Ocean waves are a fundamental component of marine dynamics, exerting significant impacts on maritime navigation and offshore operations. The wave spectrum provides an effective representation of the statistical distribution of wave energy across the frequency domain. In this study, observational data from 70 NDBC buoy stations were used to retrieve significant wave heights based on both the JONSWAP spectrum and the third-order Stokes nonlinear spectrum. Results show that, compared with the JONSWAP-based retrievals, the nonlinear spectrum achieves average improvements degree exceeding 10% in both absolute and relative errors, with maximum improvements degree of 28.54% and 22.29%, respectively, demonstrating the nonlinear spectrum’s clear advantages. Further analysis indicates that the performance degrees of the nonlinear spectrum are closely related to significant wave height, wind speed, the angle between wind and wave directions, and water depth. Specifically, the improvement degree increases with larger wave heights and stronger winds; smaller directional angle between wind and waves yields greater benefits, though wind speed exerts a much stronger influence than directional angle. In water depths shallower than 500 m, the inversion accuracies of both wave spectras are higher than those in depths deeper than 500 m. However, within the depth range of 500-5 500 m, the improvement degree in the nonlinear spectrum exceeds that in the shallow-water region, showing a linear increasing trend.