Fig. 2: Results of the optimization process using COCO and STAR from 6 different initial vertical diffusivities.
From: Deep ocean mixing mismatch between model and observational estimates
a Cost function versus iteration number during the optimization. Note that discontinuities arose at each optimization cycle (i.e., at the timing of COCO recalculation with an updated vertical diffusivity). b Vertical profile of vertical diffusivity in the global ocean at each optimization cycle. c Vertical profile of root mean square error (RMSE) for the ocean potential temperature in the global ocean at each optimization cycle. d The same as (c) but for the ocean salinity. The results are shown with black line for T00 (Tsujino et al. 2000), red line for K0.1 (constant value of 0.1), green line for K1 (constant value of 1), blue line for S02 (St. Laurent et al. 2002), yellow line for O13 (Oka and Niwa, 2013), and purple line for dL20 (de Lavergne et al. 2020).
