This work discusses two aspects of simulating catalysts in real world driving situations where frequent and random changes of engine load lead to recurrent, harsh and simultaneous changes in exhaust gas mass flow, temperature and composition. The signals at catalyst inlets attain frequencies of up to 5 Hz. Many exhaust gas species composition, mass flow and temperature sensors exhibit delays and low pass behavior that smooth out the signals to frequencies of 1 Hz and below. It was investigated how sensitive exhaust gas after-treatment models are to the sampling rate, time delay and low pass deformations of the pre-catalyst time signals used as the model input. Simulation outputs were shown to be very sensitive to these signal deformations. This sensitivity results from the fact that in most transient events, oxygen storage dynamics play a central role on prediction quality. Additionally, two common, numerically motivated simplifications in catalyst simulation and their limitations in highly transient situations are discussed. The first simplification deals with the proper oxygen balance in the system. The use of implicit formulations of differential equations may lead to inconsistencies in mass balance and thus to erroneous levels of stored oxygen. The second includes variables that may change slowly under normal combustion conditions but may change rapidly during fuel cut off or similar highly transient states. Thus, a different numerical treatment is needed. The consequences of these two aspects for real driving cycles are illustrated. As demonstrated, prediction errors of 50% or more are possible. (C) 2011 Elsevier B.V. All rights reserved.