Coarse grained particle methods significantly reduce the computation cost of large-scale fluidized bed simulation by lumping many real particles into a computation parcel. This research provides a method to estimate the errors associated with parcel size in large-scale fluidized bed simulations. This uncertainty is first quantified in small scale domains by comparing results of discrete particle method with that employing coarse parcels of different sizes. Then, this uncertainty is correlated with parcel size and simulation domains consisting of a simple homogeneous cooling system and more complex bubbling and circulating fluidized beds. These correlations allow us to accurately estimate the uncertainty in large-scale fluidized beds based solely on data obtained in smaller systems. The ability to estimate model-related uncertainty in larger systems makes this method relevant for industrial applications. (c) 2018 American Institute of Chemical Engineers AIChE J, 64: 2340-2350, 2018