This paper mainly introduces recent applications of nonlinear optimization approach to atmospheric and oceanic sciences. Emphasis is placed on the theoretical framework of the conditional nonlinear optimal perturbation (CNOP) method that is based on nonlinear optimization, and the works aiming to make it comprehensive. The application progresses of the CNOP method in atmospheric and oceanic sciences are briefly presented, including the applications for ensemble forecast, predictability of some high-impact ocean-atmospheric environmental events, recognition of model parameter sensitivities, assessments of model tendency error and boundary condition error. In addition, we also discuss the difficulties and challenges for the application of the CNOP approach and suggest directions for future development.