Tractor rollover caused by physical contact with static obstacles or harsh terrain is a major hazard in agriculture. In addition, obstacle detection systems based on cameras and lidars struggle to perform well in agricultural working environments while their success rates might also be affected by a high number of false alarms. On the other hand, modern tractors are increasingly getting equipped with embedded GNSS and real-time kinematics (RTK) technologies that allow the driver to know its position on Earth with little error: this capability, combined with precise speed estimation which is a common feature of such systems, can be used to improve the safety of agricultural vehicles if static obstacles are georeferenced and adaptive buffer zones are created to serve as danger areas. In this study, which is part of a project called “Obstacle detection and tracking system for fixed and moving obstacles in agriculture” (SIRTRAck) funded by the Italian National Institute for Insurance against Accidents at Work (INAIL), a Keyhole Markup Language (KML) file containing georeferenced information on static obstacles in an experimental field has been created and passed to an algorithm on board of a tractor equipped with RTK, which integrated the dataset with the vehicle's real-time position, direction and speed. As the tractor entered the field, risk zones have been calculated by the algorithm from real-time motion information and have been used to alert the driver if the tractor entered one of them. The possibility of relying on ground truth and on adaptive shapes for danger zones built on speed, driver’s reaction time and expected slippage proved to be a valid and low-cost method for preventing tractor rollover hazards and might be easily used to improve existing geofencing features in both manned and autonomous agricultural vehicles.