The main objective of these numerical experiments is to explore the
possible role of fine-scale topographic ripples seen in 
ANSLOPE data on downslope transport of dense overflows.   
To this end, idealized simulations are carried using
nek5000 (Fischer, 1997), which integrates the three-dimensional
Boussinesq equations using the spectral element method (SEM).
This method combines the geometrical flexibility of finite element
formalism with the numerical advantages (low spurious dissipation
and dispersion errors) of the spectral expansion. The model has been
previously employed for gravity current simulations in  
Ozgokmen et al. (2004a,b; 2006, 2007, 2008). 

The basic model configuration consists of a domain with dimensions
of Lx=6, Ly=3 and H=1 in x, y and z-directons, respectively, as scaled by
the radius of deformation. The overflow with thickness of h_o/H=0.2 is 
introduced into the interior with an constant background slope of
s_b=0.1. The overflow undergoes geostrophic adjustment (anslope26)
until reaching a final state that is characterized by eddies 
along the upper part of the slope and a slow, down-slope Ekman drain
near the bottom. Then, topographic variations ("ripples")
are introduced in the lateral (y) direction with an amplitude of  
h_r/h_o=0.5. Even ripples with along-slope wavelengths of Rd
(anslope25) seem to enhance down-slope transport. As the ripple
wavelength is decreased (anslope25-22), the character of the overflow
seems to change significantly with respect to that over perfectly
smooth topography. An obvious conclusion from these experiments
is that small-scale details of bathymetry may matter for overflow dynamics.
(Yes, the devil is in details; this is when climate modelers will
run away from this poster.)