Opened 11 months ago

Last modified 3 months ago

#701 new defect

Resolve attribution between transpiration and evaporation

Reported by: luyssaert Owned by: somebody
Priority: critical Milestone: ORCHIDEE 4.1
Component: Physical processes Version: trunc
Keywords: Cc:

Description

The reference simulations suggest that evaporation is much larger than transpiration even in tropical forests. This is in conflict with general understanding of evapotranspiration where transpiration is expected to be by far the dominant process.

Change History (3)

comment:1 Changed 11 months ago by luyssaert

  • Priority changed from major to critical

comment:2 Changed 11 months ago by luyssaert

  1. Testing xios

Albeit unlikely, the figures appear as if transpiration and bare soil evaporation were simply swapped (see ​https://forge.ipsl.jussieu.fr/orchidee/wiki/ReferenceSimulations/6616#D.Transpirationvsevaporation). Check whether the correct variables are written to the correct fields in xios_orchidee_send.

Result: the value for vevapnu was followed from the line where it is calculated to the line were it is written. No indication was found of evaporation and transpiration being swapped in between their calculation and i/o.

  1. Testing ok_bare_soil_new

r6616 comes with two options to calculate the bare soil. Read more about ​ok_bare_soil_new. When using enerbil.f90 to calculate the energy fluxes (including transpiration and evaporation) ok_bare_soil_new should be false. Check whether this is indeed the case.

Result: ok_bare_soil_new was not correctly implemented (#708). This has been corrected in r6766. Using this flag tends to decrease the evaporation compared to not using it. This flag is, therefore, not the cause of the observed problem that evaporation >> transpiration at densily forested pixels.

  1. Testing Pgap

Previously the bare soil fraction for each pft was calculated as veget_max - veget. ORCHIDEE 4.0 no longer uses Lambert-beer but uses the Pgap approach to calculate veget (as a function of the solar angle). Problems with the calculation of Pgap and thus veget may result in too high fractions of bare soil.

Result: as expected veget calculated with Pgap is smaller than veget calculated with Lambert-Beer. This smaller veget is intended and from a conceptual point closer to reality (Pgap accounts for the clumping of trees and thus the clumping of their canopies, Lambert-Beer assumes a homogeneous distribution of the leaves). Using the r6616 code but replacing the calculation of veget from Pgap back to Lambert-Beer seems to solve the core problem of this ticket.

The low interception in combination with the high bare soil evaporation suggests that Pgap simulates too many gaps which in turn could be the outcome of using too small crown diameters. Especially for older canopies with fewer individuals, small crown diameters would result in more and bigger gaps between the trees. Two options: (1) Pgap is correct - check veget and (2) veget is correct check Pgap. For the section option in which Pgap needs to be checked there are 4 parameters that could decrease Pgap: (1) increasing the canopy diameter, (2) using recruitment if FM=1, (3) tune self-thinning relationship, and (4) decrease leaf_to_shoot clumping in the Pgap calculations. If none of the parameters affects Pgap it could be checked whether by using spherical canopies the leaf mass is distributed over a too large volume. Clumping more leaves at the top of the canopy could help. This hypothesis seems inconsistent with the excellent performance of Lambert Beer where the leaf mass is distributed uniformly.

Philippe Peylin may have found another lead that contributes to this problem. In the Tag2.1 the veget of a pixel without any vegetation is 1. We changed the way we calculate veget (using Pgap) and so in the trunk the veget of a pixel without any vegetation is 0. If this is the cause, it would also be solved by the test described above. Nevertheless, this seems like a less likely hypothesis because the bare soil pixels in ORCHIDEE 4.0 look OK.

  1. Testing lai

The reference simulations suggest that the LAI is too low especially over the tropics. Too low LAI would result in too low veget and thus a too high bare soil fraction.

  1. beta coefficients

The calculation of the beta coefficients could have changed unintentionally. The reference simulations should use enerbil.f90. This module has not been revised as part of the TRUNK 4.0 developments.

Last edited 3 months ago by luyssaert (previous) (diff)

comment:3 Changed 4 months ago by luyssaert

Testing Pgap

The low interception in combination with the high bare soil evaporation suggests that Pgap simulates too many gaps which in turn could be the outcome of using too small crown diameters. Especially for older canopies with fewer individuals, small crown diameters would result in more and bigger gaps between the trees.

  • The values for Pgap, right before calculating veget are credible. There seems to be no urgent need to improve/revise the calculation of Pgap itself.
  • increasing the canopy diameter: A first approach was committed in r6922 and r6923. Crown area is calculated as a function of tree height to reduce the changes for inconsistencies between the diameter-height and diameter-crown area relationships. The parameters were externalized in r6933.
  • use recruitment if FM=1. This is the default setting for PFT2. Number of recruits could be increased but the number was set based on literature for tropical forests.
  • tune self-thinning relationship. The number of trees per m2 seems reasonable.
  • leaf_to_shoot_clumping in the Pgap calculations. For all PFTs this parameter is set to one and has no impact for the moment.
  • Clumping of the trees is assumed to be random. Which will result in the gaps being Poisson distributed. Nevertheless, a Poisson distribution is nowhere assumed in the code. Tree clumping is NOT controled by a parameter that could be adjusted.

Testing veget
Given that Pgap seems reasonable, the problem may be caused by veget

  • Note veget is calculated relative to veget_max, hence the sum of veget for all PFTs is between 0 and 1. This implies that if one PFT is not properly growing our veget will be too low. This is still the case for PFT3 in r6933.

Conclusion
For the moment we expect that the problem is caused by the absence of LAI for PFT3 (See ticket #714)

Last edited 3 months ago by luyssaert (previous) (diff)
Note: See TracTickets for help on using tickets.