Abstract: We investigate the performance of cloud convection and tracer transport in a global off-line 3-D chemical transport model. Various model simulations are performed using different meteorological (re)analyses (ERA-40, ECMWF operational and ECMWF Interim) to diagnose the updraft mass flux, convective precipitation and cloud top height.
The diagnosed upward mass flux distribution from TOMCAT agrees quite well with the ECMWF reanalysis data (ERA-40 and ERA-Interim) below 200 hPa. Inclusion of midlevel convection improves the agreement at mid-high latitudes. However, the reanalyses show strong convective transport up to 100 hPa, well into the tropical tropopause layer (TTL), which is not captured by TOMCAT. Similarly, the model captures the spatial and seasonal variation of convective cloud top height although the mean modelled value is about 2 km lower than observed.
The ERA-Interim reanalyses have smaller archived upward convective mass fluxes than ERA-40, and smaller convective precipitation, which is in better agreement with satellite-based data. TOMCAT captures these relative differences when diagnosing convection from the large-scale fields. The model also shows differences in diagnosed convection with the version of the operational analyses used, which cautions against using results of the model from one specific time period as a general evaluation.
We have tested the effect of resolution on the diagnosed modelled convection with simulations ranging from 5.6° × 5.6° to 1° × 1°. Overall, in the off-line model, the higher model resolution does not make a large change to the diagnosed convective tracer transport. Similarly, the resolution of the forcing winds in the higher resolution CTM does not make a large improvement compared to the archived mass fluxes.
Including a radon tracer in the model confirms the importance of convection for reproducing observed midlatitude profiles. The model run using archived mass fluxes transports significantly more radon to the upper troposphere but the available data does not strongly discriminate between the different model versions.