The problem of describing the whole dynamics on the globall attractor of a semidynamical system is studied. The solution is reduced to finding the function of rate of attraction to an attractor. The method of determing numerically an estimate for this function is presented.The main problem of mathematical modelling of nonstationary processes is to describe the dynamics of the system studied. In the general case this reduces to the construction of a trajectory ´Øµ Ë Ø ¡ by the initial data and the resolving operator Ë´Ø ¡µ of the problem. When solving practical problems the original operator Ë Ø ¡ ¡ and the initial data are replaced by approximate Ë Ø ¡ ¡ and . Thus, the modelled and real trajectories begin to differ with time. The standard estimate of the local rate of discrepancy has the exponential formThe coefficient tends to zero as the accuracy of approximation of the operator and the initial data increases. Thus, the accuracy of modelling can be ensured on a finite interval ¼ Ì´ µ℄. In real processes the errors of modelling, the approximation of the initial data, and the forcing are essentially unremovable. Thus, the value Ì´ µ turns out to be much less than the value of interest. This is the case, for example, when modelling unstable fluid and gas flows in the general climate theory. However, of most interest in these problems is some typical state, which can be observed in the system, rather than the behaviour of a concrete trajectory. The strict definition of typicalness depends on a problem. In this connection there arises a problem of describing all possible limiting states Å, which are realized in the system over long periods of time. Note that this statement is meaningful only for systems with compact set Å. For problems of mathematical physics which deal with noncompact spaces the existence of Å and moreover its compactness are not obvious.The studied approach for partial differential equations was first presented by O. A. Ladyzhenskaya in 1972 in [18]. For the Navier-Stokes two-dimensional problem the compact set Å was constructed which uniformly attracts the arbitrary bounded subset of the initial data in the state space À. The minimality of this set among all sets with this property and the exact invariance with respect to the resolving operator of the problem were proved. It was shown that among all exactly invariant