Flux towers provide important climatic information about surface exchange processes in the land-atmosphere system. At the Finse flux tower in alpine Norway, the inhomogeneous surface and hilly terrain influence the quality of the eddy covariance (EC) measurements. Due to the large variation in surface cover, including bare ground, different vegetation types, and open water, it is challenging to provide a representative estimation of net surface radiation around the flux tower. This study investigates the effect of non-corresponding footprints between the net surface radiation and the turbulent fluxes measured with the EC technique. This was done by measuring net surface radiation for different surface types, where the surface types were provided from a detailed vegetation map over the Finse study site. By fitting linear regression models (LMs) and training artificial neural networks, continuous time series of net surface radiation for different surface types for the period 18.08.18 – 25.08.18 where predicted. The time series from the LM were used to derive an overall value of net surface radiation, which accounted for the heterogeneity of the surface cover. Doing so, a better comparable value of net surface radiation was obtained for the evaluation of the surface energy balance closure. Based on the linear model predictions, the study found a difference, with statistical significance of 99%, in net surface radiation between almost all the alpine surface types investigated. This implies that considerable differences of albedo and properties controlling the emissions of longwave radiation, occur in the surface cover which contributes as source area to the measurements of the turbulent fluxes. The energy balance ratio was found to be 58% when evaluated with the LM predicted net surface radiation. Thus, the energy balance closure increased by 4.40 percent points when allowing for the effect of surface heterogeneity on the net surface radiation. However, when the energy balance closure was evaluated with the ordinary least squares regression technique, the closure was found to be only 37%. No improvement of the energy balance closure was found when the heterogeneity of surface cover was allowed for by this evaluation method. The work presented here is a proof-of-concept study, showing how heterogeneity at flux sites can be accounted for using the methodology presented.