This thesis is the result of research activity performed from 2005 to 2009 at the Laboratory for Atmospheres of National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (GSFC), under the guidance of Dr. Oreste Reale. The present study describes some fundamental properties occurring in global models in order to spontaneously produce tropical cyclones in the Eastern Tropical North Atlantic (ETNA). The NASA finite-volume General Circulation Model (GEOS-4 and GEOS-5), the NCEP Global Forecast System (GFS) global model, and the European Centre for Medium-Range Weather Forecast (ECMWF) Nature Run outputs were analyzed in a set of diagnostic studies to understand how different global models with different configurations can produce tropical cyclogenesis. Diagnostics performed were classified in two groups according to the experimental configurations: simulations dependent on initializations (GEOS-4, GEOS-5 - version 2.0, and GFS) and free running simulations (ECMWF Nature Run). Objectives methods were developed to identify some controlling factors in the development of tropical cyclones and were applied to all the data generated for this study, focusing mainly on the ETNA. Global model outputs of nine tropical systems during three Atlantic hurricane seasons (2004-2006) were used and compared.

The verification of the Kuo necessary condition for barotropic instability and the kinetic energy transfer across spatial scales are found to be important mechanisms by which tropical cyclogenesis can develop in global models. The main result of this work is that a vertically-aligned barotropically unstable column appears during cyclogenesis, predominantly controlled by the large scale forcing. With the exception of the Nature Run, which is not dependent on initialization, the actual roles of the Data Assimilation System (DAS) and the forecasting system cannot be rigorously separated in these experiments.

Therefore, in the third part of this thesis the impact of initialization was investigated through a diagnostic study of an Observing System Experiment, previously performed at NASA. It is shown in this work that the assimilation of Atmospheric Infrared Sounder (AIRS) temperature retrievals under partially cloudy conditions in the DAS can significantly improve a less than optimal initialization and consequently tropical representation in the forecast. In particular, AIRS retrievals allow the GEOS-5 DAS to create more confined and deeper circulations, in agreement with observations. As a consequence of the improved initialization, forecast track improves as well. The conclusion of this thesis suggests that the capabilities of data assimilation and forecasting systems have to advance side by side in order to improve tropical cyclone forecasting.