Understanding how ship characteristics and operating conditions impact underwater radiated noise (URN) is key to identifying essential avenues for noise reduction and predict associated risks for the marine life. Previous studies have revealed that current models could only explain up to 50% of the observed variability in URN levels. In this study, we used a unique dataset, collected within the Marine Acoustic Research Station (MARS) project (www.projet-mars. ca/en). The dataset is consisting of ~1000 acoustic signatures from vessels in the St. Lawrence shipping lane (eastern Canada), and 174 high-quality signatures from partner vessels following an optimized measurement protocol and considering design parameters, meteorological, and oceanographic data. Applying functional regression, as described by MacGillivray et al. (2022), we quantified the relationship between vessel characteristics, operation conditions and URN. Our findings quantify the direct effect of the speed, size, and draft on URN across a frequency range of 10-500 Hz. Using these relationships, we propose a tailored URN predictive model representative of the St. Lawrence fleet. We then compare its performance to previously published models and asses the gains linked to the active collaboration of ship owners, contributing to the wide range of available operation conditions available in our dataset.