CH-AWS / FF-202002

You are here:

Flux calculations for years 2006-2019 (except 2012).

This flux product includes eddy covariance fluxes between 2006 and 2019. It was not possible to calculate fluxes for 2012 due to timestamp issues in the raw binary files. Maybe at a later point in time we can solve this issue, but it was not possible for this product.

Before 2015, data were only recorded during the vegetation period. From 2015 onwards, data for the full year are available.


Level-1 Fluxes

  • Python script for conversion and flux calculations: FCT093_EP706
  • Script for correcting the calibration gas issue: CGC Calibration Gain Correction v0.3 (GitLab)
  • EddyPro version: 7.0.6 (18 Dec 2019)
  • For an overview of raw data formats see here.


Calibration Gas Issue

The calibration gas used to calibrate IRGA CO2 measurements during the time periods 2017_2, 2018_1, 2019_1 and 2019_2 was wrong and the CO2 measurements needed to be corrected before flux calculations. This correction cannot be done in EddyPro directly. For this run, a Python script was used to multiply the measured CO2 with a specfic gain value if 0.974 (see here for more details on this issue).

The workflow to yield correct fluxes for the affected time periods was:

  1. Original raw binary files were converted to csv using FCT093EP706.
  2. A gain of 0.974 was then applied to the two CO2 columns (CO2 in ppm and mmol) in the converted csv files, using a Python script. The columns to which the gain was applied to were stored in the same file as the originally measured values. The two new columns have the suffix “_GAIN-0.974”.
  3. The gain-corrected data csv files from (2) were then used to calculate fluxes.


Open Lag Runs

The lag time between SA and IRGA was checked in particular for the oldest fluxes IRGA75 2006-2015. This was done by searching the max covariance without using a default lag time in EddyPro, i.e. all found lag times in the EddyPro full_output file show found lag times. The peak of the distribution of all found lag times per year were:

  • 2006: peak @0.35s für CO2 and H2O
  • 2007: peak @0.30s and @0.40s für CO2 and H2O, set to 0.40s
  • 2008: peak @0.15s für CO2 and H2O
  • 2009: peak @0.15s für CO2 and H2O
  • 2010: peak @0.20s für CO2 and H2O
  • 2011: peak @0.15s für CO2 and H2O
  • 2013: peak @0.15s für CO2 and H2O, but the peak distribution of found lag times is less clear than in the other years, even seems to go into negative numbers
  • 2014: peak @0.10s for CO2, @0.15s for H2O
  • 2015: peak @0.15s für CO2 and H2O


Wind Direction

The wind direction was checked for all years. For the HS50 / IRGA72 setup, the settings to yield the correct wind direction in EddyPro was known (thanks to tests done during the ICOS labelling process for CH-DAV), i.e. the wind direction for the respective years should be correct. This wind direction was then used to adjust the SA orientation of older years. The orientation was selected so that the wind direction of the R2 / IRGA75 years looked the same as the HS50 / IRGA72 years. This was necessary because the SA orientation in the earlier years was not correctly documented, so I had to “reverse-engineer” it.


Spectral Correction

For the spectral correction a specific spectral assessment file (SAF) is created. For the HS50 / IRGA72 setup, one full year of data without interruption was available for 2018 (2018_1). All other years were interrupted at some point. Therefore, the SAF created for 2018 was used also for the other years with the same setup (2016_1, 2016_2, 2017_1, 2017_2, 2019_1, 2019_2, 2019_3) to keep the spectral correction the same between years.

In contrast, years with the R2 / IRGA75 setup each used their own respective SAF, because it made less sense to base the spectral correction on one specific year (all years only had the vegetation period covered, no full years were measured yet). One issue for these years was that the spectral correction did not work as intended when the Vickers & Mahrt 1997 raw data tests were taken into account when creating the reference spectra. In particular, the tests seem to kick out all H2O spectra, most likely due to values being above the absolute threshold that is defined in EddyPro. Therefore, for these years the V&M97 tests were disabled and then the spectral correction worked as intended.


  • Imported gas binned spectra with V&M97 enabled: CO2: 3848 / H2O: 0
  • Imported gas binned spectra with V&M97 disabled: CO2: 3848 / H2O: 3282


Off-season Carbon Uptake Correction (Burba Correction)

Due to the self-heating of the IRGA75, the Burba correction needs to be applied to calculated IRGA75 fluxes. At CH-AWS, only the year 2015 recorded data during the winter (off-season), i.e. the correction should be applied. However, past tests seemed to indicate that the correction as it is implemented in EddyPro “over-corrects” affected time periods during winter. Past tests for other stations indicated that the “true” correction is only 5% of the suggested full correction. At CH-AWS, we could test this by comparing the IRGA72 fluxes (which do not need correction) with concurrently measured IRGA75 fluxes (need correction). Although we have time periods of concurrent IRGA75 / IRGA72 measurements, the overlap is rather small (one winter season). We will re-visit this issue again and try to come up with a solution, but for this flux product the IRGA75 fluxes were not corrected. Keep in mind that is possible that the flux uptake rates are probably slightly overestimated for IRGA75. However, since this issue mainly affects the winter season, this mainly affects winter data for 2015 (no winter data are available 2006-2014).


Raw Data Files 2006-2008

Date files were converted by WE. The original data files showed the measurements in mV. WE converted the files to units usable in the flux calculations.


Level-2 Fluxes

  • Produced in Amp 0.14.0 (on 24 Feb 2020).

Amp outputs a separate flag column for each QC step. In this step, the following QC checks were considered:

  • Plausible range (outlier)
    0=good data, 2=bad, flux value outside respective plausible range
    The following ranges were considered OK:
    – CO2 flux between -50 and +50 umol m-2 s-1
    – H2O flux between -20 and +50 mmol m-2 s-1
    – ET between -1 and +3 mm hour-1
    – LE between -200 and +1000 W m-2
    – H between -200 and +800 W m-2
  • SSITC tests (EddyPro qc_* flag)
    0=good data, 1=OK, 2=bad
    This flag is directly generated by EddyPro during flux calculations.
  • Spectral correction factor (*_scf)
    0=good data, 2=bad
    scf values between 0 and +2 were considered OK.


  1. Merge all results
    All EddyPro Level-1 results (*full_output* files) from FF-202002 from all years 2006-2019 (except 2012) were merged into one file.
    Some stats:
    – CO2 flux: 129 907 values available / 107 191 missing (45.2%)
    – H2O flux: 127 681 / 109 417 (46.2%)
  2. Create quality flags
    Create QCF_ columns that contain quality flag (0-1-2), then aggregate single quality flags into one overall quality flag QCF_AGG_ column for each flux.

For ET, the qc flag for LE must be used.