Description of output variables

Gap-filling

Except for NEE the variable names of the original variables are not changed. NEE is changed to NEEorig. The gap-filled variables receive suffices that have the meaning as outlined in the following table:

Suffix

Description

_f

The gap-filled variable. E.g. NEE_f is the filled net ecosystem CO2 exchange. The unit remain as the originally submitted variable.

_fmet

The method used to fill the gap. 1 = Inferred from marginal distribution of Rg, Tair, and VPD; 2 = Inferred from Rg only; 3 = Inferred from time of the day only (mean diurnal variation). Consult the introductory page for details. 0 (zero) means that the original data was available.

_fwin

The window size used to fill the gap. Consult the introductory page for details. 0 (zero) means that the original data was available.

_fqc

A synthetic QC for the gap-filling. 1 = category A (most reliable), 2 =category B (medium), 3=category C (least reliable). 0 (zero) means that the original data was available. Consult the introductory page for details. The synthetic QC is tentative and a detailed analysis is pending. In the summary file an evaluation of the gap-filling categories for your data set is provided (using artificial gaps)

_fn

The number of values that were averaged to fill the gap. If the original data was available this is set to -9999.

_fs

The standard deviation of the values that were averaged to fill the gap. If the original data was available this is set to -9999. 

_fqcOK

A synthetic QC that is 0 when _fqc is > 1 (Cat. B or C), otherwise 1. When aggregating to daily, weekly etc. timesteps, this flag tells which fraction of the data was original or filled with category A (most reliable).

_rGAPS_*

This variable is derived by introducing random gaps (half-hourly, daily, weekly block gaps) to the original variable and treatment of the gaps in the same way as for the original variable. Consequently the suffixes represented by the star (*) have the same meaning.

_*_unc

Same as for all the variables above, but calculated for every data point, even if not a gap. E.g., you can find the uncertainties for your whole time span in _fs_unc, please see methods page for details.

side note: The filled meteorological variables are considered as side products with little theoretical justification. However it seems that empirically a large proportion of the gaps can be filled with reasonable accuracy (e.g. <30 W m-2 for Rg at Cat. A). A systematic study on this is pending.