Flood hazard datasets may be improved by accounting for uncertainty within flood models, but the additional time and cost needed to run such analyses may limit their widespread use. This study examines whether sparse Gaussian Process Regression (GPR), a machine learning model with relatively short execution times, may be used to emulate the results from more detailed flood models. Examining a 173 km² catchment in northern Texas, our analysis found that the GPR emulator was able to mimic the results from a detailed flood model while reducing computation time. These preliminary results indicate that GPR emulation is a promising approach to develop flood hazard datasets that incorporate uncertainty at regional and national scales.