yes BT can be considered as ST. you can compute it by emissivity.

If the emissivity is less than one (which it almost always is) you need to correct for the downwelling in your retrieval. The downwelling that you use in the correction will depend on the bistatic scattering distribution function of the surface. If the surface is Lambertian, the downwelling will need to be taken for a zenith angle of about 55 degrees (this is invariably a higher brightness temperature than at zenith). The problem is that you need to know the emissivity, the downwelling and the bistatic scattering distribution function in order to retrieve the surface temperature from a near-surface measurement. If your radiometer is viewing through a bit (or a lot) of the atmosphere, you'll need to correct for atmospheric transmission as well. For more on this see the following references:

Mätzler, C., 2005: On the determination of surface emissivity from satellite observations. IEEEGEOSCIENCE AND REMOTE SENSING LETTERS 2 (2): 160-163.

Mätzler, C., and P.W. Rosenkranz, 2007: Dependence of Microwave Brightness Temperature onBistatic Scattering: Model Functions and Application to AMSU-A, IEEE Trans. Geosci. RemoteSens., 45(7), 2130- 2138.

Chawn is right the atmospheric correction is to be done and tis is a routine step. However as far as emissivity is concerned you can use standard emissivity values for different materials incase it is difficult for you to do the determination. You will find it easily on net, if not let me know I can supply you with an exhaustive list.

There are few publications regarding estimation of LST using 37Ghz V polarized BT.

Holmes, T. R. H., R. A. M. De Jeu, M. Owe, and A. J. Dolman (2009), Land surface temperature from Ka band (37 GHz) passive microwave observations,J. Geophys. Res., 114, D04113, doi:10.1029/2008JD010257.

How much accurate is this single channel approach ?

The accuracy is studied fairly well in he paper. It is a function of geographic position on the globe and a function of the time of year. The situation shown in figure 8(d) is really a best case scenario with snow at a minimum in the Northern Hemisphere. Snow should induce larger bias due to uncertainties in the emissivity of snow. There are still relatively large biases in regions of the globe even neglecting the effects of snow.

Are these biases important? It depends on your application. Are you interested in long-tern climate trends? Then the biases are less important. Or are you interested in modelling the surface energy balance on an hourly or finer time scale at any of those positions in blue or red in figure 8(d)? In that case the IASIs in combination with the Geostationary IR imagers do better. But the IR instruments can't see through cloud and 37 GHz can to a great extent, you only get two overpasses a day at any one point, and IASI-A has only been around since 2006.