We describe bistatic scattering measurements at 230 GHz, in the 330-490 GHz range and, at 650 GHz on various surfaces. These include a series of eight reference targets constructed from alumina grit embedded in an absorptive epoxy matrix, and a set of conventional outdoor building materials. The samples’ surface topographies were measured by focus-variation microscopy (FVM) and their autocorrelation lengths and RMS roughness levels extracted. All bistatic measurements were performed in the principal plane, at incidence angles of 25o, 45o, and 65o, in s and p polarization. The reference samples’ normalized roughness levels cover the range 0.040 ≤ σ/λ ≤ 0.60, and their normalized autocorrelation lengths cover the range 0.086 ≤ 𝐿/λ ≤ 1.14. The measurements are described in terms of bidirection reflectance distribution function (BRDF) or normalized radar cross section (nRCS), and include regimes of both diffuse scattering and specular reflectance. The reference samples’ measurements are compared to two ab initio scattering theories, the Modified Integral Equation Method (IEM-B) of A. Fung, and the Generalized Harvey-Shack (GHS) model, that have no free parameters. Although there are several individual cases where either the IEM or GHS theory (or both) provide a good match to measurement, their overall agreement across the entire dataset is poor. In addition, the diffuse BRDF in each bistatic scan has been fit to a Lambertian (constant) dependence of scattering angle, and a purely empirical model developed for the dependence of Lambertian scattering on frequency, roughness, polarization, and incidence angle. The empirical model provides the best match to measurement across the full dataset, and can be used for reliable phenomenology studies of submillimeter imaging or wireless telecommunication. Nearly all the outdoor building materials, like the roughest of the reference samples, fall in a regime where L/σ is not large, and therefore where ab initio scattering theories can’t be expected to apply.