Abstract: Based on the cognition that numerical flow modeling is a regression process to prototype, from the correlation between model and prototype, a linear formulation that relates model’s and prototype’s Manning’s roughness coefficients(Manning’s n) is derived and a regression coefficient method for calibrating Manning’s n in shallow water equations is presented. Numerical tests show that the method does not only rely on the initial value of Manning’s n, but also has the attributes of good stability and fast convergence. Application to the complex flow simulation for the Modaomen channel indicates that the present method could improve the modeling precision, and has good adaptive, simple and practical abilities as well. It provides an effective way to determine Manning’s n in shallow water equations.