search

CH-Earth / summa

Structure for Unifying Multiple Modeling Alternatives:
http://www.ral.ucar.edu/projects/summa
GNU General Public License v3.0
79 stars 103 forks source link

Fix/restart canopy ice #481

Closed wknoben closed 2 years ago

wknoben commented 2 years ago

Make sure all the relevant boxes are checked (and only check the box if you actually completed the step):

Test checkboxes n/a here since this affects restarts only.

Issue

Canopy ice content can occasionally go above zero for above-zero temperatures as a result of frost (negative sublimation) appearing on the canopy. In a regular run, this ice typically becomes liquid content on the next timestep (assuming temperature remains above zero). See figure for results from 28 years of half-hourly point-scale simulations, showing that this happens during regular simulations. A restart file created when both scalarCanopyIce > 0 and scalarCanopyTemp > 0 cannot be used to restart the run, because this condition is found in check_icond.f90 and the run gracefully exists.

regular_simulation

Solution

The check is mainly in place to not allow the user to use unrealistic system states as initial conditions. Replacing the current graceful exit with a warning printed to screen retains similar functionality (user is aware of something being off - provided they check the log), while also allowing SUMMA to use its own restart files. In the current implementation, ice content up to 1E-3 is allowed (but flagged in the output log), whereas ice content above this threshold is still a reason for a graceful exit.

Test cases

Selected 1 day from the 28-year long run to investigate this issue further. Restart file generated on 2003-01-01 09:00 which matches the critical conditions of scalarCanopyIce > 0 and scalarCanopyTemp > 0. The short run up to restart file creation is identical to the 28-year run up to that point; see below. At this point:

On the following timestep, all remaining ice content becomes liquid and this remains in place until the liquid content goes up further on t = 20:30 h.

create_restart

Case 1: restart file + regular SUMMA

As expected, the current implementation gracefully doesn't run.

Case 2: zero canopy ice + regular SUMMA

Setting scalarCanopyIce to 0 allows the run to start but leads to differences between the regular (long) run and the restarted run, because this artificially removes some water from the system. See below.

regular_summa_no_canopy_ice

Case 3: restart file + modified SUMMA

After changes, SUMMA will notify the user that something needs checking:

Warning: canopy ice content in restart file (0.4763579693535316E-05) > 0 when canopy temperature > Tfreeze. Continuing.

but picks up from the restart file and the simulation proceeds as before. See below.

modified_summa_unchanged_restart

The unexpected difference between the long run and the restarted run at t = 20:00 h and t = 20:30 h is possibly related to variable precision. scalarCanopyLiq between long run and restarted run differ in order of magnitude of e-14 to e-13 before this point (scalarCanopyIce values are identical), which may be the cause of these differences. I don't think this is related to the code change but is a more generally related to how restart files are saved and used.

scalarCanopyLiq differences for the period 2003-01-01 09:30 to 2003-01-01 22:00: [ 4.83688237e-14], [ 7.48117897e-14], [ 9.54668710e-14], [ 1.09663395e-13], [ 1.18571253e-13], [ 1.24623781e-13], [ 1.28502033e-13], [ 1.30901178e-13], [ 1.32272187e-13], [ 1.33084567e-13], [ 1.33609771e-13], [ 1.33993237e-13], [ 1.34173232e-13], [ 1.34276440e-13], [ 1.34339499e-13], [ 1.34357162e-13], [ 1.34346957e-13], [ 1.34308945e-13], [ 1.34310946e-13], [ 1.34325587e-13], [ 1.34338748e-13], [ 1.87629619e-10], [-2.87366370e-09], [-7.17519483e-13], [-2.52301028e-14], [ 2.44843344e-14]

wknoben commented 2 years ago

Updated original comment with:

  1. Better understanding of what causes the canopy ice content to be > 0 for T > Tfreeze
  2. New implementation of the solution that allows small (< 1E-3) ice content values but still exits on ice content larger than this threshold
wknoben commented 2 years ago

Updated as requested. Example warning: 

Warning: canopy ice content in restart file (=0.4763579693535316E-05) > 0 when canopy temperature (=273.449) > Tfreeze (=273.160). Continuing.

Example error (changed canIceTol for this example because that was quicker than modifying the restart file. canIceTol = 1e-3 in code):

summa_readRestart/check_icond/canopy ice (=0.4763579693535316E-05) > canIceTol (=0.100E-05) when canopy temperature (=273.449) > Tfreeze (=273.160)