The design of low-boom supersonic aircraft requires robust uncertainty quantification frameworks to account for variabilities in flight conditions and atmospheric properties. Multi-fidelity methods provide a computationally efficient strategy by leveraging inexpensive low-fidelity simulations with expensive-togenerate high-fidelity data. However, as the dimensionality of the design space increases with additional uncertain variables, conventional multi-fidelity approaches such as Co-Kriging suffer from exponential growth in computational cost with increasing dimensionality. In this study, a multi-fidelity stochastic framework is developed to evaluate the operational robustness of a low-boom supersonic airliner configuration under operational and atmospheric uncertainties. The baseline geometry is a 40-passenger supersonic transport derived from Li et al. [1] and further modified in our previous work [2]. High-fidelity near-field pressure signatures are computed using the open-source SU2 CFD solver [3], while low-fidelity aerodynamic solutions are obtained through A502 PANAIR [4], a linearized supersonic potential flow solver. Ground-perceived sonic boom loudness is calculated using ITUBOOM [5], an atmospheric propagation code developed at the ITU AeroMDO Lab. To mitigate surrogate modeling challenges associated with shock-induced pressure discontinuities, near-field signatures are transformed into equivalent area distributions, which serve as the primary modeling parameters. Two multi-fidelity datasets are constructed: an initial lower-dimensional case considering operational uncertainties (e.g., Mach number and angle of attack), and a higher dimensional case incorporating atmospheric uncertain variables (altitude, temperature, humidity, and wind profiles in x and y directions) based on atmospheric statistics from Rallabhandi et al. [6]. For uncertainty quantification, the recently proposed two-step Multi Fidelity Support Vector Regression (2-step MFSVR) method from the authors’ previous work [7] is evaluated against conventional Co-Kriging approaches on these datasets. The 2-step MFSVR method is specifically designed to handle high-dimensional spaces while capturing nonlinear relationships between fidelity levels. Input uncertainties are propagated to obtain probability distributions of boom signature metrics, while Total Sobol’ indices quantify each parameter’s contribution to output variability and robustness.