Opened 14 months ago

Closed 9 months ago

#586 closed defect (fixed)

Large GPP difference between model configurations

Reported by: luyssaert Owned by: somebody
Priority: critical Milestone: IPSLCM6.v2
Component: Biogeochemical processes Version:
Keywords: Cc:

Description

It was observed that a configuration with only one forest PFT prescribed per pixel compared to a configuration with several forest PFTs prescribed in that pixel had about 200% of the GPP. Although competition for water between the PFTs cannot be excluded for the moment (which is intended model behavior) this looks more like a bug (= unintended model behavior).

Proposed tests: make a run.def for a single PFT (+bare soil) and prescribe this PFT to 100% of the pixel, make a run.def with two identical PFTs and prescribe each to 50% of the pixel. It is expected that in case water stress is the culprit, the GPP of this test case will be identical between both configurations. If, however, the GPP differs we are looking at a bug (possibly a problem with the weighting of the PFTs).

Change History (4)

comment:1 Changed 13 months ago by luyssaert

The GPP issue reported in #586 is not related to the absolute number of PFTs. The total GPP of the 3 runs in test1 below should be equal but is not. The model seems to be unable to grow PFTs with large cover fraction. This also seems to be the case for PFT4 (test2). This looks like a bug, possibly related to a wrong multiplication/division with veget_max somewhere (hydrol_arch is a suspect, see below). Suggested test: run without hydrol_arch?

To add to the confusion: when the first PFT has a fraction of 80% and the second of 20% the first PFT is growing (test3). It seems we can grow large cover fractions, if it is the first PFT after bare soil (but not in the PFT6 100% case?!)

Test1
Run1
PFT6 100% not growing

Run2
PFT6 20% growing
PFT6 80% not growing

Run3
PFT6 1% growing
PFT6 99% not growing

Test2
Run1
PFT4 20% growing
PFT4 80% not growing

Test3
Same as test test1 run 2, but with PFT6 80% as first PFT after baresoil, in this case PFT6 80% does grow and PFT6 20% not

Note: the following bug was fixed and could be related
Changeset 5785

DEV: bugfix for the tranpirational demand vs supply. Transpir was divided by veget_max, but since transpir is already at the PFT level, there is no need for doing so. The hydraulic resistance in roots and shoot has been added to the outputs.

Note that many changes has been made to veget_max and veget_max_cov as well as to order in which some of these variables are calculated in slowproc.

Last edited 13 months ago by luyssaert (previous) (diff)

comment:2 Changed 11 months ago by jgipsl

  • Priority changed from major to critical

comment:3 Changed 9 months ago by luyssaert

Following r6411 test cases with two times PFT6 in a 50/50 and a another run with PFT6 in an 20/80 ratio resulted in identical values for all tested variables. These tests did not consider water stress but demonstrate that all issues with scaling variables to their fraction within a pixel have been fixed.

A spatially explicit run with r6411 of two times PFT4 in a 50/50 and another run with PFT4 in a 10/90 ratio over Europe both resulted in different values. The difference is already present in LAI at the end of day 1. This hints at problem in stomate_prescribe.f90. The problem was found to be related to the way managed forests (FM=2) are currently initialized. The code works correctly for test where FM = 1 over Europe in a 10-year long test.

The fix committed in r6415 fixed an initialization issue when FM=2. This allowed us to run the same test cases as above but with FM = 2. The code works correctly for test where FM = 1 over Europe in a 10-year long test. These tests did not consider water stress but demonstrate that all issues with scaling variables to their fraction within a pixel have been fixed.

Changes to make this specific orchidee_pft.def file (3 PFTS: bare soil and two times the same PFT) that can be used for this test has been committed in r6416.

Last edited 9 months ago by luyssaert (previous) (diff)

comment:4 Changed 9 months ago by luyssaert

  • Resolution set to fixed
  • Status changed from new to closed
Note: See TracTickets for help on using tickets.