In this thesis, we first focused on error estimates for unsteady problems. We have contributed to both a posteriori and a priori error estimators for unsteady inviscid problems and viscous unsteady problems. For the first one, we have been interested on linearized methods for reducing dissipation errors. Regarding the a priori errors, a new estimator is proposed with application to viscous compressible flows. These a priori estimators have been employed for goal-oriented anisotropic mesh adaptation problems, for both Euler and laminar Navier-Stokes flows, in a joint work with Gamma3 team. We have developped a method to derive an optimal mesh to observe/improve a given output functional in an unsteady context. The weights of the interpolation error are adjoint states in this case. A new global fixed-point algorithm is proposed herein order to converge the couple mesh/solution. We have applied this algorithm for blast wave problems and acoustics, for both 2D and 3D cases.
