Systematic biases of coupled ocean-atmosphere models in the tropics directly affect their ability to simulate El Niño, Walker and Hadley circulations, thereby compromise their ability to predict climate on continental and decadal/century time scales. Among them, double ITCZ (Inter-Tropical Convergence Zone) is one of them. It refers to a systematic model bias characterized by two bands of precipitation spuriously symmetric to the equator in the central Pacific. Despite heroic efforts to correct this bias in the modeling community, its cause is still not clear and a cure does not exist. To date, it remains a prominent error in many state-of-the-art coupled models such as the Community Climate System Model (CCSM). Our proposed research aims to understand and reduce systematic model biases in the tropics with a specific focus on the double ITCZ problem. The objective is to understand the physics of double ITCZ in coupled models and then to design numerical algorithms to correct it. We propose a hypothesis on a positive feedback process that drives coupled models into a double ITCZ state in the tropical Pacific. Key elements of the feedback are the surface wind curls/divergences and the heat transport by ocean currents. To test the hypothesis, a series of questions will be investigated that are germane to the understanding and improvement of coupled atmosphere-ocean models, including development of new physical and numerical algorithms. Expected results of the proposed research include improved coupled models for climate predictions, and more fruitful exploitation of state-of-the-art computational facilities to address the greenhouse problem in the Earth’s system.