ORCHIDEE-TRUNK OPTIMIZATION
ROUND8 (11/03/2024)

Select:

Update (29/10/2024)

  1. Optimization is redone for crops (PFT12d, PFT13d) following the ORCHIDEE code corrections and few changes in parameters and control variables.
  2. Model version: ORCHIDEE_trunk revision 8698.
  3. Run.def change:
    HEIGHT_INIT__12 = 0.05
    HEIGHT_INIT__13 = 0.05
    OK_CROP_PLANTING_N_SOIL = y
    OK_CROPS_IPRESENESCENCE = n
  4. Different optimized parameters setup is used:
    ParameterPFT12PFT13
    K_LATOSA9500
    [4750 - 19000]
    11000
    [5500 - 22000]
    K_BELOWGROUND4e-7
    [2e-7 - 8e-7]
    4e-7
    [2e-7 - 8e-7]
    VMAX_UPTAKE_NH49
    [2.25 - 18]
    9
    [2.25 - 18]
    VMAX_UPTAKE_NO39
    [2.25 - 18]
    9
    [2.25 - 18]
    COEFF_MAINT_INIT0.05
    [0.025 - 0.1]
    0.08
    [0.04 - 0.16]
    SLAINIT0.02
    [0.01 - 0.04]
    0.02
    [0.01 - 0.04]
    NUE_OPT60
    [30 - 120]
    60
    [30 - 120]
    LONGEVITY_SAP480
    [240 - 960]
    480
    [240 - 960]
    DECIDUOUS_RESERVE0.3
    [0.1 - 1]
    0.3
    [0.1 - 1]
    K_LATOSA_MAX = K_LATOSA_MIN = K_LATOSA
  5. Different control variables setup is used:
    VariablePFT12PFT13
    GPP [gC/m2/y]600 - 1500700 - 1700
    LAI_MEAN_GS1.5 - 3.52.5 - 4.5
    ng_ratio0.3 - 0.75
    lr_ratio1.3 - 3.5
    bmc_ratio0.1 - 0.6
    ge_ratio1 - 31 - 4
    ng_ratio = NPP / GPP
    lr_ratio = LEAF_M_c / ROOT_M_c
    bmc_ratio = (LABILE_M_c + RESERVE_M_c + FRUIT_M_c) / (SAP_M_BE_c + SAP_M_AB_c + HEART_M_BE_c + HEART_M_AB_c + ROOT_M_c + LEAF_M_c)
    ge_ratio = GPP / EVAP

Update (06/10/2024)

A bug has been found in the calculation of error associated to each simulated variable that is further used in the cost-function calculation. As an effect the control variables were badly weighted and some variables were given much more priority during the sim-obs tuning.

Optimization is redone for crops with corrected error calculation (PFT12c, PFT13c)

Update (02/10/2024)

  1. Optimization is redone for crops (PFT12b, PFT13b) following the ORCHIDEE code improvement.
  2. Model version: ORCHIDEE_trunk revision 8678.
  3. Two new flags are activated (OK_CROP_PLANTING_N_SOIL = y and OK_CROPS_IPRESENESCENCE = y)
  4. Different optimized parameters setup is used:
    ParameterPFT12PFT13
    COEFF_MAINT_INIT0.05
    [0.005 - 0.05]
    0.08
    [0.005-0.08]
    PRE_HARV500
    [400 - 600]
    500
    [400 - 600]
    PRE_GDD1000
    [800 - 1300]
    1000
    [800 - 1300]
    HEIGHT_INIT0.2
    [0.05 - 0.2]
    0.2
    [0.05 - 0.2]
    K_LATOSA9500
    [1000 - 9500]
    11000
    [1000 - 11000]
    PIPE_DENSITY2e5
    [1e-5 - 2e6]
    2e5
    [1e5 - 2e6]
    K_BELOWGROUND4e-7
    [4e-8 - 4e-4]
    4e-7
    [4e-8 - 4e-4]
    K_SAP0.0006
    [0.0006 - 0.03]
    0.0006
    [0.0006 - 0.03]
    SLAINIT0.02
    [0.005 - 0.02]
    0.02
    [0.005 - 0.02]
    K_LATOSA_MAX = K_LATOSA_MIN = K_LATOSA

Update (29/03/2024)

  1. Simulations are reproduced with ORC3 (rev8120) at the same sites.
  2. Spinup time-series are added for ORC3 (variable: total soil carbon).
  3. Prior daily time-series and seasonal cycles across 5Y are added for ORC3 and ORC4 (variables: fluxlat, fluxsens, gpp, lai, nee).

Configuration (11/03/2024)

  1. Model version: ORCHIDEE_trunk revision 8464.
  2. Sites: selection of 5 sites from the ensemble used in SIF project (selected by C.Bacour, 254 sites across all PFTs).
  3. Forcing: global CRUJRA v2.4 (closest pixel to each site location).
  4. PFT composition: 100%-composition of a corresponding PFT.
  5. Spinup: 240Y recycling 1901-1910 forcing with fixed ATM_CO2 = 350 ppm.
  6. Transient: 100Y across 1901-2000 with fixed ATM_CO2 = 350 ppm.
  7. Optimization: 20Y (2001-2020) with yearly outputs, the last 10Y are used in optimization.
  8. Control variables with ranges of variation:
    VariablePFT10PFT11PFT12PFT13PFT14PFT15
    GPP [gC/m2/y]500 - 1000600 - 14001000 - 1400600 - 1600600 - 1400500 - 1200
    LAI_MEAN_GS0.5 - 1.50.5 - 1.50.5 - 1.50.5 - 1.50.5 - 20.5 - 1.5
    ng_ratio0.4 - 0.60.32 - 0.580.53 - 0.70.51 - 0.620.32 - 0.580.32 - 0.5
    lr_ratio0.5 - 1.5
    ge_ratio1 - 33 - 8.51.2 - 2.83 - 8.51 - 21 - 3.2
    bm_ratio0 - 0.1
    ng_ratio = NPP / GPP
    lr_ratio = LEAF_M_c / ROOT_M_c
    bm_ratio = (LABILE_M_c + RESERVE_M_c) / (SAP_M_BE_c + SAP_M_AB_c + HEART_M_BE_c + HEART_M_AB_c)
    ge_ratio = GPP / EVAP

  9. Optimized parameters with default values and ranges of variation
    (vmin = 0.5*vdef and vmax = 2*vdef for all parameters except for VMAX_UPTAKE_XXX vmin=0.25*vdef):
    ParameterPFT10PFT11PFT12PFT13PFT14PFT15
    K_LATOSA7000
    [3500 - 14000]
    5500
    [2750 - 11000]
    9500
    [4750 - 19000]
    11000
    [5500 - 22000]
    8500
    [4250 - 17000]
    10000
    [5000 - 20000]
    K_BELOWGROUND4.0E-7
    [2.0E-7 - 8.0E-7]
    4.0E-7
    [2.0E-7 - 8.0E-7]
    4.0E-7
    [2.0E-7 - 8.0E-7]
    4.0E-7
    [2.0E-7 - 8.0E-7]
    4.0E-7
    [2.0E-7 - 8.0E-7]
    4.0E-7
    [2.0E-7 - 8.0E-7]
    VMAX_UPTAKE_NH49
    [2.25 - 18]
    9
    [2.25 - 18]
    9
    [2.25 - 18]
    9
    [2.25 - 18]
    9
    [2.25 - 18]
    9
    [2.25 - 18]
    VMAX_UPTAKE_NO39
    [2.25 - 18]
    9
    [2.25 - 18]
    9
    [2.25 - 18]
    9
    [2.25 - 18]
    9
    [2.25 - 18]
    9
    [2.25 - 18]
    COEFF_MAINT_INIT0.06
    [0.03 - 0.12]
    0.05
    [0.025 - 0.1]
    0.05
    [0.025 - 0.1]
    0.08
    [0.04 - 0.16]
    0.07
    [0.035 - 0.14]
    0.13
    [0.065 - 0.26]
    SLAINIT0.031
    [0.0155 - 0.062]
    0.031
    [0.0155 - 0.062]
    0.02
    [0.01 - 0.04]
    0.02
    [0.01 - 0.04]
    0.031
    [0.0155 - 0.062]
    0.031
    [0.0155 - 0.062]
    NUE_OPT45
    [22.5 - 90]
    45
    [22.5 - 90]
    60
    [30 - 120]
    60
    [30 - 120]
    45
    [22.5 - 90]
    45
    [22.5 - 90]
    LONGEVITY_SAP180
    [90 - 360]
    180
    [90 - 360]
    480
    [240 - 960]
    4801
    [2400.5 - 9602]
    180
    [90 - 360]
    240
    [120 - 480]
    K_LATOSA_MAX = K_LATOSA_MIN = K_LATOSA

  10. Observation error (R) is set as RMSD for simulations time-series vs the mean between min/max values of range of variation.
  11. Parameter error (B) is set as 15% of the parameter range of variation.
  12. Cost-function is calculated with the standard formula, but the obs-sim difference is calculated as deviation from the prescribed range of variation and set to 0 for the time-series part falling inside the range.
  13. Minimization method: Genetic Algorithm (population = 30, number of iterations = 15).
  14. Computing platform: LSCE obelix with 15 processors.