Eddy-covariance flux measurements: Outline for the day

1
Eddy-covariance flux measurements Outline for
the day

• Morning
• ChEAS sites and current research projects
• Methods (Berger et al, 2001 Yi et al, 2000)
• Results
• Published or in press
• Yi et al 2000, Davis et al, Cook et al.
• In preparation (with hypotheses)
• Various, PSU research group members
• Future plans/proposals

2
Eddy-covariance flux measurements Outline for
the day

• Afternoon
• Eddy covariance flux calculation and Li-Cor demo
• Small group discussion. Suggested topics
• Causes of interannual variability at WLEF
• Causes of differences among ChEAS tower flux
  measurements
• Potential for instrument bias, errors, and
  improvements
• Extension of interannual variability studies
  beyond ChEAS
• Uses of sub-canopy flux and turbulence
  measurements
• Two-dimensional flux experiments and analyses
• Caterpillars observed or imagined?

3
ChEAS eddy covariance flux measurements

• I Sites and research projects

4
Chequamegon Ecosystem-Atmosphere Study (ChEAS)
flux towers
WLEF tall tower (447m) CO2 flux measurements at
30, 122 and 396 m CO2 mixing ratio
measurements at 11, 30, 76, 122, 244 and 396
m
Forest stand flux towers Mature deciduous
upland (Willow Creek) Deciduous wetland
(Lost Creek) Mixed old growth
(Sylvania) All have both CO2 flux and high
precision mixing ratio measurements.
5
Upland, wetland, and very tall flux tower.
Old growth tower to the NE. High-precision
CO2 profile at each site. Mini-mesonet,
15-20km spacing between towers.
?
? Lost Creek
Landcover key
Open water
? WLEF
Wetland
Coniferous
Mixed deciduous/coniferous
Shrubland
? Willow Creek
General Agriculture
6
View from 396m above Wisconsin WLEF TV tower
7
ChEAS web sites

• http//cheas.psu.edu - Main page.
• ftp//ftp.essc.psu.edu/pub/workgroup/davis/ -
  Data access.
• http//cheas.psu.edu/fieldsites.html - Site
  descriptions.
• Also see
• http//www.daac.ornl.gov/FLUXNET/fluxnet.html -
  Fluxnets home page, and,
• http//public.ornl.gov/ameriflux/Participants/Site
  s/Map/index.cfm - AmeriFluxs home page.

8
(No Transcript)
9
(No Transcript)
10
Data availability
Site Start date Full years of data
WLEF Spring 1995 1997-2001
Willow Creek Summer 1999 2000-2001
Lost Creek Fall 2000 2001
Sylvania Summer 2001
Ceilometer Summer 1998 1998-2001
Radar Spring-Fall, 1998 Spring-Fall, 1999 Spring-Fall, 1998 Spring-Fall, 1999

11
Research Funding

• Regional atmosphere/forest exchange and
  concentrations of carbon dioxide.
• Study of net ecosystem exchange of carbon dioxide
  via eddy covariance measurements at the WLEF
  tower in northern Wisconsin, as well as the study
  of carbon dioxide transport and distribution
  within the boundary layer.
• PI P.S. Bakwin, U. Colorado/NOAA
• Co-I K.J. Davis Penn State
• Department of Energy, National Institutes for
  Global Environmental Change
• Duration July, 1994June, 1997 July,
  1997June, 2000 July, 2000June 2003.
• Measuring and modeling component and whole-system
  CO2 fluxes at local to regional scales.
• Study of component processes which make up CO2
  fluxes in a forest ecosystem, and comparison to
  whole-ecosystem net flux measurements from small
  flux towers. Also a comparison between
  homogeneous ecosystem fluxes within the WLEF
  tower footprint and the WLEF net flux signal.
• PI P.V. Bolstad, U. Minnesota
• Co-Is K.J. Davis, Penn State, and P.B. Reich, U.
  Minnesota
• Department of Energy, National Institutes for
  Global Environmental Change
• Duration July, 1997 - June, 2000. July 2000
  June 2003.

12
Research Funding

• Quantifying carbon sequestration potential of mid
  and late successional forests in the upper
  Midwest
• Observations of CO2 exchanges in an old-growth
  forest in the upper peninsula of Michigan and
  comparison to existing flux towers in younger
  forest stands northern Wisconsin. Davis's work
  would provide tower construction,
  instrumentation, and data analysis support for a
  30m tower.
• PI Eileen Carey, U. Minnesota
• Co-Is P.V. Bolstad, U. Minnesota K.J. Davis,
  Penn State
• Department of Energy, Terrestrial Carbon
  Processes
• Duration January, 2001 - December, 2003
• Regional forest-ABL coupling Influence on CO2
  and climate
• Study of the coupling between the surface energy
  balance, boundary layer development, and net
  ecosystem exchanges of carbon dioxide, as well as
  the influence of the covariance between carbon
  dioxide fluxes and boundary layer development on
  boundary layer mixing ratios of carbon dioxide.
  Observations at the WLEF and the Walker Branch
  AmeriFlux sites using an NCAR radar.
• PI K.J. Davis, Penn State
• Co-I A.S. Denning, Colorado State University
• Department of Energy, TECO/Terrestrial Carbon
  Program
• Duration September, 1997 - August, 2002

13
ChEAS eddy covariance flux measurements

• II Methods

14
Methods Eddy covariance flux measurements in
ChEAS

• Basic theory of eddy covariance flux
  measurements.
• Tower flux instrumentation
• LI-COR calibration
• Sonic rotation
• Lag time correction
• Spectral corrections
• Random and systematic errors due to turbulence
• Preferred NEE algorithm (WLEF only)
• Filling missing data

15
Publications describing methodology

• Yi, C., K.J. Davis, P.S. Bakwin, B.W. Berger, and
  L. C. Marr, 2000. The influence of advection on
  measurements of the net ecosystem-atmosphere
  exchange of CO2 observed from a very tall tower,
  J. Geophys. Res. 105, 9991-9999.
• Berger, B.W., K.J. Davis, P.S. Bakwin, C. Yi and
  C. Zhao, 2001. Long-term carbon dioxide fluxes
  from a very tall tower in a northern forest
  Flux measurement methodology. J. Atmos. Oceanic
  Tech., 18, 529-542.
• Davis, K.J., P.S. Bakwin, B.W. Berger, C. Yi, C.
  Zhao, R.M. Teclaw and J.G. Isebrands, The annual
  cycles of CO2 and H2O exchange over a northern
  mixed forest as observed from a very tall tower.
  Global Change Biology, in press.

16
Theory
17
Theory
Reynolds averaged ( mean turbulent
components of all variables) scalar conservation
equation.
Time rate of change (e.g. CO2)
Mean transport
Turbulent transport (flux)
Source in the atmosphere
Integrate from the earths surface to the
imaginary plane defined by the level of the flux
sensor.
18
Theory
Yi et al, 2000
19
Theory What is wc?

• Prime indicates departure from the mean.
• w gt 0 is an updraft
• c gt 0 is air rich in the scalar c
• wc gt 0 is upwards transport of the scalar
• Averaging this over time sums the transport
  observed due to all updrafts and downdrafts.

20
Daily cycle of ABL depth, and CO2 fluxes and
mixing ratios
Radar ABL depth
WLEF fluxes
CO2 profile
Davis et al, in press
21
Diurnal cycle of CO2 in the ABL
Bakwin et al, 1998
22
Instruments at WLEF
Berger et al, 2001
23
Instruments at WLEF

• Two profiling LI-CORs in the trailer, one
  sampling 396m, one cycling among all 6 levels.
  Slow time response. High-precision and
  accuracy calibration (Bakwin et al, 1998). C-bar.
• Vaisala humidity and temperature sensors at 3
  levels (30, 122 and 396m). Slow Q-bar, T-bar.
• Three sonic anemometers (30, 122 and 396m). w,
  T
• Three LI-CORs in the trailer, one for each sonic
  level. Fast time response. Long tubes, big
  pumps. Measure CO2 and H2O. c, q
• Two LI-CORs on the tower (122 and 396m). Fast
  time response. Short tubes, smaller pumps.

24
Calibration of fast CO2 and H2O sensors at
ChEAS towers

• Calibration occurs using the fluctuations in the
  ambient atmospheric CO2 and H2O mixing ratios.
• Slow sensors provide absolute values of these
  mixing ratios used to calibrate the fast
  LI-CORs.
• Ideal gas law corrections to LI-COR cell
  temperature, pressure and humidity are applied.
• Calibration slope and intercept are derived every
  2 days. These values are smoothed (monthly
  running mean) to derive the long-term calibration
  factors used for the fast LI-CORs.

25
Calibration of fast CO2 and H2O sensors
Berger et al, 2001
26
Whats up? (Sonic rotations)

• Sonic anemometers are oriented perfectly in the
  vertical, (and the winds streamlines arent
  always perpendicular to gravity).
• Data is collected over a long time (about a year)
  and we define up by forcing the mean vertical
  wind speed to be zero.

27
Sonic rotations
Berger et al, 2001
28
Lag time calculation

• We must correct for the delay between the CO2 and
  H2O measurements and the vertical velocity
  measurements.
• Lag time is determined by finding the maximum in
  the lagged covariance between vertical velocity
  and CO2/H2O for every hour.

Level (m) IRGA position Tube length (m) Lag time (s) Tube inner diameter (m) Flow rate (L min-1) Reynolds number
396 Trailer 406 87 0.009 17.8 2640
122 Trailer 132 23 0.009 21.9 3250
30 Trailer 40 16 0.009 9.5 1420
396 Tower 5 1.7 0.0032 1.4 592
122 Tower 5 1.1 0.0032 2.2 915
Berger et al, 2001
29
Lag time calculation
Berger et al, 2001
30
Spectral corrections

• Flow through tubes smears out some of the
  atmospheric fluctuations, especially the small
  (high frequency) eddies.
• Obvious for H2O. Much worse than theory
  predicts.
• Not directly observed for CO2. Small effect.
• The sonic anemometer (virtual) temperature
  measurement is not smeared out, so we use
  similarity between the virtual temperature
  spectrum and the water vapor spectrum to correct
  for the loss of high frequency eddies in H2O.
• We use past studies of flow in tubes to correct
  for the loss of high frequency eddies in CO2.

31
Spectral corrections
CO2
Berger et al, 2001
Tv
H2O
32
Spectral corrections
Table shows the typical of flux lost due to
smearing of small eddies.
Level (m) IRGA position CO2 (day) CO2 (night) H2O
396 Trailer 1 7 16
122 Trailer 1.5 9 19
30 Trailer 5 12 21
396 Tower lt0.1 1 13
122 Tower lt0.1 1 11
Berger et al, 2001
33
Systematic errors getting the large and
small eddies
Berger et al, 2001
34
Random errors a finite number of eddies are
counted in one hour
Random sampling errors for any one hour can be as
large as the magnitude of the measured flux!
Berger et al, 2001, following Lenschow and
Stankov, 1986.
35
Preferred NEE (WLEF only)

• Data is taken from 30m at night and 122 or 396m
  during the day (the highest level where there is
  turbulent flow) when all data are available.
• If data are missing, any existing flux
  measurement is used.
• Data are screened out when the level of
  turbulence is very low. CO2 is probably draining
  down hill.
• Early in the morning upper level data from WLEF
  is replaced with 30m data (Yi et al, 2000)
  because the flow appears to be systematically
  2-D.
• Thus from 3 NEE measurements, one is derived as
  our preferred measurement for each hour.

36
Nighttime drainage flows?
Loss of flux at low turbulence levels at the
Willow Creek tower.
Cook et al, submitted Davis et al, in press
37
Morning advection at WLEF
Yi et al, 2000
38
Multiple level comparison at WLEF
Davis et al, in press
Comparison of all 3 levels, growing season 1997.
39
ChEAS eddy covariance flux measurements

• III Results

40
Missing data, gross fluxes, light and temperature
response

• Nighttime NEE measurements (for CO2) are fitted
  to soil or air temperature. This is assumed to
  describe the total respiration flux.
• Daytime NEE measurements are fitted to PAR after
  total respiration has been computed using the
  fits and subtracted from NEE. This fit
  describes the response of forest photosynthesis
  to sunlight.
• These fits are used to compute gross fluxes
  (respiration, photosynthesis) and to fill in
  missing NEE data needed to compute cumulative NEE.

41
Gross fluxes and functional fits
Davis et al, in press
42
Example of gross fluxes fit to temperature
and PAR at WLEF for one month. Davis et al, in
Press.
43
Hourly fluxes at WLEF for 1997, observed and
filled. Davis et al, in press.
44
Cumulative fluxes at WLEF, 1997
Davis et al, in press
45
Gross fluxes at WLEF, 1997
RE
Davis et al, in press
NEE
-GEP
46
1997 Cumulative NEE, GEP and RE vs. assumptions
and methods
(gC m-2 yr-1 tC ha-1 yr-1 100)
Davis et al, in press
47
Lack of energy balance Are turbulent fluxes
underestimated?
Davis et al, in press Cook et al, submitted
48
Monthly mean CO2 fluxes at WLEF, 1997. Davis et
al, in press
49
Monthly mean latent heat fluxes at WLEF,
1997. Davis et al, in press
50
Bowen ratio vs. time of year at WLEF,
1997. Davis et al, in press, following Cook et
al, submitted.
51
Drainage during stable conditions What goes
down must come up (somewhere).
52
Very large positive turbulent fluxes from about
150 degrees. Blocking of flow. Occur
during windy, weakly stable conditions when the
canopy is decoupled from the ABL. Cook et al,
sub.
53
Is this venting of drainage?Can we capture these
events across the landscape?
54
Cook et al, submitted to Global Change Biology
Willow Creek
55
Courtesy D. Hollinger
56
Early leaf-out, 1998, Wisconsin
57
Impact on atmospheric CO2
58
Spatial coherence of seasonal flux anomalies
A similar pattern is seen at several flux towers
in N. America and Europe. Three sites have
high-quality CO2 measurements data at
Fluxnet (NOBS, HF, WLEF). The spring 98 warm
period and a later cloudy period appear at all
3 sites.
Temp
CO2
NEE
Day of year
80
200
59
Detection of the spring 98 anomaly via oceanic
flasks?
2 Alaskan flask sites have slightly higher CO2
in the spring of 98. Mace Head, Ireland shows a
depression of CO2 in the spring of
98. Potential exists to link flux towers with
seasonal inverse studies.
60
Synoptic variability in CO2
61
North American Carbon Plan(NACP)http//www.carbo
ncyclescience.gov