Herons and egrets form roosts (R) in winter and breeding colonies (C) in summer. Both Rs and Cs are

formed annually and population size of them can range from a few birds to huge numbers. There should be network flows of individuals among Rs or Cs as well as between Rs and Cs. But it is not feasible to observe such network flows. To estimate networks among Rs or Cs, we constructed a network model based on the data of population size of Rs and Cs in Ibaraki Prefecture, Japan, from 2001 to 2011. Bipartite network between Rs and Cs can be expressed as R(t) = C(t) * A^T and C(t+1) = R(t) * A, where R(t) and C(t) are respectively 1 x r and 1 x c vectors for size distribution of Rs and Cs at year t. A is a r x c incidence matrix estimated with a genetic algorism. These equations can be transformed to R(t) = R(t-1) * AA^T and C(t+1) = C(t) * A^TA, where AA^T and A^TA are respectively r x r and c x c network matrices between Rs or Cs. Blockmodeling of these networks showed that larger and long lasting Rs or Cs had weak connections with other smaller Rs or Cs. On the contrary, smaller Rs or Cs were tightly linked with each other.