Abstract: Objectives: Global Navigation Satellite System-Reflectometry (GNSS-R), as a passive bistatic radar configuration, offers unique advantages in high spatiotemporal resolution, cost efficiency, and global signal availability. These merits have propelled its widespread applications across marine remote sensing (e.g., sea surface altimetry, wind speed retrieval) and terrestrial monitoring (e.g., soil moisture estimation, ice sheet characterization). Particularly, GNSS-R-based water surface altimetry has garnered significant scientific attention due to its capacity for non-contact, all-weather monitoring, with extensive investigations conducted globally on coastal zones, inland lakes, and river systems. While existing studies have demonstrated its technological potential, critical challenges persist in conventional shore-based single-antenna implementations, including suboptimal accuracy (typically decimeter-level) and discontinuous observations under dynamic small-scale water conditions. Methods: This study presents a real-time centimeter-level GNSS-R water surface altimetry method employing a top-bottom dual-antenna geometry based on relative positioning principles. By resolving baseline vectors through multi-GNSS constellation combinations (BDS/GPS/Galileo) and leveraging the geometric relationship between the receiver antennas and the water surface, the system achieves continuous inversion of height variations. The methodology was rigorously validated through 12- month continuous operational trials at an instrumented test site. Results: The dual-antenna system achieved a 93.97% mean ambiguity resolution success rate, with altimetric performance characterized by a root mean square error (RMSE) of 2.39 cm and standard deviation (STD) of 2.06 cm. These metrics demonstrate the system's capability for stable long-term monitoring in small-to-medium water bodies, outperforming traditional single-antenna approaches. Multi-constellation integration (BDS/GPS/Galileo) enhanced temporal continuity by 85.1% and 209.2% relative to standalone BDS and GPS+Galileo systems respectively, while simultaneously improving altimetric precision by 17.9% and 69.5%. The combined configuration further reduces environmental dependency for equipment deployment and enhances spatiotemporal resolution in spatially constrained scenarios. Conclusions: The proposed framework enables cost-effective water level monitoring through minimal adaptation of existing GNSS infrastructure. Field validation under meteorological extremes (e.g., storm events) confirms operational robustness, positioning this technology as a viable solution for precision hydrometry in reservoirs, rivers, and lakes.