Land Cover CCI

Land Cover CCI
CCI Phase 1 results
Climate research perspectives
Multi-level validation strategy
CCI-LC products
3 LC state products for the 2000, 2005 and 2010 epochs
Fate of anthropogenic CO2
emissions (2003-2012 average)
Dynamic global
vegetation model
Land component of Earth System Model
Simulates the Energy, Water and Carbon balance
LSCE – ORCHIDEE model
•  Surface description: a tile approach
•  A mosaïc of vegetation
Land cover map
•  10 to 15 different PFTs
MOHC – Joint UK Land
Environment Simulator (JULES)
•  JULES uses a tiled model of sub-grid heterogeneity
•  9 Land surface types: BL tree, NL tree, C3 grass, C4
grass, Shrub, Urban, Water, Ice, Bare Soil
MPI-M – JSBACH model
•  MPI-Earth Surface Model (ESM), with JSBACH as
land surface component which can be run independently (offline
simulations)
•  13 different PFTs
Earth system
models comparison
1.  Does the use of CCI-LC maps improve model-data comparisons?
2.  Does the use of CCI-LC maps reduce model-model differences?
3.  Does consistency in land cover improve model comparisons?
Modeling Team Offline Original LC LC_CCI LC Online Dynamic vegeta9on Original LC LC_CCI LC LSCE-­‐
ORCHIDEE V V V V V MOHC-­‐JULES V V V V V MPI-­‐JSBACH V
V
✗ V V -­‐ 
-­‐ 
-­‐ 
Same climate forcing for offline simulaAons (WATCH WFDEI) Same Epoch (2010) for LC_CCI inputs – but modeling groups define climate zones… Treatment of land use fluxes and disturbance up to the modeling teams Current land cover map • Outdated land cover inputs •  IGBP land cover (Belward et al. 1999) and, •  Olson vegetaAon map with 96 classes (1983) ESA LC_CCI PFT datasets Land Cover -­‐> PFT Conversion Tool (BEAM) LC-CCI to improve initial
model conditions
Köppen-Geiger
climate zones
Land cover
Phenology &
Physiognomy
Plant Functional Types
PFT comparison
- Bare fraction
JSBACH
JULES
Possibly too high in
high latitudes due to
85% allocation from
sparse vegetation
categories
ORCHIDEE
PFT comparison
- C3 / crops fractions
JSBACH
(crops)
JULES
(All C3)
Decrease in high
latitudes due to
replacement by
bare soil/barren
ORCHIDEE
(natural C3 fraction)
PFT comparison
- BoBS fraction
ORCHIDEE
Notable increase in E Europe at expense of evergreens
Fractional difference (-)
By-model experiment
Carbon, energy and water budget
•  LAI
•  Albedo
•  Gross Primary Production
•  Runoff
(GPP)
•  Net Primary Production (NPP)
•  Total precipitation changes
•  Validation of Dynamic Global
•  Radiation & moisture fluxes
Vegetation Model (in
•  Heat sensible flux
comparison with IGBP map)
•  Latent sensible flux
•  Surface temperature
Evapotranspiration
•  Heterotrophic respiration
LAI comparison
ORCHIDEE: LAI comparison between Olson and CCI-LC
Mean difference in LAI (Olson – CCI_LC)
Red: higher LAI with Olson-derived simulations
Blue: higher LAI with CCI_LC-derived simulations
LAI comparison
ORCHIDEE: LAI comparison between Olson and CCI-LC
•  Comparison to GIMMS3gSurface parameters
• 
AVHRR GIMMS3g LAI dataset (Zhu et al. 2013 Remote Sensing)
• 
RMSE and bias between simulations of LAI and GIMMS3g
data with Olson and CCI-LC derived PFT maps
2
0.8
0.6
1
Bias
RMSE
1.5
0.5
0.4
0.2
0
0
Olson
CCI
Olson
CCI
Carbon budget - GPP
JULES: Increase in global GPP in both online and offline simulations
Offline
Control
LC_CCI 2010
GPP (Gt C yr-1)
152.5
166.7
Global Mean Annual GPP
125
100
Source
Estimates of global annual GPP:
•  123 ± 8 Gt C yr-1 (Beer et al.
2010)
•  119.4 ± 5.9 Gt C yr-1 (Jung
et al. 2011)
•  120 Pg C yr−1 (Denman et
al. 2007)
•  109.29 Pg C yr−1 (Zhao et
al. 2005) for 2001-2003
period
•  Discrepancy between offline and
online GPP likely due to
variations in JULES
GtC yr−1
HadGEM2A: Control
HadGEM2A: LC_CCI 2000
75
HadGEM2A: LC_CCI 2005
HadGEM2A: LC_CCI 2010
MODIS GPP MOD17A3
Jung et al
50
25
0
1980
1985
1990
1995
Year
2000
2005
Carbon budget - GPP
JULES: Spatial distribution of differences
Gross Primary Production ( gC m−2 yr−1 )
LC_CCI 2010 − Control
50°N
0°
50°S
0°
100°W
−2000
−1500
−1000
−500
0
100°E
500
1000
1500
2000
Carbon budget - GPP
Gross Primary Production ( gC m−2 yr−1 )
JULES: Comparison to estimates
from Jung et al. (2011)
Model − Observations
Control
50°N
•  LC_CCI 2010 shows
reduction in biases in:
•  West Africa
•  Southern Africa
•  Amazon
•  India
•  South East Asia
0°
50°S
LC_CC:2010
0°
100°W
−3000
−2000
−1000
0
100°E
1000
2000
3000
Carbon budget - NPP
ORCHIDEE: NPP comparison (CCI-LC – Olson)
•  Large increase in NPP over croplands
•  Decrease in high latitudes due to decrease in grass cover
KgC/m2/yr
Carbon budget - Evaluation
using atm. CO2 concentration
ORCHIDEE
LMDz transport: comparison to GlobalView data
! improvement of the CO2 seasonal cycle with
reduced amplitude
Radiation & moisture
fluxes
Tropical
Bare
C3
C4
GPP
0.20
0.15
0.10
0.05
0.00
cVeg
0
−1
−2
−3
JULES: West Africa
Tropical
Increase in C4 Grass, replacing BL
tree or C3 grass in old maps
C3
value
value
-0.40
J
F M A M
J
J
A
S O N D
J
F M A M
J
J
A
Bare
-0.10
0.00
-0.04
-0.08
0.00
0.03
0.00
0.00
-0.12
0.00
J
F M A M
J
J
A
S O N D
J
F M A M
Ice
J
J
Month
A
S O N D
J
F M A M
J
J
A
S O N D
JJ FF M
M AA M
M JJ JJ AA SS O
O NN DD
J
C4, 66% (10)
C4, 71% (11)
F M A M
J
J
Month
A
S O N D
J
F M A M
J
J
r20mm
Water
r20mm
Urban
0
S O N −2D
4
2
0
−2
−4
1.0
0.5
0.0
−0.5
−1.0
C4, 45% (14)
Pr90pctl
Month
Shrub
2
Pr90pctl
−2
4
2
0
−2
−4
1.0
0.5
0.0
−0.5
−1.0
4
C4, 39% (8)
Tmax90pct
0.00
Leads to:
Increase in GPP
Reduction in veg. C
Increase in extreme
max temp in Sept
(10)
•  Extreme
precipitation earlier
in season (11)
C4, 32% (21)
LH
-0.21
• 
• 
• 
• 
SH
11
0.66
0.71
0
−25
−50
−75
Bare, 50% (9)
Alb
-0.17
SM
0.00
ET
-0.28
cVeg
10
40
20
0
−20
r20mm
Tmax90pct
C4
C4, 66% (10)
C3, 63% (13)
Pr90pctl
LH
C3
0.00
Tmax90pct
SH
NL
GPP
0.08
Alb
BL
C4, 45% (14)
C4, 71% (11)
PFT, %increase (point ID)
LH
ID
00
−25−2
−50−4
−751.0
40.5
0.0
2
−0.5
−1.0
0
C4, 32% (21)
SH
SM
0−2
4
−20 2
C3, 63% (13)
C4, 39% (8)
0.04
40 2
20 0
PFT, %increase (point ID)
Bare, 50% (9)
100
75
50
25
0
10
5
0
−5
−10
C4 grass
C4
10
ET
C3 grass
0
−1
−2
−3
Alb
BL Tree
Tropical
C3
0.20
0.15
0.10
0.05
0.00
SM
cVeg
0
1040
520
00
−5
−20
−10
0
−25
0.08
−50
0.04
−75
0.00 4
11
Bare
C4
value
Bare
ET
GPP
100
75
0.2050
0.1525
0.10 0
0.05
0.0010
5
00
−1−5
−10
−2
−3
0.08
100
75
0.04
50
0.00
25
A
S O N D
Radiation & moisture
fluxes
JULES Maximum temperature
Energy budget Visible albedo
Change in albedo (-)
ORCHIDEE (CCI-LC – Olson)
•  Increase in high lat. albedo with increased bare soil
•  Decrease in W Canada with increasing tree cover
•  Decrease in some cropland areas
By catchment experiment
Surface albedo
JSBACH
• 
Over high- and mid-latitude catchments: CCI-LC albedo higher than reference
• 
Over arid and semi-arid regions: improvement with CCI-LC albedo
By catchment experiment
WFDEI 2m temperature
JSBACH
•  Effect of albedo difference visible in estimated 2m t° (↑
albedo " ↓t°)
•  General improvement over most catchments
By catchment experiment
Uncertainty due to epoch
JSBACH
• 
Notable difference in CCI-LC maps only in Amazon and Congo (large-scale
deforestation " increased albedo) ! need for simulation with LC change
• 
For other regions, using the 2010 epoch is a reasonable assumption
Models inter-comparison
Albedo
Surface Albedo
0.4
MODIS
0.35
GlobAlbedo
0.3
WFDEI-­‐REF
0.25
WFDEI-­‐CCI
0.2
ECHAM6-­‐REF
0.15
ECHAM6-­‐CCI
Over Giorgi regions
AMZ = Amazonia
WAF = West Africa
SAF = South Africa
NAS = North Asia
SAS = South Asia
NAU = North Australia
JULES-­‐REF
0.1
JULES-­‐CCI
0.05
HadGEM2A-­‐REF
0
• 
HadGEM2A-­‐CCI
AMZ
WAF
SAF
NAS
SAS
NAU
Surface Albedo
Stronger sensitivity of
JSBACH than JULES to CCILC maps
• 
Overestimation over the
Northern Asia region (w.r.t.
GlobaAlbedo)
• 
Over Amazon and West
African, JSBACH closer to
GlobAlbedo but farther away
from MODIS
0.3
0.25
MODIS
0.2
GlobAlbedo
ORCHIDEE-­‐Olson
0.15
ORCHIDEE-­‐CCI
ORCHIDEEonl-­‐Olson
0.1
ORCHIDEEonl-­‐CCI
0.05
ORCHIDEE-­‐DGVM
0
AMZ
WAF
SAF
NAS
SAS
NAU
• 
Significant improvement of
ORCHIDEE with CCI-LC maps
Models inter-comparison
Evapotranspiration
Deviation from LandFlux Diag ET mean
60.0%
LF Min
LF Max
WFDEI-­‐REF
WFDEI-­‐CCI
ECHAM6-­‐REF
ECHAM6-­‐CCI
JULES-­‐REF
JULES-­‐CCI
HadGEM2A-­‐REF
HadGEM2A-­‐CCI
40.0%
20.0%
0.0%
-­‐20.0%
-­‐40.0%
Over Giorgi regions
AMZ = Amazonia
WAF = West Africa
SAF = South Africa
NAS = North Asia
SAS = South Asia
NAU = North Australia
• 
-­‐60.0%
AMZ
WAF
SAF
NAS
SAS
NAU
Direction of change when
using CCI-LC data:
• 
Deviation from LandFlux Diag ET mean
60.0%
• 
LF Min
40.0%
LF Max
20.0%
ORCHIDEE-­‐Olson
0.0%
ORCHIDEE-­‐CCI
ORCHIDEEonl-­‐Olson
-­‐20.0%
ORCHIDEEonl-­‐CCI
-­‐40.0%
ORCHIDEEoff-­‐DGVM
-­‐60.0%
-­‐80.0%
AMZ
WAF
SAF
NAS
SAS
NAU
• 
Is consistent between
offline & online
simulation for JULES
and JSBACH
Opposite direction for
ORCHIDEE over AMZ,
WAF and SAS
Improvement over high
latitude for 3 models
Summary
•  Significant differences are observed in the land surface description
when using CCI-LC maps w.r.t. older reference maps
• 
• 
• 
Boreal region with the continuum bare – sparse vegetation –
grassland
– 
JULES & ORCHIDEE: vegetation to bare soil è increased albedo associated with
decrease in extreme temperatures
– 
JSBACH: more forest & less bare soil è increases of LAI, biomass and surface
albedo è decrease of surface temperature and sensible heat flux
JSBAC model: less forest and more bare soil in West Africa, Sahel,
Congo
– 
LAI and biomass decrease, albedo increase
– 
reduced NPP, GPP and evapotranspiration
– 
soil moisture and runoff increases.
Globally, large reduction in C3 grass cover and increase in C4
grass cover ! generally increases GPP, but decreases cVeg,
associated with improvements in max temp bias
Summary
•  All models show high sensitivity to land cover as an initial
condition for both stocks and fluxes
• 
Mainly changes in high latitudes and tropics
• 
Improvement in high latitude NPP, in GPP online
• 
Improvement in terrestrial carbon sink estimate and seasonality
• 
For Amazon and Congo, albedo is increasing from the earliest 2000 Epoch to the
latest 2010 Epoch ! indication of large-scale deforestation of tropical rainforest
• 
Mixed response of biomass, depending on region BUT global reduction in positive
bias
• 
Improvement to atmospheric data at high latitudes
• 
Improved representation of 2m air temperature and surface albedo in some regions
• 
CCI-LC constrains the problem of where to improve existing model
processes and parameters
• 
Implementation of fully consistent changes of land cover (not only
input data but all associated parameters – e.g. PFTs distributions)
would lead to a more larger impact of CCI data
2 papers to come
Paper 1 - “Land cover uncertainty in land carbon cycle processes”
… Making sense of multiple benchmarks …
Paper 2 - “A plant functional type classification for Earth System
Models from the European Space Agency Essential Climate Variables
Program”
… Toward regular PFT classification from ESA-CCI …
For the future
•  Further discussion/development of cross-walking procedure
(LCCS class into PFTs)
•  In Phase 1: assess the impact of LC uncertainty with all models
with their own reference map
Next step: Same experiment with all models with the same
reference map
•  Model optimization for new CCI-LC maps
•  More links between C and water/energy expertise in modelling
groups
•  Create a LC change scenario (yearly) and assess its impact on the
C, water and energy fluxes (especially on on-going deforestation
and associated fluxes)
•  Use of WB product and/or LC condition products in models