|dc.description.abstract||Cloud properties and their vertical structure are important for meteorological studies due to their impact on both the Earth's radiation budget and adiabatic heating. Examination of bulk cloud properties and vertical distribution simulated by the National Center for Environmental Prediction (NCEP) Global Forecast System (GFS) using various satellite products and ground-based measurements is a main objective of this study. Cloud variables evaluated include the occurrence and fraction of clouds in three layers, cloud optical depth, liquid water path, and ice water path. Cloud vertical structure data are retrieved from both active and passive sensors that are compared with GFS model results. In general, the GFS model captures the spatial patterns of hydrometeors reasonably well and follows the general features seen in satellite measurements, but large discrepancies exist in low-level cloud properties. More boundary layer clouds over the interior continents were generated by the GFS model whereas satellite retrievals showed more low-level clouds over oceans. The GFS model simulations also missed low, shallow stratocumulus clouds along the west coast of North America, South America, and southwestern Africa and overestimated thick, large-scale clouds associated with the Asian summer monsoon. Although the frequencies of global multi-layer clouds from observations are similar to those from the model, latitudinal variations show large discrepancies in terms of structure and pattern. The modeled cloud optical depth for optically thin or intermediate clouds is less than that from passive sensor and is overestimated for optically thick clouds. The distributions of ice water path (IWP) agree better with satellite observations than do liquid water path (LWP) distributions.
Mistreatment of such stratocumulus clouds in the GFS model leads to an overestimation of upward longwave flux, and an underestimation of upward shortwave flux at the top-of-atmosphere (TOA). With respect to input data bias in cloud fields, the GFS temperature is comparable with satellite retrievals and ground-based measurements, but the GFS relative humidity shows a wet bias at 150 and 850 hPa both from satellite retrievals and ground-based measurements. Discrepancies in cloud fields between observations and the model are attributed to differences in cloud water mixing ratio and mean relative humidity fields, which are major control variables determining the formation of clouds.
To improve the simulation of cloud fields, application of other cloud parameterization scheme to the GFS model is performed. The new scheme generates a large quantity of marine stratocumulus clouds over the eastern tropical oceans as well as low cloud amounts in the other regions. High-level and middle-level clouds generated from the new scheme are more comparable with the satellite retrievals in terms of the spatial distributions and zonally averaged cloud fractions.
An application of a simple linear relationship between de-correlation lengths (Lcf) and latitudes to the GFS model is conducted in order to see how successfully the equation explains the characteristics of cloud vertical structure on the changes in cloud fraction at different vertical levels. The method to solve for Lcf is a combination of Brent (1973) approach and a stochastic cloud generator using data collected from space-borne active sensors. Cloud fractions derived from a simple linear fit are compared to those computed from Lcf values based on observations. The pattern of zonal Lcf values from a simple linear fit is quite different from that of Lcf values based on observations. An offset pattern in subtropical regions is notable. The distribution of median Lcf values calculated from observed clouds do not show much dependence on latitude. This suggests that other physics, such as convection and cloud formation mechanism rather than simply latitude, should be considered when explaining how Lcf behaves. Such findings are expected to help improve the inherent problems of the GFS cloud parameterization scheme and to gain insight into the method used in determining cloud fraction.||en_US