Network flow models were used to simulate the flow of CO 2-saturated brine in the pore networks corresponding to three different sandstones. The simulations were used to study upscaling of anorthite and kaolinite reaction rates from pore to core scales. Unique to our simulations is the use of computed tomography to capture the mineral distribution in the samples as well as the sample pore network. The upscaled reaction rates determined from these simulations incorporate mass balance principles and microscale reaction rate laws and capture the physical, mineral, and flow heterogeneities in the network. These upscaled rates were compared with upscaled rates predicted by a continuum model and by a volume-averaged-concentration method. For the anorthite reaction, which remains far from equilibrium, the volume-averaged reaction rate exceeded the reaction rate of the network model by 18% to 46%. While the continuum model rate also exceeded the network model rate by -1% to 53%, its predicted values were generally better than the volume-averaged method. The kaolinite reaction is near equilibrium and is heavily influenced by the form of the microscale rate law in the precipitation regime. Three alternate rate laws were tested, which produced significantly different predictions for the bulk reaction rates. For all three rate laws, continuum and volume-averaged reaction rates incorrectly predicted the magnitude of the kaolinite reaction rate (disagreements of -700% to 55%), and the predicted reaction type, dissolution versus precipitation, was also often opposite to that of the network model. Finally, for both anorthite and kaolinite, all upscaled reaction rates showed significant flow rate dependence.