Computer modeling is critical for catalyst layer (CL) design in polymer electrolyte membrane fuel cells. Water-filled and ionomer-filled agglomerate models have been suggested as representations of the CL microstructure. In this thesis, improved water-filled and ionomer-filled agglomerate models are developed. Results indicate that the agglomerates provide identical current densities at low and high overpotentials, but differ at mid-range values. These models are integrated in a multiscale simulation of a 2D membrane electrode assembly (MEA) model. A comparative analysis shows that the choice of agglomerate alters the reaction distribution in the CL but does not significantly change the model's performance. Lastly, it is proposed that the CL microstructure be characterized by optimization-based parameter estimation, which matches MEA model predictions to experimental data. Results suggest that experimental data is not readily characterized by an agglomerate model; the MEA model requires more detail to describe the phenomena across a range of operating conditions.