Why are there hydrodynamic differences?
Recap
Yesterday, I shared that the N-less run’s hydrodynamics are different than the long hindcast run’s.
As we discussed in our meeting, I have run a few experiments in an attempt to identify the cause of the hydrodynamic deviations.
First, I checked the hydrodynamic deviations at the end of day 1 of the N-less run and verified that I could observe some differences.
Then, I ran the N-less model for one day (2013.01.01) using three different initialization combinations:
- perfect restart (on 5 nodes)
- 10 nodes (using continuation)
- perfect restart and 10 nodes (match long hindcast)
Recall that I initialized the OG N-less run using continuation and 5 nodes and the long hindcast used perfect restart and 10 nodes.
Results
At the end of day 1 in the OG N-less run, it was difficult to see any hydrodynamic differences on a scale of -1 to 1 m/s. However, a scale from -0.1 to 0.1 m/s made the differences very apparent. Thus, -0.1 to 0.1 m/s is the scale I use for all figures on this post.
Also, since u and v tend to have deviations of the same magnitude in the same region, I am only showing u.
Furthermore, the hydrodynamic deviations are most apparent in the surface layer, so I am only showing results from the surface.
Thus, these figures are showing differences in surface u compared to the long hindcast run on a scale from -0.1 to 0.1 m/s.
Figure 1 shows the hydrodynamic difference in Puget Sound, and Figure 2 shows the hydrodynamic difference in the Salish Sea.
Fig 1. Hydrodynamic difference between long hindcast and N-less runs in Puget Sound.
Fig 2. Hydrodynamic difference between long hindcast and N-less runs in Salish Sea.
Summary
While the perfect restart does not fully eliminate hydrodynamic differences, it does significantly reduce them compared to the OG N-less run. However, it seems that both a perfect restart and the same number of nodes (i.e. 10 nodes) are required to reproduce the hydrodynamics of the long hindcast run.