Abstract/Summary

Accurate representation of near-tropopause fields is important for the forecast skill of numerical weather prediction models, yet there remain significant systematic forecast errors in the region of the tropopause. Although extratropical near-tropopause humidity, temperature, and wind model biases have been documented from several models, more knowledge of their causes is required as a step towards reducing these biases. Typically, a moist bias is present in the lowermost stratosphere in the analyses used to initialise forecasts, which leads to a growing cold bias in the lowermost stratosphere over the course of the forecasts due to long-wave radiative cooling. Experiments are conducted with the European Centre of Medium-Range Weather Forecasts global forecast system where the humidity in a layer 0–4 km above the tropopause in the extratropics is reduced to correct the moist bias in the initial conditions. In these experiments, the lowermost stratosphere cold bias growth is halved compared with the control and gradually remoistens, returning to typical analysis values with a half-life of around 8–9 days. The reduction in cooling in the lowermost stratosphere is due to a reduction in long-wave radiative emission from the water vapour above the tropopause. The main contributors to the remoistening are resolved advective transport and parametrised turbulent mixing in the model, the cloud microphysical process rates being similar in the modified and control experiments. The biases in near-tropopause moisture transport are almost independent of horizontal resolution. These results show that the temperature bias in the extratropical lower stratosphere can be reduced by correcting the collocated moist bias, but the moist bias cannot be fixed by solely correcting the initial conditions and further model improvements are also required to reduce cross-tropopause moisture transport.