In this study, we use infrared thermography to calculate local heat transfer coefficients of top and bottom heated flows of near-critical carbon dioxide in an array of parallel microchannels. These data are used to evaluate the relative importance of buoyancy for different flow arrangements. A Joule heated thin wall made of Inconel 718 applies a uniform heat flux either above or below the horizontal flow. A Torlon PAI test section consists of three parallel microchannels with a hydraulic diameter of 923 μm. The reduced inlet temperature (TR = 1.006) and reduced pressure (PR = 1.03) are held constant. For each heater orientation, the mass flux (520 kgm−2s−2 ≤ G ≤ 800 kgm−2s−2) and heat flux (4.7 Wcm−2 ≤ q″ ≤ 11.1 Wcm−2) are varied. A 2D resistance network analysis method calculates the bulk temperatures and heat transfer coefficients. In this analysis, we divide the test section into approximately 250 segments along the stream-wise direction. We then calculate the bulk temperatures using the enthalpy from the upstream segment, the heat flux in a segment, and the pressure. To isolate the effect of buoyancy, we screen the data to omit conditions where flow acceleration may be important or where relaminarization may occur. In the developed region of the channel, there was a 10 to 15 percent reduction of the local heat transfer coefficients for the upward heating mode compared to downward heating with the same mass and heat fluxes. Thus buoyancy effects should be considered when developing correlations for these types of flow.