Managing within field variability is promising to achieve European objectives of sustainability in crop production. Technological development has allowed to precisely characterize fields heterogeneity in space and time. However, learnings from low adoption of yield maps in west-European context have highlighted the importance of reliable methods to support decisions. Blackmore et al. designed a delineation method considering yield as an integrative variable that reflects spatial and temporal variations within fields. Aiming to divide the field into high and stable, low and stable and unstable yield zones, this method matches principles of parsimony and intelligibility. Indeed, zonation is straightforward to adapt agronomic practices to each productive potential. This delineation method has been used to assess and optimize Nitrogen Use Efficiency (NUE) in north American context, revealing large potential of improvement. In Europe where pollution due to nitrogen fertilization remains high, implementation of this method faces different problematics. Firstly, agronomic context is different with smaller fields, higher yields and more diversified crop rotation than in the US. These issues enhance the necessity to evaluate the method in west-European context on a large scale. Secondly, use of yield maps is low in Belgium and France and involve a substitution technology to implement Blackmore’s method. Vegetation indices can be calculated from satellite images on large areas and have been extensively studied to predict crop yield at regional or national scales. Nonetheless, few studies analyzed the relationship between within-field yield and vegetation indices distributions and their sensitivity to retrieve yield-based management zones. This interrogate the possibility of delineating management zones applying Blackmore’s methodology with vegetation indices sensed at specific periods.

The first step of the analysis was conducted over fifty fields distributed in Belgium and France. It aims to identify on each field and year available the optimal sensing window and the best vegetation index to retrieve yield distribution. Different indices calculated from Sentinel-2 images sensed during the period 1^st of April – 30^th of June were compared to the different yield maps.

The second step of the analysis consists of delineating management zones applying Blackmore’s method on vegetation index. This part of the analysis has been applied on six fields where yield maps were available for at least three years. The optimal period and best vegetation index are used to implement the method. A confusion matrix between the management zones obtained with yield maps and the one obtained with vegetation index is constructed to detail differences in zones delineation. 

From the correlation analysis, NDVI got the highest and most constant results. Two sensing periods stood out for the different indices but the last one, during flowering stage, was chosen to reduce climate effect on yield pattern prediction. Delineation of management zones through the original method reveals few unstable zones in the six fields. Detection with vegetation indices is further limited considering a simple transposal of Blackmore’s methodology over indices. Upcoming step should focus on the stability threshold used with vegetation indices to enhance zone prediction.