Particle tracking through stochastically generated networks of discrete fractures provides an alternative to the conventional advection-dispersion description of transport in fractured rock. However, discrete fracture network simulations are computationally intensive and usually limited to small scales. An approach for direct upscaling of discrete fracture simulations is described. Trajectories for nonreacting tracer particles are first extracted from relatively small discrete fracture network simulations. Tracer-rock interaction is represented by also calculating a cumulative reactivity parameter along each trajectory. The residence time/reactivity information is then used in a Monte Carlo simulation to construct artificial particle trajectories of any length, thereby achieving the upscaling objective. In its simplest form the procedure has the form of a random walk evolving in a two-dimensional space. Tests using site-specific and generic networks show that it is necessary to modify the random walk to produce sequential correlation along the trajectories. We achieve this by using a discrete state Markov process to direct the random walk. The procedure is computationally efficient, easily implemented, and compares well with the network simulations.