Enhancing Synthetic Rating Curve Development Through Empirical Roughness Built for Hydrofabric Datasets