Conversation
AlexisRenchon
force-pushed
the
ar/fix_gpp_pmodel
branch
from
c883c0c to
ea04e98
Compare
|
@AlexisRenchon it looks like GPP is still biased high everywhere, but if this is what you see in their model, let's make the change anyways? what values does this model code use for quantum yield (temperature dependent)? |
I'll keep working on this. This is what they do in pyrealm, the package the optimal lai model used |
| - ϕa0_c4 = 0.352*0.087 (unitless) - constant term in quadratic intrinsic quantum yield (Scott and Smith, 2022) | ||
| - ϕa1_c4 = 0.022*0.087 (K^-1) - first order term in quadratic intrinsic quantum yield (Scott and Smith, 2022) | ||
| - ϕa2_c4 = -0.00034*0.087 (K^-2) - second order term in quadratic intrinsic quantum yield (Scott and Smith, 2022) | ||
| - ϕa0_c3 = 0.044 (unitless) - constant term in quadratic intrinsic quantum yield (pyrealm/Bernacchi et al., 2003) |
There was a problem hiding this comment.
I think these all need to be updated in both parameter toml files
| - ϕa0_c3 = 0.044 (unitless) - constant term in quadratic intrinsic quantum yield (pyrealm/Bernacchi et al., 2003) | ||
| - ϕa1_c3 = 0.00275 (K^-1) - first order term in quadratic intrinsic quantum yield (pyrealm/Bernacchi et al., 2003) | ||
| - ϕa2_c3 = -0.0000425 (K^-2) - second order term in quadratic intrinsic quantum yield (pyrealm/Bernacchi et al., 2003) | ||
| - ϕa0_c4 = -0.008 (unitless) - constant term in quadratic intrinsic quantum yield (pyrealm/Cai and Prentice, 2020) |
There was a problem hiding this comment.
this value is very odd to me - it means when T = 0C, quantum yield is negative...but I guess it is from an empirical fit with uncertainty...
2 participants
This PR makes changes to our pmodel implementation to match pyrealm implementation which is used, for example, by the Zhou optimal LAI model code
To do about Amazon / c4 grassland bias in south America (LHF and GPP too low in Amazon, GPP too high in c4):
run single column at coordinates of Amazon, South America grassland c4. Check what is suspicious: does PAR look good? APAR? Root depth? Moisture stress? LAI? Are things distinct by c3 / c4 (e.g., root depth, and therefore, moisture stress for c3 and c4...), is RAI = 1 reasonable (or should we try RAI = f*maxLAI)?
try calibrating GPP + LHF on the first commit of this PR (params adjusted). need to add a dataset to calibrate on (for gpp) on derecho. It seems GPP is overestimated everywhere, so possibly calibration can help (just offset globally)