Crop nitrogen (N) management in the field have to be adapted to the crop need to avoid economic and environmental losses by denitrification, volatilization in the atmosphere and leaching in the ground water. The fertilisation adaptation to the crops needs by the fractioning is then a major issue, especially for crops with high N needs as the cereals. The short revisit cycle, the large swath and the open access of Sentinel-2 (S2) observation provide images over the whole Belgium every 10 days (up to 5 days by the summer 2017) which represents a valuable data source to map the N absorbed by the crops for each cereals field at national scale. There are two main approaches to estimate the N absorbed by a crop with satellite imagery; (i) direct methods relating directly the reflectance with the absorbed N via vegetation indices (Clevers J. et Gitelson A., 2013; Fei L. et al., 2014; Wei F. et al., 2016) and (ii) indirect methods linking the N to the retrieved chlorophyll concentration which are highly correlated (Cartelat, A et al., 2005; Zhou X. et al., 2016). Houlès V. et al. (2007) showed that the relationship between absorbed N and chlorophyll is more stable when considering the chlorophyll at the canopy level rather than at the leaf level. This study focuses on the application of an Artificial Neural Network (ANN) on S2 images to retrieve the two biophysical variables (BV) included in the estimation of the canopy chlorophyll content, the Green Area Index (GAI) and chlorophyll (Cab) for the winter wheat in Belgium. The choice of an ANN was driven by its integration of the knowledge on physical processes involved in the photon transport within vegetation canopies (Verger et al., 2011). The two main questions addressed here are, firstly, the performance and impact of S2 new red-edge bands (centered at 705 and 740 nm) to retrieve these BV, and secondly, the impact of intra-field spatial heterogeneity on the accuracy of the retrieved BV. These analyses are based on the one hand on synthetic data generated with a Radiative Transfer Model and, on the other hand, on a comprehensive in situ dataset to assess the retrieval performances and the S2 bands contribution. The GAI is sampled over the whole country with digital hemispherical photographs (DHP) in 18 fields spread over 5 provinces between March and June 2016, while the Cab content is measured during the key period of the third N application with the Dualex scientific Force A in 6 fields, 5 conventional and one organic heterogeneous field. The results highlight three major positive effects of the S-2 red-edge bands: (i) the compensation of the saturation effect around a GAI value of four observed with SPOT-5 Take5 time series, (ii) the reduction of data dispersion of low GAI value (below 2) increasing the determination coefficient (r²=0.85; RRMSE=27%) and (iii) the expansion of the range of the Cab value retrieved by satellite observation leading to a better distinction of the difference in high Cab value (r²=0.70; RRMSE=33%). While the effect of intra-field heterogeneity is highly impacting the GAI estimated with the 20 meters bands (underestimation of the GAI retrieved), the retrieval of the Cab is insensitive. Finally, the combination of GAI x Cab representing the canopy chlorophyll content appears to be the most appropriate BV to study the absorbed N because it expands the value range observed and represents an agronomical variable more contrasted between fields leading to a more accurate advice to farmers.