Evaluating Tropical Upper-tropospheric Water in Climate Models Using Satellite Data

Abstract: Measuring and simulating moist processes in the tropical upper troposphere are difficult tasks. Humidity in this region of the atmosphere is mainly supplied by deep convection and, problems with simulated convection areknown to be a major contributor to uncertainties in climate model projections. Observations within this region of the atmosphere are hampered by the lowabsolute humidity as well as by the presence of clouds.This thesis examines the seasonal changes in and the effects tropicaldeep convection have on upper-tropospheric water, in addition to its effecton outgoing longwave radiation (OLR). Multiple satellite observations areassessed and used to evaluate the climate models EC-Earth, CAM5, andECHAM6. The data are analysed using two main methods: longterm averagesand compositing. Compositing represents an improvement over climatologiesbecause it brings the comparison closer to the processes associated with deepconvection. The compositing method is adapted from Zelinka and Hartmann[2009], improved, and applied for the first time to climate models.Upper-tropospheric humidity (UTH) undergoes large seasonal and regionalchanges in the tropics. Over land areas, convection is more intense, producinggreater amounts of water at higher heights, and having a greater effect on theOLR. Corresponding model simulations capture the large-scale and seasonalchanges, however there are significant inconsistencies when compared withthe observations, especially over land regions. Simulated mean UTH in areaswhere DC systems develop are consistently higher than observed over bothland and ocean. However, the direct response of UTH to DC systems is foundto be similar to the observations. Modeled cloud fractions near the tropopauseare tend to be overestimated, whereas ice water content is often too low. Theobserved OLR can, regionally, differ from the simulated results by as muchas 20 W m −1 . Moreover, above and around deep convection systems, thelocal decrease of OLR is throughout underestimated. Further, the modelsall demonstrate a lack of spatial variability indicated by a diurnal repetitionof convection at the same location over land. These results obtained by thecomposite method reveal details that could not have been obtained using atraditional climatology based comparison.