LAI, FAPAR and FCOVER variables are required for the monitoring, understanding and modelling of land surfaces at the global scale. While several products were already developed from the current medium resolution sensors, the few validation exercises achieved demonstrated that significant discrepancies and inconsistencies were observed. The objective of this study was to develop new global estimates of LAI, FAPAR and FCOVER that will build on the pros and minimize cons of already existing products. In a first step, the performances of the MODIS, CYCLOPES, GLOBACRBON and JRC‐FAPAR products were reviewed. The MODIS and CYCLOPES products were selected since they provide higher level of consistency. These products were then fused to generate the ‘best estimate’ of LAI, FAPAR and FCOVER that were later scaled to better match their expected range of variation. Finally, neural networks were trained to estimate these best estimates products from SPOT‐VEGETATION top of canopy directionally normalized reflectance values. Performances of the derived products called GEOV1 were evaluated, showing significant improvements as compared to previous products. These products will be extended back to 1981 using the AVHRR series of observation, and continued after the VEGETATION era thanks to AVHRR‐METOP, PROBA‐V and Sentinel3 future missions