Reference and mean models in the perfect scenario.

Top row: Reference model (Panel (a)) used to generate observations (cf. Fig 13(a)), and the mean model (Panel (b)) of the initial ensemble. Bottom row: mean of the final ensemble obtained through data assimilation without any model-error correction (MEC) (Panel (c)), and the corresponding mean when MEC is still adopted (Panel (d)) even though the forward simulator is perfect.