Transcript Slide 1

Virtual Tall Towers and Inversions
How to Make Productive Use
of Continental CO2 Measurements
in Global Inversions
Martha Butler
The Pennsylvania State University
Department of Meteorology
April 24, 2007
TransCom Meeting
[email protected]
• The Problem with Continental CO2 data
• Virtual Tall Tower Overview
• Inversion Setup
• Perfect Data Results
• Continuing Work
Problem Definition
• Continental data are messy: big signal with lots of
• Confounded by boundary layer processes (diurnal,
synoptic, seasonal)
• Transport models have problems representing the
continental boundary layer, especially near the land
• Most continental observations are made in this
surface layer.
• Observation site density does not permit global
estimation of fluxes at the resolution we would like,
unless we use continental data.
What About Flux Towers?
• CO2 is measured continuously at every continental flux
tower, but…
• These data are not required to be either accurate or precise
to support CO2 flux calculations.
• What if?
• Flux tower CO2 data could be high precision and calibrated
with the global standards (at reasonable cost).
• The data could be “adjusted” (using meteorological and flux
data available at the site) to represent a mid-day CO2 mixing
ratio value above the surface layer.
• This adjustment adds to the uncertainty of the observation;
is it worth it?
What Is This Adjustment?
Following the mixed layer similarity theory of Wyngaard & Brost
[1984] and Moeng & Wyngaard [1989], the vertical gradient of a
scalar in the boundary layer:
  gb 
 zi
 wc0
 gt 
 w* zi
 zi
 wczi
 w* zi
gb and gt are bottom-up and top-down gradient functions scaled by
boundary layer depth zi
w* is the convective velocity scale
wc0 and wczi are the surface and entrainment fluxes of the scalar C
Example Adjusted Time Series
Hourly (midday hours)
adjusted CO2
mixing ratios
Old Black
Courtesy of David Tyndall
See also for high-precision, calibrated measurements at flux towers
and for the Rocky Mountains sites.
Inversion Setup
• Forward transport model: PCTM, 2.5° longitude x 2° latitude using
GEOS-4 met data (3-hr) for 2000, 2001, 2002
• Background fluxes:
– Seasonal fossil with 1995 spatial pattern (ORNL)
– SiB3 hourly flux for 2000, 2001, 2002 (CSU)
– Takahashi 2002 ocean
• Matrix inversion in the TransCom3 style solving for monthly fluxes in
36 land regions and 11 ocean regions in 2002
• Forward model sampling strategy:
– Entire atmospheric column sampled hourly in 2002 at each observation
– Ten-day means of the atmospheric grid saved for analysis.
Networks Tested (so far)
• 85 station network: NOAA GMD sites active in 2002
• 159 station network: NOAA GMD + GAW
• 145 station network:
– Remove some profiles and added VTT sites active in 2002
• 165 station network:
– Add high-precision, well-calibrated sites active after 2002
Flux tower/VTT locations
Britt Stephen’s mountain sites
Bev Law’s Oregon transect
2 established sites in Africa
• 174 station network:
– Add a few other sites in under-represented regions
85 Station Network
159 Station Network
145 Station Network
165 Station Network
174 Station Network
Tests of the Inversion Setup
• Sensitivity to Choice of Priors & Prior Uncertainty
Priors derived from TransCom3 Seasonal Mean Inversion
At least 0 ± 2 Gt C/yr for each region
0 ± 5 Gt C/yr for each region
0 ± 10 Gt C/yr for each region
At least 0 ± 3 Gt C/yr for each land region, TransCom for ocean regions
• Sensitivity of Posterior Flux Uncertainty to Choice of Networks
– Boreal and Temperate North America
– South America
– Africa
Global Aggregated Results
Thick colored lines show
posterior flux (and the prior
contained in the background
Dashed lines show the
posterior uncertainty for the
aggregated area for the 85
station network.
Error bars show the posterior
uncertainty for the 174 station
Note that flux scales differ for
land and ocean.
June Posterior Flux Uncertainties
Oregon transect
Flux Towers
Flux Towers
Boreal North America and Tropical America
Flux Towers
Examples for the African Regions
Continuing Work
Noisy data tests with simulated observations
Model sampling to match observation times
Preparation of observation data
“The real deal”
Daytime forward pulse runs from land regions
• People
Davis Group at Penn State
Denning Group at CSU
Randy Kawa at NASA GSFC
Everyone, everywhere in the CO2 measurement network
• Funding Agencies
– PSU, College of EMS Centennial Travel Grant
Comments and suggestions are welcome!
And thank you!