Dynamical Prediction of the Terrestrial
Ecosystem
and the Global Carbon Cycle:
a 25-year Hindcast Experiment
Jin-Ho Yoon
Dept. Atmospheric and Oceanic Science
University of Maryland
Co-Authors: Ning Zeng, Augustin Vintzileos, G. James Collatz, Eugenia Kalnay,
Annarita Mariotti, Arun Kumar, Antonio Busalacchi, Stephen Lord
Collaborators: C. Roedenbeck, P Wetzel, E. Maier-Reimer,
Brian Cook, H. Qian, R. Iacono, E. Munoz
March 5, 2008 CPASW
‘Breathing’ of the biosphere: CO2 as a
major indicator of climate
Modeled land-atmo flux vs. MLO CO2 growth rate
Terrestrial carbon model forced by observed climate variability
ENSO, Pinatubo, drought and other signals
Seasonal-interannual Prediction
of Ecosystem and Carbon Cycle
Two strands of recent research made this a real possibility
• Significantly improved skill in atmosphere-ocean prediction
system, such as CFS at NCEP
• Development of dynamic ecosystem and carbon cycle models
that are capable of capturing major interannual variabilities,
when forced by realistic climate anomalies
A pilot study at U Maryland:
 Feasibility study using a prototype eco-carbon prediction system
dynamical vs. statistical
Targets
• Predict spatial patterns and temporal variability of carbon
fluxes and pool size  Example: Reduced biosphere
productivity and enhanced fire and CO2 from Amazon.
• Predict atmospheric CO2 concentration and growth rate.
Atmospheric CO2 can be a ‘climate index’ indicating
anomalies in the global ecosystem
The NCEP Climate Forecast System
(CFS, Saha et al. 2006)
CFS captures major ENSO and other seasonal-interannual variability
The VEgetation-Global Atmosphere-Soil Model (VEGAS)
Atmospheric
CO2
Photosynthesis
NEE = Rh – NPP = + 3 (Interannual)
Autotrophic
respiration
NPP=60 Pg/y
Carbon
allocation
Turnover
Rh=60Pg/y
Heterotrophic
respiration
4 Plant Functional Types:
Broadleaf tree
Needleleaf tree
C3 Grass (cold)
C4 Grass (warm)
3 Vegetation carbon pools:
Leaf
Root
Wood
3 Soil carbon pools:
Fast
Intermediate
Slow
Forecasting Procedure I
Climate
Predition
Spinup
Ecosystem+
Carbon Model
CFS (9mon, 15
CFS (9mon, 15
members)
members)
Precip
Precip
Temp
Temp
VEGAS
Initialization
I
VEGAS
1 mo forecast
ensemble mean
Predicted
Eco-carbon
Output
9mon, 15 members
Month 1
Output
9mon, 15 members
Month 2
Forecasting procedure II
L=3
L=2
L=1
Ensemble
mean
L=0
t-1
t
t+1
NEE(‘validation') and MLO CO2
NEE (land-atmo C flux): VEGAS forced by observed climate (Precip, T
This will be called ‘validation’ as there is no true observation available
Ocean contribution smaller, so NEE can be compared with atmo CO2
Using regression of inversion/OCMIP with Nino3.4/MEI?
NEE(‘validation') and Inversion
(from MPI)
First look: Productivity (NPP)
Predicted global cabon flux (Fta)
'Observed'
Lead time from 0 to 8 months
1. CFS/VEGAS captures most of the interannual variability, but
2. Amplitude is underestimated
Anomaly Correlation Fta
High skills in
• South America
• Indonesia
• southern Africa
• eastern Australia
• western US
• central Asia
Summary of skill for anomaly correlation
Correlation .vs. Regression (Amplitude)
Correlation
Regression
Benchmark Forecast:
Do we need dynamical forecast?
Relaxation or Damping of climate forcing
Anomaly at L=0 will persist or
damped to zero with decorrelation time scale.
Persistence
Damping
L=0
L=1
L=2
Benchmark Forecast
Do we need dynamic forecast system?
Benchmark Forecast
Beyond ENSO:
Drought during 1998-2002
Beyond ENSO: Fire in the US
Natural and anthropogenic factors
Model
Observation
Conclusions: prediction
• Encouraging results (better than expected)
Seasonal prediction of Prec/Tas(CFS) +
Dynamic ecosystem model (VEGAS)
Parallel project on ‘Fire danger (potential)
prediction’.
Working on ‘operational mode’.
Conclusions: prediction
Issues
• Overestimation at midlatitude
Implications of prediction
• Applications to ecosystem and carbon cycle
• A new framework for study eco-carbon
response and feedback to climate
• Identifying ways of incorporating eco-carbon
dynamics in the next generation of climate
prediction models
Thank you!
Questions?
The NCEP Climate Forecast System
(CFS, Saha et al. 2006)
Summary of skill for anomaly correlation