[en] Calculations using both mechanistic and empirical (i.e., Kd) solute-attenuation submodels in an explicitly coupled reaction-transport computer code can be compared to identify conditions for which the empirical models are adequate for assessing the performance of geologic repositories for nuclear wastes. The approach is illustrated using a planned performance assessment strategy in which Kd values for Cs are calculated on the basis of the modal abundances and relative Cs-sorption affinities of zeolites, smectite, and glass present in tuffs underlying the candidate nuclear waste repository at Yucca Mountain, Nevada. Thermodynamic considerations and site data characterizing the mineralogy along possible flow paths at Yucca Mountain indicate that zeolite metasomatism may result from minor variations in groundwater composition. A coupled reaction-transport code, CTM, is used to simulate the effects of these replacement reactions on Cs transport, and the results are compared to calculations using a Kd-attenuation model in which zeolite metasomatism is not considered. Calculations using the mechanistic model predict a cycle of attenuation and remobilization of Cs as sorptive clinoptilolite is progressively replaced by less sorptive mordenite. This secondary reaction front migrates more slowly than the fluid, but total Cs concentrations in the aqueous plus solid phases at the front increase continuously. The secondary front is not predicted in the Kd-model calculations, which as a consequence significantly underpredict total Cs concentrations in the flow path. 15 refs., 3 figs., 3 tabs