A posteriori estimation of error norm via approximation error 
and adjoint parameters.

An estimation of the global norm of solution error is addressed using
 local approximation error, tangent linear problem and especial adjoint 
problem loaded by the information from tangent problem. This approach 
is aimed to provide not only a value of solution error norm but although
 the local sensitivity of this norm to approximation error. The local 
approximation error used in analysis is obtained by an action of high 
order finite-difference stencil on the field that is previously computed 
by main low order algorithm. The numerical tests for diffusion equation 
are presented.