Network Real-Time Kinematic (NRTK) positioning is widely recognized for centimetre-level accuracy and operational efficiency, but conventional NRTK systems rely on reference stations anchored to bedrock infrastructure and are therefore limited to terrestrial areas. To overcome this constraint, this study proposes an NRTK framework for mobile platforms featuring simultaneous estimation of atmospheric delays and baseline dynamics, regularization to decorrelate positional and atmospheric parameters with coefficients optimized by mean square error minimization, and integration of precise point positioning at a main base station to maintain an absolute position reference. Validation using Hong Kong's terrestrial continuously operating reference station network demonstrates positioning accuracy comparable to conventional bedrock-based NRTK, with three-dimensional east-north-up errors of (2.90, 3.22, 4.32) cm and (2.90, 2.88, 6.70) cm in two operational scenarios. The technique enables buoy-based NRTK systems for marine applications such as port traffic management, fishing fleet navigation, and offshore resource exploration.