inversion plan and current progress on mci andrew schuh colorado state university mci workshop june...

30
Inversion plan and current progress on MCI Andrew Schuh Colorado State University MCI Workshop June 17, 2009

Post on 19-Dec-2015

219 views

Category:

Documents


2 download

TRANSCRIPT

Inversion plan and current progress on MCI

Andrew Schuh

Colorado State University

MCI Workshop

June 17, 2009

Inverse Modeling of CO2

Air Parcel Air Parcel

Air Parcel

Sources Sinks

wind wind

SampleSample

Changes in CO2 in the air tell us about sources and sinks

What is an atmospheric CO2 inversion?

Inverse Modeling of CO2

Air Parcel Air Parcel

Air Parcel

Sources Sinks

wind wind

SampleSample

Changes in CO2 in the air tell us about sources and sinks

What is an atmospheric CO2 inversion?

Prior Guess for sources and sinks (SiB3)

Prior Guess for NEE (SiB3)

Inverse Modeling of CO2

Air Parcel Air Parcel

Air Parcel

Sources Sinks

wind wind

SampleSample

Changes in CO2 in the air tell us about sources and sinks

What is an atmospheric CO2 inversion?

LPDM particle transport model

Black: air parcels in contact with surfaceRed: air parcels reach lateral boundaries

(20

days

of “

upst

ream

” tr

ansp

ort i

n 50

sec

onds

) SiB-RAM

S-LPDM

Back-Trajectories from WLEF Tower

(400 m TV tower near Park Falls, WI)

Inverse Modeling of CO2

Air Parcel Air Parcel

Air Parcel

Sources Sinks

wind wind

SampleSample

Changes in CO2 in the air tell us about sources and sinks

What is an atmospheric CO2 inversion?

High precision CO2 observations (NOAA, PSU, Ameriflux, EnvironCanada)

Seasonal cycle • 31-day running mean• Strong coherent

seasonal cycle across stations

• West Branch (wbi) and Centerville (ce) differ significantly from 2007 to 2008

• Large variance in seasonal drawdown, despite being separated by, at most, 550 km

Mauna Loawbi aircraft

Inverse Modeling of CO2

Air Parcel Air Parcel

Air Parcel

Sources Sinks

wind wind

SampleSample

Changes in CO2 in the air tell us about sources and sinks

What is an atmospheric CO2 inversion?

“Update” source/sink estimates based upon regressing residuals against upstream sources and sinks

Inversion Portion

• Easiest way to think of the inversion is as a fancy regression of the tower CO 2 residuals on the upstream flux regions

• Regression, geostatistical regression, Kalman Filters, EnKF, MLEF,….

• At the end of the day…. data appears to be the limiting factor for all of these models

CarbonTracker Inversion Model, Net Terrestrial Annual Flux (gC/m2/yr)

2002 (Drought) 2004 (Non-Drought)

CarbonTracker (inter-annual variability)

Contributing InversionsContributor Structure Transport Prior Resolution Notes

NOAA(CarbonTracker)

EnKF TM5 CASA Ecoregion Nested Global, possibly higher resolution MCI nest

CSU1 (Ravi L.) EnKF PCTM SiB3 6° by 10° Global

CSU2 (A. Schuh) KF RAMS SiB3 100km/10km * Regional w/ boundary conditions provided by NOAA CT and/or CSU PCTM

UofMichigan (Michalak)

Geostat. Inv. WRF? NA ?

Leftover questions: Fossil? (Vulcan, Andres, etc?), fire fluxes ?

Inversion Results Protocol?

• Follow Regional-Continental Level Protocol developed by Jacobsen/Post? Contribute as CF-compliant netcdf file, etc.

• Supplement NEE data w/ information on priors (if relevant) and covariance estimates

Outline of a particular inversion

Forward Model Run (Prior estimates of GPP/Re/NEE and

Transport)

Lagrangian Particle Model (Footprints)

Inversion Results

My CSU inversionEnsemble of MCI Inversions

Inversion Results

Form Meta-inversion via combination of all

inversions

Compare individual inversions to MCI

inventory

Some Initial Results from Forward Model

• Forward SiB3/RAMS coupled run

• Comparison of simulated CO2 to observed CO2

Some Initial Results from Forward Model

• Forward SiB3/RAMS coupled run

• Comparison of simulated CO2 to observed CO2

Conclusions• Anomalies are likely a result of the coupling

between RAMS and SiB3 and are not necessarily evident in either independently

• Transport (particles) can still be calculated and post-hoc SiB3 fluxes “attached” as meta-data in order to run inversions.

• Any forward runs of optimized fluxes would be affected by this anomalous behavior and thus these would need to be corrected to form optimized wall-to-wall CO2 fields.

Some Initial Results from Forward Model

• Forward SiB3/RAMS coupled run

• Comparison of simulated CO2 to observed CO2

Some Initial Results from Forward Model

• Forward SiB3/RAMS coupled run

• Comparison of simulated CO2 to observed CO2

… using day time minimum comparison for now

CO2 Network

Conclusions on northern sites….

• Seasonal cycle (amplitude) of biosphere fluxes are likely too strong in SiB3

• Seasonal amplitude in residuals appears to be function of latitude (more northern implies larger amplitude) as well as longitude (eastern site vs. western site).

Conclusions on MCI sites….

• Caveat: irrigation was not handled in this particular forward run, water stress will be artificially reduced in crop locations to compensate for the fact that most crops do no experience extreme water stress.

• Seasonal cycle (amplitude) of biosphere fluxes may be too strong in SiB3Crop and/or issues with vertical transport, PBL height, etc.

• Seasonal cycle in residuals at LEF and WBI are very weak implying that large seasonal cycle in residuals over MCI is likely a local effect

Conclusions on “the rest”….• CO2 at Colorado sites (Niwot and BAO)

appear to be captured well by model potentially indicating an inflection point in amplitude of seasonal residuals with latitude

• TX tower (WKT) residuals indicate that southern boundary inflow might be biased high

• Relatively weak seasonal cycle in residuals in Texas, AZ, OR, and CO indicate the residuals are likely an effect of local fluxes and stress due to unrealistic water stress in irrigated crops in the more arid portions of the Midwest.

Left to do…….• Run nested simulation at 10km for MCI (has been tested and

is ready to run).

• Run particles (LPDM) for available towers for 2007

• Test out potential new “bells and whistles” for inversion including adaptive inflation scheme, different decompositions of inversion correction factors and figure out nesting of the continental and MCI inversion regions.

• Hope to have initial inversion results by end of summer 2009 (for use by Cooley/Breidt) and have uncertainty better characterized by AGU 2009.

• Driving data and spinup results are “in the can” for 2003-2008 awaiting time and disk space.