All operational forecast systems at ECMWF suffer from lower tropical stratosphere cold bias, which has a distinctive resolution dependency. At typical vertical resolutions of the ECMWF Integrated Forecasting System (?450m or ?350m in the lower stratosphere), the lower stratospheric cold bias is increased when the horizontal resolution is increased, while the upper stratospheric warm bias is concomitantly reduced. This is because the stratosphere cools in the global-mean when horizontal resolution is increased. The cooling is due to discretization errors in the vertical advection, associated with inadequate representation of resolved gravity waves in the vertical direction. Although for typical climate model resolutions this problem is negligible, for high resolution numerical weather
prediction systems this is a serious problem. It is shown that an increase in i) the vertical resolution
and/or 2) the order of semi-Lagrangian vertical interpolation reduce the temperature sensitivity to
horizontal resolution via better representation of gravity waves in the vertical direction. It is alternatively shown that filtering grid-point temperature oscillations in the vertical direction alleviates the global-mean cooling of the stratosphere at high horizontal resolutions.
All operational forecast systems at ECMWF suffer from lower tropical stratosphere cold bias, which has a distinctive resolution dependency. At typical vertical resolutions of the ECMWF Integrated Forecasting System (?450m or ?350m in the lower stratosphere), the lower stratospheric cold bias is increased when the horizontal resolution is increased, while the upper stratospheric warm bias is concomitantly reduced. This is because the stratosphere cools in the global-mean when horizontal resolution is increased. The cooling is due to discretization errors in the vertical advection, associated with inadequate representation of resolved gravity waves in the vertical direction. Although for typical climate model resolutions this problem is negligible, for high resolution numerical weather
prediction systems this is a serious problem. It is shown that an increase in i) the vertical resolution
and/or 2) the order of semi-Lagrangian vertical interpolation reduce the temperature sensitivity to
horizontal resolution via better representation of gravity waves in the vertical direction. It is alternatively shown that filtering grid-point temperature oscillations in the vertical direction alleviates the global-mean cooling of the stratosphere at high horizontal resolutions.