A community benchmark for 2-D Cartesian compressible convection in the Earth's mantle

Benchmark comparisons are an essential tool to verify the accuracy and validity of computational approaches to mantle convection. Six 2-D Cartesian compressible convection codes are compared for steady-state constant and temperature-dependent viscosity cases as well as time-dependent constant viscosity cases. In general we find good agreement between all codes when comparing average flow characteristics such as Nusselt number and rms velocity. At Rayleigh numbers near 106 and dissipation numbers between 0 and 2, the results differ by approximately 1 per cent. Differences in discretization and use of finite volumes versus finite elements dominate the differences. There is a small systematic difference between the use of the anelastic liquid approximation (ALA) compared to that of the truncated ALA. In determining the onset of time-dependence, there was less agreement between the codes with a spread in the Rayleigh number where the first bifurcation occurs ranging from 7.79 × 10^5 to 1.05 × 10^6.