Deep-Sea Research II 49 (2002) 363–401 Oceanic vertical exchange and new production: a comparison between models and observations Anand Gnanadesikan a, *, Richard D. Slater a , Nicolas Gruber b , Jorge L. Sarmiento a a AOS Program, Princeton University, P.O. Box CN710, Princeton, NJ 08544, USA b Department of Geosciences, University of California, Los Angeles, CA, USA Accepted 22 June 2001 Abstract This paper explores the relationship between large-scale vertical exchange and the cycling of biologically active nutrients within the ocean. It considers how the parameterization of vertical and lateral mixing effectsestimatesofnewproduction(definedasthenetuptakeofphosphate).Abaselinecaseisrunwithlow vertical mixing in the pycnocline and a relatively low lateral diffusion coefficient. The magnitude of the diapycnaldiffusioncoefficientisthenincreasedwithinthepycnocline,withinthepycnoclineoftheSouthern Ocean, and in the top 50 m; while the lateral diffusion coefficient is increased throughout the ocean. It is shownthatitispossibletochangelateralandverticaldiffusioncoefficientssoastopreservethestructureof the pycnocline while changing the pathways of vertical exchange and hence the cycling of nutrients. Comparisons between the different models reveal that new production is very sensitive to the level of vertical mixing within the pycnocline, but only weakly sensitive to the level of lateral and upper ocean diffusion. The results are compared with two estimates of new production based on ocean color and the annual cycle of nutrients. On a global scale, the observational estimates are most consistent with the circulationproducedwithalowdiffusioncoefficientwithinthepycnocline,resultinginanewproductionof around 10 GtC yr 1 : On a regional level, however, large differences appear between observational and model based estimates. In the tropics, the models yield systematically higher levels of new production than the observational estimates. Evidence from the Eastern Equatorial Pacific suggests that this is due to both biasesinthedatausedtogeneratetheobservationalestimatesandproblemswiththemodels.IntheNorth Atlantic, the observational estimates vary more than the models, due in part to the methodology by which the nutrient-based climatology is constructed. In the North Pacific, the modelled values of new production areallmuchlowerthantheobservationalestimates,probablyasaresultofthefailuretoformintermediate *Corresponding author. E-mail addresses: gnana@splash.princeton.edu (A. Gnanadesikan), rdslater@splash.princeton.edu (R.D. Slater), jls@splash.princeton.edu (J.L. Sarmiento). 0967-0645/01/$-see front matter r 2001 Published by Elsevier Science Ltd. PII:S0967-0645(01)00107-2