Abstract—In this study, an efficient Monte Carlo like method integrated with a Kriging surrogate model is proposed to estimate the uncertainties of underwater sound propagation loss with respect to multiple uncertain environmental parameters. The Kriging model is trained to simulate and replace the classical Kraken underwater acoustic propagation model in order to significantly reduce the computational cost but ensuring the accuracy. Monte Carlo process is performed by means of continuously sampling from the trained Kriging surrogate model, by which we finally achieve a quantification of the sound propagation loss. The Kriging models are respectively employed within different segments of a full calculation range in order to overcome the ill-conditioning problem of Kriging algorithm and increase the spatial resolution. 90% confidence bands are also calculated to illustrate the spatial distribution of uncertainties of the sound propagation loss.