--- Welcome to the official ADCIRCWiki site! The site is currently under construction, with limited information. ---

For general information, see the ADCIRC site, ADCIRC FAQ, or the ADCIRC Wikipedia page. For model documentation not yet available on the wiki, see the ADCIRC site. New content is being continuously added to the ADCIRCWiki, and material from the main ADCIRC site will be gradually transitioned over to the wiki.

Horizontal eddy viscosity: Difference between revisions

From ADCIRCWiki
Jump to navigation Jump to search
(→‎Smagorinsky Eddy Viscosity: revised language on derivative...it appears both spatial and temporal derivatives of eddy viscosity are neglected)
(clarification on negative ESLM, added notes on internal flags)
 
(2 intermediate revisions by the same user not shown)
Line 9: Line 9:
== Smagorinsky Eddy Viscosity ==
== Smagorinsky Eddy Viscosity ==
ADCIRC also allows a [https://en.wikipedia.org/wiki/Turbulence_modeling#Smagorinsky_model_for_the_sub-grid_scale_eddy_viscosity Smagorinsky-type] turbulence closure model<ref name=smag1>Smagorinsky, J. “General Circulation Experiments with the Primitive Equations.” Monthly Weather Review 91, no. 3 (March 1, 1963): 99–164. https://doi.org/10.1175/1520-0493(1963)091<0099:GCEWTP>2.3.CO;2.
ADCIRC also allows a [https://en.wikipedia.org/wiki/Turbulence_modeling#Smagorinsky_model_for_the_sub-grid_scale_eddy_viscosity Smagorinsky-type] turbulence closure model<ref name=smag1>Smagorinsky, J. “General Circulation Experiments with the Primitive Equations.” Monthly Weather Review 91, no. 3 (March 1, 1963): 99–164. https://doi.org/10.1175/1520-0493(1963)091<0099:GCEWTP>2.3.CO;2.
</ref>.  There are currently two slightly different ways of enabling this in the model.  One is by specifying a negative value for <code>[[ESLM]]</code> in the [[fort.15 file]].  If this is done, the Smagorinsky turbulence closure is enabled and the absolute value of <code>ESLM</code> is used as the coefficient.  Alternatively, the <code>Smag_Control</code> namelist can be used, e.g.<br/>
</ref>.  There are currently two slightly different ways of enabling this in the model.  One is by specifying a negative value for <code>[[ESLM]]</code> in the [[fort.15 file]].  If this is done, the Smagorinsky turbulence closure is enabled and the absolute value of <code>ESLM</code> is used as the coefficient.  Alternatively, the <code>Smag_Control</code> namelist can be used, e.g. by adding the following to the end of the fort.15 file <br/>
<code>&Smag_Control SMAG_LOWER_LIM=1.0d-8, SMAG_UPPER_LIM=100 /</code><br/>
<code>&Smag_Control SMAG_LOWER_LIM=1.0e-8, SMAG_UPPER_LIM=100 /</code><br/>
When this is done, the value of <code>ESLM</code> is taken as the coefficient.  This also allows the user to optionally specify lower and upper bounds on the actual eddy viscosity determined via the Smagorinsky formulation.  The values given in the above example of 1E-8 m<sup>2</sup>/s and 100 m<sup>2</sup>/s are the default values.  There is no way to supply upper and lower bounds with the other method.   
Here again, the absolute value of <code>ESLM</code> is taken as the coefficient, though a negative <code>ESLM</code> value isn't necessary if the namelist is usedThe namelist also allows the user to optionally specify lower and upper bounds on the actual eddy viscosity determined via the Smagorinsky formulation.  The values given in the above example of 1E-8 m<sup>2</sup>/s and 100 m<sup>2</sup>/s are the default values.  There is no way to supply upper and lower bounds with the other method.   


If the [[fort.13_file#Eddy_Viscosity|<code>average_horizontal_eddy_viscosity_in_sea_water_wrt_depth</code>]] nodal attribute is used when the Smagorinsky eddy viscosity is employed, then the nodal attribute values are taken to be the Smagorinsky coefficient values.   
If the [[fort.13_file#Eddy_Viscosity|<code>average_horizontal_eddy_viscosity_in_sea_water_wrt_depth</code>]] nodal attribute is used when the Smagorinsky eddy viscosity is employed, then the nodal attribute values are taken to be the Smagorinsky coefficient values.   


=== Usage Notes ===
=== Usage Notes ===
The two methods currently result in different flags being set in the code, which have slightly inconsistent behavior.  Specifically, the "negative <code>ESLM</code> method" results in a check being run at model initialization to verify that the default (Kolar-Gray) formulation of the lateral stresses in the GWCE is not in use, and the model errors out if it is.  That check is not done in the namelist method.  The lateral stress formulation is specified by the first digit in a [[IM#Six-digit_IM_Codes|six-digit IM code]] being set to 1.  It is also set for any regular IM value for 2D ADCIRC.  Mathematically, the Kolar-Gray formulation of the lateral stress does not take into account a derivative of the eddy viscosity terms, meaning it is technically inconsistent with a Smagorinsky turbulence closure model.  However, it is not clear whether the time derivative term is ever large enough to be consequential, and therefore merit departure from ADCIRC's default GWCE formulation.
The two methods currently result in different flags being set in the code, which have slightly inconsistent behavior.  Specifically, the "negative <code>ESLM</code> method" sets <code>CSmag_Eh=.TRUE.</code> and results in a check being run at model initialization to verify that the default (Kolar-Gray) formulation of the lateral stresses in the GWCE is not in use, and the model errors out if it is.  That check is not done in the namelist method, which sets <code>Smag_Comp_Flag=.TRUE.</code>.  The lateral stress formulation is specified by the first digit in a [[IM#Six-digit_IM_Codes|six-digit IM code]] being set to 1.  It is also set for any regular IM value for 2D ADCIRC.  Mathematically, the Kolar-Gray formulation of the lateral stress does not take into account a derivative of the eddy viscosity terms, meaning it is technically inconsistent with a Smagorinsky turbulence closure model.  However, it is not clear whether the time derivative term is ever large enough to be consequential, and therefore merit departure from ADCIRC's default GWCE formulation.


==References==
==References==
<references />
<references />

Latest revision as of 11:23, 6 July 2020

Horizontal eddy viscosity is a term in the momentum equations to cover the well-known turbulence closure problem. The term is also useful numerically for increased stability. For background on the turbulence closure problem, see, e.g. the AMS glossary definition, the Wikipedia entry on turbulence modeling, or a fluid mechanics textbook such as Kundu's[1]. Readers unfamiliar with these concepts should note that, contrary to the name, eddy viscosity is unrelated to true viscosity, even though the units (length squared over time) are the same. Typically, one should expect eddy viscosity to be many orders of magnitude larger than the viscosity of actual water (around 1x10-6 meters2/second for seawater). In ADCIRC, the user can either specify a constant in time horizontal eddy viscosity or employ the Smagorinsky model to update the horizontal eddy viscosity based on local flow conditions with each time step. In both cases the horizontal eddy viscosity or Smagorinsky coefficient can be either spatially constant or spatially varying.

User-Specified Constant in Time Eddy Viscosity

A spatially constant eddy viscosity can be supplied via the ESLM parameter in the fort.15 file. Alternatively, the average_horizontal_eddy_viscosity_in_sea_water_wrt_depth nodal attribute permits a spatially variable eddy viscosity.

Usage Notes

Commonly used values range from 1 to 50 meters2/second, with 10 being a good starting point for many coastal ocean modeling scenarios. By the nature of turbulence closure, one might expect the value to be smaller as mesh resolution increases. However, this may not always be the case since areas with high resolution may also be more turbulent. In particular, some modelers use a larger value in overland areas, which can also improve stability. However, specifying a value that is too large can induce instabilities, as well.

Smagorinsky Eddy Viscosity

ADCIRC also allows a Smagorinsky-type turbulence closure model[2]. There are currently two slightly different ways of enabling this in the model. One is by specifying a negative value for ESLM in the fort.15 file. If this is done, the Smagorinsky turbulence closure is enabled and the absolute value of ESLM is used as the coefficient. Alternatively, the Smag_Control namelist can be used, e.g. by adding the following to the end of the fort.15 file
&Smag_Control SMAG_LOWER_LIM=1.0e-8, SMAG_UPPER_LIM=100 /
Here again, the absolute value of ESLM is taken as the coefficient, though a negative ESLM value isn't necessary if the namelist is used. The namelist also allows the user to optionally specify lower and upper bounds on the actual eddy viscosity determined via the Smagorinsky formulation. The values given in the above example of 1E-8 m2/s and 100 m2/s are the default values. There is no way to supply upper and lower bounds with the other method.

If the average_horizontal_eddy_viscosity_in_sea_water_wrt_depth nodal attribute is used when the Smagorinsky eddy viscosity is employed, then the nodal attribute values are taken to be the Smagorinsky coefficient values.

Usage Notes

The two methods currently result in different flags being set in the code, which have slightly inconsistent behavior. Specifically, the "negative ESLM method" sets CSmag_Eh=.TRUE. and results in a check being run at model initialization to verify that the default (Kolar-Gray) formulation of the lateral stresses in the GWCE is not in use, and the model errors out if it is. That check is not done in the namelist method, which sets Smag_Comp_Flag=.TRUE.. The lateral stress formulation is specified by the first digit in a six-digit IM code being set to 1. It is also set for any regular IM value for 2D ADCIRC. Mathematically, the Kolar-Gray formulation of the lateral stress does not take into account a derivative of the eddy viscosity terms, meaning it is technically inconsistent with a Smagorinsky turbulence closure model. However, it is not clear whether the time derivative term is ever large enough to be consequential, and therefore merit departure from ADCIRC's default GWCE formulation.

References

  1. Pijush K. Kundu, Ira M. Cohen, David R. Dowling (2012). Fluid Mechanics.
  2. Smagorinsky, J. “General Circulation Experiments with the Primitive Equations.” Monthly Weather Review 91, no. 3 (March 1, 1963): 99–164. https://doi.org/10.1175/1520-0493(1963)091<0099:GCEWTP>2.3.CO;2.