Abstract: The traditional covariance matrix estimators ignore both the change of statistical consistency of sea clutter with radar range resolution and the limitation of integrable pulse by the number of available secondary cells. To solve the problems, a new covariance matrix estimator based on the diagonal loading is proposed. The loading coefficient changes with the statistical consistency of clutter data adaptively. The new estimator has less relative errors and a good condition number. Moreover, the loading coefficient is approximately optimal. Experimental results to real sea clutter data show that the detector’s performance is highly improved by using the new estimation.