Atrial Fibrillation (AF) is the most common type of cardiac arrhythmia. Most AF-related thrombi originate within the left atrial appendage (LAA). This study investigated the key factors influencing thrombus formation in the LAA using global sensitivity analysis (GSA) based on computational fluid dynamics (CFD) simulations. GSA was conducted to assess the effects of four physiological input parameters: initial thrombin location within the LAA, fibrinogen (Fg) concentration in the blood, sensitivity to activated protein C (K3 constant), and inlet velocity. A total of 160 CFD simulations were performed using a 2D idealized left atrial geometry with the most common LAA morphologies: Cactus (CA), Chickenwing (CW), Windsock (WS), and Broccoli (BR). The area under the curve (AUC) of fibrin, which is a precursor of thrombus formation, was computed in the LAA to quantify net fibrin formation over time. Gaussian Process Emulators (GPE) were trained using the simulations' results to predict the Sobol indices from the input parameters. Fg concentration, initial thrombin location, and their interaction exhibited the largest Sobol indices in all LAA morphologies, impacting both average and maximum AUC. Inlet velocity affected the average AUC in BR, and its interaction with the initial thrombus location was significant for this morphology. Additionally, K3 contributed to the output variance in CW and BR. These findings emphasize the overall significance of Fg concentration and initial thrombin location, along with their interaction, in thrombus formation. The impacts of inlet velocity and K3 concentration appear to be morphology-specific. The distinct values obtained from maximum and average fibrin AUC provide complementary insights into thrombus formation.