Abstract
Introduction To enhance the simulation performance of wave numerical models, high-precision ocean models are widely utilized. However, the low efficiency of high-precision numerical computation remains one of the key bottlenecks hindering the advancement of wave forecasting. Methods To address this issue, this study introduces a mixed-precision framework based on variable-specific precision allocation, applied to the MArine Science and Numerical Modeling (MASNUM) ocean wave model, considering physical sensitivities. Results The results demonstrate that by strategically reducing the precision of non-critical variables to single-precision (float32) or half-precision (float16), the mixed-precision scheme significantly improves computational efficiency while maintaining the accuracy of the simulation results. Specifically, compared to the double-precision baseline, the mixed-precision approach results in minimal accuracy loss, with SMAPE values for significant wave height ranging between 0.12% and 0.43%, and RMSE ranging from 0.01 m to 0.02 m. Discussion In terms of computational performance, combined structural and precision optimizations yield a 2.97–3.39× speedup over double-precision. The findings robustly demonstrate the potential of mixed-precision computing for high-resolution, real-time ocean forecasting applications, providing valuable insights for balancing computational efficiency and simulation accuracy.