Difference between revisions of "IM"
Taylorgasher (talk  contribs) (finished adding flag variables) 
(→Sixdigit IM Codes) 

Line 94:  Line 94:  
      
      
−    +   A value of 46 does the same as 13 (same order) but in baroclinic mode<br/><code>CBaroclinic=.TRUE.</code> {{ADC versionversion=55relation=eq}} 
    
 5   5  
Line 102:  Line 102:  
      
      
−    +    
    
 6   6  
Line 113:  Line 113:  
}  }  
−  A common code combination is <code>IM=111112</code>, which is identical to the default <code>111111</code> (same as <code>IM=0</code>), but simulates in explicit masslumping mode. Note that <code>[[A00, B00, C00]]</code> must be set to <code>0.0 1.0 0.0</code> when in this mode. Lumped explicit mode is a useful alternative to the (default) semiimplicit consistent GWCE mass matrix mode, because the latter requires a matrix solve that increases computational time and memory. By comparison, the explicit masslumping mode is about twice as fast and scales to fewer grid nodes per computational core.<ref>S. Tanaka, S. Bunya, J.J. Westerink, C. Dawson, R.A. Luettich, Scalability of an Unstructured Grid Continuous Galerkin Based Hurricane Storm Surge Model, J. Sci. Comput. 46 (2011) 329–358. doi:10.1007/s1091501094021</ref> Moreover, for model setups that are sufficiently resolved in space and time, differences in the solution between approaches should be small. Though, many users have reported somewhat lower stability in lumped explicit mode.  +  A common code combination is <code>IM=111112</code>, which is identical to the default <code>111111</code> (same as <code>IM=0</code>), but simulates in explicit masslumping mode. Note that <code>[[A00, B00, C00]]</code> must be set to <code>0.0 1.0 0.0</code> when in this mode. Lumped explicit mode is a useful alternative to the (default) semiimplicit consistent GWCE mass matrix mode, because the latter requires a matrix solve that increases computational time and memory. By comparison, the explicit masslumping mode is about twice as fast and scales to fewer grid nodes per computational core.<ref>S. Tanaka, S. Bunya, J.J. Westerink, C. Dawson, R.A. Luettich, Scalability of an Unstructured Grid Continuous Galerkin Based Hurricane Storm Surge Model, J. Sci. Comput. 46 (2011) 329–358. doi:10.1007/s1091501094021</ref> Moreover, for model setups that are sufficiently resolved in space and time, differences in the solution between approaches should be small. Though, many users have reported somewhat lower stability in lumped explicit mode. 
== References ==  == References ==  
<references />  <references /> 
Revision as of 14:07, 10 July 2020
IM
is an important parameter in the fort.15 file that defines numerical model formulation and dimension. Among other things, IM
specifies whether ADCIRC is solved in twodimensional depthintegrated (2DDI) or in threedimensions (3D), solution of the governing equations is semiimplicit or explicit in time, and whether the model formulation is barotropic or baroclinic. Popular values for 2D barotropic ADCIRC include IM=0
and IM=111112
, though the latter also requires modifying A00, B00, C00
.
Default IM Values
Default simulation option combinations can be specified through single or double digit values, some of which are shortcuts to the sixdigit codes described in the next heading.
The available IM
values are specified in the table below and in the following section on 6digit values:
IM Value  Sixdigit Equivalent  Description 

0  111111  Barotropic 2DDI 
1  611111  Barotropic 3D velocitybased momentum 
2    Barotropic 3D stressbased momentum 
10    Barotropic 2DDI with passive scalar transport 
11    Barotropic 3D velocitybased momentum with passive scalar transport 
20  111113  Baroclinic 2DDI 
21  611113  Baroclinic 3D velocitybased momentum 
30    Baroclinic 2DDI with passive scalar transport 
31    Baroclinic 3D velocitybased momentum with passive scalar transport 
Note that all default IM
values employ the semiimplicit consistent GWCE mass matrix solver. It has less numerical error and tends to be more stable than the explicit masslumping approach at the expense of computational time and memory.
Sixdigit IM Codes
For finegrained control of various options sixdigit codes for IM
can be specified. Each digit represents a specific option regarding the dimension and the formulation of certain terms or integration methods in the GWCE or momentum equations.
The available options for each digit are specified below, with the first digit being the leftmost. The internal flags that are set are listed to help users dig through the code.
Value  Digit 1: 2DDI/3D, Lateral Stress in GWCE^{[1]}  Digit 2: Advection in GWCE^{[2]}  Digit 3: Lateral Stress in Momentum^{[1]}  Digit 4: Advection in Momentum^{[2]}  Digit 5: Area Integration in Momentum  Digit 6: GWCE Mass Matrix, Barotropic/Baroclinic  

1 (default)  2DDI, KolarGray fluxbasedCGWCE_LS_KGQ=.TRUE.

Non conservativeCGWCE_Advec_NC=.TRUE.

Integration by parts, velocitybasedCME_LS_IBPV=.TRUE.

Non conservativeCME_New_NC=.TRUE.

Corrected CME_AreaInt_Corr=.TRUE.

Consistent (implicit for linear part of gravity wave term), barotropicILump=0
 
2  2DDI, 2part fluxbasedCGWCE_LS_2PartQ=.TRUE.

Conservative form 1CGWCE_Advec_C1=.TRUE.

Integration by parts, fluxbasedCME_LS_IBPQ=.TRUE.

Conservative form 1CME_New_C1=.TRUE.

Original CME_AreaInt_Orig=.TRUE.

Lumped (explicit), barotropicCGWCE_Lump=.TRUE. , ILump=1
 
3  2DDI, 2part velocitybasedCGWCE_LS_2PartV=.TRUE.

Conservative form 2CGWCE_Advec_C2=.TRUE.

Integration by parts, velocitybased symmetricalCME_LS_IBPSV=.TRUE.

Conservative form 2CME_New_C2=.TRUE.

  Consistent (implicit for full gravity wave term), barotropicCGWCE_HDP=.TRUE. , IFNL_HDP=1 , ILump=0
 
4  2DDI, 2part fluxbased symmetricalCGWCE_LS_2PartSQ=.TRUE.

  Integration by parts, fluxbased symmetricalCME_LS_IBPSQ=.TRUE.

    A value of 46 does the same as 13 (same order) but in baroclinic modeCBaroclinic=.TRUE.
 
5  2DDI, 2part velocitybased symmetricalCGWCE_LS_2PartSV=.TRUE.

  2 Part, velocitybased (not implemented)CME_LS_2PartV=.TRUE.

      
6  3D, KolarGray fluxbasedC2DDI=.FALSE. , CGWCE_LS_KGQ=.TRUE. , C3D=.TRUE. , C3DVS=.TRUE. , ILump=0

  2 Part, fluxbased (not implemented)CME_LS_2PartQ=.TRUE.

    See above 
A common code combination is IM=111112
, which is identical to the default 111111
(same as IM=0
), but simulates in explicit masslumping mode. Note that A00, B00, C00
must be set to 0.0 1.0 0.0
when in this mode. Lumped explicit mode is a useful alternative to the (default) semiimplicit consistent GWCE mass matrix mode, because the latter requires a matrix solve that increases computational time and memory. By comparison, the explicit masslumping mode is about twice as fast and scales to fewer grid nodes per computational core.^{[3]} Moreover, for model setups that are sufficiently resolved in space and time, differences in the solution between approaches should be small. Though, many users have reported somewhat lower stability in lumped explicit mode.
References
 ↑ ^{1.0} ^{1.1} K.M. Dresback, R.L. Kolar, R.A. Luettich, Jr. (2005). On the Form of the Momentum Equation and Lateral Stress Closure Law in Shallow Water Modeling, in: Estuar. Coast. Model., American Society of Civil Engineers, Reston, VA, 399–418. doi:10.1061/40876(209)23
 ↑ ^{2.0} ^{2.1} K.M. Dresback, R.L. Kolar, J.C. Dietrich (2005). On the Form of the Momentum Equation for Shallow Water Models Based on the Generalized Wave Continuity Equation: Conservative vs. NonConservative. Advances in Water Resources, 28(4), 345358. doi:10.1016/j.advwatres.2004.11.011
 ↑ S. Tanaka, S. Bunya, J.J. Westerink, C. Dawson, R.A. Luettich, Scalability of an Unstructured Grid Continuous Galerkin Based Hurricane Storm Surge Model, J. Sci. Comput. 46 (2011) 329–358. doi:10.1007/s1091501094021