Abstract. The POM-SAT model for comparing air and ground temperatures is based on the supposition that surface air temperature (SAT) records provide a good prediction of thermal transients in the shallow subsurface of the Earth. This model consists of two components, the forcing function and an initial condition, termed the pre-observational mean (POM). I explore the sensitivity of this model as a function of forcing periods at time scales appropriate for climate reconstructions. Synthetic models are designed to replicate comparisons between borehole temperatures contained in the global database of temperature profiles for climate reconstructions and gridded SAT data. I find that the root mean square (RMS) misfit between forcing functions and transient temperature profiles in the subsurface are sensitive to periods longer than about 50 years, are a maximum when the period and the 150-year time series are equal and then decreases for longer periods. The magnitude of the POM is a robust parameter for periods equal to or shorter than the length of this time series. At longer periods there is a tradeoff between the amplitude of the forcing function and the POM. These tests provide guidelines for assessing comparisons between air and ground temperatures at periods appropriate for climate reconstructions. The sensitivity of comparisons between the average Northern Hemisphere gridded SAT record and subsurface temperature-depth profile as a function of forcing period is assessed. This analysis indicates that the Northern Hemisphere extratropical average SAT and reduced temperature-depth profile are in good agreement. By adding modest heat to the subsurface at intermediate periods some improvement in misfit can be made, but this extra heat has negligible influence on the POM. The joint analysis of borehole temperatures and SAT records indicate warming of about 1.1°C over the last 500 years, consistent with previous studies.