The development and the release of sensors capable of providing data with high spatial resolution (> 4 000 points ha'1) in agriculture raises new questions as to how to represent this spatial information. The objective of this study was to propose a methodology to help define the optimal grid size to map high resolution data in agriculture. The geostatistical method finds the grid size which maximizes the sum of two components: (i) the proportion of nugget variance that is removed, and (ii) the proportion of sill variance that remains in the data. The optimum grid size was found to be dependent on the resolution of the available information and the spatial structure of the raw data. Experiments on simulated datasets with varying data resolution (from 500 to 2 000 pts.ha'1) and spatial structure (range of variogram between 10 and 45 m) showed that the proposed methodology was able to define varying optimal grid sizes (from 5 to 12 m). The proposed geostatistical approach was then applied on a real dataset of total soluble solids/sugar content of table grape so that the optimal mapping grid size could be found. Once it was defined, two interpolation methods: simple averaging over blocks and block kriging, were applied to mapping the data. Results show that both methods help depict the within-field variability in the data. While the averaging procedure is easier to automate, the block kriging approach provides users with a level of uncertainty in the aggregated data. Both mapping approaches significantly impacted the within-field spatial structure: (i) the small-scale variations were ten times lower than in the raw data, and (ii) the signal-to-noise ratio of the aggregated data with the optimal grid was twice as high as that of the raw data. As the proposed geostatistical methodology is a first attempt to define the optimal grid size to map high resolution spatial data, areas for future development applications are also proposed. © 2018 Springer Science+Business Media, LLC, part of Springer Nature