Using the MIT 2D climate model, we simulate the climate system response to changes in greenhouse gas, aerosol, and ozone concentrations (i.e., the GSO forcing) to explore the model parameter space for climate sensitivity, deep-ocean heat uptake, and net aerosol forcing (imposed using the sulfate aerosol distribution). These simulations are then used to assess which combinations yield results that are inconsistent with observations. In this work, we examine the model's response in both upper air and surface temperature change patterns using the climate-change detection algorithm and find that a set of simulations with strong aerosol forcing (about 1.5 W/m$^2$ cooling in 1986) and in the strong response region (high climate sensitivity and low ocean heat uptake) are consistent with the radiosonde record but not with the surface record. This inconsistency was further explored. If the net aerosol forcing were this large (i.e., the combined direct and indirect sulfate plus additional aerosols), the net global radiative forcing would remain unchanged (or even decline) until near 1960 and then it would rise roughly 1 W/m$^2$ between 1960 and 1995. Thus, for a sufficiently sensitive model (climate sensitivity = 3.0~K and effective ocean diffusivity = 0.8~cm$^2$/s), the modeled changes in upper-air temperatures, in response to this forcing, remain consistent with the observational radiosonde record for 1961-95. These results suggest that the upper-air temperature record has limited value as a diagnostic for detection and attribution studies aimed towards assessing uncertainties in key model properties. Both the record length and spatial-distribution of the surface record provide important additional information as a diagnostic. As an attribution diagnostic, the upper-air record is best used for the understanding of ozone effects on temperature change.