Interannual Variability and a Persistent Source of CO2 at WLEF

1
Interannual Variability and a Persistent Source
of CO2 at WLEF

• Daniel M. Ricciuto
• ChEAS meeting VII
• June 23, 2004

2
WLEF continues to be a source of CO2
NEE and gross fluxes at WLEF using standardized
filling technique
Year NEE RE GPP
WLEF 1997 44 1074 -1030
WLEF 1998 75 1099 -1024
WLEF 1999 102 1148 -1046
WLEF 2000 95 1136 -1041
WLEF 2001 171 1197 -1026
WLEF 2002 50 xxx xxx
WLEF 2003 80 1090 -1010
3
NEE at WLEF and nearby ChEAS towers
4
Higher respiration than other ChEAS sites
Growing season R-T fits (u screened)

• Respiration during the growing season at WLEF
  is higher than that of any of the other ChEAS
  sites at 18oC
• Most similar to Sylvania (old growth flux
  tower)
• WLEF base respiration twice that of Lost Creek,
  higher Q10 than Willow Creek.
• Upscaling hypothesis fails for respiration.

5
Photosynthesis at WLEF A combination of wetland
and upland values

• Magnitude of uptake is lower than that of
  Willow Creek, higher than Lost Creek.
• Again, most similar to Sylvania (old growth
  site)
• Upscaling hypothesis works here WLEF
  photosynthesis is a combination of upland and
  wetland photosynthesis

6
Questions about error at WLEF

• Two major questions
• Is systematic error causing WLEF to look like a
  source when it is, in fact, in balance or a small
  sink?
• Are the ranges of NEE and gross fluxes larger
  than random error?
• Interannual range of NEE 130 gC m-2
• Interannual range of RE 120 gC m-2
• Interannual range of GPP 50 gC m-2
• Major sources of error in WLEF annual flux
  estimates
• Largest systematic errors affecting annual WLEF
  NEE average
• Nighttime drainage during low turbulence
  conditions
• Measurement level bias grassy clearing at 30
  meters
• Random errors affecting annual WLEF NEE
  variability
• Interannual variations in footprint
• Interannual variations in level selection for
  preferred NEE (30,122,396)
• Variability from sampling error

7
Fluxes down the drain U screening bias?

• To produce the annual NEE estimates previously
  shown, we used a u cutoff value of 0.2 ms-1

Average annual NEE as a function of u cutoff

• This screens about 50 of nighttime growing
  season data.
• Increases annual NEE by about 60 gC m-2 yr-1.
• We dont think biological fluxes should be a
  function of U. However, NEE continues to
  increase to about 110 gC m-2 yr-1 as u cutoff
  approaches 0.4.
• If this possible bias is taken into account,
  WLEF would be an even larger source.

Annual NEE (gC m-2yr-1)
1997-2001 average
U cutoff (ms-1)
8
That pesky grassy clearing level bias

• Preliminary analysis by Wang grassy clearing
  is a significant part of daytime 30m footprint
  but not of 122m or 396m footprints.
• Difference between 30m and 122m implies that
  using 30m may cause daytime fluxes to be
  underestimated by 8-10
• Translates to error of 35gC m-2yr-1
  (opposite sign)
• Daytime 396m fluxes 33 larger than 30m. Cant
  explain 122-396 meter difference. ???

9
Conclusions about error at WLEF

• Random error of an annual NEE measurement is
  about 25 gC m-2.
• This is smaller than the range of variability
  (130 gC m-2).
• Random error of gross fluxes is larger due to
  inaccuracies in flux decomposition model.
• Errors in respiration smaller than range of
  variability but errors in photosynthesis
  comparable.
• Averaging time bias at higher levels due to low
  frequency loss is negligible.
• The 2 major systematic errors (U and level bias)
  partially cancel. Remaining error (20 gC m-2)
  actually makes WLEF a larger source and does not
  change the interpretation of results (WLEF is
  still a persistent source).

10
Explaining monthly to interannual variability

• Now that we know the variability is real, what
  is causing it? Climate? Disturbance? Both?
• Over hourly to daily timescales, T and PAR
  explain most variability.
• Over longer timescales, other factors become
  more important Soil moisture, precip,
  phenology.
• Anomalies from this mean year driven by both
  climate and disturbance

14-day average NEE 1997-2001
11
Explaining monthly to interannual variability

• Figure on the left shows monthly deviations from
  5-year mean (1997-2001)
• One important factor is growing season
  timing/length, illustrated on the left.
• May 1997 late leafout
• May 1998 early leafout
• Other factors become clearer when fluxes are
  broken down into gross components.

12

• Temperature and precipitation are important
  driving factors at monthly to annual timescales.
• Reduced precipitation causes reduced uptake and
  reduced respiration (see summer 1998). Net
  effect lower NEE.
• Increased precipitation causes increased
  respiration and increased uptake (see July 1999).
  Net effect higher NEE
• Non-climatic factors also important.
  Caterpillars in 2001.
• Cant explain anomalous uptake in summer 1997.

13
Climate Variables Explain Most Growing Season
Respiration Variability
Observed vs. Modeled monthly averaged nighttime
NEE RE (a0SWC2 a1SWC a2)Q10((Tair-10)/10
)

• Hourly growing season values (1997-2001) of NEE
  are fit to the above 4-parameter equation.
• Monthly average model-observation comparison
  shown.
• Growing season average model correlates very
  well with observations (R2 gt 0.98)
• Much weaker monthly correlation if no SWC
  dependence included in model (R2Tair 0.65)

14
Soil Moisture Dependence of Respiration

• Sharply reduced respiration at low soil
  moisture (lt0.14)
• Such conditions occur in 1998 (drought) and
  near the end of the growing season
• Optimal soil water content 0.19
• Slightly decreased respiration during very
  moist conditions
• Effect of soil saturation (more anaerobic) or
  time of year

Modeled RE as a function of SWC, Tair15oC
15
Climate Variables Explain Some Growing Season
Photosynthesis Variability

• A similar 4-parameter fit to hourly values of
  GPPNEE-RE using either VPD or soil moisture and
  PAR produces a monthly average model-observation
  correlation 0.7 (not shown)
• Also a weaker correlation with monthly
  precipitation
• Lagged climate variables and/or disturbance more
  important?

16
What we know and dont know

• We know
• The interannual variability signal in NEE is
  larger than random error.
• WLEF is a persistent source, even when
  considering systematic error.
• But probably would be in balance if we used only
  396m in daytime
• Growing season respiration at WLEF is largely
  controlled by temperature and soil moisture on
  monthly to annual timescales.
• Drying of the soil in the growing season reduces
  respiration, and to some degree, photosynthesis.
  (see dry years 1998, 2003)
• We dont know
• Why WLEF is a source. Drying wetlands?
  Persistent disturbance?
• All of the controlling factors of photosynthesis.
  Why so much in 1997?

17
(No Transcript)
18
ChEAS Site U Dependences (Jun-Aug)
19
U screening error

• How do we determine the error resulting from our
  choice of U cutoff?
• As u cutoff increases, data availability is
  reduced causing an increase in random error. (see
  error bars below)
• As u cutoff decreases, associated systematic
  error increases.

• Methodology determining u screening error
• Assume 0.4 ms-1 is ideal u cutoff
• deviation of RE values and associated random
  error are determined as a function of u cutoff.
• fractional error RE (u gt 0.4) / RE (u gt
  0.2) -1
• yearly resp (1err) RE(yearly)
• standard deviation of systematic error
  determined using error bars on left.

20
Determining Level Bias Error

• Use of 30m daytime fluxes generally causes
  underestimation of daytime uptake, therefore
  overestimation of NEE.
• Simulate NEE determined using ideal hourly level
  selection using monthly diurnal means.
• Determine error for each year by using actual
  level selections and compare to ideal case.

• Error ranges from 0-20 gC m-2

21
Water Table Levels near WLEF

• Is the region around WLEF drying out? Or was
  1997 just really wet?
• Explanation for anomalous uptake in 1997? Why?
• Are drying wetland margins responsible for the
  source of CO2 at WLEF? We dont observe higher
  respiration in 1998 (dry) or less in 1997.