Diapycnal mixing in Gulf Stream meanders

Citation
A. Rodriguez-santana et al., Diapycnal mixing in Gulf Stream meanders, J GEO RES-O, 104(C11), 1999, pp. 25891-25912
Citations number
57
Categorie Soggetti
Earth Sciences
Journal title
JOURNAL OF GEOPHYSICAL RESEARCH-OCEANS
ISSN journal
21699275 → ACNP
Volume
104
Issue
C11
Year of publication
1999
Pages
25891 - 25912
Database
ISI
SICI code
0148-0227(19991115)104:C11<25891:DMIGSM>2.0.ZU;2-E
Abstract
We use historical data [Bane ei al., 1981], interpolated to isopycnic coord inates, to examine the possibility of significant diapycnal mixing within t he upper thermocline layers of the Gulf Stream. The data consist of 28 air- dropped expendable bathythermograph (AXBT) sections in the northern Blake P lateau distributed in five different surveys done within 8 days. From the d ata we obtain a separation index between isopycnals j = p partial derivativ e z/partial derivative rho, where z is the depth of an isopycnal and p is t he density; the diapycnal shear partial derivative v/d rho, where v is the geostrophic velocity; and the gradient Richardson number Ri. Following Pele gri and Csanady [1994], we postulate that the material derivative of the de nsity, or density tendency omega(rho) = D rho/Dt, is the result of small-sc ale instabilities related to near-critical Ri values. We present the distri bution of these quantities (j, partial derivative v/partial derivative rho, Ri, and w(rho)) and the diapycnal convergence/divergence (partial derivati ve w(rho)/partial derivative rho) over isopycnals and on sections normal to the coast. The results show the passage of steep meanders being related to anomalously low j values (strong density gradients) within the upper therm ocline layers and the cyclonic filaments of the stream. The statically Stab le upper thermocline layers, however, are concurrent with large diapycnal s hear: and turn out to be dynamically unstable, characterized by low Ri and high-density tendencies and diapycnal convergence/divergence, The errors in volved in calculating the dependant variables from AXBT data are assessed u sing both an error propagation approach and a Monte Carlo error simulation. These errors, although significant, are not large enough to modify the obs erved patterns substantially.