State-of-the-art methods for modeling wind farm energy production follow a two-step approach: modeling the wind resource and then the farm’s conversion of wind into power. These models carry uncertainty due to both accuracy limitations and the stochastic nature of wind. Probabilistic modeling addresses this by assigning probabilities to different production levels by propagating error distributions. We extend this probabilistic framework to two dimensions by examining the joint probability distribution of energy production at two wind farms in the same region but with different configurations. When operational data from one farm is available, it can inform the expected production of the second, quantifying the uncertainty reduction. This is particularly relevant for repowering projects or developments near existing farms. Existing approaches focus on short-term prediction and rely on operational data, which is unavailable for future farms. Instead, joint production must be estimated using a physics-based a priori approach. While joint distributions of some error sources exist, others remain unassessed, and no model propagates these joint errors to energy production. We address this gap by developing a physics-based framework to propagate joint uncertainties and assess its impact through a case study.