A new comprehensive calibration methodology for a population balance model (PBM) is presented using experimental size distribution data obtained on-line during flocculation tests. It was found that the moving-pivot discretization technique should be used to accurately solve the PBM. A new methodology for grid transformation and volume to number transformation, based on the sludge concentration and the densities of liquid, solids, and flocs, is presented. Model fits were performed on four different fitting variables. Volume-based fitting (vol % and D[4, 3]) tends to fit the middle part of both the vol % and number distribution better than the tails of smaller and larger particle sizes. Severe underpredictions (up to a factor of 30) of the lower-size classes are observed at steady state. Number-based fitting [(weighted) number distribution] tends to fit the small-size classes better, but still underpredicts them by a factor of 2 at steady state. (c) 2005 American Institute of Chemical Engineers.