An investigation of the cloud and convection scheme impacts on tropical clouds and radiation

Description

To understand why GCMs predict up to one order of magnitude different ice water path while maintaining top of atmosphere (TOA) radiation balance, AMIP simulations using GFDL AM2 with high frequency output (every 3 hours) are used to investigate the effect of different cloud and convection schemes on the tropical clouds, precipitation, and TOA radiation balance. Model cloud fraction and ice water content are evaluated using ISCCP, CloudSat, and ARM observations. Simulations using the University of Washington (UW) scheme predict ~4 times larger IWC over the tropics than that using the relaxed Arakawa Schubert (RAS) and compare well with CloudSat and ARM IWC retrievals. By compositing the tropical cloud properties (cloud fraction, IWC, and LWC) by large-scale and convective precipitation respectively, we found the Tiedtke scheme underestimates the cloud fraction near freezing level and around 250 mb. IWC associated with convective precipitation is much smaller than that associated with large-scale precipitation. As a result, the partition of tropical precipitation between large scale and convection significantly regulates the tropical IWC. Using the standalone radiation calculation, the relative contribution of cloud fraction and IWC to TOA radiation flux is evaluated. For example, a doubling of IWC in AM2 RAS simulation induces a net TOA radiation flux of ~1.5 w m-2 in the tropics due to the cancellation of shortwave and longwave radiation. Large scale precipitation efficiency (defined as precipitation divided by the column total condensed water content) and its variation with cloud phase is also investigated. Impacts of prognostic or diagnostic cloud fraction in the Tiedtke scheme are also investigated. Compared with simulations using prognosed cloud fraction, simulations diagnosing cloud fraction using a fixed variance PDF for total water reduce the middle and low-level cloud fraction and large-scale precipitation over the deep tropics. Alternative experiments with the total water PDF variance dependent on surface subgrid inhomogeneity, vertical motion, and height are also conducted to explore the sensitivity of model tropical cloud and hydrological cycle to cloud fraction parameterization.