QCQA operations data processing intercomparison

1
QC/QA operations data processing intercomparison

• A. Brut, N. Jarosz, E. Ceschia?, J. Elbers?, B.
  Piguet,J.M. Bonnefond, O. Traullé

CNRM INRA ? ALTERRA ? CESBIO
2
Context and Objectives

• Importance of data processing and flux
  computation in the framework of CarboEurope-IP
  and FLUXNET to improve turbulent flux estimates
  obtained with the eddy covariance method
• To evaluate the differences in flux computation
  softwares
• To detect  weaknesses  in data processing and
  improve the reliability of flux estimates

3
Data and methods

• 3 summer days from Le Bray Tower (thanks to
  Nathalie Jarosz)
•  Golden files  from 5 CarboEurope sites
  intercomparison still in progress (thanks to Jan
  Elbers)
• Different sites, different instruments (sonic
  anemometers, fast sensors for H20/CO2)
• Eddy covariance method Fc ltwcgt
• But different  ways  to obtain fluctuations and
  to correct for errors (instrumental/theoretical)

4
Steps of flux computation

• 1. Identification of spikes
• 2. Corrections for cross-wind, time lag between
  sensors
• 3. Calculation of rotations and averages
• 4. Covariance calculation
• 5. Spectral corrections after Moore (1986)
• 6. Density corrections for scalar fluxes (Webb et
  al., 1980)
• 7. Final unit conversions
• Gap filling
• Quality control

5
Results intercomparison

• A few statistics

TEAMS INRA CESBIO ALTERRA CNRM U
R² 0.99 R² 0.99 R²0.99 R²0.99 slope a
1.01 a 0.93 a 1.01 a 1.04 Hs
R² 0.99 R² 0.99 R²0.96 R² 0.98
a 1.02 a 0.9 a 1.01 a 1.05
Le R²0.99 R²0.98 R²0.98 R²
0.96 a 0.94 a 0.87 a
1.05 a 0.97 Fc R² 0.97 R²0.95
R² 0.91 R² 0.93 a 0.99 a
0.84 a 1.06 a 1.10
6
Results friction velocity
7
Results friction velocity

• - mean values over the 3 days
• INRA 0.453 m/s
• CESBIO 0.429 m/s
• ALTERRA 0.451 m/s
• CNRM 0.457 m/s
• total mean value 0.45 0.012 m/s (3)

8
Results sensible heat flux

• mean values over the 3 days
• INRA 48.9 W/m²
• CESBIO 39.3 W/m²
• ALTERRA 51.6 W/m²
• CNRM 49.1 W/m²
• total mean value 47.2 5.4 W/m² (10)

9
Results latent heat flux

• - more noise on data
• - mean values over the 3 days
• INRA 100.5 W/m²
• CESBIO 93.6 W/m²
• ALTERRA 109W/m²
• CNRM 97.9 W/m²
• total mean value 100.3 6.6 W/m² (6)

10
Results CO2 flux

• - mean values over the 3 days
• INRA -3.5 µmol/m²/s
• CESBIO -2.6 µmol/m²/s
• ALTERRA -4.7 µmol/m²/s
• CNRM -3.4 µmol/m²/s
• total mean value -3.55 0.86 µmol/m²/s (25)

11
Results energy budget

• Latent sensible heat fluxes versus net
  radiation
• high scatter for CNRM data
• low value for the slope of linear fit for CESBIO

12
Conclusions

• On average, good results for friction velocity
  but rather large scatter for CO2 fluxes
• For a particular meteo event different flux
  estimates large discrepancies between flux
  algorithms lead to discrepancies of turbulent
  fluxes
• In general - high scatter for CNRM data
• - weaker fluxes for CESBIO
• - higher flux values (on average) for
  ALTERRA

13
Suggestions

• Follow all the recommandations of the 1st
  CARBOEUROPE QC/QA meeting to calculate the
  turbulent fluxes
• Go on the investigation with the Golden files
• Validation of the data using quality control
  tools (ogive functions, turbulent parameters etc
  )
• Others ???