Transcript Slide 1

Using Simulated OCO Measurements for
Assessing Terrestrial Carbon Pools in the
Southern United States
PI: Nick Younan
Roger King, Surya Durbha, Fengxiang Han
Zhiling Long, Narendra Rongali, Haiqing Zhu
Orbiting Carbon Observatory (OCO)
Introduction




Estimated global total net flux of carbon
from changes in land use increased
from 503 Tg C (1012 g) in 1850 to 2376
Tg C in 1991 and then declined to 2081
Tg C in 2000.
The global net flux during the period
1850-2000 was 156 Pg C (1015 g), about
63% of which was from the tropics.
The US estimated flux is a net source to
the atmosphere of 7 Pg C for the period
1850-2000, but a net sink of 1.2 Pg C for
the 1980s and 1.1 Pg C for the 1990s.
Hence, better estimates at regional level
are required to understand and reduce
the uncertainties in the sink/source
estimations
Source:http://www.netl.doe.gov/technologies/carbon_seq/overview/imag
es/carbon-flux-diagram.gif
Data Source: Houghton, R.A, 1999. The annual net flux of carbon to the atmosphere from the
changes in land use 1850-1990. Tellus 51B:298-313
Currently funded DOE project for leverage





What are the current annual rates of terrestrial carbon
sequestration in each state of the region?
What's the overall contribution of terrestrial carbon
sequestration in each state of the region to mitigating its
total greenhouse gas emission?
What's the current baseline for possible carbon trading in
the region?
What's the potential of further enhancing terrestrial
carbon sequestration in the region?
What are the overall economic impacts of current and
potential terrestrial carbon sequestration on the region?
County-level Surface Soil organic C Density
(0-30 cm, kg C/m2)
7.0
C Density, kg C/m2
6.0
5.0
4.0
3.0
2.0
1.0
0.0
State
SW
SE
Central Delta
North
Total Soil Organic C Density
(kg C/m2)
C Density, kg C/m2
25
20
15
10
5
0
State
SW
SE
Central
Delta
North
County-level MS Forest C density (kg C/m2)
C Density, kg C/m2
10
8
5
3
0
State
SW
SE
Central
Delta
North
Comparison of Soil C and Forest C
Storage in regions of MS
Total Forest C: 392 Tg C
Total Soil C: 809 Tg C
SW
SE
Centr
al
Delta
Housing/Furniture C:
661 Tg C, 3.0%
Crop C: 85 Tg C,
Pasture C:
27.8 Tg C,
0.13%
Forest
C:
Soil Organic
Total terrestrial carbon storage
and pools in the Study Area
Total Terrestrial C Storage: 21762
Focus Areas of the Project (Plan B)

The RPC experiment seeks to address the following
questions:





What information about carbon exchange can be obtained from
OCO high-precision column measurements of CO2?
How can we integrate top-down OCO measurements with ground
based measurements, atmospheric and terrestrial ecosystem
models to quantify carbon exchange over different ecosystems?
What are the current annual rates of terrestrial carbon
sequestration in each state of the Southeast and Southcentral U.S.?
What is the current baseline in the region for possible
carbon trading?
What is the potential for enhancing terrestrial carbon
sequestration?
NASA-CASA Model


NASA-CASA (Carnegie Ames Stanford Approach) model
is designed to estimate monthly patterns in carbon
fixation, plant biomass, nutrient allocation, litter fall, soil
nutrient mineralization, and CO2 exchange, including
carbon emissions from soils world-wide.
Assimilates satellite NDVI data from the MODIS sensor
into the NASA-CASA model to estimate

Spatial variability in monthly net primary production (NPP),


biomass accumulation,
and litter fall inputs to soil carbon pools
CASA Model-Inputs/Outputs

Data

Inputs:


Outputs:


NDVI ( MODIS) , Soil (SSURGO), Precipitation (PRISM), Air
Temperature (PRISM), Land Mask, Solar Radiation (NARR),Vegetation
type.
Carbon pools, LAI, NPP, NEP, AET,APAR , FAPR, LEAFFR, NBP, NPP
moist, NPP temp, PET, resp, rootfr, soilc, stemfr.
Other:

Soil, Land cover, Parameters.
Soil Types (SSURGO)
Precipitation (PRISM)
CASA output fits/reflects well with the combination of
Soil C and forest C in county-level of MS
Total Soil Carbon
Soil Microbial Respiration source of Carbon
Leaf Area Index (LAI)-2002
May
Jun
July
Net Primary Productivity (NPP)-2002
May
Jun
Monthly NPP was estimated in CASA as :
NPP=f(NDVI)x PAR x LUE x g(T) x h(W)
July
Net Ecosystem Productivity (NEP)-2002
May
Jun
July
RPC Experimental Design (Modified)
• Fossil Fuels
• Assimilation of aircraft
measurements, satellite
data (precipitable water,
surface winds)
Meteorology
(e.g. GOES
data analysis)
• Winds, cloud
mass fluxes,
model
Parameters
• Forward Transport Model
Transport
Model
•
•
•
•
Vegetation Indices
Biome type
Soil properties
Weather Reanalysis
Land Surface
Model (CASA)
• 1 year spinup
• Monthly
• OCO, Networks
[CO2] OBS
1 year spinup
(2002)
• Terrestrial
CO2 surface
flux
Inversion
Design of Simulation Experiments
Surface
Fluxes
CASA
Model
Transport
Model
Perturbation
With Errors
Simulated OCO
Observations
Perturbation
With Errors


Simulated Priors
Ensemble
Based
Inversion
Estimated
Fluxes
Simulated OCO data not available from NASA yet.
Currently use data generated on our own.
Evaluation
Kalman Filter
Observations
Initial
Estimates
Forecast
Background
Estimates
Update
Updated Estimates
 Bayesian data assimilation is conceptually
simple but computationally prohibitive
for application on large problems.
 Kalman filter is a simplified
approximation to the Bayesian
estimation, which assumes:
 Normality of error statistics, and
 Linearity of error growth.
Two main approaches can be
followed to handle observations
(Mathieu et al, 2008):
1.A Filter, whereby the analysis is
only influenced by observations
made in the past, which is the
case for real-time applications
and forecasting.
2. A smoother, where the analysis
is influenced by all observation
available over a given period
“T” ( assimilation window)
Ensemble Based Assimilation
Ensemble-based Update
Initial
Ensemble
Forecast
Model
Errors
Addition
Background
Ensemble
Observations
Mean
Error Covariance
Statistical
Analysis
Update
Kalman Gain
Reduced Kalman Gain
Updated Ensemble



Ensemble based approaches combine the Kalman filter concept with Monte-Carlo
techniques.
More accurate than the Kalman filter because there are no assumptions about the
normality and linearity of errors.
Investigated two methods for the update process: deterministic (EnSRF) and
stochastic (EnKF).
Example Assimilation Results (I)
Ground Truth Fluxes
Observations
source
100
75
sink
Assimilation Results
50
Assimilation Errors
25
0
-25
-50


The synthetic ground truth fluxes simulate one source area and one sink area.
The ensemble based technique was able to assimilate the observations to generate
flux estimates with small errors.
Example Assimilation Results (II)
Error Statistics Obtained from a 10-Step Assimilation Experiment
Standard
Deviation
Mean
Time Steps



Time Steps
Errors are consistent throughout all time steps.
Results are similar in this case for both the deterministic (EnSRF) and the stochastic (EnKF)
methods.
Working on Implementing the covariance localization technique for the update process.
 Estimates for background error covariance may be inaccurate when small ensembles
are utilized. This technique helps to improve the accuracy for such estimation based on
small ensembles.
Tasks Completed/Ongoing







Input data sets for the CASA model conditioned ( written several
scripts, ArcMap models) for the southern United States
CASA model simulations for the entire Southern United states in
progress.
Sensitivity studies of CASA model outputs with NASA-CQUEST is
being performed.
In situ soil carbon studies completed for Southern United States
Explored several transport models for suitability for carbon fluxes
transport. Currently working on WRF-CHEM for this purpose.
Assimilation Code-based on Ensemble Kalman filter(both stochastic
and deterministic update methods) developed in Matlab.
Participated in 2008 Carbon Cycle and Ecosystems Joint Science
Workshop to be held April 28 - May 2, 2008
Publications







Younan, N. H. , Durbha, S. S., King, R. L., Han, F. X, Long, Z., Rongali, N., Zhu, H.,
(2009) . "Data Assimilation for Assessing Terrestrial Carbon Pools in the Southern
United States”. 33rd International Symposium on Remote Sensing of Environment
(ISRSE), Italy.
Younan, N. H., King, R. L., Durbha, S. S., Han, F. X, Long, Z., Chen, J. (2007). “Using
Simulated OCO Measurements for Assessing Terrestrial Carbon Pools in the
Southern United States”. American Geophysical Union ( AGU) , Fall Meeting .
Durbha, S. S., Younan, N., King, R., Han, F. X., Long, Z. (2008). A Rapid Prototyping
Capability Experiment to Assess Terrestrial Carbon Pools in Southern United States.
2008 NASA Carbon Cycle and Ecosystems Joint Science Workshop, Maryland, USA.
Nutrient fertilizer requirements for sustainable biomass supply to meet U.S.
bioenergy goal (In revision).
County-level distribution of soil and forest carbon storage in Mississippi ( under
preparation)
Validation of NASA-CASA model for terrestrial carbon pools in Mississippi.
( under preparation)
Questions?
Source :http://earthobservatory.nasa.gov/Features/CarbonCycle/Images/carbon_cycle_diagram.jpg