AbstractA large number of processes and factors control the quality of simulations with a numerical weather prediction model and especially with mesoscale models; identification and optimization of these processes are critical for improving forecast skill. The importance of cumulus parameterization schemes in simulation of tropical cyclones was recognized early, and a large number of studies have addressed this issue. However, certain other aspects have received relatively less attention. In particular, unlike simulation with a global circulation model, a mesoscale simulation is characterized by a limited domain and hence inhomogeneous lateral boundary conditions that strongly affect the quality of the simulation. In this work, we investigate the relative impact of size of the model domain and the cumulus parameterization scheme on simulation of 10 cyclones over the Bay of Bengal during the period 1999–2009. For five domains with different spatial extents, simulations were carried out for three different cumulus parameterization schemes (Anthes‐Kuo, Grell, and Kain‐Fritsch2) for each of the 10 events (using the mesoscale model MM5). Our results show that the size of the domain also plays an equally critical role as the parameterization scheme in simulation of maximum intensity, track, and spatial structure of the cyclones. The impact of domain size is not linear; while each domain chosen is large enough, neither the largest nor the smallest domain provides the best simulation. However, there is consistency in the sense that a single domain emerges as best for intensity and track among the five considered. While the specific conclusions may depend on the ocean basin, the methodology is generic and can be applied to any ocean basin of cyclogenesis.