Stripe patterns in wind forecasts induced by physics-dynamics coupling on a staggered grid in CMA-GFS 3.0

Abstract. An unphysical stripe pattern is identified in low-level wind field in China Meteorological Administration Global Forecast System (CMA-GFS), characterized by meridional stripes in u-component and zonal stripes in v-component. This stripe noise is primarily confined to the planetary boundary layer over land. The structural mismatch between static field variations and the observed 2Δx noise amplitude suggests that locally forced mechanisms from surface inhomogeneity alone cannot explain the wind stripe patterns. Meanwhile, pure dynamical core simulations exhibit no such noise, confirming that the dynamical core itself does not generate these patterns. These results suggest that staggered-grid mismatch in physics-dynamics coupling is likely the primary mechanism. Idealized two-dimensional experiments demonstrate that combining one-dimensional dynamic-core advection and physics-based vertical diffusion on a staggered grid generates 2Δx-wavelength spurious waves when surface friction is non-uniform. One-dimensional linear wave analysis further confirms that staggered-grid coupling between dynamic advection and inhomogeneous damping forcing induces dispersion errors in wave solutions. Sensitivity tests validate that eliminating grid mismatch in physics-dynamics coupling removes this stripe noise. These findings collectively indicate that while staggered grids benefit the dynamic core's numerical stability and accuracy, their inherent grid mismatch with physics parameterizations requires specialized coupling strategies to avoid spurious noise. Potential solutions to remedy this issue are discussed.