Summer Arctic sea ice forms a heterogeneous mosaic of melting ice, melt ponds, leads, and pressure ridges whose thermodynamic and aerodynamic contrasts can modify surface–atmosphere exchange. Direct observations quantifying these effects during the melt season remain scarce, limiting evaluation of how sea-ice heterogeneity should be represented in larger-scale models. During the CONTRASTS expedition in summer 2025, we performed eddy-covariance measurements at approximately 1m and 2m height on two mobile towers, complemented by spatially resolved near-surface air-temperature measurements, to compare atmospheric turbulent exchange across leads, melt ponds, pressure ridges, and the adjacent sea ice. Median sensible and latent heat fluxes were close to zero W m^-2, although individual periods reached several tens of W m^-2 in both directions. The near-surface atmospheric stability was mainly neutral, with modifications towards slightly stable (unstable) related to a warmer (colder) background air-mass temperature. Total turbulent flux variability reached about one third of the variability of simultaneously measured net radiation fluxes. At lead-ice transitions, a larger lead contribution to the flux footprint was associated with more negative sensible heat fluxes, because the saline lead water remained about 1.7°C colder than surrounding melting ice. In contrast, no systematic pond--ice flux difference was detected, consistent with both surfaces remaining near 0°C and without a detectable difference in moisture availability. Pressure-ridge wakes disrupted the vertical coherence of heat and momentum fluxes, with the strongest disturbances for near-perpendicular flow across the ridge. Because the mean turbulent heat exchange between the surface and the air was close to zero, even modest surface-induced perturbations could reverse its sign and thereby alter the turbulent contribution to the surface energy balance. Under the sampled summer conditions, aerodynamic perturbations associated with pressure ridges produced stronger local modifications of turbulent exchange than the comparatively weak thermodynamic differences between the sampled surface types. The thermodynamic influence of leads, however, depends critically on water temperature: the near-freezing saline leads observed during CONTRASTS acted as a local heat sink, whereas warmer leads as observed in other campaigns reverse the surface–air temperature difference and therefore the direction of turbulent heat exchange. These observations show that summer sea-ice heterogeneity cannot be characterized by surface morphology alone: the atmospheric response depends on the actual thermodynamic difference and dynamically induced flow distortion, and that footprint-resolved measurements are essential for quantifying and parameterizing turbulent exchange over heterogeneous sea ice.