Variables Quick Start Guide

FLUXNET2015 Variables Quick Start Guide

(A version of this page originally appears in the supplementary materials for the FLUXNET2015 reference paper1)

This document is designed to guide non-expert users to quickly get started selecting variables from the FLUXNET2015 dataset. Key points are explained here, but we strongly recommend looking at documentation of the dataset to understand the different versions of variables, as the choice of variable can impact analysis results. See the online documentation and the FLUXNET2015 reference paper1 for more information, and for in depth questions, contact us ([email protected]), regional network teams, or site teams.

FLUXNET2015 data products. Non-expert users can get started with the FLUXNET2015 SUBSET data product. Compared to the FULLSET product, SUBSET includes all the same sites and the complete temporal record, but only includes a selection of the variables, which should fit the required data for most users. All variables mentioned in this guide are available in the SUBSET data product.

Quality flags and gap-filled data. The *_QC variables are quality control flags for the records of associated variables, identifying originally measured or gap-filled values, along with a quality indicator in the latter case. For instance, TA_F_QC is the quality flag for the gap-filled air temperature variable TA_F. At the half-hourly or hourly resolution (HH), the _QC variable indicates if the corresponding record is a measured value (*_QC=0), or the quality level of the gap-filling that was used for that record (*_QC=1 better, *_QC=3 worse quality). At coarser temporal resolutions, i.e., daily (DD) through yearly (YY), the quality flag indicates the percentage of measured (*_QC=0) or good quality gap-filled (*_QC=1) records aggregated from finer temporal resolutions.

CO2 Flux Variables. CO2 fluxes are key in the dataset, being the variable with most related variables and versions produced. The main versions of CO2 variables discussed in this guide cover Net Ecosystem Exchange (NEE), Ecosystem Respiration (RECO), and Gross Primary Production (GPP).

NEE: the variable proposed in the SUBSET product is NEE_VUT_REF since it maintains the temporal variability (as opposed to the MEAN NEE), it is representative of the ensemble, and the VUT method is sensitive to possible changes of the canopy (density and height) and site setup. The FULLSET includes other versions such the CUT where the same USTAR threshold is used for all the years, removing possible variability in the NEE due to the threshold value or the MEAN that is the average of the ensambles (which smooths over the variability). The other NEE variables in SUBSET represent uncertainty estimates: random uncertainty from measurements (NEE_VUT_REF_RANDUNC) and the uncertainty due to the USTAR threshold-based filtering (NEE_VUT_XX; see USTAR uncertainty details below).

RECO and GPP:  in the SUBSET product there are two main versions of GPP and RECO. They originate from two CO2 flux partitioning methods adopted for FLUXNET2015: nighttime (NT) and daytime (DT). The two methods are independent, making their consistency an indicator of the robustness of the estimates. Without a context in which they are being used, it is impossible to give a-priori preference to one or the other. Our suggestion is to use both daytime (DT) and nighttime (NT) variables and consider their difference as uncertainty. Alternatively, users can filter the sites to use in the analysis based on the consistency between the two products. Being highly dependent on site and time-aggregation, the difference between the two partitioning methods can reach, at an annual time resolution, over 500 gC m-2 yr-1. The RECO and GPP products in SUBSET are calculated from the corresponding NEE variables filtered with the VUT method, generating RECO_NT_VUT_REF  and  RECO_DT_VUT_REF for RECO, and GPP_NT_VUT_REF  and  GPP_DT_VUT_REF for GPP. As for the NEE variable above, GPP and RECO also include variables describing the uncertainty due to USTAR threshold-based filtering (see USTAR uncertainty details next).

USTAR Uncertainty: the uncertainty stemming from the USTAR threshold estimation is the main source of uncertainty for  this dataset. An ensemble of USTAR thresholds are applied to filter NEE and the resulting versions of NEE are represented through percentiles (NEE_VUT_XX). The effect of the USTAR threshold uncertainty is also site and time-aggregation dependent, with an interquartile range at annual scale that can be up to 200 gC m-2 year-1. The ensemble of NEE versions originated using the different thresholds are all put through the partitioning with the two methods, resulting in versions of RECO and GPP also represented by percentiles (RECO_NT_VUT_XX/RECO_DT_VUT_XX and GPP_NT_VUT_XX/GPP_DT_VUT_XX).

Energy and Water Flux Variables. The main variables for Latent and Sensible Heat fluxes are LE_F_MDS and H_F_MDS, respectively, both gap-filled and with quality flags. Similarly from NEE, random uncertainty from measurements are estimated (LE_RANDUNC and H_RANDUNC). If the intended use requires ensuring the energy-balance closure, a version of LE and H is provided for which the closure is enforced using the Bowen ratio method (see paper for details): LE_CORR and H_CORR. These versions also include uncertainty estimation for the half-hourly and daily time resolution (LE_CORR_25/H_CORR_25 and LE_CORR_75/H_CORR_75).

1Pastorello, G., Trotta, C., Canfora, E. et al. The FLUXNET2015 dataset and the ONEFlux processing pipeline for eddy covariance data. Sci Data 7, 225 (2020).