The implementation of centralized filtering theory incurs computational difficulties especially in applications to large scale systems. These difficulties can be overcome by the aid of hierarchical system theory. Based on this concept, a simple decentralized estimation algorithm is developed which improves the algorithm proposed by Sage, et al. The diagonalized decomposition of a system has made an important role for reducing the complexity of an estimation algorithm. The proposed hierarchical estimation scheme is divided into two levels whose are the coordination part of Lagrange multipliers and the decentralized estimation part. The procedure may be computationally attractive for nonlinear systems as well as linear systems. A numerical example is used to illustrate the approach for simultaneous estimation of states and parameters, which compose the nonlinear system. It is seen that the new method leads to a saving in computation time without lossing estimation accuracy.