Green roofs are considered urban ecosystems that provide stormwater management, reduce local air temperature through evapotranspiration, and enhance the aesthetic qualities of buildings. Green roof growing mediums must have a good water retention capacity, low organic content, and a suitable particle size distribution. Flow attenuation and water retention within the green roof soil can be understood by employing a mechanistic approach based on Richards' unsaturated flow model coupled with the Van Genuchten-Mualem soil hydraulic functions. Several studies have evaluated such an approach to predict the hydrologic response of green roof systems assuming constant soil parameters for numerical simulation purposes. In this paper, we report a framework for modeling 2D flow through a green roof soil, addressing parameter uncertainties by implementing a Bayesian statistical approach. Four, 1 m2 green roofs, with different soil mixture and vegetation (fern and sedum species), were examined in a side-by-side fashion over a one-year period and for a wide range of rainfall events. Parameter uncertainties were evaluated by estimating the posterior probability distributions of the soil parameters. The estimated posterior distributions settle around the most likely value for all the parameters (hydraulic conductivity k, saturated water content θs, residual water content θr, dimensionless soil moisture characteristic parameter n, and the empirical soil moisture characteristic parameter α); however, large dispersion was observed for some distributions. Most outflow prediction uncertainty ranges tracked the observed data.