.In numerical computations of fluid flows, the resolution is always limited and more than often, the Kolmogorov's dissipation scale is not resolved. Thus we face a need for parameterizing the scale that are not resolved. This problem is commonly called the "parameterization". The parameterization problem is further involved, when the fluid consists of multiple components, chemically react each other, and some of the components go through the phase changes. Furthermore, the resulting suspended particles (liquid or solid) within a fluid make its thermodynamics further involved, as called microphysics in atmospheric sciences. The turbulent combustion flow and the atmospheric-climate system may be considered the two major examples of complex the fluid-dynamic systems that require the extensive parameterization of unresolved physical and chemical processes. The work for this talk is motivated from the latter aspect, but it intends to be general including the former system. The basic idea behind this talk is based on our knowledge that extensive part of those subgrid-scale processes are associated with coherencies. Thus, an efficient representation of these subgrid-scale coherencies become a key for developing subgrid-scale parameterization in systematic manner. Based on this general idea, the segmentally-constant approximation (SCA) reduces the full subgrid-scale processes (as a PDE system) into a parameterization level. A link of this approach to POD (proper orthogonal decomposition) will be mentioned.