A Traceability Framework to Facilitate
Evaluation and Improvement of
Global Carbon Cycle Models
Yiqi Luo, Jianyang Xia, Oleksandra Hararuk, Rashid Rafique
University of Oklahoma, OK, USA
Ying-ping Wang
CSIRO Marine and Atmospheric Research, Australia
yluo@ou.edu
http://ecolab.ou.edu
19th CESM conference, Breckenridge, June 18 2014
Challenge
Biome-BGC
CASA
CASA’
SDGVM
1980
Roth-C
1990
CENTURY
Forest BGC
TEM
NCAR LSM
ORCHIDEE
JULES
CoLM
LPJ
TECO
CLM4.5
DayCent
CLM3.5
2000
2010
TRIPLEX
DLEM
TRIFFID
O-CN
CLM4CN
CASACNP
Soil microbial dynamics
Soil depth
Priming effect
Wetland
Land use
Erosion
DOC
Fire
……
Low
Complexity
High
High
Tractability
Low
IPCC AR5
Chapter 6
Challenge to explain model
outputs
Why models behave so differently?
Forcings
Biome-BGC
CENTURY
DayCent
TRIPLEX
JULES
ORCHIDEE
Roth-C
CASA
CLM3.5
CoLM
TECO
CLM4.5
DLEM
LPJ
NCAR LSM
TRIFFID
CASACNP
CLM4CN
Forest BGC
SDGVM
TEM
CASA’
O-CN
……
Model Outputs
Traceability to diagnose model differences
Forcings
Biome-BGC
Roth-C
DLEM
Forest BGC
CENTURY
DayCent
TRIPLEX
JULES
CASA
CoLM
CLM3.5 components
TECO
Traceable
shared
among
models
NCAR
LSM
CASACNP
TRIFFID
LPJ
SDGVM
TEM
CASA’
Model Outputs
e.g., Ecosystem C stock
O-CN
ORCHIDEE
CLM4.5
CLM4CN
……
Common properties of terrestrial C cycle
1. Photosynthesis is the primary
carbon influx pathway;
2. Compartmentalization of C
pools
3. C partitioning and transfer
among pools;
4. donor-pool dominated carbon
transfer;
5. first-order linear transfer from
the donor pool.
Luo & Weng 2011 TREE
Model representation of ecosystem carbon cycle
Canopy Photosynthesis
CO2
b1
dX (t )
= BU (t ) − ξ (t ) ACX (t )
dt
X ( 0) = X 0
Luo et al. 2003
Luo & Weng 2011 TREE
Variations among models:
b2
Leaf
(X1)
Root
(X2)
a41 a42
Metabolic
Litter (X4)
CO2
Woody
(X3)
a51 a52
Structural
Litter (X5)
a74
Fast SOM
(X7)
CO2
CO2
CWD
(X6)
a75
CO2
a76
a87
Slow SOM a85
a86
(X8)
(1) Pool number
(2) B, ξ, A, and C
b3
a98
a97
CO2
Passive
SOM
(X9)
Diagram of C process of CABLE.
Xia et al. 2012 GMD
The “traceability” of terrestrial carbon cycle is mathematically solved as:
dX (t )
= BU (t ) − ξ (t ) ACX (t )
dt
(1)
According to equation (1), when an ecosystem at steady state, the steady-state ecosystem carbon pool size
(i.e., ecosystem carbon storage capacity; Xss) is:
'
X ss = ξ −1 ( AC ) −1 BU ss = ξ −1τ E
U ss = τ EU ss
(2)
where τE' represents the baseline residence times of different carbon pools which are determined by the
partitioning and transfer coefficients in equation 1 as:
τ E' = ( AC ) −1 B
(3)
The actual residence time (τE) of an ecosystem in the equation 2 is modified from τE' by the environmental
scalar (ξ) as:
τ E = ξ −1τ E'
(4)
For litter and soil carbon pools, ξ usually is calculated from temperature ξT and water ξW as:
ξ = ξT ξW
(5)
Xia et al. 2013 Global Change Biol.
A traceability framework for terrestrial C cycle
Environmental space Climate forcing
Precipitation
Temperature
ξ
W
Environmental
scalars linking
environmental and
carbon spaces
Preset Residence times
Litter lignin fraction
τE’
ξT
Xss
Baseline
residence times
of carbon pools
ξ
τE
NPP (Uss)
Soil texture
Determinants of ecosystem
carbon influx and actual
residence time on carbon
storage capacity.
Xia et al. 2013 Global Change Biol.
Application 1: Inter-biome differences in CABLE
Differential determinants on carbon storage capacity among biomes
160
500
400
Residence time (Year)
Residence time (Year)
(a)
300
200
120
(b)
ENF
EBF
DNF
DBF
Shrub
C3G
C4G
Tundra
Barren
80
40
40
100
0
0
0
500
1000
1500
NPP (g C m-2 year-1)
2000
10
1
0
500
1000
20
30
1500
NPP (g C m-2 year-1)
Based on spin-up results from CABLE with 1990 forcings.
Xia et al. 2013 Global Change Biol.
Application 1: Inter-biome differences in CABLE
Differential determinants on carbon storage capacity among biomes
160
500
Long τE but
low NPP.
400
Residence time (Year)
Residence time (Year)
(a)
300
200
120
(b)
ENF
EBF
DNF
DBF
Shrub
C3G
C4G
Tundra
Barren
80
High NPP but
short τE.
40
40
100
0
0
0
500
1000
1500
NPP (g C m-2 year-1)
2000
10
1
0
500
1000
20
30
1500
NPP (g C m-2 year-1)
Based on spin-up results from CABLE with 1990 forcings.
Xia et al. 2013 Global Change Biol.
Application 1: Inter-biome differences in CABLE
Modification of environmental scalars on baseline carbon residence times
Xia et al. 2013 Global Change Biol.
Application 1: Inter-biome differences in CABLE
Modification of environmental scalars on baseline carbon
residence times
Tundra:
Moderate τE’ but
very long τE.
Xia et al. 2013 Global Change Biol.
Application 1: Inter-biome differences in CABLE
Temperature and water scalars link environmental space (air temperature
and precipitation) into the C space.
Cropland is excluded in this study. Input forcing in 1990.
Xia et al. 2013 Global Change Biol.
Application 1: Inter-biome differences in CABLE
Temperature and water scalars link environmental space (air temperature
and precipitation) into the C space.
•
•
In CABLE model, the differences in environmental scalars among biomes are more determined
by the temperature scalar.
The environmental limitation on τE’ is largest in Tundra and needleleaf forests.
Xia et al. 2013 Global Change Biol.
Application 2: Inter-model difference between CABLE and CLM-CASA’
Canopy Photosynthesis
CO2
Leaf
(X1)
Root
(X2)
Metabolic
Litter (X4)
CO2
Woody
(X3)
Structural
Litter (X5)
Fast SOM
(X7)
CWD
(X6)
CO2
CO2
CO2
Slow SOM
(X8)
CO2
Passive SOM
(X9)
CABLE
CLM-CASA’
Rafique et al. In Prep.
Application 2: Inter-model difference between CABLE and CLM-CASA’
Carbon storage capacity is differently determined by τE and NPP
between CABLE and CLM-CASA’
Residence time (year)
160
120
ENF
EBF
DNF
DBF
Shrub
C3G
C4G
Tundra
CABLE
CLM-CASA
80
Global mean
40
40
30
10
1
20
0
0
500
1000
1500
2000
Ecosystem C influx (NPP; g C m-2 yr-1)
Rafique et al. In Prep.
Carbon storage differences between CABLE and CLM-CASA’
Similar
Climate forcing
Precipitation
Temperature
ξw
ξT
CLM3.5 > CABLE
Preset Residence times
Litter lignin fraction
CABLE > CLM3.5
ξ
NPP ( U ss )
CLM3.5 > CABLE
X ss CLM3.5 > CABLE
τE
CABLE > CLM3.5
Soil texture
τ E'
Application 2: Inter-model difference between CABLE and CLM-CASA’
Longer residence times in CABLE than CLM-CASA’
Actual residence time (year)
316
100
32
10
CABLE
CLM-CASA’
3
ENF
EBF
DNF
DBF
Shrub
C3G
C4G
Tundra
1
1
3
10
32
100
316
Baseline residence time (year)
Rafique et al. In Prep.
Application 2: Inter-model difference between CABLE and CLM-CASA’
Temperature and water scalars
1.0
Water scalar (ξW)
0.8
ξ=0.6
0.6
ξ=0.4
0.4
0.2
CABLE
CLM-CASA’
ξ=0.2
ENF
EBF
DNF
DBF
Shrub
C3G
C4G
Tundra
ξ=0.05
ξ=0.1
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Temperature scalar (ξT)
Rafique et al. In Prep.
Application 3: Adding nitrogen processes into CABLE
Canopy Photosynthesis
CO2
Leaf
(X1)
Root
(X2)
Metabolic
Litter (X4)
CO2
Woody
(X3)
Structural
Litter (X5)
Fast SOM
(X7)
CWD
(X6)
CO2
CO2
CO2
Slow SOM
(X8)
CO2
Passive SOM
(X9)
CABLE Carbon-Nitrogen-Phosphorus
Working group
Traceability of transient simulation
FACE model-data intercomparison project
Site name: PHAC
Location: West of Cheyenne, Wyoming
Veg: Grassland
Period: 2006-2012
CO2 treatment: 594 ppm
Site name: RHIN
Location: Harshaw Experimental
Forest, Wisconsin
Veg: Broadleaf Deciduous Forest
Period: 1998-2008
CO2 treatment: 543 ppm
Site name: NDFF
Location: Nevada Test Site, Mojave Desert
Veg: Desert
Period: 1997-2008
CO2 treatment: 507 ppm
Site name: KSCO
Location: Cape
Canaveral, Florida
Veg: Broadleaf
Evergreen Forest
Period: 1996-2006
CO2 treatment: 673 ppm
Summary
• The carbon cycle can be decomposed into a few
traceable components.
• Traceability of land models is important to
improve our understanding of the different
behaviors among models;
• The traceability framework is a diagnostics tool
with wide applications for global carbon cycle
modeling