In this study we developed and examined a hybrid modeling approach integrating physically-based equations and statistical downscaling to estimate fine-scale daily-mean surface turbulent fluxes (i.e., sensible and latent heat fluxes) for a region of southern California that is extensively covered by varied vegetation types over a complex terrain. The selection of model predictors is guided by physical parameterizations of surface flux used in land surface models and analysis showing net shortwave radiation that is a major source of variability in the surface energy budget. Through a structure of multivariable regression processes with an application of near-surface wind estimates from a previous study, we successfully reproduce dynamically-downscaled 3 km resolution surface flux data. The overall error in our estimates is less than 20 % for both sensible and latent heat fluxes, while slightly larger errors are seen in high-altitude regions. The major sources of error in estimates include the limited information provided in coarse reanalysis data, the accuracy of near-surface wind estimates, and an ignorance of the nonlinear diurnal cycle of surface fluxes when using daily-mean data. However, with reasonable and acceptable errors, this hybrid modeling approach provides promising, fine-scale products of surface fluxes that are much more accurate than reanalysis data, without performing intensive dynamical simulations.