202 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 35 The Impact of a Bottom Boundary Layer Scheme on the North Atlantic Ocean in a Global Coupled Climate Model YONG MING TANG AND MALCOLM J. ROBERTS Met Office, Hadley Centre for Climate Prediction and Research, Exeter, United Kingdom (Manuscript received 8 September 2003, in final form 7 July 2004) ABSTRACT Although the overflow and descent of cold, dense water across the Greenland–Iceland–Scotland ridge is the principal means for the maintenance of the thermohaline circulation in the North Atlantic Ocean, this feature is not adequately treated in global ocean numerical models. In this paper, a bottom boundary layer scheme is introduced into the HadCM3 coupled atmosphere–ocean–sea ice general circulation climate model, in order to give an improved representation of cold water formation in the North Atlantic Ocean. The scheme uses a simple terrain-following bottom boundary layer incorporated into the ocean general circulation model; only the tracer tendencies are evaluated in the bottom boundary layer, with the velocities taken from the near-bottom interior values. It is found that with the bottom boundary layer scheme, there are several significant effects on the deep water formation and flow. The overflow of dense water from the Nordic Seas into the North Atlantic Seas is improved with the introduction of the authors’ bottom boundary layer scheme. Further, the thermohaline circulation is reduced in strength, but is also deeper, when com- pared with simulations without any bottom boundary layer scheme. There is also a stronger flow along the northwestern boundary, a more southerly location of the North Atlantic Current, and a stronger and larger subpolar gyre. Overall, these effects are an improvement when compared with climatology, although some differences remain. 1. Introduction the northern Atlantic. Unfortunately, global ocean nu- merical models currently being used for climate studies The thermohaline circulation (THC) in the North do not yet deal with this process adequately (e.g., Hirst Atlantic Ocean plays a fundamental role in the global and McDougall 1996), primarily because the model climate system (Broecker 1991). The THC transports resolution does not sufficiently resolve the continental warm water from low to high latitudes, where it is slopes (Roberts and Wood 1997). In addition, the z- cooled, releasing heat into the atmosphere. The cooling coordinate models cannot properly represent a down- of the water makes it sufficiently dense that it mixes to slope flow, and even when dense overflow water is pro- great depth, and then returns equatorward as a deep duced, it is artificially diluted through convective ho- and cold current. One consequence of this is the rela- mogenization instead of being transported downslope tively mild climate of the high-latitude northern regions (Winton et al. 1998). relative to the analogous regions elsewhere. It has been suggested that these problems can be The overflow and descent of cold, dense waters alleviated by incorporating a bottom boundary layer across the Greenland–Iceland–Scotland (GIS) ridge is (BBL) parameterization scheme (e.g., Hirst and Mc- the principal means by which the deep North Atlantic Dougall 1996). While the importance of bottom bound- Ocean is ventilated and renewed. This overflow is an ary layers has been recognized for many years, it is only important control point for the THC and the conse- recently that consideration has been given to incorpo- quent production and spreading of deep water with the rating parameterizations of them into ocean general cir- correct properties is of critical importance for a valid culation models (OGCM). We note here, in particular, description of the large-scale thermohaline circulation, the approaches adopted by Beckmann and Döscher and thus for an understanding of the climate system in (1997), A. Gnanadesikan (1997, unpublished manu- script), Killworth and Edwards (1999), Döscher and Beckmann (2000), Song and Chao (2000), and Nakano Corresponding author address: Yong Ming Tang, Met Office, and Suginohara (2002). Each of these couple a BBL Joint Centre for Mesoscale Meteorology (JCMM), Meteorology Building, University of Reading, P.O. Box 243, Earley Gate, model to a z-coordinate OGCM, and find various im- Reading, Berkshire RG6 6BB, United Kingdom. provement over simulations with no BBL parameter- E-mail: [email protected] ization present. Unauthenticated | Downloaded 09/30/21 12:11 AM UTC JPO2671 FEBRUARY 2005 T A N G A N D ROBERTS 203 Beckmann and Döscher (1997, hereinafter BD97) the BBL allowed strong downslope flows to develop. couple a terrain-following (␴ coordinate) BBL to a z- However, the code requires a Rayleigh friction term in coordinate OGCM, but simplify the full BBL equations the momentum equations, with a coefficient equal to by evaluating only the tracer tendencies, with the ve- the Coriolis parameter f in order to ensure that the locity field taken from the near-bottom interior values. bottom boundary layer thickness was of the same order The BBL depth hBBL cannot evolve in time, being de- as the Ekman depth. termined by the lowest ␴ level. They also introduce an Song and Chao (2000) use the KE99 formulation, but empirical parameter ␥ to measure the relative coupling with a different method of evaluating the pressure gra- between the BBL and the interior, 0.0 Յ ␥ Յ 1.0, where dient in the BBL, in which they apparently assume that ␥ ϭ 0.0 represents a simple z-coordinate system, and the pressure gradient is constant across the BBL; this ␥ ϭ 1.0 means that the BBL is fully implemented in the differs from KE99 who calculate a depth-averaged ␴-coordinate system; there is a corresponding modifi- pressure gradient. They also use a terrain-following co- cation to the BBL tracer equations. This approach is ordinate, which takes account of the bottom slope in a appealingly simple, although it does not give the correct different way. They report some results from idealized velocity field in the BBL and uses a fixed bottom model simulations that are broadly similar to the results boundary layer depth. BD97 applied their scheme to an of KE99, but have some differences. idealized box model, and showed that including a BBL The approach of Nakano and Suginohara (2002, resulted in enhanced downslope tracer transport. Loh- hereinafter NS02) is similar to G97, but they include the mann (1998, hereinafter L98) inserted the BD97 advection terms in the momentum equations. Again scheme into a coupled atmosphere–ocean–sea ice hBBL is a constant, set equal to 100 m, and the code, like model with idealized geometry for the North Atlantic G97, requires a Rayleigh friction term with a coefficient Ocean, and with climatological fixed wind forcing, in a equal to the Coriolis parameter f. They incorporate sensitivity study of the thermohaline circulation. Later their BBL scheme into an OGCM, and find that they Döscher and Beckmann (2000, hereinafter DB00) ap- can produce a realistic overslope/downslope flow in the plied a slight modification of their method to the for- Northern Atlantic and around Antarctica. mation of North Atlantic Deep Water (NADW) and As was pointed out in Winton et al. (1998) and G97, obtained some improvement over an OGCM without level coordinate models in particular have trouble rep- any BBL. resenting the thin boundary layer flows down sloping Killworth and Edwards (1999, hereinafter KE99) topography. This is because of 1) an inability to repre- represent the BBL as a two-dimensional slab boundary sent the correct pressure gradient for a thin slope; 2) an layer, which occupies some temporally and spatially inability to resolve the Ekman layer, which breaks geo- varying layer at the ocean floor in a continuous model. strophy to the point where the downslope flow can oc- The BBL equations consist of the tracer equations, cur; and 3) excessive convective entrainment, which di- depth-integrated momentum equations with an aver- lutes the properties of the overflow water. Some of the aged pressure gradient term, and an equation for the above schemes attempt to address all of these prob- evolution of hBBL, the bottom boundary layer depth. lems. However, we choose to implement the BD97 for- Entrainment/detrainment is implemented through a mulation into a fully coupled atmosphere–ocean–sea turbulence closure parameterization. The numerical ice general circulation model as a first step in investi- version of this KE99 scheme is presented for a z-coor- gating the impact of a BBL model on the coupled cli- dinate system, but it has yet to be fully implemented in mate. This formulation only addresses point 3 above, an OGCM, although some idealized box model simu- particularly given the resolution limitations imposed by lations were reported (see also Song and Chao 2000, long climate integrations, but it is possible that this is discussed briefly below). sufficient to represent the most important aspects of Worker A. Gnanadesikan (1997, unpublished manu- deep overflows for climate. script available online at http://www.gfdl.noaa.gov/ Therefore the purpose of this paper is to describe the ϳa1g/bbl.html, hereinafter G97) proposes a generaliza- impact of the Beckmann and Döscher (1997) BBL tion of BD97 in which the BBL momentum equations model in our climate model. Since the atmosphere and are reinstated, albeit without the advective terms, and ocean are fully coupled, we only need to specify the driven by a pressure gradient computed within the initial state of the atmosphere, in contrast to simula- BBL, thus in essence following the KE99 approach. tions with an OGCM alone where the atmospheric forc- However, hBBL is not allowed to evolve in time and ing is always specified. There is also no need to provide instead is fixed at a constant value of 50 m. Conse- any artificial supply of dense water at the northern quently, turbulence is modeled with a vertical turbulent boundary of the model, as would be the case for a eddy coefficient, rather than with a entrainment/ regional version of an OGCM.