Search

We examine simulations of today's climate performed with a global atmospheric general circulation model run at spectral truncations of T42, T170, and T239, corresponding to grid cell sizes of roughly 310 km, 75 km, and 55 km, respectively. The simulations were forced with observed sea-surface temperatures and sea-ice concentrations. The T42 simulations and initial simulations at T170 and T239 were performed using a model version that was carefully "tuned" to optimize results at T42; subsequent simulations at T170 and T239 used a model version that was partly re-tuned to improve results at T170. On the scales of a T42 grid cell and larger, nearly all quantities we examined in all the T170 and T239 simulations agree better with observations, at least in terms of spatial patterns, than in the T42 simulations. In some cases the improvements are very substantial. Improvements are seen in all-season, global domain results, and in results pertaining to most seasons and latitude bands. Increasing the model resolution from T42 introduces biases (errors in the mean) into some simulated quantities; the worst of these were removed by the partial retuning we performed at T170. This retuning has little effect on the spatial patterns of results, except in Northern Hemisphere winter at T170, where it tends to bring improvements. We discuss aspects of simulated regional climates, and their dependence on model resolution. [ABSTRACT FROM AUTHOR]