Basic radiative transfer theory reveals that the physical temperature gradients related to heat flow (HF) within the lunar regolith below the diurnal-varying layer can be constrained by measurements of brightness temperature (TB) variations over a selected wavelength range, coupled with knowledge of the regolith electrical loss properties. We present a methodology illustrating the processing using recalibrated data from the Chang’E-2 (CE2) 10- and 3.85-cm wavelength microwave radiometer (MRM) data (channels 1 and 2). The utilized data are contained in latitude bins of 1° width centered at latitudes of $0, 5, \ldots, 60$ , north and south. Results, presented as retrieved HF gradients (dT/dz in K/m) versus longitude for each of the 25 latitude bins, provide useful constraints on the overall variability of the HF gradient within the 60S–60N latitude range. In particular, no evidence is found that HF variations exceeding twice the Apollo value occur widely in the extensive highland regions of low electrical loss. Limitations related primarily to signal-to-noise ratio (SNR) issues in regions of elevated electrical loss severely reduce thermal gradient retrieval accuracy needed for correlation with enhanced thorium concentrations observed in the frontside mare. Analyses of the CE2 data reveal a noise uncertainty of ~0.5 K in the TB10-TB3.85 measurements, equivalent to ~0.7–1.4 (loss tangent dependent) times the Apollo-measured mean HF value of 1.8e−6 W/cm2. The limitations preclude meaningful identification of regional variations within a precision of less than one Apollo HF unit. An analysis is presented demonstrating that a threefold increase in SNR would be attained by addition of a 30-cm wavelength channel to the CE2 MRMs design, providing sufficient resolution to map regional HF variations to ~0.5e−6 W/cm2 precision, ~30% of the averaged Apollo sites’ measured HF value.