wiki:DevelopmentActivities/ORCHIDEE-DOFOCO

Version 179 (modified by luyssaert, 5 years ago) (diff)

--

ORCHIDEE-CN-CAN

ORCHIDEE-CN-CAN is based on the latest version of the trunk, the latest version of the ORCHIDEE-CN branch and some key functionality of ORCHIDEE-CAN (aka ORCHIDEE-DOFOCO on the svn server). The committed version can be considered a light version of ORCHIDEE because some of the available functionality was not yet committed (for example, the DGVM). Nevertheless, this light version is believed to contain all functionality that affects the parameterization of the model. The functionality that is not yet included depends on vegetation growth, soil hydrology or the energy budget but does not affect these processes at the pixel level.

1 - The available functionality is described in more detail below

2 - Known conflicts with the trunk: tot_bare_soil, VIS/NIR snow albedo

3 - Missing functionalities compared to the trunk: DGVM, and fire disturbances

4 - Missing functionalities compared to ORCHIDEE-CAN (add these and we can close the branch): wind throw and species change

5 - Better functionalities than those available in the trunk, ORCHIDEE-CN or ORCHIDEE-CN-CAN: spitfire for fire disturbances and gross land cover changes

6 - Coupling: LMDZ coupled with ORCHIDEE-CAN has been used only with nudging and zoomed over Europe. Given that several key parameters for the coupling have changed: albedo, roughness is now dynamic, more landscape heterogeneity when using age classes, transpiration, … it is expected that lots of testing will be required to come to a satisfying result with LMDzOR-CN-CAN.

What is the future of the version?

Now that the AR6v0 is becoming stable, a robust well tested version of ORCHIDEE can soon be committed and tagged. This gives the ORCHIDEE-team the opportunity to prepare for the future by making a series of far reaching changes to the code. The first ambition is to add CN into AR6v0 to obtain AR6v1. In parallel we will start testing and discussing AR6v2 before making this the trunk for both coupled and off-line simulations. The basis of AR6v2 is the current ORCHIDEE trunk (AR6v0) into which CN and CAN functionality is merged. At this stage key functionality from other branches may be merged as well (to be discussed in tandem with the fair use policy). In addition to the recent improvements to the trunk, the new trunk (AR6v2) should have an increased functionality (see below), increased internal consistency (see below, for example albedo and photosynthesis) and some long awaited fixes (for example no longer using tot_bare_soil).

How can I contribute to this version?

Given the large number of code changes, there is no guarantee that the current version of AR6v2 will run flawless for all set-ups and possible combinations of flags currently being used by ORCHIDEE users and developers. (1) All users and developers are invited to install AR6v2 and set-up their favorite simulation. Use one of the attached run.def as a starting point for your favorite set-up. Report on the outcome of your tests during the ORCHIDEE meetings. (2) All developers are invited to contribute bug fixes, and code for improved functionality to AR6v2. (3) adjust the interface such that AR6v2 can be coupled to the latest version of LMDz and can be used for regional simulations with WRF.

How do I run this version?

For the moment, the instructions for ORCHIDEE-DOFOCO can be used to download and extract the model (see https://forge.ipsl.jussieu.fr/orchidee/wiki/Documentation/UserGuide/ORCHIDEEDOFOCOInstall). Use this run.def for a set-up with 13 MTCs and no age classes (as the trunk) or this run.def (run_r5006.def) to make use of 13 MTC and 4 age classes for each forest MTC, thus resulting in 37 PFTs.

When running the model, the default setting should be XIOS. Thus, compile the model with gmake with_xios. Note that there currently is a bug on obelix when compiling XIOS. It keeps re-compiling, even though no changes has been made to XIOS. To avoid this and reduce compiling time, change the following line in the Makefile from

with_xios : xios ioipsl driver_xios verif

to

with_xios : ioipsl driver_xios verif

Once the model is compiled, make sure that XIOS=y in orchidee_ol.card.

Moreover, new output files for ORCHIDEE have been specified within the code.

Below is given an example for stomate_history_4dim.
In stomate.driver, add the following within the shown if statement:

    # Modify file_def_orchidee.def if XIOS is activated
    if [ X${orchidee_ol_UserChoices_XIOS} = Xy ] ; then
        # Modify file_def_orchidee.xml
        IGCM_comp_modifyXmlFile nonblocker file_def_orchidee.xml stomate3 enabled ${stomate_enabled}
        IGCM_comp_modifyXmlFile nonblocker file_def_orchidee.xml stomate3 output_freq ${stomate_freq}
    fi

In stomate.card add the new file to list of output files:

[OutputFiles]
List=   (stomate_history.nc, ${R_OUT_SBG_O_M}/${PREFIX}_1M_stomate_history.nc, Post_1M_stomate_history) \
        (stomate_ipcc_history.nc, ${R_OUT_SBG_O_M}/${PREFIX}_1M_stomate_ipcc_history.nc, Post_1M_stomate_ipcc_history),\
        (stomate_history_4dim.nc, ${R_OUT_SBG_O_M}$ {PREFIX}_1M_stomate_history_4dim.nc, NONE)



For sechiba, the following changes are needed in sechiba.driver:

    if [ X${orchidee_ol_UserChoices_XIOS} = Xy ] ; then
      # Modify file_def_orchidee.xml file
      IGCM_comp_modifyXmlFile nonblocker file_def_orchidee.xml sechiba1 enabled ${sechiba1_enabled}
      IGCM_comp_modifyXmlFile nonblocker file_def_orchidee.xml sechiba1 output_freq ${sechiba1_freq}
      IGCM_comp_modifyXmlFile nonblocker file_def_orchidee.xml sechiba3 enabled ${sechiba1_enabled}
      IGCM_comp_modifyXmlFile nonblocker file_def_orchidee.xml sechiba3 output_freq ${sechiba1_freq}
      IGCM_comp_modifyXmlFile nonblocker file_def_orchidee.xml sechiba2 enabled ${sechiba2_enabled}
      IGCM_comp_modifyXmlFile nonblocker file_def_orchidee.xml sechiba2 output_freq ${sechiba1_freq}
    fi

If you are seeing the following error when you run:

Error [StdIStream& operator>>(StdIStream& in , CDuration& duration)] : In file '/home/users/mmcgrath/ORCHIDEE-CN-CAN/modeles/XIOS/src/duration.cpp', line 74 -> Bad duration format: impossible to read a pair (value, unit).

then it is possible because you have not added the needed lines in stomate.card, stomate.driver, sechiba.card, and sechiba.driver. This error arises because the .xml files, when they have been copied to the working directory, cannot have values like "_AUTO_" in them. libIGCM is supposed to replace such values, which exist in the template .xml files in modeles/ORCHIDEE/src_xml/, when it copies files from the run directory to the working directory. If there is something missing in the .driver files, this replacement doesn't happen and XIOS throws the parsing error above.

Functionalities (alphabetical order)

Age classes

Age classes were introduced to better handle heterogeneity at the landscape level. The feature allows us to distinguish between different successional stages of the same PFT. Age classes are independent of the number of diameter classes. Using age classes adds a lot of details in both the biophysics and the biogeochemistry following natural disturbances, forest management and land cover change. If half of a grassland is afforested with a PFT that already exists in the pixel, previous versions of ORCHIDEE will combine this newly forest land and the existing forest in a single PFT. This will result in a low albedo, a high roughness, ... When age classes are used, the newly afforested and the existing forest will end up in separate PFTs. One will have a high albedo, the other a low, ...

Age classes were defined as separate PFTs and if wanted different age classes of the same PFT could be run with different parameters. This option has not been tested yet because it is expected to result in discontinuities when the biomass is moved from one age class to another. The number of age classes is fixed but for each PFT it can be decided whether age classes are used or not. This adds a lot of flexibility to the model. ORCHIDEE-CAN, for example, has been run with 64 PFTs, using age classes for European forest and using no age classes for all forests outside the domain of interest. Setting-up a simulation with age classes will require some thinking when setting-up the run.def. A python-script was written to create this kind of run.def. Increasing the number of PFTs has important consequences for the speed of the model and the memory use. Because a single run can contain PFTs with and PFTs without age classes, processing of the simulation output needs to account for the relationship between PFTs of the same species but a different age class.

The number of age classes is defined by the parameter NAGEC. Setting this parameter to 1 is a good start unless you have a special interest in using age classes. When NAGEC is set to more than 1, PFT_TO_MTC', AGEC_GROUP and PFT_NAME will all need to be carefully defined. See the attached run.def for a functional example. See below for some principles: NAGEC = 4 Assume we want to use four age classes for all forests. We will end-up with 37 PFTs, the 1 for bare soil, C3 grass, C4 grass, C3 crop and C4 crop and 4 times 8 for the 8 forest PFTs. Thus NVM = 37

Because we still use the 13 default MTC we can use the default maps. Let the model know how many MTCs it should find on the maps: NVMAP=13

If you want to use IMPOSE_VEG=y then only vegetation should be added to the youngest age class. ORCHIDEE will update the vegetation fractions during the simulations

SECHIBA_VEG_01=0.0769230769231
SECHIBA_VEG_02=0.0769230769231
SECHIBA_VEG_03=0.0
SECHIBA_VEG_04=0.0
SECHIBA_VEG_05=0.0
...

SECHIBA_VEGMAX_01=0.0769230769231
SECHIBA_VEGMAX_02=0.0769230769231
SECHIBA_VEGMAX_03=0.0
SECHIBA_VEGMAX_04=0.0
SECHIBA_VEGMAX_05=0.0
...

Link PFTs to MTCs

PFT_TO_MTC_01=1
PFT_TO_MTC_02=2
PFT_TO_MTC_03=2
PFT_TO_MTC_04=2
PFT_TO_MTC_05=2
...

Tell ORCHIDEE which PFTs have a successional relationship

AGEC_GROUP_01=1
AGEC_GROUP_02=2
AGEC_GROUP_03=2
AGEC_GROUP_04=2
AGEC_GROUP_05=2
...

Give all PFTs a name for clarity

PFT_NAME__01=Soilbare
PFT_NAME__02=Broadleavedevergreentropical_age01
PFT_NAME__03=Broadleavedevergreentropical_age02
PFT_NAME__04=Broadleavedevergreentropical_age03
PFT_NAME__05=Broadleavedevergreentropical_age04
...

Albedo

ORCHIDEE-CN-CAN makes use of a two stream radiative transfer scheme through the canopy. The scheme is based on Pinty et al 2006. This approach accounts not only for the leaf mass but also for the vertical and horizontal distribution of the leaf mass (=canopy structure). In ORCHIDEE-CN-CAN the same scheme is used to simulate the reflected, transmitted and absorbed light. This implies that albedo and photosynthesis are now fully consistent as well as the light reaching the forest floor (the latter is used in for example recruitment). ORCHIDEE-CN-CAN cannot revert to previous approaches for calculating albedo. The radiative transfer through the canopy is controlled by 3 parameters for each wavelength/band: single leaf scattering leaf_ssa_xxx, forward scattering leaf_psd_xxx and background reflectance bgrd_ref_xxx. At present VIS and NIR have been parameterized. Parameterization is based on running an inverse radiation scheme on the MODIS albedo product while accounting for the different land cover types. The inverted parameters are provided by the JRC as the JRC TIP product.

Allocation

ORCHIDEE-CN-CAN uses the allometric allocation as developed in OCN. In ORCHIDEE-CAN the approach was adjusted to work for more than one diameter class. Since it was developed this allocation has been used in ORCHIDEE-CN, ORCHIDEE-CNP and ORCHIDEE-MICT. In those three branches only a single diameter class was used. Except for the way the reserves and labile pools are calculated, the allocation scheme is identical between all aforementioned versions. The model is, however, very sensitive to the way the reserves and labile pools are calculated. The allocation makes use of a labile pool for which the activity is calculated based on the temperature. As such the model addresses the sink/source discussion initiated by Körner.

ORCHIDEE-CN-CAN calculates the number of individuals and uses this as a criterion to initiate a stand replacing disturbance. This approach, guided by the self-thinning relationship, avoids the need for a stand-level turnover time. ORCHIDEE-CN, ORCHIDEE-CNP and ORCHIDEE-MICT still make use of stand-level turnover.

There are no options to revert to the allocation based on resource limitation. All references and parameters for allocation based on resource limitation have been removed from the code. Allometric allocation makes use of the following PFT-specific parameters: sla, tau_root, tau_leaf, tau_sap, pipe_density, tree_ff, pipe_tune_x, k_latosa_max, and k_latosa_min. In addition to this set of parameters that mainly describe the allometric relationships and the longevity of the different tissues, the calculation of the allocation coefficients makes use PFT-specific tissue conductivities, i.e., k_sap, k_root, and k_leaf (see also plant water stress). As such there is a functional link between C and N-allocation and the hydraulic architecture of a plant. Details on the parameters can be found in the SI of Naudts et al 2015 in GMD or in src_parameters/constantes_mtc.f90.

Background albedo

If covered by snow, the background albedo is calculated by the snow module and accounts for snow age and snow density (needs to be checked – last time snow did not account for NIR). If not covered by snow the background albedo is not simulated but prescribed by the parameters bgrd_ref_vis and bgrd_ref_nir. In deciduous forest, grasslands and croplands, the background albedo is known to be strongly affected by the phenology and senescence of the understory vegetation. ORCHIDEE-CN-CAN has two options to prescribe the background albedo:

1 - The background albedo is prescribed per PFT but is constant throughout the year. This is the option that has been used in ORCHIDEE-CAN and is the option that has been validated over Europe. Set alb_bg_modis = n.

2 - The background albedo is constant across PFTs. This option reads background maps. Given that those maps are based on the JRC TIP product, they should be compatible with the new albedo scheme. This option, however, has not been validated yet. Set alb_bg_modis = y.

Anthropogenic species change

Following a disturbance (which could be a clear cut), tree species changes and forest management change can be prescribed or read from a map in ORCHIDEE-CAN. Set lchange_species = y, read_species_change_map = y, and read_desired_fm_map = y and specify the paths of those maps in the COMP/stomate.card. This functionality replaces the DGVM in areas where humans rather than nature govern species distribution, for example, Europe. Note that there are some constraints on the possible species changes. If the forest is unmanaged (fm=1), the code assumes that nature will determine the species rather than humans. Anthropogenic species change has not been developed to work together with land cover change. For the moment it is one or the other.

Bare soil (under development)

The flag ok_bare_soil_new controls how the bare soil is perceived and calculated. If set to FALSE the total bare soil is still calculated as a function of veget. When a deciduous PFT sheds its leaves, the gaps in the forest will contribute to bare soil fraction in the grid. Although this approach was introduced a long time ago to get acceptable evaporation estimates from forest, the approach also resulted in using the albedo of deserts as the background albedo of forest gaps. The new albedo scheme considers a specific background albedo for each PFT and calculates the albedo of the PFT including the canopy gaps. Moving gaps to the bare soil is no longer needed. So, if ok_bare_soil_newis set to TRUE, canopy gaps no longer contribute to the bare soil. It needs to be tested what will happen with the evaporation in the single-layer model. The multi-layer energy budget should be able to correctly deal with the gaps in the canopy because the diffusivity will increase when the canopy is becoming sparser.

C13

The concentration of C13 in the leaves can be calculated by setting ok_c13 to y in the run.def. This calculation follows Farquhar's approach and is currently limited to the fractionation in the leaves.

Coming soon

  1. The calculation of wind storm damage can be activated by setting ok_windthrow to y in the run.def. This module calculated the critical wind speed of a stand taking stand and soil properties into account. If the stand is managed, the damaged trees are salvaged logged. If the stand is unmanaged the damaged trees are left on-site to decompose. The default setting for ok_windthrow is n.
  1. Bark beetle outbreaks. The core of this module has been developed but needs to integrated in ORCHIDEE-CN-CAN. The bark beetle module was developed such that it interacts with the windthrow module.

Consistency checks

The code distinguishes between three options to check for mass balance problems. These options are controlled by the parameter err_act. Always use err_act = 3 when developing and testing the code. Note that in addition to checking for mass balance closure ORCHIDEE-CN-CAN will also check for the preservation of veget_max. This is useful to make sure no surface area is lost when moving biomass from one PFT to another following natural disturbances, forest management, land cover changes and when using age classes. In some parts of the code, for example, modules that deal with disturbances, it is assumed that the tallest trees are stored in the last diameter class. When the difference in diameter between diameter classes becomes very small, this assumption could be violated. Therefore, the diameter classes are sorted to enforce the assumed order and where needed the order is checked.

1 - err_act = 1 is recommended when running global long-term simulations. Under this option, mass balance closure is checked for all biogeochemical processes but only at the highest level thus stomate.f90 and stomate_lpj.f90. Although the mass balance checks are not very expensive in terms of computer time, skipping the numerous lower level checks is expected to save some time. Under this option the mass balance error is only written to the history file. No information is provided in which subroutine the problem occurred.

2 - err_act = 2 is recommended when developing and testing the model. Now the mass balance is explicitly checked in stomate.f90, stomate_lpj.f90 and all its subroutines. Under this option the mass balance error is written to the history file and if the mass balance is not closed, the warning message will indicate in which subroutine the problem likely originated.

3 - arr_act = 3 is recommended when having a problem with mass balance closure. The mass balance is explicitly checked in stomate.f90, stomate_lpj.f90 and all its subroutines. If a mass balance error occurs, the model is stopped.

CWRR vs Choinel

ORCHIDEE-CN-CAN was developed and tested with CWRR. Set hydrol_cwrr to y. The Choinel code is still available. The hydrological schemes were not touched during the development of ORCHIDEE-CN-CAN.

Diameter classes

Diameter classes were introduced to better simulate the canopy structure, they are a tool to simulate heterogeneity within a single PFT. Given that the canopy is the interface between the land and the atmosphere this feature has effects well beyond forest management. Stand structure was observed to affect albedo, transpiration, photosynthesis, soil temperature, roughness length, and recruitment. Using diameter classes adds very little complexity to setting-up the simulations as well as to the output files. The complexity is mostly within the code.

The computational cost of using diameter classes is negligible and when a reasonable low number of diameter classes is used, the memory cost remains very small as the dimensions of only two variables are increased. The number of diameter classes is the same for all PFTs and is set by the parameter NCIRC. ORCHIDEE-CN, ORCHIDEE-CNP, and ORCHIDEE-MICT are all coded and used for NCIRC = 1. ORCHIDEE-CAN and ORCHIDEE-CN-CAN are coded and tested for NCIRC = 3. Until further testing, three diameter classes are considered sufficient.

Given earlier choices in ORCHIDEE, we either need to define the boundaries of each diameter class or the diameter distribution. While developing the code, we considered the second approach the most flexible. To allow maximal flexibility, each diameter class needs to be defined separately by the variable CIRC_CLASS_DIST_0000X, where X is the number of the diameter class. The smallest number presents the smallest diameter class. From literature it is known that a truncated exponential distribution is a good first guess:

CIRC_CLASS_DIST_00001=9
CIRC_CLASS_DIST_00002=3
CIRC_CLASS_DIST_00003=1 

The above declaration implies that 9/13th of the trees will always be in the smallest diameter class, 3/13th will be in the medium class and 1 tree out of 13 will be in the largest diameter class. These ratios are kept throughout the simulations and the boundaries of the diameter classes are adjusted to respect this constraint. Consequently, an even-aged stand will be simulated with three diameter classes where the diameter of the first class may be, for example, 20.3 cm, the diameter of the second class 20.4 cm and the diameter of the third class 20.5 cm. The same code and set-up allows to simulate, in the same simulation, an uneven-aged stand for the same PFT but in a different pixel with, for example, the smallest diameter 7 cm, the medium diameter 25 cm and the largest diameter 45 cm.

Forest management (change)

70% of the global forest are managed invalidating the assumption in previous versions of ORCHIDEE that forests are long-lived natural vegetation. Forest management, inspired by ORCHIDEE-FM was implemented in ORCHIDEE-CAN. Owing to the allometric allocation scheme, the introduction of diameter classes and a canopy structure only the principles, i.e., Deleuze and Dhote and RDI based management were retained. If the forest management strategy is not specified the default value "unmanaged" (FM = 1) is used. This implies that there are no thinning or harvest. Once the stand density drops below the threshold or the tree diameter exceeds another threshold a stand replacing disturbance is applied and a new stand is prescribed in the next time step. Therefore, the biomass pools in ORCHIDEE-CN-CAN no longer depend on a prescribed longevity.

When developing and testing the model, a single forest management strategy can be applied for all pixels and PFTs. ORCHIDEE-CN-CAN distinguishes 4 different strategies:

1 – FM=1 unmanaged

2 – FM=2 high stand management: with RDI based thinnings and density/diameter based final harvest

3 – FM=3 coppice

4 – FM=4 short rotation coppice with willow or poplar

Set read_fm_map to n and specify the desired management strategy (1-4) through forest_managed_forced.

For applications that focus on forestry or require landscape heterogeneity, a PFT-specific management strategy can be read from a spatially explicit map. Thus, the same PFT in different pixels can be assigned a different management strategy. However, within a pixel a single PFT can only have one management strategy. Unless, one wants to run forest management over Europe the user will have to create his/her forest management maps first. Set read_fm_map to y and specify the location of the forest management map in COMP/stomate.card. Check the existing forest management maps for Europe for an example of how the map should be defined.

When prescribing a forest stand (independent of forest management) the Initial density nmaxtrees, and the range of the initial tree height of the seedlings needs to be specified height_init_min and height_init_max. Irrespective of the management strategy the maximum carrying capacity needs to be described. Carrying capacity was formalized through the self-thinning relationship which makes use of two parameters alpha_self_thinning and beta_self_thinning. As a fail-safe option the longevity of a stand is still defined but should only be used when all other criteria fail to kill the stand (not observed). Longevity is defined by the parameter residence_time.

The details of each of the 4 management strategies can be refined through a set of PFT-specific parameters. Note that not every management strategy makes use of all parameters. For more details see the SI of Naudts et al 2015 (last table). The different management strategies require parameter values for : first thinning height h_first, stand replacing density ntrees_dia_profit, harvest diameter max_harvest_dia, ccppice diameter coppice_diameter, rotation length src_rot_length, number of rotations src_nrots, fuelwood diameter fuelwood_diameter and the minimum and maximum alpha and beta (thus 4 parameters) specifying the RDI range alpha_rdi_upper, alpha_rdi_lower, beta_rdi_upper and beta_rdi_lower.

According to economic theory, high-stand forest are harvested when the actual growth drops below the long-term growth. This has been implemented in ORCHIDEE-CAN and ORCHIDEE-CN-CAN. This feature was found to be very sensitive to the time frame for which actual increment was calculated. This option can be by-passed by setting this period unrealistically high, for example, n_pai =1000. Persons interested in further testing/developing this feature should set this parameter (unit: years) to 5 or 10.

While developing the code some conflicts were encountered between RDI and self-thinning. As a first solution an additional threshold was introduced rdi_limit_upper. When debugging progressed this threshold was set to 0.99 (if set to 1.00 there is no correction any more). The initial problem was resolved but the initial fix has not been removed yet. For the time being set rdi_limit_upper to 0.99.

CHECK: MAX_HARVEST_DIA

Following a disturbance (which could be a clear cut), tree species changes and forest management change can be prescribed or read from a map in ORCHIDEE-CAN. Set lchange_species = y, read_species_change_map = y, and read_desired_fm_map = y and specify the paths of those maps in the COMP/stomate.card. This functionality replaces the DGVM in areas where humans rather than nature govern species distribution, for example, Europe. Note that there are some constraints on the possible management changes. If the species is a conifer, for example, the new management strategy cannot be coppicing (fm=3). Anthropogenic species change has not been developed to work together with land cover change (this would require some good bookkeeping for veget_max). Forest management change has not been tested together with land cover change but because they affect different variables, i.e., forest_managed and veget_max combining both functionalities seems not overly complex.

Land cover change (with age classes)

Land cover change now accounts for age classes. It is controlled by the flags land_cover_change and veget_update. Set land_cover_change = n and veget_update = 0Y if land cover change should be disabled.

Litter decomposition (needs editing)

After large-scale dieback events (with a closed n-cycle, i.e., impose_cn = n), so much soil mineral N becomes immobilized to decompose litter that too little N is left for plant regrowth. To address this, we implicitly represent the action of fungivores, which eat the decomposing fungi and release N for the plants and increase N turnover rates. We set aside a fraction of qd (stomate_litter.f90) which becomes available for plant uptake in nitrogen_dynamics. This fraction is prescribed by fungivores and is allocated to n_fungivores which represents the nitrogen that is released by the fungivores and that is moved into the variable plant_uptake. This parameter is not very well constrained and has been calibrated to find a good compromise between plant regrowth and litter decomposition. The share of the n contained in the decomposing fungi that is released as an excrement from the fungivores should be between 0 and 1 and is set in the run.def:

FUNGIVORES=0.3

After a die-back (with a closed n-cycle, i.e., impose_cn = n) too little N is available to reach the target C:N ratio of the receiving soil pool, meaning that carbon litter can not decompose. If the C:N ratio is not controlled it spins out of control reaching values of more than 600. The target C:N ratio, which is a variable in stomate_litter.f90, can now vary between 1 and max_cn where the latter can be set in the run.def. Good results were obtained by setting max_cn to 250. max_cn is bound between 1 and 400, the C:N ratio of wood which is the pool with the highest C:N ratio (and biomass that contributes to the litter). During the simulation the target C:N ratio varies and a couple of decades after the dieback reaches the expected value between 5 and 20. The ecological justification behind a max_cn above the C:N ratio of the fungi is that mycorrhizae receive their C from the plant photosynthesis products. Thus, litter can still decompose even though C is not incorporated into the plant biomass. max_cn can be set in the run.def:

MAX_CN = 250

ORCHIDEE-CN allows negative n_mineralisation values to be passed to stomate_soilcarbon where they were treated as immobilization. If we have persistent negative n_mineralisation, however, we can reach a situation where we completely deplete our soil_n_min pools to satisfy the demand for immobilization. The model cannot recover from such a condition. This tends to happen when there is no plant recruitment, in which case a generation of trees will all reach their maximum diameter and die at the same time. This creates an enormous amount of litter, meaning our som_input values become relatively large and when we subtract som_input from n_mineralisation, the values become negative. To remedy this, we'll truncate som_input based on the current mineralisation pool, but we'll also limit the occurrence of negative values. We'll set a minimum min_n for n_mineralisation so we can keep some N in the soil (and so the simulation can recover). This problem was discovered and fixed before adding the fungivores and max_cn. It should be tested whether this fix is still needed. min_n can be set in the run.def:

MIN_N = .0001

Litter raking

Tree litter was collected from the forest and used in the winter stables instead of straw. In spring the litter and manure was spread on the croplands. This lateral flow of C and N between PFTS in the same pixel can be accounted for in ORCHIDEE-CN-CAN by setting use_litter_raking = y. If litter raking is to be used, the model will search for annual maps. The path of these maps needs to be specified in COMP/stomate.card. Litter raking maps were prepared for Europe. Unless liter raking is your research topic set use_litter_raking = n.

Mortality

ORCHIDEE-CN-CAN distinguished 3 types of natural mortality. The first two options are similar to those in previous version of ORCHIDEE and are set by the flag constant_mortality. If constant_mortality = y, the background mortality of a forests is calculated as a constant, prescribed fraction. In ORCHIDEE-CN-CAN, this fraction is given by residence_time (see also forest management). . If constant_mortality = y, the background mortality of a forest is a function of its net primary production (npp). If npp decreases, mortality will increase.

Both options have been developed, tested and can be used in ORCHIDEE-CN-CAN. However, because of the introduction of self-thinning mortality in ORCHIDEE-CN-CAN, constant_mortality = y soon became the default setting. In ORCHIDEE-CN-CAN, the total mortality is the maximum of the background mortality and the mortality from self-thinning. Only if self-thinning is absent or too low, background mortality will play a role. This approach implies that when constant_mortality = y is used in combination with self-thinning, background mortality will only play a role in the first years to decade before self-thinning starts. Despite its limited use, it represents an essential process: owing to background mortality, the number of individuals decreases, the remaining individuals grow faster and thus manage to reach self-thinning in a reasonable amount of time. It needs to be tested how the interplay between background mortality and self-thinning will work out when constant_mortality = n is used.

Nitrogen cycle

ORCHIDEE-CN-CAN strictly follows ORCHIDEE-CN where it concerns the implementation of the N-cycle. Following mass balance problems caused by negative N mineralization and followed by negative immobilization, the code has been slightly adjusted to ensure mass balance closure. First the flag stomate_ok_ncycle needs to be set to y, to run the model with a N-cycle. Subsequently the parameter impose_cn is used to control the N-cycle calculations. If set to y, C/N ratios are calculated but whenever N appears to be limiting, it is taken from the atmosphere to satisfy this need. This is the preferred setting when testing/developing the code without a proper spin-up. N-limitation will only be accounted for when setting impose_cn = n. With this setting the N-cycle is closed (checked when checking for mass balance closure) it will require a spin-up to produce reasonale results.

The paths of the Nammonium_FILE, Nnitrate_FILE, Nfert_FILE and Nbnf_FILE will need to be set to prescribe the N-inputs from fertilization, biological nitrogen fixation and atmospheric N-deposition.

Photosynthesis

Photosynthesis is calculated as in ORCHIDEE-CAN, and ORCHIDEE-CN but the way the levels are defined has changed. The reason of this change is in the albedo and energy budget for which physical layers are required. For this reason the space between the bottom and the top of the canopy is divided into x layers. X is set by the parameter nlevels_photo. Applications with ORCHIDEE-CAN used nlevels_photo = 10. This values was retained in ORCHIDEE-CN-CAN. Contrary to ORCHIDEE-CN and previous versions of ORCHIDEE where each canopy layer contained 0.5 units of LAI, in ORCHIDEE-CN-CAN, each canopy layer will have the same depth (in m) but will contain a different amount of LAI because tree canopies are assumed spherical.

Plant water stress

With ORCHIDEE-CN-CAN there is two option to calculate waters stress.

1 - Same as in the trunk, where a soil moisture stress functions limits C assimilation through constraints on the carboxylation capacity.

2 - The second possibility takes hydraulic architecture of plants into account, when calculating the plant water supply. This scheme, based on Hickler et al (2006), calculates the water supply as the ratio of the pressure difference between the soil and leaves. The plant water supply is the amount of water the plant can transport from the soil to the stomate accounting for resistance of water transport in the roots, sapwood and leaves. The resistances are inversely proportional to the conductivities in these different tissues, with the sapwood conductivity decreasing when cavitation occurs. If transpiration rates exceeds plant water supply, stomatal conductance is reduced.

Moreover, as a further refinement to the hydraulic architecture, a more detailed description of the soil to root resistance has been included (now the default setting) and is activated with the flag OK_SOIL_ROOT. In the original scheme of the hydraulic architecture a tuning factor was used to calculate the soil to root resistance (PSI_SOIL_TUNE). When OK_SOIL_ROOT = y, an adjustment of the soil water potential (psi_soilroot) due to soil to root resistance is allowed. Pssi_soil is weighted by the fraction of evapotranspiration supplied by each soil layer, which depends on the soil to root resistance per layer that accounts for different root properties and hydraulic conductivity.

If you wish to make simulations with the hydraulic architecture the CWRR hydrology scheme needs to be activated (HYDROL_CWRR=y). Moreover, both mleb and ok_hydrol_arch needs to be true. These can be controlled by the overall flag multi_layer_control (see section on Single layer vs. multi-layer energy budget for more elaboration on these flags).

The following PFT dependent parameters are needed for the calculations accounting for plant hydraulic architecture; minimal leaf water potential PSI_LEAF_xx, sapwood leaf water potential that causes 50 % loss of xylem PSI_50_xx, additive tuning parameter to account for soil-root interactions PSI_SOIL_TUNE_xx, maximum sapwood conductivity K_SAP_xx, root conductivity K_ROOT_xx, leaf conductivity K_LEAF_xx, specific root lenght SRL_xx, fine root radius R_FROOT_xx, minimum root water potential PSI_ROOT_xx.

Prescribe initial vegetation

At the start of the model run or after a die-back or clear-cut new vegetation needs to be planted as ORCHIDEE does not grow vegetation from seeds. The initial dimensions of the vegetation is thus prescribed. Given that the allocation follows allometric relationships, any of the tree dimensions or any mass of any component could have been used to prescribe. The variable height was chosen because it is easy to (mentally) visualize the prescribed vegetation. In the run.def HEIGHT_INIT_MIN and HEIGHT_INIT_MAX need to be prescribed. Typical values are 2 to 5 meter. If more than one diameter class is used, HEIGHT_INIT_MAX needs to larger than HEIGHT_INIT_MIN. The larger the difference between the min and max, the more vegetation layers the canopy will be composed from.

In addition the initial number of seedlings needs to be prescribed as well. For this the parameter NMAXTREES needs to be set. NMAXTREES is a critical parameter to obtain acceptable model behavior. If it is too high, lai saturates but the stand-level GPP will be distributed over too many individuals, each individual will grow very little and so it will take very long before self-thinning is reached. If it is set too low, LAI will be too low resulting in a too low GPP and thus very slow growth. A good starting values is a bit below self-thinning. That way the vegetation starts growing, individual are killed thanks to the background mortality and within 10 to 20 years self-thinning is reached. Why not starting at self-thinning? During code development it was tried to have the model start at the exact number of trees at which self-thinning will start given the diameter of the tree. One issues was that when prescribing small individuals (2-3 m) the calculated number of trees could in the millions and so the GPP had to be distributed over too many individuals.

The way the initial vegetation is prescribed can also be used to prescribe a mature vegetation right at the start of the simulation. One could set HEIGHT_INIT_MIN and HEIGHT_INIT_MAX to for example 15 and 20. If NMAXTREES is not adjusted, massive self-thinning will take place on the first day. It is better to set NMAXTREES to a value just below the exact self-thinning value. Prescribing mature vegetation has been tested and works but it is very sensitive. If the combination of NMAXTREES, HEIGHT_INIT_MIN and HEIGHT_INIT_MAX is far from realistic the model may crash (the change in KF following the change in gap probability could result in problems in allocation) and N-feedbacks will become apparent in for example the leaf mass (likely because when mature vegetation is prescribed at the start of a run, there is not enough soil N to maintain a mature vegetation).

Recruitment

When stands grow old the tree density decreases and under certain conditions the LAI can no longer cover the ground area. When this happens productivity will start to decrease. In nature the decrease in LAI comes with an increase in the amount of light reaching the forest floor which enables seedlings to grow and to restore the LAI. This process is known as recruitment. Note that in managed forest and forest with a lot of stand replacing disturbances (for example, fire) the forest may never reach the stage where the canopy sufficiently opens up to enable recruitment.

ORCHIDEE-CN-CAN can simulate recruitment for each PFT separatly by setting recruitment_pft_xx to true or false. When using age classes it makes sense to have the same setting for all age classes of the same species. When developing and testing the code it was considered too time consuming to change the settings at the PFT level so a flag ignoring the setting for recruitment_pft_xx was introduced. If the ok_recruitment flag is set to true, the settings of recruitment_pft_xx will be used. If the ok_recruitment flag is set to false, the settings of recruitment_pft_xx will be ignored and the model runs without recruitment.

Recruitment has been developed, tested and validated for tropical forests. There is no reason why it shouldn't work for other forests but that needs to be confirmed. At present recruitment was introduced at the PFT level. It probably makes more sense to link it to the management strategy than to the PFT. This needs to be checked.

River routing

The river routing code was not touched during the development of ORCHIDEE-CN-CAN. This functionality has not been tested yet for ORCHIDEE-CN-CAN. Unless rivers are your main interest, set river_routing to n.

Soil maps

ORCHIDEE-CN-CAN runs for two soil maps: zobler (3 soil types) and usda (12 soil types). The usda map will become the default setting in the near future. The code had to be adjusted in sechiba_hydrol_architect.f90 and stomate_windthrow.f90. The soil map can be selected in the run.def by setting the value for soiltype_classif:

SOILTYPE_CLASSIF= usda 

or

SOILTYPE_CLASSIF= zobler

Single layer vs multi layer energy budget

There are still some issues with the multi-layer energy budget, and it is currently only possible to run for one PFT. Thus, we suggest you use the single layer energy budget. This can, however, be done by two different methods that gives the exact same results:

A: use the energy scheme as found in the original enerbil.f90

B: use the multi-layer energy scheme for a single canopy layer. This collapses the multi-layer energy scheme to the original enerbil code, but if you wish to take hydraulic water stress with feedback on stomatal conductance into account in your simulation, you need to choose this method.

Several flags exist to control these choices. For simplicity, a flag has been made to automatically control several of the flags related to the multi-layering, called multi_layer_control. Multi_layer_control has 4 possibilities

1 – single layer, using the multi layer energy scheme (i.e. choice B above)

2 – multi-layer energy budget

3 – user specific as set in the run.def

4 – single layer, using the original enerbil scheme (i.e. choice A above)

The default setting of multi_layer_control in ORHIDEE-CN-CAN is 1. This is well tested and considered save to use. More details of the flags controlling the multi-layer can be found below and within the model code.

The flags controlled by the multi_layer_control are:

ok_hydrol_arch: Flag that activates the hydraulic architecture routine (true/false). The trunk version of ORCHIDEE (false) uses soil water as a proxy for water stress and applies the stress to Vcmax. When set to true the hydraulic architecture of the vegetation is accounted for to calculate the amount of water that can be transported through the plant given the soil and leaf potential and the conductivities of the roots, wood and leaves. Water supply through the plant is compared against the atmospheric demand for water. If the supply is smaller then the demand, the plant experiences water stress and the stomata will be closed (water stress is now on gs rather than Vcmax). Note that whether stomatal regulation is used or not is controled by a separate flag: ok_gs_feedback. ok_gs_feedback: Flag that activates water stress on stomata (true/false). This flag is for debugging only! It allows developers to calculate GPP without any water stress. If the model is used in production mode and ok_hydrol_arch is true this flag should be true as well.

ok_mleb: Flag that activates the multilayer energy budget (true/false). The model uses 10 (default) canopy layers to calculate the albedo, transmittance, absorbance and GPP. These canopy layers can be combined with 10 (default) layers below and 9 layers above the canopy to calculate the energy budget (ok_mleb=y). If set to no, this flag will make the model use 10 layers for the canopy albedo, transmittance, absorbance and GPP and just a single layer for the energy budget. Be aware that if you wish to run with hydraulic architechture ok_mleb needs to be se to true as well. Furthermore, if you wish to run with the original energy scheme (enerbil), set the layers for mleb to 1.

ok_impose_can_structure: This flag is for debugging only! It allows developers to use a prescribed canopy structure rather then the structure calculate by ORCHIDEE. The flag activates the sections of code which directly link the energy budget scheme to the the size and LAI profile of the canopy for the respective PFT and age class that is calculated in stomate, for the albedo. If set to TRUE and the multi-layer budget is activated the model takes LAI profile information and canopy level heights from the run.def. If set to FALSE, and the multi-layer energy budget is used the profile information and canopy levels heights comes from the PGap-based processes for calculation of stand profile information in stomate.

ok_mleb_history_file: Flag that controls the writing of an output file with the multi-layer energy simulations (true/false). Note that this is a large file and writing slows down the code.

The multi_layer_control set these flags in the following manner:

1 – single layer, using the multi layer energy scheme

       ok_hydrol_arch = .TRUE.;  
       ok_gs_feedback = .TRUE.; 
       ok_mleb = .TRUE.; 
       ok_impose_can_structure = .TRUE.; 
       ok_mleb_history_file = .FALSE.

2 – multi-layer energy budget

       ok_hydrol_arch = .TRUE.; 
       ok_gs_feedback = .TRUE.; 
       ok_mleb = .TRUE.; 
       ok_impose_can_structure = .FALSE.; 
       ok_mleb_history_file = .FALSE.

3 – user specific as set in the run.def

       All read from the run.def

4 – single layer, using the original enerbil scheme

       ok_hydrol_arch = .FALSE.; 
       ok_gs_feedback = .FALSE.; 
       ok_mleb = .FALSE.; 
       ok_impose_can_structure = .FALSE.; 
       ok_mleb_history_file = .FALSE.

Specific leaf area

Similar to ORCHIDEE-CN, ORCHIDEE-CN-CAN users can choose to use a constant specific leaf area (SLA) or a dynamic (SLA) by setting the flag dyn_sla. SLA is a fundamental parameter in the allocation scheme and it is well established that SLA is dynamic in nature especially during leaf onset. The current implementation of dynamic SLA is basic and there is room to enhance consistency in the calculations, for example, by recalculation the allocation factors KF and LF as a function of SLA.

Spin-up

The analytical spin-up works with ORCHIDEE-CN-CAN. To ensure growth at the onset of the analytic spin-up for all PFTs initial values are needed for the SOM pools, and that is irrespective of whether IMPOSE_CN is y or n. The initial pools are set in stomate_io.f90, and they can be specified in the run.def by:

SOM_INIT_ACTIVE = 1000
SOM_INIT_SLOW = 3000
SOM_INIT_PASSIVE = 3000
SOM_INIT_SURFACE = 1000

The initial values of carbon for the four SOM pools are used globally. Sensitivity test have shown that the analytic spin-up is not sensitive to the actual size of the initial values. The different system will after a while settle in to there own state (N limited or saturated) in spite of having the same initial state.

Start and restart (under development)

The table below shows typical configurations to start and restart ORCHIDEE-CN-CAN

Sechiba stomate
Restart impveg impsoil laimap ok_stomate restart Comment
- - - - + - 1. Typical for a spin-up from scratch
- - - + - n.a. 2. Typical for a spin-up with sechiba only
- + + - + - 3. Typical for a site-level spin-up
- + - - + - 4a. Typical for a test run with prescribed PFTs
- - + - + - 4b. Typical for a test run with prescribed soil properties
+ - - - + + Typical for a simulation following the spinup described in 1
+ - - + - n.a. Typical for a simulation following the spin-up described in 2
+ + + - + + Typical for a simulation that will be compared against site-level observations and that was spun-up with 3
+ + - - + + Typical for restarting a test run following a test run described in 4a
+ - + - + + Typical for restarting a test run following a test run described in 4a

The code allows for many more combinations than described above. Some of those configuration could be useful in specific cases but be aware of the following:

  • if the restart file for sechiba does not come from the same simulation/year as the restart file for stomate, the values for the variables that are stored in both sechiba and stomate will be overwritten by the values stored in the stomate restart file because that file is read later in the initialization phase. Some variables are duplicated because that adds the flexibility to the configuration that can be used to restart the model.
  • When read_lai is true, ok_stomate should be false. It may work but it is flawed from a conceptual point of view. read_lai was developed for cases where the model cannot simulate its own canopy. ok_stomate was developed so the model would simulate its own canopy.
  • Combining a sechiba restart from with impveg, impsoil or laimap may result in values from the restart file being overwritten because impveg, impsoil and laimap are checked after reading the sechiba restart file.
  • NCO commands could be used to hack into the restart file. This could be powerful when stomate is used and one wants to restart the model after a spin-up but also want that the forest starts growing the first year of the simulation. In that case circ_class_biomass and circ_class_n should be set to zero. This approach is straightforward when no age classes are used. When age classes are used many more variables need to be moved to the correct PFT to avoid conflicts.

Stomate

ORCHIDEE-CN-CAN strengthen the links between sechiba and stomate. As in previous versions, stomate makes use of variables calculated in sechiba but in ORCHIDEE-CAN and ORCHIDEE-CN-CAN, sechiba requires information from stomate to work properly. For the moment set stomate_ok_stomate to y (_AUTOBLOCKER_). For the future it seems possible to prescribe LAI and assume a canopy structure but this code still needs to be restored and tested. For the time being set lai_map = n.

Parameterization and Evaluation

Workflow

The following outlines the strategy for parameterizing and evaluating the performance of ORCHIDEE-CN-CAN in simulating several key forest ecosystem processes including tree growth dynamics, energy exchange, C-N cycling and plant hydrology.

The work will likely require several iteration containing all or just a coupled of the following: 1) running the model, 2) tuning key parameters, 3) re-running the model 4) spin-up and 5) evaluating to model, to discern parameter values that allow us to reproduce as closely as possible the global patterns and trends for several key processes. The tests will move across scales starting at pixel scale moving to longitudinal band, European scale and ending at global scale. Although the longitudinal bands are of little use in the evaluation itself they can be considered pre-tests before global tests are run. It is hoped that longitudinal bands would speed up the tests and reduce the computational cost.

CLARIFICATION: Is the reduction of computational cost due to the longitudinal tests finding bugs that would otherwise only appear in global runs?

Table 1. Workflow of the parameterization and evaluation of ORCHIDEE-CN-CAN.

Phase Work to be done
Prepare and check Check whether the model runs, the key functionality, and the spin-up are bug free (Table 2). Check whether the scripts, tools and data are available and still working (Table 3).
Initial evaluation Run longitudinal spin-up and check order of magnitude in soil carbon pools. Use the same runs to check response gradients to temperature, N-deposition, precipitation and management (§5.4).
Parameterization Use FLUXNET and some additional data sources to parameterize the model at the site level (Table 4).
Final evaluation Run a spin-up + transient simulation over Europe compare the simulation against the spatially explicit data (Table 5).

Apparent bug-free

Before starting the work proposed in Table 1 it needs to be confirmed that the model is technically capable of the tasks presented in Table 2.

QUESTION: Are we aiming for 500-1000 years with full debug flags on both obelix and irene to increase our confidence? That might be too expensive even at 2 degree resolution for a global run, but we could check.

Table 2. Essential technical capabilities before evaluating ORCHIDEE-CN-CAN.

Description of the task Status
Is the model stable? Can it be run for 500 - 1000 years? Can it be run for all PFTs? Recent problems with grasslands
Is the N-cycle working? Can we run simulations with IMPOSE_CN=n and IMPOSE_CN=y? OK
Is land cover change working? Can we run simulations with (changing) land cover maps (impose_veg=no)? OK
Is forest management working? Can we read forest management maps, are the temporal trends in biomass what is expected for unmanaged, high-stand and coppice management? Still need to run idealized set-ups
Is litter raking working? Can we read litter raking maps and is the litter pool in forest decreasing and the litter pool in croplands increasing? Still needs to be checked
Does the model run in parallel? OK
Is the analytical spin-up working? OK
Do we have all the driver maps for the years 1600 to 2000 for European simulations? Climate, land cover change, forest management, litter raking, and N-deposition? OK
Do we have all the driver maps for the years 1600 to 2000 for global simulations? Still need litter raking and FM
Read N-deposition maps through COMP In progress

Data availability

Observational data products for model-data comparison can be found at: /home/satellites5/maignan/ORCHIDEE/EVALUATION.

  • Phenology

GIMMS (LAI and FPAR3g)

  • Forest structure

◦ Remote sensing products of biomass (temperate and boreal maps, i.e., Turner) ◦ Biomass of EU forest from JRC (Europe) ◦ Global 1-degree Maps of Forest Area, Carbon Stocks, and Biomass, 1950-2010 (ORNL DAAC) ◦ Avitabile product (Global forest biomass) ◦ Forest basal area (Europe) ◦ Canopy height (Global)

  • NPP

◦ Site-level NPP database Luyssaert et al 2007

  • NEP

◦ FLUXNET site-level data

  • TER

◦ FLUXNET site-level data

  • GPP

◦ FLUXNET site-level data ◦ EC-based upscaled GPP, i.e., Jung

  • NPP/GPP

◦ site-level data and regional patterns, i.e., Campioli et al 2015

  • Soil hydrology

◦ ESA CCI ECV ◦ measurements from Brazil (ABRACOS product) ◦ River discharge records from selected gauges (Global coverage)

  • Albedo

◦ MODIS or GlobAlbedo? for albedo

  • Energy (sensible and latent heat)

◦ GLEAM for evapotranspiration ◦ EC-based upscaled evapotranspiration, i.e., Jung

  • Tree ring width (not on /home/satellites5/)

◦ ITRDB

  • LCC

◦ Luyssaert et al 2014 – FLUXNET changes in Rn, LE, H, G, albedo ◦ Duveiller et al 2018 – Remote sensing changes in Rn, LE, H+G, albedo

  • NFI

◦ France, Spain, Germany and Sweden (/home/satellites5/) ◦ EU-wide data through the VERIFY project?

There are several data tools (ATLAS, Mapper and Jerome’s) to help compare model outputs with observation, which we might be able to use. If we will not be using the available tools for comparison, we need to preprocess the observational data products to produce global means, time series, decadal averages of spatial patterns etc. The analyses presented in Naudts et al 2015 are a good starting point for the evaluation of ORCHIDEE-CN-CAN. The following scripts are available and have been tested and re-activated

Table 3. Model evaluations scripts available per November 2018

Description Status Path
Extract species-level productivity from French NFI data and compare against simulated productivity – looking to replace the French data by EU-wide data through the VERIFY project OK dofoco/dofoco/SCRIPTS/XXX
Extract Jung GPP and compare against spatially explicit simulations for 8 regions in Europe OK dofoco/dofoco/ SCRIPTS/XXX
Extract Jung evapotranspiration and compare against spatially explicit simulations for 8 regions in Europe OK dofoco/dofoco/ SCRIPTS/XXX
Extract MODIS albedo (NIR & VIS) and compare against spatially explicit simulations for 8 regions in Europe In progress dofoco/dofoco/ SCRIPTS/XXX
Extract height and compare against spatially explicit simulations for 8 regions in Europe OK dofoco/dofoco/SCRIPTS/XXX
Extract effective LAI and compare against spatially explicit simulations for 8 regions in Europe OK dofoco/dofoco/ SCRIPTS/XXX
Extract BA and compare against spatially explicit simulations for 8 regions in Europe OK dofoco/dofoco/ SCRIPTS/XXX
Extract half-hourly EC observations and compare with half-hourly site level simulations for NEP, GPP, TER, and evapotranspiration ??? ENSEMBLE/FLUXNET

Parameterization

As we will compare the simulations to observations, the simulations need to be our best shot to resemble reality. Hence, all site-level simulations made for the parameterization will have the following configuration:

  • Start from a spin-up
  • impose_cn = no (i.e., accounting for N-deposition )
  • Forest management = 2 (i.e., accounting for a thin and fell type of forest management)

QUESTION: Do we consider litter raking on site level? Do we consider litter raking for FLUXNET evaluations?

European simulations will have the following configuration:

  • Start from a spin-up
  • impose_cn = no (i.e., accounting for N-deposition )
  • impose_veg = no (i.e., accounting for PFT distribution)
  • Forest management from map

The following parameterization approach – making use of parameters that were already shown to be sensitive to tuning – is proposed:

Table 4. Proposed order, parameters and tools to parameterize ORCHIDEE-CN-CAN. Once all steps has been performed, the NPP to GPP ratio should be re-evaluated, possibly resulting in another tuning cycle for some of the PFTs.

Process Parameter(s) Tool
Onset of growing season and start of senescence thresholds for phenology FLUXNET
NPP/GPP ratio coeff for maintenance respiration FLUXNET + Campioli
Magnitude of LAI k_latosa_min and k_latosa_max NFI + Luyssaert et al 2007
Magnitude of GPP LL_alpha, Vcmax, J_max FLUXNET
Magnitude of NPP Implicit through NPP/GPP and GPP Luyssaert et al 2007
Evapotranspiration LAI_top, water stress FLUXNET
Water stress To be discovered FLUXNET
Magnitude of Rh To be checked - Rh Luyssaert et al 2007
Magnitude of TER To be checked - Rh FLUXNET
Diameter, height, and biomass form factor NFI + Luyssaert et al 2007
Density and biomass self-thinning parameters NFI + Luyssaert et al 2007
Harvest and biomass Self-thinning and RDI parameters NFI + ???
Albedo Implicit through LAI and forest structure FLUXNET?
Tree ring width Self-thinning, recruitment ITRDB or VERIFY Fig 4

Settings for the FLUXNET comparison

The parameterization starts with several 1-pixel test cases coinciding with long-term flux-net sites to test whether the model captures the growth dynamics such as phenology, max LAI, GPP, etc. These tests require a spin-up. The 1-pixel test cases will allow for both parameter tuning and changes in the code to improve the model behavior. The majority of the data represent mature forests, hence, the modelled forests should be mature as well. The model will be run for 80 years, before any output will be compared to the FLUXNET measurements. An iterative process is be planned:

  • 80 years to reach mature forest → parameterize
  • Re-run the 80 years to reach mature forest with the new parameters → parameterize
  • Re-run spin-up and 80 year simulation to reach mature forest with the new parameters → parameterize
  • Continue until satisfied

Only if we experience too many difficulties with manual tuning (if there are too many non-linearities in the model), we will use the multi-site optimization tool developed by Vlad . When the simulated growth dynamics are satisfying, 140 years long tests will be performed to check cumulative variables such as basal area, tree height, tree diameter, stand density, standing biomass, and harvest. To evaluate net ecosystem exchange of carbon and soil carbon and nitrogen pools a spin-up is required. Note that the spin-up depends on the parameters used in ORCHIDEE and that the sensitivity of parameters in ORCHIDEE depends on the spin-up. There is no easy way to break this dependency. We should avoid to ‘over-tune’ the 1-pixel FLUXNET comparisons. Instead, we will continue evaluating the model over longitudinal bands.

DISCUSSION POINT: The current scripts for FLUXNET evaluation are broken down into three parts. 1) An initial looping over the driver file (1-10 years, depending on the file), 2) 500 years of spinup (regardless of the length of driver file), 3) one final loop over the driver file, for production. I like how the process is described above, and agree that makes a lot of sense for forests. In particular, I think we need 80 years from planting after spinup plus a production run over the forcing file to avoid trees dying and biasing our results. It's not clear how much time it will take to coax the scripts to do this. I think the grassland and cropland sites can easily use the existing setup. We also may have an issue with age classes, as that changes the PFT of interest (but should be straightforward to fix).

Settings for longitudinal bands

Test with longitudinal bands will be used to ensure that we have the expected gradients across regions (response to: climate, precipitation, management, and nitrogen). Longitudinal bands will not be (formally) compared against data, they should be considered as intermediate tests. If the expected responses are not present in the longitudinal bands, there is no reason to expect that they will be present in the European/global runs. Therefore, the issue should be investigated/fixed before launching large scale simulations. Three longitudinal tests will be conducted covering 2 pixels wide bands from N-Canada to Southern Chili, N-Norway to S-Africa and one band N-E Russia to Australia (including W-Australia → only place where something grows in Australia). If the evaluation is limited to checking whether response gradients are indeed present in the simulations, the configuration could be limited to running the analytical spin-up until equilibrium. Once we moved on to the longitudinal bands the focus will be on tuning of parameters. Any problems with the model and its functionality should have been caught during the 1-pixel and longitudinal tests.

Longitudinal bands can be run largely independent from the 1-pixel tests. After concluding the 1-pixels tests it probably makes sense to run the longitudinal bands to check whether the latest parameters still produce the expected response gradients.

Settings for European tests

Once happy with the pixel and longitudinal tests we will move on to the European scale. Ultimately, the aim is to produce the control run for future simulation experiments. The model should be spin-up for 1600 by making use of the 1901-1920 climatology (because those are the coldest decades in the climate forcing). During the spin-up the forest management map and N-deposition maps for 1600 should be used. As a consequence, equilibrium soil carbon should be tested at the regional scale rather than the individual pixels.

Subsequently, a transient spin-up will be run starting from the 1600 spin-up. During this spin-up, changes in forest management, CO2, N-deposition, litter-raking, and land cover change will be accounted for. The climate data will still have to cycle over 1901-1920. The transient simulation will run until 1750 and, pending on successful evaluation, extended until 1950 and 2015. 1750, 1950 and 2015 are common starting points for simulation experiments.

Whether we take 20 or 30 years climate-cycles depends on the exact length of the simulation; the target is that by the time the simulation reaches the year 1901, a cycle has been completed and so the cycle in the climate forcing is synchronized with the simulation years.

The European control should be run from the year 1601 to 2015 includes:

  • A spin-up as initial condition to make a European 1° x 1° CONTROL simulation
  • 64 PFTs (some with 4 age classes)
  • With forest management
  • Litter raking
  • Increasing atmospheric CO2 concentrations read from file
  • LCC
  • CRU-NCEP meteorological forcing, cyclic meteorology (1901-1920) until 1900, then use the corresponding year
  • A dynamic N-cycle (impose_cn=n)
  • N-deposition, we have N-deposition files from approximately 1860 until present.

DISCUSSION POINT: We are not guaranteed that the European control will be able to be used as a restart for other runs. For example, we expect VERIFY to be at 11 km resolution suing completely different forcing data. Does this change at all how we want to proceed? Or is that a separate issue? I think it makes sense to do the evaluation as stated, and then do what is needed for the other runs separately.

Table 5. Observational data products to evaluate European control simulation

Observations Tool
EU-wide NFI data (Fig. 3) through the VERIFY project ???
Jung GPP – relies on FLUXNET thus partly circular XXX
Jung evapotranspiration – relies on FLUXNET thus partly circular XXX
MODIS albedo NIR & VIS XXX
Remote sensing tree height XXX
Extract effective LAI from MODIS XXX
BA from JRC – relies on NFI thus partly circular XXX
Biomass from JRC – relies on NFI thus partly circular ???
Biomass from Turner – Remote sensing? ???
Harvest from Schelhaas - data through the VERIFY project ???
LCC – Luyssaert et al 2014 XXX
LCC – Duveiller et al 2018 ???

Settings for spin-up and re-parameterization

As a spin-up is costly in both time and computer resources, we need a strategy to avoid wasting these resources. Thus, the spin-up will like the parameterization, be done across scale moving from pixel to global scale. The spin-up will be done in parallel with the parameterization. Often the problems with the spin-up have a technical characters and show up for the pixels with extreme climate conditions. Before launching a longitudinal, regional or global spin-up, we should agree on the model version to use, because structural changes to the code will necessitate re-running the spin-up. The model version to use, will most likely be the version ready once the parameterization at the 1-pixel level is satisfying, and no more changes to the code need be added.

  • Identify the variables that are targeted by the spin-up (such as NEP, heterotrophic respiration, decomposition etc). The spin-up will reveal whether parameters affecting these variables need to be tuned. The spin-up itself is also an interesting test case that could be loosely compared against data.
  • We could compare the spin-ups to maps of soil carbon stocks to check the order of magnitude. Soil carbon maps should only be formally compared with the control run (spin-up + transient simulation) for the year 2000 because that run includes the simulated effects of N-deposition, management, litter raking and land cover change. The more simple configurations of the spin-up do not account for these processes or do not account for the right sequence of processes

Attachments (8)