In appropriate situations, large populations of geese exhibit dynamical rearrangements by repeated mergers and splits among the groups. We describe the grouping process in terms of a mean-field model based on the Smoluchowski equation of coagulation with fragmentation and observationally plausible kernels. To verify our model, we conducted field observations on skeins of airborne geese, noting both the group-size distribution and the group-forming processes. We found that the group-size distribution we obtained in our field measurements could be represented by a fractional power function with an exponential cutoff. This function matches the asymptotic form of the steady-state solution of our model. Furthermore, we estimated the effective number of individuals involved in interactions by comparison of the model to our field data.