Comparing the overflow of dense water in isopycnic and cartesian models with tracer observations in the eastern Mediterranean

Isopycnic and cartesian model simulations for the overflow and spreading of dense water are compared with each other and with independent transient-tracer observations. This case study is performed for Adriatic dense water overflowing into the deep eastern Mediterranean with chlorofluoromethane (CFC-12) observations used to test the model simulations. The realism of both types of model simulation depends on the representation of diapycnal mixing. In the cartesian model, convective adjustment and mixing dilute the overflow of Adriatic dense water and lead to unrealistic vertical homogenization. Incorporating a modified convection scheme emphasizing the sinking of dense fluid, rather than its mixing, leads to a more realistic penetration of the dense overflow. In the isopycnic model, there is an improved simulation of the overflow, which leads to the density contrast of the deep Mediterranean waters being maintained. However, there is too low a CFC-12 concentration at mid-depths unless explicit diapycnal mixing is incorporated. In each model, the different spreading of dense water is associated with a different bottom pressure torque and depth-integrated transport, and hence with contrasting tracer distributions throughout the water column.