Summary: In 1877 Boussinesq (and others) put forward the basic analogy between the mixing effects of turbulent fluctuations and molecular diffusion: \(-\nabla\cdot\overline{(u'u')}\sim -\nabla\cdot\bigl(\nu_{\text{T}}(\nabla\bar{u}+\nabla\bar{u}^t)\bigr)\). This assumption lies at the heart of essentially all turbulence models and subgridscale models. By revisiting the original arguments of Boussinesq, Saint-Venant, Kelvin, Reynolds and others, we give three new approximations for the turbulent viscosity coefficient \(\nu_T\) in terms of the mean flow based on approximation for the distribution of kinetic energy in \(u'\) in terms of the mean flow \(\bar{u}\). We prove existence of weak solutions for the resulting system (NSE plus the proposed subgridscale term). Finite difference implementations of the new eddy viscosity/subgrid-scale model are transparent. We show how it can be implemented in finite element procedures and prove that its action is no larger than that of the popular Smagorinski-subgrid-scale model.
Part II, cf. G. P. Galdi and W. J. Layton, Math. Models Methods Appl. Sci. 10, No. 3, 343–350 (2000; Zbl 1077.76522). Further parts have been reviewed in (V) Zbl 1042.76537.