The consistency of eddy-covariance measurements is often evaluated in terms of the degree of energy balance closure. Even over sloping terrain, instrumentation for measuring energy balance components is commonly installed horizontally, i.e. perpendicular to the geo-potential gradient. Subsequently, turbulent fluxes of sensible and latent heat are rotated perpendicular to the mean streamlines using tilt-correction algorithms. However, net radiation (Rn) and soil heat fluxes (G) are treated differently, and typically only Rn is corrected to account for slope. With an applied case study, we show and argue several advantages of installing sensors surface-parallel to measure surface-normal Rn and G. For a 17 % south-west-facing slope, our results show that horizontal installation results in hysteresis in the energy balance closure and errors of up to 25 %. Finally, we propose an approximation to estimate the surface-normal Rn, when only vertical Rn measurements are available.