Abstract Row crops like vineyards undergo various and significant manipulations of training system and cultural practices, which strongly influence the quality of products. Variations of water vapour fluxes from the soil surface and the leaves in the row volume are closely linked to the ratio of energy available to each compartment. A physically realistic model of available energy partition between the rows and the soil surface is therefore a key factor towards optimization of such systems, and must be included in canopy models. A number of available models were not directly validated. The purpose of the study was therefore to design a model of net radiation partition and check it directly. The model of net radiation partition between rows (Rnv), considered as a whole, and intervening soil surface (Rns) of a row-crop canopy was developed from physically realistic yet simple assumptions: - global solar (short wave) radiation partition was calculated by a previously validated geometric model; - long-wave radiative fluxes between the soil surface, the rows and the atmosphere were calculated from the corresponding view factors, which only depended on canopy geometry; - atmospheric radiation was estimated by a simple empirical relation based on air temperature as the only input variable; - air temperature in the vicinity of leaves replaced leaf surface temperatures as a more convenient input variable, with little loss of information. The input variables were incoming direct and diffuse solar radiation, soil surface mean temperature and air temperature near the leaves. The main parameters were soil and leaf albedos, row porosity and dimensions. A direct validation of the model was attempted by measuring net radiation above the canopy and at five positions above the soil surface in a vineyard of the Bordeaux area. The reliability of soil surface net radiation measurements was estimated by thorough error propagation analysis. When found significant, errors were corrected and finally soil surface net radiation data were corrected only for delay in direct downward solar radiation striking net radiometers, because canopy was discontinuous and height of net radiometers was not negligible compared to canopy height. In these conditions, model calculations were in agreement with measurements, although the model slightly underestimated Rns and therefore overestimated Rnv. As the mean error was about 20 W m −2, and therefore compatible with instrument accuracy, the results were considered satisfactory. This available energy partition model is able to estimate radiative balance in various canopy systems and in various thermal environment conditions, leading to easier simulations of energy balance and water fluxes. It could therefore be a useful tool for optimizing row-crop canopies, taking fully into account any kind of present or future thermal environment.