The Markov chain Monte Carlo (MCMC) method is used to estimate the unknown parameters of air-sea oscillator system, which is a model of Eastern Pacific sea surface temperature (SST). Firstly,the posterior probability density function for unknown parameters of air-sea oscillator system is deduced with the Bayesian formula. Secondly, the Adaptive Metropolis algorithm is used to construct the Markov Chains of unknown parameters. And the converged samples are used to calculate the mathematic expectation. The results of numerical experiments show that parameters estimated by the new method have high precision and the noise is filtered effectively from the observations.