Perfec Restart

[arango]

The perfect restart option (PERFECT_RESTART) still needs some work. It is now working in 2D applications. For the soliton example, I get the following errors when compared with the control run:

Perfect restart activated:

Variable       RMS          Correlation

 zeta      1.72573635e-10  1.00000000e-00
 ubar      9.82710975e-11  1.00000000e+00
 vbar      8.24326868e-11  1.00000000e-00

NO perfect restart:

Variable       RMS          Correlation
  
 zeta      2.98162633e-06  9.99999987e-01
 ubar      2.44486068e-06  9.99999992e-01
 vbar      1.91947734e-06  9.99999942e-01

However, it is still not working in 3D applications like the upwelling test case.

John Warner fixed a bug in ini_fields.F and now we really get a perfect restart in 2D applicatioons. The errors in the soliton test case are now:

Variable       RMS          Correlation

 zeta      0.00000000e+00  1.00000000e+00
 ubar      0.00000000e+00  1.00000000e+00
 vbar      0.00000000e+00  1.00000000e+00

We are still hunting for a bug in the perfect restart for 3D applications.

The restart is now working for the basic 3D kernel. The problem was that we were writing output after calling rhs3d. This overwrites the values of ru and rv. We moved output above of rhs3d and created a new routine set_zeta to assign the free-surface to its time averaged value. This affects the adjoint, so we are supporting both options for now. Many thanks to John Warner for his persistance.

We now get zero error in the 3D upwelling test case for either ANA_VMIX or GLS_MIXING:

Variable    RMS          Correlation
 
 zeta   0.00000000e+00  1.00000000e+00
 ubar   0.00000000e+00  1.00000000e-00
 vbar   0.00000000e+00  1.00000000e+00
 u      0.00000000e+00  1.00000000e+00
 v      0.00000000e+00  1.00000000e+00
 temp   0.00000000e+00  1.00000000e-00
 salt   0.00000000e+00  1.00000000e-00

We are still working on the perfect restart for sediment, biology, and other algorithms.