Each of these tests focused on a local region of the ocean where observations may be compared in a fair way to a column ocean model and avoid non-local fluxes and

long-term/large-scale system adjustments. For these tests, there was a great deal of uncertainty in the ability learn more to observe each experiment’s forcings, and flux corrections were needed in order to attain a reasonable agreement to data. With that caveat, it was found that the KPP provides excellent predictions of the time evolving structure of the observed response. For low latitudes, an alternate strategy was employed to test the KPP. Out of concern that uncertainties in wind forcing were too large, Large Eddy Simulations (LES) were used to simulate observations (Large and Gent, 1999). This is reasonable as LES resolve much of the length and time scales involved in turbulent processes whose net effects KPP is supposed to represent. Because of computational limitations, LES simulations of Large and Gent

(1999) DNA Synthesis inhibitor were limited in space and time and therefore would not capture the longer-term structures and feedbacks that could likely be present in the world’s ocean. In this present study we want to revisit the question of the potential for observational data to constrain uncertainties in KPP mixing physics. In particular we focus in on the issue of how to make a fair comparison between the output of the MITgcm and 65 moored buoys in the TAO/TRITON array in the Tropical Pacific on short (i.e. less than seasonal) time

scales. Later we will use this short-term metric, in addition to metrics that we have devised for longer time scales (Zedler et al., submitted for publication), as a basis for using Cyclooxygenase (COX) Bayesian inference to explore parameter space of the KPP within the MITgcm. Our particular approach for sampling does not require the construction of a statistical surrogate model (Jackson et al., 2004 and Jackson et al., 2008), but the success of its search depends on the reasonableness of how candidate model configurations are tested against data. In this case, there is a certain danger that a close match to observational data could be attained for spurious reasons, perhaps related to errors in our knowledge of the wind forcing, or the many ways a model can exploit compensating errors to get a good match to observational data. We therefore are motivated to create a metric that involves a more direct test of KPP mixing physics by focusing on short time scales and the relationships between wind forcing and the response of sea surface temperatures.