1. Ensemble Forecasting from LPJ simulations over the Mediterranean basin: Preliminary Results
Dr. Monia Santini
University of Tuscia and CMCC
CMCC Annual Meeting
Marina di Ugento (LE)
June 9th-11st, 2010 1

2. Aim
To analyze uncertainties in projections of Med-basin natural vegetation trends for the 21st century from the DGVM-LPJ, based on future global and regional climate simulations, on model parameterization and formulations.

3. Ensemble Forecasting
(Araujo and New, 2007)
“…each individual outlook to the future in a scenario-study will not necessarily contain the most likely prospects, but, as a whole, the simulations provide the bandwidth of possible changes.” (Dammers, 2000).
Initial condition perturbation
The ensemble spread gives information about the prediction errors

4. LPJ inputs/outputs Spatialized inputs Soil: texture
Climate: monthly temperature, precipitation, cloud coverage, number of wet days
Aggregated inputs
CO2 concentration
Average grid-cell basis outputs:
Monthly: carbon and water fluxes
Annually: carbon pools , fire return period, and competition among 9 Plant Functional Types (PFTs) + barren areas

5. Sources of LPJ uncertainties
Model inputs: future climate projections
Representation of driving processes and/or model formulation
Parameters within the model (e.g. Zaehle et al ; Anav, 2009) regarding soil and/or PFTs

6. Sources of LPJ uncertainties
Future climate projections
A1B

7. Sources of LPJ uncertainties
Model formulation
Daily precipitation generation:
Interpolation from monthly data
Random generation from precipitation amount and number of wet days
DAY OF MONTH
PRECIPITATION (mm)
Monthly precipitation =256 mm
Number of wet days = 12
Monthly precipitation next month =20mm

8. Sources of LPJ uncertainties
PFT parameters
TNE = Temperate Needleleaved Evergreen
TBE = Temperate Broadleaved Evergreen
TBS = Temperate Broadleaved Summergreen
BNE = Boreal Needleleaved Evergreen
C3 = C3 grass

9. Input data processing Soil (HSWD) 0.0083333° (≈1 km)
LPJ soil parameterization
COARSE
MEDIUM COARSE
MEDIUM
FINE COARSE
FINE MEDIUM
FINE MEDIUM COARSE
FINE NONVERTISOL
FINE VERTISOL
ORGANIC
(Harmonized Soil World Database; IIASA, 2009)

10. Input data processing
Soil (HSWD) ° (≈1 km)  ° (≈10 km)
Climate interpolated to aggregated soil grid (natural neighbouring)
CMCC grid

11. Input data processing
Soil (HSWD) ° (≈1 km)  ° (≈10 km)
Climate interpolated to aggregated soil grid (natural neighbouring)
ENEA grid

12. Input data processing
Soil (HSWD) ° (≈1 km)  ° (≈10 km)
Climate interpolated to aggregated soil grid (natural neighbouring)
CNRS-IPSL grid

13. Input data processing
Soil (HSWD) ° (≈1 km)  ° (≈10 km)
Climate interpolated to aggregated soil grid (natural neighbouring)
METEOFRANCE grid (regular)

14. Input data processing
Soil (HSWD) ° (≈1 km)  ° (≈10 km)
Climate interpolated to aggregated soil grid (natural neighbouring)
METEOFRANCE grid (irregular)

15. Input data processing
Soil (HSWD) ° (≈1 km)  ° (≈10 km)
Climate interpolated to aggregated soil grid (natural neighbouring)
MPI grid

16. Input data processing
Soil (HSWD) ° (≈1 km)  ° (≈10 km)
Climate interpolated to aggregated soil grid (natural neighbouring)
CO2-5 years interpolated to 1 year
IMAGE Team, 2001

17. LPJ simulation protocol
Perform 24 separated runs of LPJ
Spin-up period of 1000 years (from 971 to 1970)
Transient simulation from 1971 to 2050

18. Whole-domain climate input anomalies
‘Temperature’ anomaly
YEARS
ANOMALY
Relative to
(all model average)
‘Number of wet days’ anomaly
YEARS
ANOMALY

19. LPJ outputs For each grid cell LPJ produces annual values for:
Plant Functional Types
Vegetation carbon
Soil carbon
Litter carbon
Carbon loss by fires
Fire return period
...and monthly values for:
Net Primary Production
Heterotrophic Respiration
Run-off
Evaporation
Interception
Transpiration
Soil moisture (2 layers)
NEE
Blue water fluxes
White water fluxes (unproductive)
Green water fluxes (productive)

20. PFT percent distribution
TEMPERATE NEEDLELEAVED EVERGREEN
2000

21. PFT percent distribution
TEMPERATE NEEDLELEAVED EVERGREEN
2000
2050

22. PFT percent distribution
C3 PERENNIAL GRASS
2000

23. PFT percent distribution
C3 PERENNIAL GRASS
2050
2000

24. Litter C - aggregated average
gC/m2
YEARS
YEARS

25. Soil C - aggregated average
gC/m2
YEARS
YEARS

26. Vegetation C - aggregated average
gC/m2
YEARS
YEARS

27. C loss by fires - aggregated average
gC/m2
YEARS
YEARS

28. Fire probability - aggregated average
YEARS
YEARS

29. Spatial “litter C” anomalies
minus , whole ensemble average

30. Spatial “soil C” anomalies
minus , whole ensemble average

31. Spatial “vegetation C” anomalies
minus , whole ensemble average

32. Spatial “C loss by fires” anomalies
minus , whole ensemble average

33. Spatial yearly Blue Water anomalies
minus , whole ensemble average

34. Spatial yearly Green Water anomalies
minus , whole ensemble average

35. Spatial yearly NPP anomalies
gC/m2
minus , whole ensemble average

36. First conclusions
At aggregated level, different ensemble members have a variable influence in function of the considered model output.
At spatialized level:
two different behaviors between N and S Med-basin for C-related fields
Med-surrounding areas the most critical in terms of water resources and fires

37. in progress and future work
Validation (last decade) at eddy sites (other EU-projects)

38. in progress and future work
Forcing with land use change simulations to separate natural vegetation from croplands focusing on fires
Input to
Historical analysis of LUC
Land use change scenarios from coupled
Input to
Climate projections
drive
Input to
LPJ model
result
LU/LC maps:
URBAN
CROPLAND detail from
FORESTLAND/GRASSLAND detail from LPJ model
RUI = Rural-Urban Interface
Carbon loss by fires
Fire return interval
Output for
Impacts at the RUI

39. in progress and future work
Ensemble updating using other GCMs-RCMs and/or emission scenarios (e.g. AR4 and AR5)
e.g. COSMO-CLM domain (14 km res.)
Temporal coverage

40. in progress and future work
Forcing with land use change simulations to separate natural vegetation from croplands focusing on White, Green & Blue water fluxes in SSA
WGB = White Green Blue
SSA = Sub-Saharian Africa
CLIMAFRICA

41. Thank you!