The aim of this study is to propose an empirical spatial model to estimate the spatial variability of grapevine phenology at the within-field scale. This spatial model allows the characterization of the spatial variability of the fields through a single measurement performed in the field (reference site) and a combination of site-specific coefficients calculated through historical data. This approach was compared to classical approaches requiring extensive sampling and phenology models based on climatic data, which do not consider the spatial variability of the field. The study was conducted on two fields, one of cv Cabernet Sauvignon (CS, 1.56 ha) and the other one of cv Chardonnay (CH, 1.66 ha) located in the Maule Valley, Chile. Measurements of the date of occurrence of grapevine phenology (budburst, flowering and veraison) were observed at the within field level following a regular sampling grid (18 sites for cv CS and 19 sites for cv CH) during 4 seasons for cv CS and 2 seasons for cv CH. The spatial model was calibrated using data collected during the first 3 seasons of cv CS and the 2 seasons of cv CH, while the last season of cv CS was used for the validation process. Regarding the quality of estimation, the best results were obtained with the spatial model in almost all cases, with a Root Mean Square Errors (RMSE) lower than 3 days. However, if the variability of phenology is low, the traditional method of sampling could lead to better results. This study is the first step towards a modeling of the spatial variability of grapevine phenology at the within-field scale. To be fully operational in commercial vineyards, the calibration process needs simplification, for example, using low cost, inexpensive ancillary information to zone vineyards according to grapevine phenology. This study opens up the opportunity to combine classical models of phenology based on climatic data with spatial models, with the aim of predicting the grapevine phenology both in time and space.