Check global and regional GPP

[luyssaert]

Check: (1) the global GPP after an analytical spinup, transient simulation and FG2. The global GPP should be between 100 and 150 Pg C y-1; (2) the regional distribution (as defined in the mapper). It should be in line with tag 2.1.; (3) time series of LAI should be screened (LAI script); and (4) spatial patterns of the LAI time series should be screened (visual).

Notes for check 4:
In r6826 MTC4 grows well along the Pacific West Coast which is a hotspot for the MTC but it grows poorly in the Rocky Mountains. Too much water stress?

For r7322 the mapper shows some clear deficiencies in the relationship between LAI and precipitation. ORC4 has too little LAI and thus too little GPP and interception in the humid tropics. These are the regions where ORC4 reduces the GPP due to N-limitation (which in turn can be seen in the sugar_loading).

[luyssaert]

• PFT 2 and 3: GPP in the tropical core should go up to 4000 gC y-1 for both tropical forest PFTs. Current extreme values just approach 3000 gC y-1. See ​figure

Proposed changes:

• PFT 6: Temperate deciduous trees should have a GPP going up to 1500 gC y-1 in central Europe. Current extreme values just approach 1100 gC y-1. See ​figure

Proposed changes:

• PFT 7: Boreal eevrgreen trees should have a GPP going up to 1100 gC y-1 in southern Scandinavia. Current values in that region are below 700 gC y-1. See ​figure

Proposed changes:

• PFT 8 and 9: cannot sustain their LAI and die continuously from low productivity. See ​figure. See also ticket #677.

Proposed changes: double k_latosa, reduce psi_leaf to -3

• PFT 10: C3 temperate grasslands should have a GPP going up to 1500 gC y-1 in central Europe. Current values in that region are below approach 800 gC y-1. See ​figure

Proposed changes:

• PFT 12: C3 temperate croplands should have a GPP going up to 1500 gC y-1 in central Europe. Current values in that region are around 1100 gC y-1. See ​figure

Proposed changes:

• PFT 14: Tropical C3 grasslands grow in drier regions. What is an acceptable GPP? Current simulations are below 700 gC y-1. See ​figure

Proposed changes:

[luyssaert]

One remaining issue up to r7113 is that a reasonable LAI often comes with a too low GPP. diffuco.f90 was turned inside out and two inconsistency were found: (1) two different nitrogen profiles were used in this subroutine. At low GPP this inconsistency had little impact but at higher GPP it could result in a 10-20% increase in LAI. This has been corrected in r7121. (2) The light used in photosynthesis was not calculated correctly. This bug resulted in unlogic light profiles in the canopy. The impact on low GPPs was rather low as well. A correction has been committed in r7114. Both of these changes add to a considerable increase in GPP (given a fixed LAI) especially after correcting a bug in the calculation of target_labile in stomate_growth_fun_all.f90. This correction changes sugar loading but requires new parameters. The code correction was committed in r7124, the parameter changes were committed in r7125.

[luyssaert]

Following manual tuning or r7272 for the model, performance of r7272 is getting close to ORC2 and ORC3. Global totals, regional time series and spatial patterns were checked with the mapper. The parameters from this satisfying test were committed in r7278 and will be subjected to further manual refining in the upcoming tests. In between r7125 and r7272 several aspects of LAI and GPP were checked. For the moment we have no indications that are the code is bugged. The response of the model to changes in parameter values follows our understanding of the model. Spikes at the beginning of the spinup are related to soil N mineralization and justify the need of a spinup.

[luyssaert]

This ticket was re-opened because their are still some differences between ORC2/ORC3 and ORC4 in terms of its spatial pattern in LAI and the relationship between LAI~precipitation. These differences are also reflected in gpp, transpiration, interception and bare soil evaporation. The basis of the differences seems to be N-limitations in the soil (resulting in sugar loading -> lower gpp -> lower lai). Following additional checks nothing suspicious was found in the model results or the code. This ticket will be closed. Formal optimization will be used to better explore the parameter space. This is no longer considered a blocking issue to proceed with milestone 4.1.