We develop a numerical model to understand the evolution of fracture permeability in hydrothermal upflow zones resulting from the combined effects of thermoelastic stresses and precipitation of silica as high-temperature, reactive fluid traverses temperature and pressure gradients. Because we test the model by comparing the results with those from previously published laboratory experiments on cyclindrical granite cores, we solve the problem of radial flow under an applied pressure difference of a silica-saturated fluid through vertical, initially parallel-walled cracks distributed evenly about the core sample. We emplace a steady state logarithmic temperature profile and assume that silica precipitation occurs so that the silica concentration in the fluid remains at equilibrium with local temperature and pressure along the flow path. The model results show a rapid initial decrease in permeability resulting from thermoelastic stresses, followed by a further decrease resulting from silica precipitation. The greater the initial temperature gradient, initial permeability, and/or initial crack width (at a given permeability), the greater the permeability decrease resulting from thermoelastic stresses. As a result of silica precipitation, the permeability eventually declines as t-3/2. The model results agree with the general trends in the laboratory data, thus confirming that silica precipitation is the main cause of the observed decrease in permeability during the experiments. Disagreement between the model and laboratory data in detail suggests that complications such as reaction kinetics, precipitation of other minerals and nonhomogeneous crack distributions need to be considered in the model. Thermoelastic stresses, though not important at the laboratory scale, may be important at the field scale.¿ 1997 American Geophysical Union