo Without Collisions
0dev.to
Before Giving an AI Agent Access to Your Codebase, Check Its Permissions
0dev.to
The Chainalysis Breach: From "No Credentials Accessed" to "All Users Exposed" in Five Days
0dev.to

Comment trouver des extraits forts dans une transcription avec une fenêtre glissante en Python
0dev.to
The hardest websites to scrape in 2026: what the benchmarks show
0dev.to