On the relationship between tropical convection and clear sky upper troposphere moisture

Upper troposphere water vapor is a crucial factor in the climate system. The moisture abundance in the clear sky upper troposphere is strongly influenced by tropical deep convection. TOVS upper troposphere water vapor measurements and ISCCP deep convective clouds are used to investigate the relationship between upper troposphere humidity (UTH) and frequency and area coverage of tropical deep convection. The study shows that a more diffused distribution of convection in the tropics is associated with a higher UTH in the free troposphere, while a more concentrated convection distribution is associated with a drier free upper troposphere. It is also shown that increased convective area coverage is accompanied with an increased area coverage of less frequent convection and a decreased area coverage of highly frequent convection, and vice versa. The identification of the geographical distribution of convection perturbation associated with a potential climate change is therefore important to the upper troposphere water vapor feedback.; The mechanism behind the observed relationship is interpreted using trajectory analysis. The results show that a more diffused distribution of convection gives rise to a shorter average horizontal distance and a less vertical displacement between convection and clear regions, and hence a moister clear sky condition. Similarly, when deep convection is more concentrated, the average traveling distance is longer, and the clear sky upper troposphere is drier.; This process is verified using a simple heating driven circulation model with prescribed distributions of convective latent heating. The NCAR CCM is then used to examine the relationship between UTH and convection. The model is found to capture the relationship between water vapor concentration and total convective area reasonably, but it fails to reproduce the respective contributions from highly and less frequent convections. The deficiency lies in the fact that convection frequency is broadly overestimated in the model. The area coverages of different convective regimes are not simulated correctly. The increase of total spatial coverage of deep convection in the model is mainly reflected in the increased area coverage of highly frequent convection, leading to an opposite relationship between UTH and highly frequent convection compared with the observations.