As a new procedure for modeling geothermal energy extraction systems, a two-dimensional modeling technique for subsurface fracture networks on the basis of ''fractal geometry'' is presented. Models of fracture networks are generated by distribution fractures randomly in space and by using the fractal relation between fracture length r and the number of fractures N expressed with a fractal dimension D as N=Cr-D, where C is a constant that signifies the fracture density within the rock mass. This procedure makes it possible to characterize geothermal reservoirs by parameters measured from field data, such as from core sampling. In this characterization the fracture density parameter C of a geothermal reservoir is used as a parameter to model the subsurface fracture network. Using this fracture network model, the connectivities of the water flow paths between wells are calculated by means of a Monte Carlo simulation, and the result is then compared with that derived from a percolation model. We show that many fewer fractures are required to connect two wells for the fracture network model than for the percolation model. The transmissivities beween wells for the fracture network model are also obtained as a function of the fracture density parameter C. The results show that the transmissivities in geothermal reservoirs are significantly dependent upon the fracture density of rock mass, and they can be predicted from the fracture density parameter C of the reservoirs. ¿ American Geophysical Union 1995