The effect of buoyancy caused by heat generation from a microelectromechanical system (MEMS)-based thermal shear stress sensor is investigated. Due to the small size and relatively low power consumption of such sensors, the buoyancy effect on the overall flow structure is generally negligible. However, its impact on the flow variables such as shear stresses can be significant because such quantities are local and depend on the gradients of the velocity profile next to the sensor. Due to the small dimension of the MEMS sensor, a multiscale modeling approach is adopted to examine the effect of buoyancy on the velocity and wall shear stress profiles. Full-length channel computations are initially performed with finer resolution near the sensor region. Using the boundary conditions derived from the full-length computations, another simulation is performed concentrating on a small region near the shear stress sensor. Based on the temperature distribution in the region of the sensor, the effective thermal length scale is several times the streamwise dimension of the sensor. For a state-of-the-art MEMS sensor dimension of 200mm, the effect of buoyancy on the accuracy of shear stress measurement can be noticeable.