Il CMCC e la sfida della Scienza del Clima

1
Vegetation-Land Surface modelling at CMCC-ANS
A. Alessandri
CMCC Workshop 10 June 2009 Iberotel, Marina di
Ugento - Lecce
2
Outline
Biosphere initialization-spin up
Boreal Summer climate sensitivity to improved
land-vegetation
Relationship between vegetation and precipitation
variability
3
SILVA Surface Interactive Land VegetAtion
(Alessandri, 2006. PhD)

• SILVA (Surface Interactive Land VegetAtion
  Alessandri, 2006) is a new Land Surface Model
  developed at INGV-CMCC
• SILVA includes a tiled computation of surface
  fluxes following the SECHIBA
• type approach (Docoudré et al., 1993).
• The core parameterizations coming from the VEGAS
  model (Zeng et al., 2004)
• has been included in SILVA in order to
  represent the vegetation and Carbon
• dynamics.

Grid and sub-grid model discretization
4
SILVA Surface Interactive Land
VegetAtion (2)
Surface fluxes
Implicit Coupling with a Neumann closure at the
interface
AtmosphereECHAM

• Hydrology module with 2 soil layers
• 7 layers for soil diffusion
• Computes implicitly fluxes and solves surface
  energy and water balance
• Surface parameters (albedo, roughness and
  surface conductances) computed interactively.
• The model can be run with fixed observed
  vegetation distribution or with vegetation-carbon
  dynamics activated

Lower boundary Atmospheric variables and
radiative fluxes
SILVA
Land model
Soil Temperature humidity, ...
LAI, Vegetation cover height, stomatal
conductance, ...
Natural Vegetation and Carbon model

• 4 PFTs Broadleaf and needleleaf trees and
  cold/warm grass
• 3 vegetation carbon pools leaf, root and wood
• 3 carbon pools in soils slow, intermediate and
  fast degradable

Alessandri, 2006. Ph.D
5
Land Surface and Biosphere modelling
SILVA Vegetation and Carbon dynamics
(3)

• Implements 4 Natural PFTs Broadleaf and
  needleleaf trees and cold and warm grasses (C3
  and C4 photosynthetic pathways are distinguished)

• Photosynthesis rate is computed through a Jarvis
  type empirical approach
• (Jarvis, 1976 Lhomme, 2001 Zeng et al.
  2004)

• PFTs seasonal and interannual variability
  simulated dinamically as the
• balance between growth and
  respiration/turnover

• Competition between PFTs is determined by
  climatic constraints and resource allocation
  strategies

• Natural Fire disturbance parametrization (very
  simple!) included

• Temperature and moisture dependent soil carbon
  decomposition. Fast, Intermediate and
  Slow carbon pools considered.

6
Land Surface and Biosphere modelling
SILVA Carbon cycle representation (4)
Natural Fire
Autotrophic Respiration
Limiting Factors
Light Temperature Soil Water Atmospheric Carbon
Photosynthesis
Leaf Carbon
Li
Carbon allocation
Wood Carbon
Wi
Root Carbon
Ri
Heterotrophic Respiration
Carbon turnover into Fast Soil pool
Soil Oxygen content
Medium soil Carbon
Slow soil Carbon
Fast soil Carbon
Topography
Sf
Sm
Ss
Soil Carbon Erosion
Carbon pools
CO2 flux
Solid carbon flux
7
Land-Atmosphere coupled simulations
8
Models setup

• Atmosphere Echam4.6
• Developed at MPI in Hamburg
• Spectral model
• Horizontal resolution T30, T42, T106
• Vertical resolution 19 sigma layers
• (Roeckner et al., 1996)

Implicit Coupling with a Neumann closure at the
interface
9
Experiments performed
-Prescribed Sea Surface Temperature
(SST) -Greenhouse gases and aerosols fixed. -No
CO2 feedback allowed.
-Echam4-SILVA
Vegetation and carbon dynamics included
Observed SST and Sea Ice for the period
1982-1998.
Spin-up simulation with Climatological SST and
Sea Ice
Prescribed vegetation (IGBP map) - No dynamics
included
observed sea surface temperature (SST) and Sea
Ice 1955-1999
-Echam4 uncoupled
observed sea surface temperature (SST) and Sea
Ice 1955-1999
10
Initialization of the Vegetation and Soil
Carbon pools (Climatological SSTs for
present day climate)
11
Interactive Land Biosphere experiment Vegetati
on initialization
X 5
Tree growth has been accelerated by 5 times
LAI (shading) Tree height (vectors)
Observed NDVIx10 pattern
PRESENT DAY CLIMATE RADIATIVE AND SST
CLIMATOLOGICAL FORCING
12
Interactive Biosphere experiment
Initialization of Carbon-vegetation
Accelerated spin up gt10 times for the Soil
carbon pools 2 times for root carbon 5 times for
wood carbon
Alessandri, PhD
13
Climate sensitivity to Land Surface-Vegetation
Comparison with control simulation

• Echam4-SILVA vs Echam4 (Amip-type experiments
  (1956-1999 T30 T106)
• Simulation of Boreal summer surface climate
• Vegetation cover obtained from IGBP map-No
  vegetation dynamics included.

Alessandri et al., 2007. J. Climate
14
Echam4-SILVA vs Echam4 Comparison with
control simulation (1)
JJA Climatology (1956-1999) T30 RUNS
Echam4
Echam4-SILVA
(Echam4-SILVA) Echam4 (t-test 5)
NCEP-II Reanalysis
Alessandri et al., 2007. J. Climate
15
Echam4-SILVA vs Echam4 Comparison
with control simulation (2)
JJA Climatology (1956-1999) T30 RUNS
Echam4
Echam4-SILVA
CMAP
(Echam4-SILVA) Echam4 (t-test 5)
Alessandri et al., 2007. J. Climate
16

• Present day climate simulation (1982-1998
  prescribed
• observed SSTs)
• An Observational Evidence for Land Atmoshere
  interaction through the coupled manifold (Navarra
  and Tribbia, 2006)
• (Alessandri and Navarra 2008)
• Model vs Observations Relationship between
  vegetation and precipitation variability
  (Alessandri et al.
  In preparation)

17
Observational Evidence for Land
Vegetation-Atmosphere Interaction
Alessandri and Navarra, GRL, 2008 Nature
Geoscience -Research highlight- March 2008
Through the coupled manifold technique (Navarra
and Tribbia, 2005) We assessed the reciprocal
forcing between rainfall and vegetation (NDVI) on
land areas at the 1 significance
level. seasonal mean interannual anomalies
(period 1982-1998). 19 of the vegetation
variability has been assessed to be forced by
rainfall. We estimated that 12 of the rainfall
variability is forced by vegetation.
Fractions of total variance
Montecarlo method for significance
18
Observational Evidence for Land Atmosphere
interaction (2)
Standardized PCs vs. Nino3
Our analysis reveals that the dominant component
of the vegetation-forced rainfall variability is
a delayed response to ENSO cycles.
(22)
Lag correlations
Alessandri and Navarra, 2008 (GRL)
19
Vegetation Forcing on rainfall
Model vs Observations

The Coupled Manifold of seasonal mean anomalies
LAI/NDVI vs Precipitation (1982-1998)
Z Precipitation S LAI/NDVI
Z Zfor 0.13 Zfree 0.88
Z Zfor 0.12 Zfree 0.88
S Sfor 0.18 Sfree 0.82
S Sfor 0.19 Sfree 0.81
Precfor forced variance (frac) 0.12
Precfor forced variance (frac) 0.12
significance at the 1 level
Alessandri et al., In preparation
20
ENSO-related Vegetation Forced Precipitation
Model vs Observations
(lag-correlations) (2)
Symbolsseasons Lines3-points running means
LAG (Seasons)
1 YEAR
X Marks stands for significance at the 1 level
21
Discussion and Future Plans (1)
A new Land Surface Model (SILVA) has been
developed

• SILVA considerably impacts and improves the AGCM
  simulated surface Climate

• SILVA simulates realistic vegetation cover and
  Land Carbon pools for present time climate
  conditions.

• The coupled manifold technique has evidenced a
  relationship between vegetation and rainfall
  interannual variability
• In most regions, the observed relationship is
  quite well captured by the AGCM once coupled to
  SILVA.

22
Discussion and Future Plans (2)

• Ongoing analysis SILVA has been included in the
  INGV-CMCC ESM (ELISA TALK) which has been run
  for historical and scenario simulations (EU
  project ENSEMBLES)

So far SILVA can handle only NATURAL VEGETATION
anthrophic land use and land management
representation needed

• Land cover changes (and related carbon pools) for
  historical and future projections

• Water management (i.e Irrigation, Energy
  production, Urban use)

23
END
24
Land Surface and Biosphere modelling (3)
SILVA Water and Heat storages and
fluxes

• Contains an hydrology module with 2 soil layers

• Soil thermodinamics module (7 layers soil
  diffusion)

Snow melt runoff
Snow melt runoff
Snow melt runoff
Fouriers law for heat conduction
hs
ht
Rain
Evap
Snow melt
25
Sensitivity to improved land surface-vegetation
in ECHAM4

• Simulation of Boreal summer surface climate
• Amip-type experiments (1956-1999 T30 T106)
• with both Echam4 and Echam4-SILVA
• No vegetation dynamics included

Vegetation distribution obtained from IGBP map
Alessandri et al., 2007. J. Climate
26
Experiments performed
Fixed observed vegetation experiments
Experiment with vegetation-carbon dynamics
included
Echam4-SILVA forced with prescribed vegetation
distribution (IGBP map) No vegetation dynamics
observed sea surface temperature (SST) and Sea Ice
Echam4-SILVA at T42 resolution with the
Vegetation and carbon dynamics included. Spin-up
simulation with Climatological SST and Sea Ice
observed SST and Sea Ice for the period
1982-1998. Greenhouse gases and aerosols fixed.
No CO2 feedback allowed.

• Two different horizontal resolution
• 1956-1999 at T30 horizontal resolution(3 members
  ensemble)
• 1979-1999 at T106 horizontal resolution

Compared with control simulations performed with
Echam4 in stand alone mode
27
The CMCC-INGV ESM
Land SurfaceSILVA (T31)
AtmosphereECHAM5 (T31)
CouplerOASIS
OceanOPA (ORCA2)
Sea IceLIM (ORCA2)
MarineBiogeochemistryPELAGOS (ORCA2)
Developers PG Fogli, M Vichi, A Alessandri, E
Manzini, L Patara
28
Prescribed vegetation experiment regional
impacts (1)
Relative contribution to the increased rainfall
in four selected areas (Land only) P area
averaged JJA precipitation E area averaged
JJA Evapotranspiration -DivQ area averaged JJA
total atmospheric column moisture convergence
Alessandri et al., 2007. J. Climate
29
Prescribed vegetation experiment
Comparison with control simulation ISM (2)
The IMI index is defined as the 850 mb zonal
wind averaged over (5 ºN-15 ºN 40 ºE-80 ºE) minus
that averaged over (20 ºN-30 ºN, 60 ºE-90 ºE)
Alessandri et al., 2007. J. Climate
30
Discussion and Future Plans (2)

• So far SILVA can handle only NATURAL VEGETATION
• How to include land use change (covers and
  anthropic vegetation carbon pools or directly
  emission fluxes) for future scenario studies?

CO2
CO2
Anthropic
Agricolture Extensive grass
Deforestation Forest regrowth Forest Timber
Natural vegetation
CO2
Natural
Bare soil
31
Needed land cover and carbon pools/fluxes Can
the grey fields be produced as boundary
conditions for ESM scenario simulations? (3)
Anthropic
Natural
32
Fully interactive vegetation experiments forced
variance
MODEL (LAI vs Rainfall)
Observed (NDVI vs Rainfall)
Vegetation variance forced by rainfall
Rainfall variance forced by vegetation
significance at the 1 level
Alessandri et al., In preparation
33
Forced precipitation pattern
Alessandri et al., In preparation
34

35
The Coupled Manifold tecnique (4) It is a new
method to detect the portion of variability
connected between two climatic fields following
the Procrustes formulation, that is a method to
analyze covariation between fields.
Mathematical formulation
Z f(S) Z AS the relation between the
atmospheric fields Z and S is some linear
combination of S
minZ-ASF² A(ZS)(SS) -1
Z Zfor Zfree
Using the operator A the field Z can be separated
into 2 parts
Zfor AS Forced Manifold is the subspace where
variation of Z are connected to variations of S
Zfree Z-AS Free Manifold is the subspace where
variations of Z are indipendent to S
S Sfor Sfree
Similarly
Sfree S - BZ
Sfor BZ
From Navarra and Tribbia, 2005
36
How to include the IMAGE Land use scenarios into
SILVA? Could the 20th century land cover be
included in the same way?
CO2
CO2
Anthropic
Agricolture Extensive grass Forest
regrowth Forest Timber
Natural vegetation
CO2
Natural
Bare soil
37
ESM spin up The Land Biosphere-Atmosphere
components
Green shading -gtAccelerated spin up gt10 times
for the Soil carbon pools 2 times for root
carbon 5 times for wood carbon