CiBisection error

[jennykowalcz]

I'm trying to run FATES-Hydro with competition (2 PFTs) regionally over tropical South America with API 40, but am getting this error almost immediately after the run starts:

61:  Ci bisection during photosynthesis failed
61:  try increasing tolerance or widening the starting points
61:  ci_h=  5.020355635535013E+038 fval_h= -1.511157274518286E+023 ci_l=
61:   5.020355635535013E+038 fval_l=  7.555786372591432E+022
61:  ft=            2 is c3psn: T
61:  vcmax=   14.6162229177636      jmax=   32.6669495860880      kp=
61:   -9999.00000000000     
61:  co2_cpoint=   2.62023428376122      mm_kco2=   14.5358915391505      mm_ko2=
61:    17384.8354414828     
61:  can_co2_ppress=   39.8768644531250      can_o2_ppress=   20835.6616767578      
61:  can_press=   99692.1611328126     
61:  can_vpress=   1952.30392789691      lmr=  0.461213894838821      par_abs=
61:    17.1838658465566      gb=   320025.724718468     
61:  veg_tempk=   288.939848823615      gs0=   9984.75311314055      gs1=
61:    8.00000000000000      gs2=   1.00000000000000      ci_tol=
61:   0.500000000000000     
61:  ENDRUN: 
61:  ERROR: ERROR in LeafBiophysicsMod.F90 at line 1213

(I edited LeafBiophysicsMod.F90 to print out more info upon running into the error -- ed5cffc )

@rgknox can you help me diagnose what's happening here?

I am using FATES sci.1.86.3_api.40.0.0, edited to set nclmax=3, with:

use_fates_nocomp=.false.
use_fates_planthydro = .true.

!! non-default:
fates_leafresp_model = 'atkin2017'
fates_cstarvation_model = 'exponential'
fates_hydro_solver = '2D_Picard'

!!! the following are default values:
use_fates_daylength_factor = .true.
fates_photosynth_acclimation = 'nonacclimating'
fates_stomatal_model = 'ballberry1987'
fates_stomatal_assimilation = 'net'
fates_regeneration_model = 'default'
fates_radiation_model = 'norman'

[jennykowalcz]

I see the CO2 values are partial pressure (Pa). So atmospheric CO2 (Ca) is (can_co2_ppress/can_press) x 10^ 6 = 400 ppm. So the Ci values here are >> Ca. I see Ryan's documentation says this means the plant is respiring.

I have fates_leaf_vcmax25top for PFT 2 set to 35 umol/m2/s for this case. I was previously running a similar case with a branch based on API 34.

[jennykowalcz]

Interestingly, I do not encounter this error running the same case for 1 year with the default parameter file. 🤔

[rgknox]

I find this error incredibly odd @jennykowalcz . The inputs to the solve all seem reasonable. The failure is happening at low light levels; there is not a strong btran limitation, because the conductance constants are all near their maximums (gs0,gs1 and gs2). The vcmax is low compared to the reference value at 25, since btran is not low it must be do to the temperature...

For some reason it it making a terrible estimate of the bisection bounds: ci_h= 5.020355635535013E+038 For some reason it is crashing bisection, the bisection bounds should only tighten in on the solution and not exceed the original bounds when the iterative solver failed.

I think the thing to do is 1) I can put this through the unit testing and see why exactly it is failing, and 2) for the time being we need to give you a fall-back solution to use...

[glemieux]

@rgknox and @alistairrogers discussing simplifying code for setting the Ci bounds.

[rgknox]

@jennykowalcz provided me with her parameter file. While trying to reproduce the problem I found that the fates_leaf_fnps parameter was undefined.

I think the solution here is to make sure that we have a check on the read-in that fnps is greater than 0 and less than 1.

[jennykowalcz]

Thanks for debugging this @rgknox! I can confirm that the case ran without error after I corrected fates_leaf_fnps in my parameter file. Do you want to leave this issue open as you make the code more idiot proof robust to user error, or shall I close it?

[rgknox]

Great!. We all do stuff like that, no worries, its tough to keep tabs on so many parameters. Yup, lets close this up and open a new issue.