LBL logo Lawrence Berkeley National Laboratory » Earth Sciences Division
Questions & Comments | Privacy & Security Notice

CO2 Flux Measurement Systems (SGP) (SGPCO2flx)

This precision gas system collects high accuracy, high precision measurements of CO2 concentration at 2, 4, 25, and 60 m above the ground. Based at the 60-m tower at the central facility, Southern Great Plains.

General Purpose


The SGP carbon dioxide flux (CO2flux) measurement systems provide from 1/2 to 4 hr mean estimates of the fluxes of CO2, H2O (latent heat), and sensible heat. The fluxes are obtained by the eddy covariance technique, which computes the flux as the mean product of the vertical wind component with CO2 and H2O densities, or estimated virtual temperature. A three-dimensional sonic anemometer is used to obtain the orthogonal wind components and the virtual (sonic) temperature. An infrared gas analyzer is used to obtain the CO2 and H2O densities.

Primary Quantities Measured with System

The CO2 flux systems measurement systems provide from 1/2 to 4 hr mean estimates of the fluxes of CO2, H2O (latent heat), and sensible heat from a variable area (footprint) of the land surface upwind of the instrument. In rough terms, the extent of the footprint, which depends on the mean wind speed and the degree of turbulent mixing in the atmosphere, varies from 5 - 500 times the height of the sensors above the land surface. For the instrument located at 60 m on the Central Facility (CF) tower this translates into distances between approximately 0.3 - 30 km from the tower.

The fluxes are computed from the following directly measured data. Orthogonal components of the wind velocity, u, v, and w (m s-1), and sonic temperature (K), which is approximately equal to virtual temperature are measured by the sonic anemometer. CO2 and H2O densities (mmol m-3) are measured by an infrared gas analyzer (IRGA), sent to the sonic as analog signals, and digitized with the wind velocity measurements at 10 Hz.

[↑ Return to the top]

Overall Uncertainties for Primary Quantities Measured

Uncertainties in the measurements are typically dominated by random noise from the measurement instruments. However, the data processing software is designed to identify other sources of interference that affect the measurements. The most frequent source of interference is airborne material (e.g. rain) that briefly obscures the sound or light path of the sensors. See description of processing algorithms given below.

In normal operation instrument noise limits measurements as follows:

  • CO2 flux: detection limit ~ 0.1 umol/m2/s, gain uncertainty ~ 3%

  • H2O flux: detection limit ~ 10 W/m2/s, gain uncertainty ~ 3%

  • Sensible heat: detection limit ~ 10 W/m2/s, gain uncertainty ~ 3%

[↑ Return to the top]

Detailed Description

List of Components

3-D Sonic Anemometer, Gill Solent Windmaster Pro

  • Orthogonal wind velocities u, v, and w Range: +/-20 m/s Accuracy: u,v =1.5% RMS error, w =3% RMS error Resolution: 0.01 m/s
  • Sonic temperature (from speed of sound (SOS)) Range: -40 to +60 deg C (307-367 m s-1) Accuracy: 3% RMS error in SOS Resolution: 0.02 deg C

Infrared Gas Analyzer, Licor Inc. LI-7500 (see

  • CO2 density Range: 0 to 110 mmol/m3; Accuracy: ~ 1% (limited by calibration procedure) Precision: ~ 4 umol/m3 (typical RMS instrument noise)
  • H20 density Range: 0 to 2000 mmol m-3 Accuracy: ~ 1% (limited by calibration) Precision: 0.14 mmol/m3 (typical RMS instrument noise)

Data collection system

  • 266 - 600 MHz Intel Pentium III PC clone; operating system Windows NT or 2000
  • Data collection software: 1/1/2001 - 12/18/2001 Gillsonic (courtesy Dennis Baldocchi, UC Berkeley) 1/19/2001 - present: WinfluxWMP (courtesy Joe Verfaillie, CSU San Diego)

Description of System Configuration and Measurement Methods

The anemometer and IRGA are located on the South East boom of the CF 60m tower. Data from the anemometer is transmitted to a PC in an instrument shed at the base of the tower. The PC collects and stores the serial binary data stream from the sonic anemometer connection using WinfluxWMP software (courtesy of Joe Verfaillie at CSU San Diego). The raw data is transferred to LBNL, processed into the ARM archive format, and inspected for problems on a daily basis. Processed files are sent to ARM Archive using the Site Transfer Suite on a weekly basis.

[↑ Return to the top]

Theory of Operations

Turbulent fluxes are calculated using standard methods in biometeorology. See references for discussions.

The 3-D sonic anemometer uses three pairs of orthogonal ultrasonic transmit/receive transducers to measure the transit time of sound signals traveling between the transducer pairs. The wind speed along each transducer axis is determined from the difference in transit times. The sonic temperature is computed from the speed of sound which is determined from the average transit time along the vertical axis. A pair of measurements are made along each axis 100 times per second. Ten measurements are averaged to produce 10 wind measurements along each axis and 10 temperatures each second.

The infrared gas analyzer measures CO2 and H2O densities by detecting the absorption of infrared radiation by water vapor in the light path. Details of the IRGA operation and performance can be obtained from Licor Env. Inc. (

Data collection is accomplished using WinfluxWMP software on a standard Intel based personal computer running Microsoft Windows NT or 2000. Data is collected in 1/2 hour intervals, using the computer clock start as a time reference. Each 1/2 hour data file has a time stamp reflecting the start time of the file. The computer clock is updated on a regular basis using the D4 time server software. The daily collection of 48 raw data files is downloaded from the data collection computer to a processing computer at the Lawrence Berkeley National Laboratory on a daily basis and reduced to produce eddy covariance estimates of turbulent fluxes. A set of data processing algorithms is used to create files suitable for inspection and ingest into the ARM data archive.

Data Processing Algorithms

The first program processes the raw (a0) data to produce intermediate (a1) data files. The averaging time for calculations can be varied to produce EC averages on 1/2, 1, 2, and 4 hr time scales. The calculation is performed as follows:

  1. Read in raw data and convert to engineering units (u,v,w (m/s), T sonic (C), CO2 and H20 (Volts)).
  2. Shift the CO2 and H20 signals back by 0.2 seconds (2 samples) to correct for a fixed time lag in the LI-7500 analyzer.

Identify and remove spikes from data using 100 second running mean filter. Spikes are identified as data points with values more than a set number of standard deviations away from running mean. Spike data are given value of running mean and are not used to update mean. Spikes are counted and the mean value of the spikes is calculated. A QC flag is raised if more than 100 data points in a given interval are flagged as spikes.

  1. Calculate statistics (mean, variance, skewness, and kurtosis) of each variable and covariances between all signal pairs.
  2. Calculate 2-D coordinate rotation to zero mean w and v and apply to vector and covariance quantities.
  3. Write out results.

The second program processes intermediate (a1) files to produce estimates of turbulent fluxes with initial QC flags as follows:

  1. Read in a1 file and apply IRGA output settings to change output voltages (Volts) to densities (mmol/m-3).
  2. Compute turbulent fluxes of CO2 and H20 including appropriate Webb-Pearmann-Leuning corrections (Webb et al, 1980) for sensible and latent heat (Webb et al, 1980). Use sonic temperature and H20 densities to estimate air density and specific heat. Note that these calculations assume a constant pressure of 1 atmosphere and may be slightly in error in cases where pressure is significantly different. Also note that no corrections for loss of spectral energy due to sensor separation are performed (e.g. Moore, 1986).
  3. Inspect and flag data falling outside of acceptable limits based on variance, spike counts, and turbulence conditions u*.
  4. Write out results.

[↑ Return to the top]

Current Status and Locations

Currently, there is an operational system located on the south east boom at 60 m on the Central Facility (CF) tower.

Data Quality

Data quality is judged by inspecting QC flags and variables in processed data. Algorithms for improved quality assurance are being developed.

Current Health and Status

Fully operational

Data User Notes


Automated Quality Control/Flagging Contained within NETCDF Files

Output files include QC flags as described below. NETCDF files not produced at present.

Raw data QA/QC

Spike count for u,v,w,T, q, c

This is a summary of the qc flags in a1 and b1 files.

******Flags present in _A1_ files:

qc_u number of samples out of range u

speed > 40m/s deviation from mean > 6*(std dev)

qc_v number of samples out of range v

speed > 40m/s deviation from mean > 6*(std dev)

qc_w number of samples out of range w

speed > 40m/s deviation from mean > 6*(std dev)

qc_t number of samples out of range t

deviation from mean > 5*(std dev)

qc_q number of samples out of range q

value > .1 value < 4.98 deviation from mean > 6*(std dev)

qc_c number of samples out of range c

value > .1 value < 4.98 deviation from mean > 6*(std dev)

nspk_u number of samples removed due to spikes u

= qc_u

nspk_v number of samples removed due to spikes v

= qc_v

nspk_w number of samples removed due to spikes w

= qc_w

nspk_t number of samples removed due to spikes t

= qc_t

nspk_q number of samples removed due to spikes q

= qc_q

nspk_c number of samples removed due to spikes c

= qc_c

Processed Data Checks

******Flags present in _B1_ files:

qc_flag_w QC flag on variable w

0= ok 1= out of range: w < -10 or w > 10 (currently not used) 2= spike: nspk (num. spikes) > 100

qc_flag_t QC flag on variable t

0= ok 1= out of range: mean_t < -10 or mean_t > 50 (currently not used) 2= spike: nspk (num.spikes) > 100 3= large variance: (mean_t / sqrt(variance_t)) < 2

qc_flag_q QC flag on variable q

0= ok 1= out of range: mean_q < 0 or mean_q > 2000 (currently not used) 2= spike: nspk (num. spikes) > 100 3= large variance: rho_q/sqrt(variance_q) < 2

note: rho_q = calib.low_h2o + mean_q * ((calib.high_h2o-calib.low_h2o)/5.)

qc_flag_c QC flag on variable c

0= ok 1= out of range: mean_c < 1 or mean_c > 20 (currently not used) 2= spike: nspk (num. spikes) > 100 3= large variance: rho_c/sqrt(variance_c) < 40

note: rho_c = calib.low_co2 + mean_c * ((calib.high_co2-calib.low_co2)/5.)

qc_flag_ustar QC flag on variable ustar

0= ok 1= too low: ustar < .15 2= positive u'w': u'w' > 0

[↑ Return to the top]

Instrument Mentor Quality Control Checks

Visual QC frequency: daily to weekly

QC delay: typically 1-2 days

QC type: -

Instrument mentor Marc Fischer and data processing assistant Igor Pesenson routinely view graphical displays produced at LBNL. The displays include graphs of CO2, H20, sensible fluxes, mean and variance of CO2 concentration (not corrected for barometric pressure) and wind speed.

Inputs: raw data


Reference: none

Value Added Procedures

None at present

Quality Measurement Experiments

None at present

Examples of Data

See data quick look at

Data Quicklooks/Near Realtime

See data quick look at

[↑ Return to the top]

Calibration and Maintenance

Calibration Theory

The sonic anemometer does not require maintenance or calibration. The IRGA offset and gain need to be calibrated on a periodic basis. The IRGA is calibrated by introducing gas of know concentration into a calibration hood that surrounds the light path over which infrared absorption is measured. The offset is typically calibrated using dry N2 from a gas bottle. The gain of the CO2 and H2O channels is calibrated using a bottle with a known concentration of CO2 and flow from a H2O vapor generator (e.g. Licor Inc. LI-610 Dew Point Generator).

Calibration History

October, 18, 2000 July, 13, 2001 December, 18, 2001

Maintenance Procedures

The sonic anemometer does not require maintenance or calibration. The IRGA offset and gain are calibrated on a biannual basis following the manufacturers recommended procedure.

Online Maintenance Documentation

None yet.

Supplemental Assessment of Instrument Calibration and Maintenance Procedures


[↑ Return to the top]

Frequestly Asked Questions-FAQs

Where do I get more information?

Contact the instrument mentor.

Software Documentation for this Instrument

General description of data processing and data product formats can be found in the birth of data stream file (e.g. 60mflx_bods_mlf-r20020516.xls). Please request most recent version from instrument mentor.

[↑ Return to the top]


Instrument Mentor

Marc L. Fischer, Staff Scientist

Atmospheric Science Department Environmental Energy Technologies Division Mail Stop 51-208 E.O. Lawrence Berkeley National Laboratory 1 Cyclotron Rd. Berkeley, CA 94720

Tel 510-486-5539 • FAX 510-486-6658 email

web page

Vendor/Instrument Developer

Gill Solent, UK (US dist. Texas Electronics, Tel 800-424-5651)

Citable References

  • Kaimal, J.C., Finnigan, J.J., 1994 Atmospheric Boundary Layer Flows: Their Structure and Measurement. Oxford University Press, New York
  • Moore, C.J., 1986. Frequency Response Corrections for Eddy Correlation Systems. Boundary-Layer Meteorol. 37, 17-35
  • Paw U, K.T., Baldocchi, D.D., Meyers, T.P., Wilson, K.B., Correction of Eddy-Covariance Measurements Incorporating Both Advective Effects and Density Fluxes. Boundary-Layer Meteorol. 97,487-511
  • Webb, E.K., Pearman, G.I., and Leuning, R., 1980. Correction of Flux Measurements for Density Effects due to Heat and Water Vapour Transfer. Quart. J. Roy. Meteorol. Soc. 106, 85-100

[↑ Return to the top]