AbstractThe paper addresses the question of whether the modeling practice of summing separate land-cover tiles to give urban fluxes at the neighborhood scale has merit. A central-city site in Basel, Switzerland, was instrumented to measure turbulent sensible heat fluxes Q_H from the two main land-cover types (roofs and canyons) separately and from the whole neighborhood. Path-averaged Q_H values were measured in the roughness sublayer (RSL) using scintillometry, and the spatially averaged Q_H neighborhood-scale flux was measured in the inertial sublayer (ISL) by an eddy-covariance system. The roof and canyon flux results are combined and weighted according to the respective plan-area abundance of each to give an estimated value of the neighborhood flux. The results show that this “bottom up” approach underestimates the measured ISL values by about 25% when averaged across all periods and wind directions. This finding led to consideration of possible errors from instrumentation, inappropriate turbulent source areas, failure to sample representative surfaces, and inability to fully capture RSL heat exchange. Sorting data by the two main wind directions revealed significant differences. The measured fluxes in the ISL and across the canyon top depend little upon wind direction, but daytime roof values show a marked sensitivity to wind direction. Qualitative analysis suggests this might be caused by systematic controls such as solar angle, site morphometry, and observational setup. The comparison of bottom up versus ISL is inconclusive; in some conditions agreement appears promising, and in others it does not. The question has not been proven or disproven. It may be too ambitious to test the concept at a real-world site.