<p>Natural fractures have a strong impact on flow and storage properties of reservoirs. Their distribution in the subsurface is largely unknown mainly due to their sub-seismic scale and to the scarcity of available data sampling them (borehole). Outcrop can be considered as analogues where natural fracture characteristics can be extracted. However, acquiring fracture data on outcrops may produce a large amount of information that needs to be processed and efficiently interpreted to capture the key parameters defining fracture network geometry. Outcrops thus become a natural laboratory where the interpreted fracture network can be tested mechanically (fracture aperture, distribution of strain/stress) and dynamically (fluid flow simulations (Bisdom et al., 2017). </p> <p>The goal of this paper is to propose the multiple point statistics (MPS) method as a new tool to quickly predict the geometry of a fracture network in both surface and subsurface conditions. This sequential simulation method is based on the creation of small and synthetic training images representing fracture distribution parameters observe in the field. These training images represent the complexity of the geological object or processes to be simulated and can be simply designed by the user. In this paper we chose to use multiple training images and a probability map to represent the fracture network geometry and its potential variability in a non-stationary manner. The method was tested on a fracture pavement (2D flat surface) acquired using a drone in the Apodi area in Brazil. Fractures were traced manually on images of the outcrop and constitute the reference on which the fracture network simulations will be based. A sensitivity analysis emphasizing the influence of the conditioning data, the simulation parameters and the used training images was conducted on the obtained simulations. Stress-induced fracture aperture calculations were performed on the best realisations and on the original outcrop fracture interpretation to qualitatively evaluate the accuracy of our simulations. </p> <p>The method proposed here is innovative and adaptable. It can be used on any type of rocks containing natural fractures in any kind of tectonic context. This workflow can also be applied to the subsurface to predict the fracture arrangement and its fluid flow efficiency in water, heat or hydrocarbon reservoirs.</p>