The atmospheric layer adjacent to the earth's surface is of crucial importance for weather models due to the exchange of energy between the surface and the atmosphere. This exchange is dependent on the various surface properties and influences the state of the atmosphere at all scales from small-scale turbulence to large-scale weather patterns. In spring, the melting snow cover becomes patchy, resulting in pronounced surface heterogeneity and significant changes to the overlying atmosphere. This interaction leads to the development of shallow thermal internal boundary layers, profoundly impacting turbulent exchange processes. Our objective is to experimentally access this highly dynamic part of the near-surface atmosphere and gain a more thorough understanding of the sub-meter scale heat advection processes. For this aim, we developed a novel experimental method. We vertically deploy thin synthetic screens across the transition from bare ground to snow. The screens quickly adapt to ambient temperature and, thus, their surface temperature serves as a proxy for the local air temperature. By filming them with a thermal infrared camera at 30Hz, we obtain two-dimensional visualizations of the near-surface atmospheric dynamics at a spatial resolution of 0.5cm. Additionally, we developed an algorithm to estimate two-dimensional wind speeds by tracking patterns of air temperature. We demonstrate the capabilities of the method by investigating the spatio-temporal dynamics of the stable internal boundary layer (SIBL) adjacent to the snow surface. We complement these measurements with data from conventional eddy covariance (EC) sensors in a comprehensive field campaign in spring. Analysis of the EC data indicates periods when the stratification above the snow is strong enough to dampen and partly shut down near-surface turbulence. On the contrary, two Föhn events induced pronounced shear-generated turbulent kinetic energy and strong downward heat fluxes. Throughout the melt-out, we observed a decrease in the SIBL depths, resulting in a sign change in the heat fluxes at a given height. Apart from heat fluxes exhibiting opposing signs within and above the SIBL, measurements at different levels reveal a uniform pattern of turbulence within the first few meters above the surface. As more bare ground emerged, a shift towards larger time scales in measured temperature variance occurred, with thermal infrared observations unveiling intermittent advection of plumes of warm air over snow-covered areas. Finally, we set up a centimeter-scale large eddy simulation over an idealized transition from bare ground to snow. We force the simulation with 20Hz wind speed measurements. We validate the model by comparing it to EC measurements. While the simulated horizontal wind speed and air temperature exhibit very good agreement with the measurements, there is a lack of representation of vertical motion, especially for higher frequencies. Nevertheless, we show by comparison with screen measurements that the model is capable of resolving important flow features such as the intermittent advection of warm air over snow. We show that higher wind speeds lead to shallower SIBLs. Moreover, the local surface topography has an influence on the heat fluxes through local flow modification and induces the decoupling of a thin stable layer adjacent to the snow surface. The work opens the way towards a better representation of surface exchange in larger scale models.