Yes. It is still possible to use THERM just as with previous versions of DIRSIG. It is also possible to mix-and-match thermal models, where some materials have temperatures predicted using THERM while others use Balfour.
Balfour is an empirical model, not a physics-based one. These parameters are encapsulated inside the Balfour model's governing equation.
The Balfour model assumes that the material being modeled is in a steady state. It also assumes that there is no phase change occurring. (e.g. no rain, snow, etc).
The THERM model is a physics based model while Balfour is empirically based. The inputs for the Balfour model are much easier to acquire or derive. The model itself also runs significantly faster than THERM.
The easiest technique for computing the model coefficients is to use a least squares fit on known measurements. The measurements need to include atmospheric and solar irradiance, windspeed, air temperature, and surface temperature. Since there are five coefficients, five measurements are needed for a unique least squares fit.