A new method called the localized weighted ensemble Kalman filter (LWEnKF) was developed to improve ocean data assimilation. It was tested on a model and showed promise for real complex models. The LWEnKF performed well with only 50 particles, accurately assimilating sea-surface temperature, height, and ocean profiles. Compared to other filters, the LWEnKF provided better forecasts for unobserved variables by considering higher-order moments. The local particle filter did not perform as well and requires further investigation.