Abstract: The mismatch of model parameter would lead to the inaccuracy of the estimation of states for nonlinear systems. This paper proposes a joint estimation approach based on Unscented Kalman Filter, in which both parameters and states are simultaneously estimated by means of the argument state vector composed of the unknown model parameters and states. By utilizing the Monte Carlo simulation for a discrete nonlinear stochastic system, the influence of the filter parameters on the performance of the joint estimator is analyzed and the updating rule of filter parameters is obtained. Finally, this approach is successfully applied to a typical chemical reaction process.