We address ocean modeling capability that has grown exponentially while ocean observation growth has not maintained pace, a situation leading to seemingly degraded forecast skill when model resolution is increased. Skill in predicting ocean instabilities such as mesoscale eddies requires satellite and in situ observations continually correcting numerical model conditions. Observations constrain positions of larger ocean model features, while smaller features are unconstrained. By means of an Observation System Simulation Experiment (OSSE), we show that time–space observation coverage controls the separation of constrained and unconstrained feature scales. Using 1000 independent surface drifters, we show constrained scales have deterministic prediction skill and unconstrained scales predict areas of higher expected errors. The results are shown to be consistent with ensemble forecasts. Separating constrained and unconstrained features, and using information within each appropriately, allows us to manage the present gap between observation and model resolution.

我们解决了海洋建模能力呈指数增长的趋势，而海洋观测的增长却没有保持同步，这种情况会导致在提高模型分辨率时似乎降低了预报技能。预测海洋不稳定性（如中尺度涡旋）的技能需要卫星和原位观测不断校正数值模型条件。观测值会约束较大的海洋模型要素的位置，而较小的要素则不受约束。通过观察系统模拟实验（OSSE），我们表明时空观察覆盖范围控制着受约束和不受约束的特征尺度的分离。使用1000个独立的表面漂移器，我们显示出受约束的尺度具有确定性的预测技能，而不受约束的尺度则可以预测较高预期误差的区域。结果表明与总体预报一致。分离受约束和不受约束的特征，并在每个特征中适当地使用信息，使我们能够管理观测值与模型分辨率之间的当前差距。