Stratospheric Analysis and Forecast Errors Using Hybrid and Sigma Coordinates

Abstract

Abstract Past investigations have documented large divergent wind anomalies in stratospheric reanalyses over steep terrain, which were attributed to discretization errors produced by the terrain-following (sigma) vertical coordinate in the forecast model. However, forecasting experiments have reported negligible differences in skill between sigma- and hybrid-coordinate models. This leads to the paradoxical conclusion that discretization errors in the forecast model yield significant stratospheric analysis errors, but insignificant stratospheric forecast errors. The authors reexamine this issue by performing two forecast-assimilation experiments that are identical except for the vertical coordinate: one uses a sigma coordinate and the other uses a hybrid coordinate. The sigma-coordinate analyses exhibit large divergent wind anomalies over terrain that extend from the surface to the model top and distort explicitly resolved orographic gravity waves. Above the tropopause, divergent wind errors are suppressed by an order of magnitude or more in the hybrid-coordinate analyses. Over a 3-month period, stratospheric skill scores in the hybrid experiment show statistically significant improvements relative to the sigma experiment. Previous studies, which found no such differences, all used forecasts initialized from a common archived analysis. The results show that the dominant pathway for error growth and net skill impacts is via 0–9-h forecast backgrounds cycling successively through the data assimilation phase without significant observational correction. The skill impacts noted here should further motivate weather and climate models to adopt a hybrid coordinate with the best error suppression characteristics for a given modeling application.