Recent developments in
ORCHIDEE : toward 3rd generation ?
Philippe Peylin, Sebastiaan Luyssaert, James
Ryder, Yiying Chen, Tao Wang, Bertrand
Guenet, Nicolas Viovy, Marc Paucelle, Nicolas
Vuichard, Catherine Ottle, Patricia Cadule,
Philippe Ciais, Yue Chao, Xinfeng Wang,
Xuichen Wu, Dan Zhu, Gerhard Krinner,
Fabienne Maignan, Josefine Ghattas, Sushi
Peng, Natasha MacBean, Nathalie De Noblet,
Sarah Dantec Nedelec,
and all the ORCHIDEE Team
ORCHIDEE: 20-yr of development
History :
Global GCM:
W / E (SECHIBA)
(Laval
et al., 1981)
80s
Inclusion C
(ORCHIDEE)
(Ducoudré (Viovy et al., (Polcher
et al., 1998)
et al., 1993)
1997)
90s
Focus
on Physic..
(Krinner et al.,
2005)
2000
New dev
(biogeochemical)
…….
2009
ORCHIDEE: 20-yr of development
History :
Global GCM:
W / E (SECHIBA)
(Laval
et al., 1981)
Inclusion C
(ORCHIDEE)
(Ducoudré (Viovy et al., (Polcher
et al., 1998)
et al., 1993)
1997)
80s
90s
Focus
on Physic..
(Krinner et al.,
2005)
2000
New dev
(biogeochemical)
…….
2009
Energie budget
complex
feedbacks
Water
budget
Carbone / Nitrogen
budgets
Challenge:
maintain coherence &
describe feedback
between Physic and
Biogeochemistry
Overview
• Introduction: recent model developments
• A multi-layers soil hydrology scheme
• New snow scheme & High latitude processes
• Improved soil carbon decomposition
• A new multi-layers canopy energy scheme
• Towards the use of Plant functional traits
• Conclusions
NPP Rh
Land
NPP Rh F Harvest
Land
Recent improvements of ORCHIDEE
- Age related decline in NPP
- Age related limitation of LAI
- Age related allocation between stem and roots
- Branch mortality
- Coarse woody litter compartment
- Individual growth of trees
- Generic management
Forest management
Nitrogen
Cycle
(soon P)
Fires
Temperate
Crops
Artic
Processes
(permafrost)
grassland
- Generalization of PFT concept
(unlimited, currently 13)
- Analytical soil C spin-up
Multi-layer soil hydrology
• Why a “new” physically-based scheme
(vs old double-bucket scheme) ?
– Better represent Infiltration vs Runoff processes
– Plant W uptake:
• Different plants have different root profiles
• Compute hydraulic lift : from soil to leaf W potential
– SOM decomposition is a function of W, T,..
2 buckets scheme
11 layers with diffusion
P
P
Rs
R
D
Soil moisture evolution
 Comparison with SMOS data
Normalised soil moisture
Guadalquivir area:
lon: -6:-4, lat: 37.2:38.
3 days average to reduce
instrument noise
2010
2011
OBS (smos)
0-5 cm ORC
5-20 cm ORC
20 – 200 cm ORC

The general annual cycle is rather well captured.

The amplitude of the response to the rainfall events is more
spiked in SMOS than the 0-5cm layer in ORCHIDEE.
Multi-layer snow scheme
• 3 layers scheme to improve:
– Snow dynamic (spring)
– Snow – vegetation interactions (Shrub, grass, ..)
Evaluation on new snow scheme
Model snow depth (m)
Daily snow depth (density, SWE) for Northern Eurasia,
165 stations HSDSD (1979-1992)
New ORC
Old ORC
Obs snow depth (m)
Corr: 0.78  0.83
RMSE: 0.12  0.10 m
MBE: -0.05  0
Wang et al., JGR, 2013
Evaluation at high latitudes: soil moisture
Normalized Surface soil moisture
 Obs : ESA – CCI summer soil moisture (1978-2010); 0.25°
 11-layer hydrology + soil freezing + new snow description
Obs
Old ORC
Dantec-Nédélec et al. (2014)
New ORC
Obs
Old ORC
New ORC
New soil carbon decomposition scheme
• Motivations
– Current model (century) simple
missing processes (i.e. priming)
– Effect of temperature and moisture
still relatively simple
New soil carbon decomposition scheme
• Motivations
– Current model (century) simple
missing processes (i.e. priming)
– Effect of temperature and moisture
still relatively simple
?
A NEW SOIL CARBON MODULE
• Discretized soil carbon (11 layers) + new pools introduced (DOC)
• New decomposition scheme (priming): ¶SOC
= I - k ´ SOC ´ (1 - e
) ´q ´t
¶t
-c ´FOC
SOC
0 Depth (mm)
1.96
5.87
13.69
29.33
CO2
60.61
Runoff x DOC
For each
layer
123
248
498
Between
layers
999
2000
Drainage x DOC
Decomposition
Heterotrophic respiration
(De) sorption
Advection/liquid
phase transport
Diffusion/ Bioturbation
400
300
As
As
200
N-Am
Eu
S-Am
N-Am
Af
S-Am
Af
Eu
100
Modelled stock (Pg-C)
1500
Slope = 0.76
NSD = 0.09
Pearson's corr. = 0.95
New ORC
500
1000
Global
Slope = 0.95
NSD = 0.07
Pearson's corr. = 0.95
Old ORC
Au
100
0
Soil Carbon stock (Gt C)
2000
Impact of new scheme on total SOM
200
300
HWSD stock (Pg-C)
400
Preliminary evaluation
at site level
10 cm
Obs
DOC (mg/l)
Ex: Braaschat site
(Belgium forest)
35 cm
Model Ref
75 cm
Model sensitivity tests
(varying 5 parameters)
A new multi-layer energy balance scheme
• Why a multi-layer energy canopy scheme ?
Ecosystem structure
Model
representation
Big leaf model
=
 DGVM still poorly represent site-level heat fluxes
 Canopy space and Trunk crown have different behaviours
 Under-storey vs over-storey representation ?
 Link to ecosystem services: forest canopy climate
Multi-layer scheme implementation
• Free number of layers
• E / W / C exchange
at each level
• Turbulance mixing
within air canopy
• Light penetration
following Pgap model
Implementation constraints :
• Coupling with plant growth / harvesting module (variable plant height)
• Implicit coupling with Atmospheric model (30’ step)
• Parametrisation of intra-canopy turbulence
Several sites to evaluate the model…
 Availability of vertical profiles for Temp, Wind, Rh is crucial
0:00 – 6:00
Temperature
profile at
Tumbarumba site
6:00 – 12:00
-8.
0.
0.
3.5
Observations
18:00 – 24:00
12:00 – 18:00
Model
0.
4. -6.
Daily temperature
0.
Multi-Layer Latent Heat Flux
Obs
FI-Hyy
22:00
FR-LBr
14:00
NL-Loo
DE-Bay
*10
CA-Oas
AUTum
DE-Hai
BE-Vie
Monthly average flux (months of each site are random selected)
Multi-Layer Sensible Heat Flux
Obs
FI-Hyy
22:00
FR-LBr
14:00
NL-Loo
DE-Bay
*10
CA-Oas
AUTum
DE-Hai
BE-Vie
Monthly average flux (months of each site are random selected)
performance single- vs multi- layer scheme
Taylor skill score (fraction between 0 - 1)
Le Bray
Tumba
autumn
autumn
summer
summer
spring
spring
winter
winter
Hyytia
ROCCAR
autumn
autumn
summer
summer
spring
spring
winter
winter
0
1
General improvement of Multi-layer
But not systematic…
H
Rn LE G
single-layer score
multi-layer score
What can we learn from Data Assimilation ?
Optimization of
ORC parameters
 FluxNet data
(70 sites)
Parameter errors can nearly as
large as Structural errors
Large param. error correlations
Highlight model deficiencies
Latent Heat (W/m2)
 ≈ 25 optimized
parameters / PFT
Puechabon Fluxnet site (2004)
Bacour et al. sub
Inclusion of plant trait plasticity
• Motivations:
– Plant traits likely evolve with time & climate change
– Relations between plant traits exists
Fixed traits
associated to PFT
Trait 1 Trait n
Trait 2
PFT 1
Trait 1
Trait 2
Set oftraits
Trait
Trait 1 Trait
n 1 Trait n
Trait 2 Trait 2
Trait n
Set of traits
Set of traits
PFT N
Trait 1
Trait 2
Trait n
No more PFTs but
set of variables traits
Different approaches followed at LSCE
1) Use empirical relationship between traits and
environnement:
Ex: Van Bodegom et al. 2014
2) Use optimization hypothesis
Ex: coordination of photosynthesis
3) Use correlation between traits
Ex: the leaf economic spectrum
 The 3 approaches can be combined
Inclusion of plant trait plasticity
Multi-linear relationships between VCMAX / JMAX / SLA
AND climate, following « TRY data base »
Temperature: Mean annual, Warmest month, Coldest month, Tmax – Tmin, ..
Precipitation: Mean annual, Relative humidity, driest,..
Radiation:
SW downward
Change in VCMAX between 2100 and 2010
Paucelle et al.,
in prep
Conclusion
• Soil physic (W, C, E), snow are critical..
• Escaping from big leaf can be part of 3rd gen. DGVM
• Parametrisation is crucial (may depend on scale..)
better use of data-base, plant traits,…
• Biogeochemistry & Biophysics should be developed
together..
• Difficult to maintain coherence between various
component!
Thanks for your attention..
ORCHIDEE
ORCHIDEE
yesterday
tomorrow
(many branches)
All developments
into main Trunk
.?.