Unmanned underwater vehicles (UUVs) share information through underwater acoustic communication to realize cooperative localization, which can expand the sensing range and accomplish complex tasks, e.g., underwater exploration and reconnaissance, marine engineering, and underwater operations. However, due to the lack of underwater observation information, the configuration will directly affect the accuracy of UUVs cooperative localization. The former researches are based on algebraic derivation to analyze the observability of UUVs configuration, which is not intuitive and convenient in practical application. A new analysis method based on error ellipse for UUVs configuration and its geometric interpretation are proposed. On one hand, it proves the consistency of error ellipse index with traditional Fisher Information Matrix (FIM) and Cramer Rao lower bound (CRLB) index. On the other hand, it improves the intuitiveness and convenience for analyzing the observability of UUVs configuration. In order to validate the effectiveness of the proposed method, three UUVs are simulated, including one slave UUV and two master UUVs. Five different configurations of UUVs with separation angles 30°, 60°, 90°, 120°and 150° are selected. The experimental results show that the error ellipse index can characterize the accuracy of UUVs cooperative localization, and improve the intuitiveness of the observability of UUVs configurations.