NEE PROCESSING RESULTS DESCRIPTION This file includes daily NEE estimates and related uncertainties, plus nighttime and daytime NEE averages and related uncertainties. In the starting dataset, NEE is already corrected for storage and de-spiked (method described in Papale et al 2006 Biogeosciences, z=5). The USTAR filtering is based on thresholds calculated using two methods (Barr et al 2013 and a modified version of Reichstein et al 2005). For each of the two methods, 100 bootstrapped datasets are used (for a total of 200 USTAR thresholds estimates for each year). Two different methods have been used to extract the USTAR thresholds to be applied: - Variable Ustar Threshold (VUT) for each year (identified in the variables names with "_VUT"): the thresholds found for each year and the years before and after have been put together and from their join population the final threshold extracted (USTAR thresholds varying among years). - Constant Ustar Threshold (CUT) across years (identified in the variables names with "_CUT"): all the thresholds found in the different years have been put together and final thresholds extracted from this dataset (each year filtered with the same USTAR threshold). Note that if the dataset includes up to two years of data the two methods give the same result and only the _VUT is provided. In both the _VUT and _CUT versions, 40 NEE datasets have been created filtering the original NEE using 40 different USTAR values extracted from the thresholds datasets at percentiles [1.25:2.5:98.75]. These 40 NEE versions have been first aggregated to daily time resolution and then used as basis for all the derived variables provided. The random uncertainty (RANDUNC) in the measurements is estimated using two hierarchical methods. RANDUNC Method 1 requires measured values with similar meteorological conditions within the sliding window. If the sliding window for RANDUNC Method 1 has gapfilled half-hours or fewer than five measured half-hours with similar meteorological conditions, RANDUNC Method 2 is used instead. - RANDUNC Method 1 (direct SD method): For a sliding window of +/- 7 days and +/- 1 hour of the current timestamp, RANDUNC is calculated as the standard deviation of the measured fluxes measured. The meteorological conditions must also be sufficiently similar, i.e., TA +/- 2.5 deg C, VPD +/- 5 hPa, SW_IN +/- 50 W m-2 (if radiation is higher than 50 W m-2) or SW_IN +/-20 (if lower than 50 W m-2). - RANDUNC Method 2 (median SD method): For a sliding window of +/- 5 days and +/- 1 hour of the current timestamp, RANDUNC is calculated as the median of the random uncertainty (calculated with RANDUNC Method 1) of similar fluxes, i.e., within the range of +/- 20% (but not less than 10 W m-2). VARIABLES DEFINITION: LEGEND: HH (half-hourly or hourly), DD (daily), WW (weekly), MM (monthly), YY (yearly) TIMESTAMP (YYYYMMDD): ISO timestamp - short format DOY (DDD): Day of Year NIGHT (adimensional) Flag indicating nighttime interval based on SW_IN_POT HH: 0 = daytime, 1 = nighttime DD-YY: not produced NIGHT_D (adimensional) Number of half hours classified as nighttime in the period, i.e., when SW_IN_POT is 0 HH: not produced DD: number of half-hours WW-MM: number of halfhours (average of the daily data) YY: not produced DAY_D (adimensional) Number of half hours classified as daytime in the period, i.e., when SW_IN_POT is greater than 0 HH: not produced DD: number of half-hours WW-MM: number of halfhours (average of the daily data) YY: not produced NIGHT_RANDUNC_N (adimensional) Number of half hours classified as nighttime and used to calculate the aggregated random uncertainty HH: not produced DD: number of half-hours WW-YY: number of halfhours (average of the daily data) DAY_RANDUNC_N (adimensional) Number of half hours classified as daytime and used to calculate the aggregated random uncertainty HH: not produced DD: number of half-hours WW-YY: number of halfhours (average of the daily data) NEE_CUT_REF (see temporal resolution for units) Net Ecosystem Exchange, using Constant Ustar Threshold (CUT) across years, reference selected on the basis of the model efficiency HH: (umolCO2 m-2 s-1) DD: calculated from half-hourly data (gC m-2 d-1) WW-MM: average from daily data (gC m-2 d-1) YY: sum from daily data (gC m-2 y-1) NEE_VUT_REF (see temporal resolution for units) Net Ecosystem Exchange, using Variable Ustar Threshold (VUT) for each year, reference selected on the basis of the model efficiency HH: (umolCO2 m-2 s-1) DD: calculated from half-hourly data (gC m-2 d-1) WW-MM: average from daily data (gC m-2 d-1) YY: sum from daily data (gC m-2 y-1) NEE_CUT_REF_QC (adimensional) Quality flag for NEE_CUT_REF HH: 0 = measured; 1 = good quality gapfill; 2 = medium; 3 = poor DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_REF_QC (adimensional) Quality flag for NEE_VUT_REF HH: 0 = measured; 1 = good quality gapfill; 2 = medium; 3 = poor DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_CUT_REF_RANDUNC (see temporal resolution for units) Random uncertainty for NEE_CUT_REF, from measured only data HH: uses only data point where NEE_CUT_REF_QC is 0 (umolCO2 m-2 s-1) and two hierarchical methods (see header and NEE_CUT_REF_RANDUNC_METHOD) DD-MM: from random uncertainty of individual half-hours (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used (gc m-2 d-1) YY: from random uncertainty of individual half-hours (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used (gC m-2 y-1) NEE_VUT_REF_RANDUNC (see temporal resolution for units) Random uncertainty for NEE_VUT_REF, from measured only data HH: uses only data point where NEE_VUT_REF_QC is 0 (umolCO2 m-2 s-1) and two hierarchical methods (see header and NEE_VUT_REF_RANDUNC_METHOD) DD-MM: from random uncertainty of individual half-hours (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used (gc m-2 d-1) YY: from random uncertainty of individual half-hours (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used (gC m-2 y-1) NEE_CUT_REF_RANDUNC_METHOD (adimensional) Method used to estimate the random uncertainty of NEE_CUT_REF HH: 1 = RANDUNC Method 1 (direct SD method), 2 = RANDUNC Method 2 (median SD method) DD-YY: not produced NEE_VUT_REF_RANDUNC_METHOD (adimensional) Method used to estimate the random uncertainty of NEE_VUT_REF HH: 1 = RANDUNC Method 1 (direct SD method), 2 = RANDUNC Method 2 (median SD method) DD-YY: not produced NEE_CUT_REF_RANDUNC_N (adimensional) Number of data points used to estimate the random uncertainty of NEE_CUT_REF DD-YY: not produced NEE_VUT_REF_RANDUNC_N (adimensional) Number of data points used to estimate the random uncertainty of NEE_VUT_REF DD-YY: not produced NEE_CUT_REF_JOINTUNC (see temporal resolution for units) Joint uncertainty estimation for NEE_CUT_REF, including random uncertainty and USTAR filtering uncertainty HH: [SQRT(NEE_CUT_REF_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each half-hour (umolCO2 m-2 s-1) DD: [SQRT(NEE_CUT_REF_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each day (gC m-2 d-1) WW: [SQRT(NEE_CUT_REF_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each week (gC m-2 d-1) MM: [SQRT(NEE_CUT_REF_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each month (gC m-2 d-1) YY: [SQRT(NEE_CUT_REF_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each year (gC m-2 y-1) NEE_VUT_REF_JOINTUNC (see temporal resolution for units) Joint uncertainty estimation for NEE_VUT_REF, including random uncertainty and USTAR filtering uncertainty HH: [SQRT(NEE_VUT_REF_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each half-hour (umolCO2 m-2 s-1) DD: [SQRT(NEE_VUT_REF_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each day (gC m-2 d-1) WW: [SQRT(NEE_VUT_REF_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each week (gC m-2 d-1) MM: [SQRT(NEE_VUT_REF_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each month (gC m-2 d-1) YY: [SQRT(NEE_VUT_REF_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each year (gC m-2 y-1) NEE_CUT_USTAR50 (see temporal resolution for units) Net Ecosystem Exchange, using Constant Ustar Threshold (CUT) across years, from 50 percentile of USTAR threshold HH: (umolCO2 m-2 s-1) DD: calculated from half-hourly data (gC m-2 d-1) WW-MM: average from daily data (gC m-2 d-1) YY: sum from daily data (gC m-2 y-1) NEE_VUT_USTAR50 (see temporal resolution for units) Net Ecosystem Exchange, using Variable Ustar Threshold (VUT) for each year, from 50 percentile of USTAR threshold HH: (umolCO2 m-2 s-1) DD: calculated from half-hourly data (gC m-2 d-1) WW-MM: average from daily data (gC m-2 d-1) YY: sum from daily data (gC m-2 y-1) NEE_CUT_USTAR50_QC (adimensional) Quality flag for NEE_CUT_USTAR50 HH: 0 = measured; 1 = good quality gapfill; 2 = medium; 3 = poor DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_USTAR50_QC (adimensional) Quality flag for NEE_VUT_USTAR50 HH: 0 = measured; 1 = good quality gapfill; 2 = medium; 3 = poor DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_CUT_USTAR50_RANDUNC (see temporal resolution for units) Random uncertainty for NEE_CUT_USTAR50, from measured only data HH: uses only data point where NEE_CUT_USTAR50_QC is 0 (umolCO2 m-2 s-1) and two hierarchical methods (see header and NEE_CUT_USTAR50_RANDUNC_METHOD) DD-MM: from random uncertainty of individual half-hours (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used (gc m-2 d-1) YY: from random uncertainty of individual half-hours (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used (gC m-2 y-1) NEE_VUT_USTAR50_RANDUNC (see temporal resolution for units) Random uncertainty for NEE_VUT_USTAR50, from measured only data HH: uses only data point where NEE_VUT_USTAR50_QC is 0 (umolCO2 m-2 s-1) and two hierarchical methods (see header and NEE_VUT_USTAR50_RANDUNC_METHOD) DD-MM: from random uncertainty of individual half-hours (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used (gc m-2 d-1) YY: from random uncertainty of individual half-hours (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used (gC m-2 y-1) NEE_CUT_USTAR50_RANDUNC_METHOD (adimensional) Method used to estimate the random uncertainty of NEE_CUT_USTAR50 HH: 1 = RANDUNC Method 1 (direct SD method), 2 = RANDUNC Method 2 (median SD method) DD-YY: not produced NEE_VUT_USTAR50_RANDUNC_METHOD (adimensional) Method used to estimate the random uncertainty of NEE_VUT_USTAR50 HH: 1 = RANDUNC Method 1 (direct SD method), 2 = RANDUNC Method 2 (median SD method) DD-YY: not produced NEE_CUT_USTAR50_RANDUNC_N (adimensional) Number of half-hour data points used to estimate the random uncertainty of NEE_CUT_USTAR50 DD-YY: not produced NEE_VUT_USTAR50_RANDUNC_N (adimensional) Number of half-hour data points used to estimate the random uncertainty of NEE_VUT_USTAR50 DD-YY: not produced NEE_CUT_USTAR50_JOINTUNC (see temporal resolution for units) Joint uncertainty estimation for NEE_CUT_USTAR50, including random uncertainty and USTAR filtering uncertainty HH: [SQRT(NEE_CUT_USTAR50_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each half-hour (umolCO2 m-2 s-1) DD: [SQRT(NEE_CUT_USTAR50_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each day (gC m-2 d-1) WW: [SQRT(NEE_CUT_USTAR50_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each week (gC m-2 d-1) MM: [SQRT(NEE_CUT_USTAR50_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each month (gC m-2 d-1) YY: [SQRT(NEE_CUT_USTAR50_RANDUNC^2 + ((NEE_CUT_84 - NEE_CUT_16) / 2)^2)] for each year (gC m-2 y-1) NEE_VUT_USTAR50_JOINTUNC (see temporal resolution for units) Joint uncertainty estimation for NEE_VUT_USTAR50, including random uncertainty and USTAR filtering uncertainty HH: [SQRT(NEE_VUT_USTAR50_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each half-hour (umolCO2 m-2 s-1) DD: [SQRT(NEE_VUT_USTAR50_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each day (gC m-2 d-1) WW: [SQRT(NEE_VUT_USTAR50_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each week (gC m-2 d-1) MM: [SQRT(NEE_VUT_USTAR50_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each month (gC m-2 d-1) YY: [SQRT(NEE_VUT_USTAR50_RANDUNC^2 + ((NEE_VUT_84 - NEE_VUT_16) / 2)^2)] for each year (gC m-2 y-1) NEE_CUT_MEAN (see temporal resolution for units) Net Ecosystem Exchange, using Constant Ustar Threshold (CUT) across years, average from 40 NEE_CUT_XX versions HH: average from 40 half-hourly NEE_CUT_XX (umolCO2 m-2 s-1) DD: average from 40 daily NEE_CUT_XX (gC m-2 d-1) WW: average from 40 weekly NEE_CUT_XX (gC m-2 d-1) MM: average from 40 monthly NEE_CUT_XX (gC m-2 d-1) YY: average from 40 yearly NEE_CUT_XX (gC m-2 y-1) NEE_VUT_MEAN (see temporal resolution for units) Net Ecosystem Exchange, using Variable Ustar Threshold (VUT) for each year, average from 40 NEE_VUT_XX versions HH: average from 40 half-hourly NEE_CUT_XX (umolCO2 m-2 s-1) DD: average from 40 daily NEE_CUT_XX (gC m-2 d-1) WW: average from 40 weekly NEE_CUT_XX (gC m-2 d-1) MM: average from 40 monthly NEE_CUT_XX (gC m-2 d-1) YY: average from 40 yearly NEE_CUT_XX (gC m-2 y-1) NEE_CUT_MEAN_QC (adimensional) Quality flag for NEE_CUT_MEAN, fraction between 0-1 indicating percentage of good quality data HH: average of percentages of good data (NEE_CUT_XX_QC is 0 or 1) from 40 NEE_CUT_XX_QC DD-YY: average of 40 NEE_CUT_XX_QC for the period NEE_VUT_MEAN_QC (adimensional) Quality flag for NEE_VUT_MEAN, fraction between 0-1 indicating percentage of good quality data HH: average of percentages of good data (NEE_VUT_XX_QC is 0 or 1) from 40 NEE_VUT_XX_QC DD-YY: average of 40 NEE_VUT_XX_QC for the period NEE_CUT_SE (see temporal resolution for units) Standard Error for NEE_CUT, calculated as SD(NEE_CUT_XX) / SQRT(40) HH: SE from 40 half-hourly NEE_CUT_XX (umolCO2 m-2 s-1) DD: SE from 40 daily NEE_CUT_XX (gC m-2 d-1) WW: SE from 40 weekly NEE_CUT_XX (gC m-2 d-1) MM: SE from 40 monthly NEE_CUT_XX (gC m-2 d-1) YY: SE from 40 yearly NEE_CUT_XX (gC m-2 y-1) NEE_VUT_SE (see temporal resolution for units) Standard Error for NEE_VUT, calculated as SD(NEE_VUT_XX) / SQRT(40) HH: SE from 40 half-hourly NEE_CUT_XX (umolCO2 m-2 s-1) DD: SE from 40 daily NEE_CUT_XX (gC m-2 d-1) WW: SE from 40 weekly NEE_CUT_XX (gC m-2 d-1) MM: SE from 40 monthly NEE_CUT_XX (gC m-2 d-1) YY: SE from 40 yearly NEE_CUT_XX (gC m-2 y-1) NEE_CUT_XX (see temporal resolution for units) NEE CUT percentiles (indicated by XX) calculated from the 40 estimates for each period -- XX = 05, 16, 25, 50, 75, 84, 95 HH: XXth percentile from 40 half-hourly NEE_CUT_XX (umolCO2 m-2 s-1) DD: XXth percentile from 40 daily NEE_CUT_XX (gC m-2 d-1) WW: XXth percentile from 40 weekly NEE_CUT_XX (gC m-2 d-1) MM: XXth percentile from 40 monthly NEE_CUT_XX (gC m-2 d-1) YY: XXth percentile from 40 yearly NEE_CUT_XX (gC m-2 y-1) NEE_VUT_XX (see temporal resolution for units) NEE VUT percentiles (indicated by XX) calculated from the 40 estimates for each period -- XX = 05, 16, 25, 50, 75, 84, 95 HH: XXth percentile from 40 half-hourly NEE_VUT_XX (umolCO2 m-2 s-1) DD: XXth percentile from 40 daily NEE_VUT_XX (gC m-2 d-1) WW: XXth percentile from 40 weekly NEE_VUT_XX (gC m-2 d-1) MM: XXth percentile from 40 monthly NEE_VUT_XX (gC m-2 d-1) YY: XXth percentile from 40 yearly NEE_VUT_XX (gC m-2 y-1) NEE_CUT_XX_QC (adimensional) Quality flag for NEE_CUT_XX -- XX = 05, 16, 25, 50, 75, 84, 95 HH: 0 = measured; 1 = good quality gapfill; 2 = medium; 3 = poor DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_XX_QC (adimensional) Quality flag for NEE_VUT_XX -- XX = 05, 16, 25, 50, 75, 84, 95 HH: 0 = measured; 1 = good quality gapfill; 2 = medium; 3 = poor DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_CUT_REF_NIGHT (umolCO2 m-2 s-1) Average nighttime NEE, from NEE_CUT_REF HH: not produced DD: average from half-hourly data (where NIGHT is 1) WW-YY: average from daily data NEE_VUT_REF_NIGHT (umolCO2 m-2 s-1) Average nighttime NEE, from NEE_VUT_REF HH: not produced DD: average from half-hourly data (where NIGHT is 1) WW-YY: average from daily data NEE_CUT_REF_NIGHT_SD (umolCO2 m-2 s-1) Standard Deviation of the nighttime NEE, from the NEE_CUT_REF HH: not produced DD: from half-hourly data (where NIGHT is 1) WW-YY: from daily data NEE_VUT_REF_NIGHT_SD (umolCO2 m-2 s-1) Standard Deviation of the nighttime NEE, from the NEE_VUT_REF HH: not produced DD: from half-hourly data (where NIGHT is 1) WW-YY: from daily data NEE_CUT_REF_NIGHT_QC (adimensional) Quality flag for NEE_CUT_REF_NIGHT HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_REF_NIGHT_QC (adimensional) Quality flag for NEE_VUT_REF_NIGHT HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_CUT_REF_NIGHT_RANDUNC (umolCO2 m-2 s-1) Random uncertainty of NEE_CUT_REF_NIGHT, from the random uncertainty of the single nighttime half-hours HH: not produced DD-YY: from random uncertainty of individual half-hours where NIGHT is 1 (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used to calculate the nighttime aggregation in the day. NEE_VUT_REF_NIGHT_RANDUNC (umolCO2 m-2 s-1) Random uncertainty of NEE_VUT_REF_NIGHT, from the random uncertainty of the single nighttime half-hours HH: not produced DD-YY: from random uncertainty of individual half-hours where NIGHT is 1 (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used to calculate the nighttime aggregation in the day. NEE_CUT_REF_NIGHT_JOINTUNC (umolCO2 m-2 s-1) Joint uncertainty estimation for NEE_CUT_REF_NIGHT, including random uncertainty and USTAR filtering uncertainty HH: not produced DD: [SQRT(NEE_CUT_REF_NIGHT_RANDUNC^2 + ((NEE_CUT_84_NIGHT - NEE_CUT_16_NIGHT) / 2)^2)] for each day WW: [SQRT(NEE_CUT_REF_NIGHT_RANDUNC^2 + ((NEE_CUT_84_NIGHT - NEE_CUT_16_NIGHT) / 2)^2)] for each week MM: [SQRT(NEE_CUT_REF_NIGHT_RANDUNC^2 + ((NEE_CUT_84_NIGHT - NEE_CUT_16_NIGHT) / 2)^2)] for each month YY: [SQRT(NEE_CUT_REF_NIGHT_RANDUNC^2 + ((NEE_CUT_84_NIGHT - NEE_CUT_16_NIGHT) / 2)^2)] for each year NEE_VUT_REF_NIGHT_JOINTUNC (umolCO2 m-2 s-1) Joint uncertainty estimation for NEE_VUT_REF_NIGHT, including random uncertainty and USTAR filtering uncertainty HH: not produced DD: [SQRT(NEE_VUT_REF_NIGHT_RANDUNC^2 + ((NEE_VUT_84_NIGHT - NEE_VUT_16_NIGHT) / 2)^2)] for each day WW: [SQRT(NEE_VUT_REF_NIGHT_RANDUNC^2 + ((NEE_VUT_84_NIGHT - NEE_VUT_16_NIGHT) / 2)^2)] for each week MM: [SQRT(NEE_VUT_REF_NIGHT_RANDUNC^2 + ((NEE_VUT_84_NIGHT - NEE_VUT_16_NIGHT) / 2)^2)] for each month YY: [SQRT(NEE_VUT_REF_NIGHT_RANDUNC^2 + ((NEE_VUT_84_NIGHT - NEE_VUT_16_NIGHT) / 2)^2)] for each year NEE_CUT_REF_DAY (umolCO2 m-2 s-1) Average daytime NEE, from NEE_CUT_REF HH: not produced DD: average from half-hourly data (where NIGHT is 0) WW-YY: average from daily data NEE_VUT_REF_DAY (umolCO2 m-2 s-1) Average daytime NEE, from NEE_VUT_REF HH: not produced DD: average from half-hourly data (where NIGHT is 0) WW-YY: average from daily data NEE_CUT_REF_DAY_SD (umolCO2 m-2 s-1) Standard Deviation of the daytime NEE, from the NEE_CUT_REF HH: not produced DD: from half-hourly data (where NIGHT is 0) WW-YY: from daily data NEE_VUT_REF_DAY_SD (umolCO2 m-2 s-1) Standard Deviation of the daytime NEE, from the NEE_VUT_REF HH: not produced DD: from half-hourly data (where NIGHT is 0) WW-YY: from daily data NEE_CUT_REF_DAY_QC (adimensional) Quality flag for NEE_CUT_REF_DAY HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_REF_DAY_QC (adimensional) Quality flag for NEE_VUT_REF_DAY HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_CUT_REF_DAY_RANDUNC (umolCO2 m-2 s-1) Random uncertainty of NEE_CUT_REF_DAY, from the random uncertainty of the single daytime half-hours HH: not produced DD-YY: from random uncertainty of individual half-hours where NIGHT is 0 (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used to calculate the daytime aggregation in the day. NEE_VUT_REF_DAY_RANDUNC (umolCO2 m-2 s-1) Random uncertainty of NEE_VUT_REF_DAY, from the random uncertainty of the single daytime half-hours HH: not produced DD-YY: from random uncertainty of individual half-hours where NIGHT is 0 (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used to calculate the daytime aggregation in the day. NEE_CUT_REF_DAY_JOINTUNC (umolCO2 m-2 s-1) Joint uncertainty estimation for NEE_CUT_REF_DAY, including random uncertainty and USTAR filtering uncertainty HH: not produced DD: [SQRT(NEE_CUT_REF_DAY_RANDUNC^2 + ((NEE_CUT_84_DAY - NEE_CUT_16_DAY) / 2)^2)] for each day WW: [SQRT(NEE_CUT_REF_DAY_RANDUNC^2 + ((NEE_CUT_84_DAY - NEE_CUT_16_DAY) / 2)^2)] for each week MM: [SQRT(NEE_CUT_REF_DAY_RANDUNC^2 + ((NEE_CUT_84_DAY - NEE_CUT_16_DAY) / 2)^2)] for each month YY: [SQRT(NEE_CUT_REF_DAY_RANDUNC^2 + ((NEE_CUT_84_DAY - NEE_CUT_16_DAY) / 2)^2)] for each year NEE_VUT_REF_DAY_JOINTUNC (umolCO2 m-2 s-1) Joint uncertainty estimation for NEE_VUT_REF_DAY, including random uncertainty and USTAR filtering uncertainty HH: not produced DD: [SQRT(NEE_VUT_REF_DAY_RANDUNC^2 + ((NEE_VUT_84_DAY - NEE_VUT_16_DAY) / 2)^2)] for each day WW: [SQRT(NEE_VUT_REF_DAY_RANDUNC^2 + ((NEE_VUT_84_DAY - NEE_VUT_16_DAY) / 2)^2)] for each week MM: [SQRT(NEE_VUT_REF_DAY_RANDUNC^2 + ((NEE_VUT_84_DAY - NEE_VUT_16_DAY) / 2)^2)] for each month YY: [SQRT(NEE_VUT_REF_DAY_RANDUNC^2 + ((NEE_VUT_84_DAY - NEE_VUT_16_DAY) / 2)^2)] for each year NEE_CUT_USTAR50_NIGHT (umolCO2 m-2 s-1) Average nighttime NEE, from NEE_CUT_USTAR50 HH: not produced DD: average from half-hourly data (where NIGHT is 1) WW-YY: average from daily data NEE_VUT_USTAR50_NIGHT (umolCO2 m-2 s-1) Average nighttime NEE, from NEE_VUT_USTAR50 HH: not produced DD: average from half-hourly data (where NIGHT is 1) WW-YY: average from daily data NEE_CUT_USTAR50_NIGHT_SD (umolCO2 m-2 s-1) Standard Deviation of the nighttime NEE, from the NEE_CUT_USTAR50 HH: not produced DD: from half-hourly data (where NIGHT is 1) WW-YY: from daily data NEE_VUT_USTAR50_NIGHT_SD (umolCO2 m-2 s-1) Standard Deviation of the nighttime NEE, from the NEE_VUT_USTAR50 HH: not produced DD: from half-hourly data (where NIGHT is 1) WW-YY: from daily data NEE_CUT_USTAR50_NIGHT_QC (adimensional) Quality flag for NEE_CUT_USTAR50_NIGHT HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_USTAR50_NIGHT_QC (adimensional) Quality flag for NEE_VUT_USTAR50_NIGHT HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_CUT_USTAR50_NIGHT_RANDUNC (umolCO2 m-2 s-1) Random uncertainty of NEE_CUT_USTAR50_NIGHT, from the random uncertainty of the single nighttime half-hours HH: not produced DD-YY: from random uncertainty of individual half-hours where NIGHT is 1 (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used to calculate the nighttime aggregation in the day. NEE_VUT_USTAR50_NIGHT_RANDUNC (umolCO2 m-2 s-1) Random uncertainty of NEE_VUT_USTAR50_NIGHT, from the random uncertainty of the single nighttime half-hours HH: not produced DD-YY: from random uncertainty of individual half-hours where NIGHT is 1 (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used to calculate the nighttime aggregation in the day. NEE_CUT_USTAR50_NIGHT_JOINTUNC (umolCO2 m-2 s-1) Joint uncertainty estimation for NEE_CUT_USTAR50_NIGHT, including random uncertainty and USTAR filtering uncertainty HH: not produced DD: [SQRT(NEE_CUT_USTAR50_NIGHT_RANDUNC^2 + ((NEE_CUT_84_NIGHT - NEE_CUT_16_NIGHT) / 2)^2)] for each day WW: [SQRT(NEE_CUT_USTAR50_NIGHT_RANDUNC^2 + ((NEE_CUT_84_NIGHT - NEE_CUT_16_NIGHT) / 2)^2)] for each week MM: [SQRT(NEE_CUT_USTAR50_NIGHT_RANDUNC^2 + ((NEE_CUT_84_NIGHT - NEE_CUT_16_NIGHT) / 2)^2)] for each month YY: [SQRT(NEE_CUT_USTAR50_NIGHT_RANDUNC^2 + ((NEE_CUT_84_NIGHT - NEE_CUT_16_NIGHT) / 2)^2)] for each year NEE_VUT_USTAR50_NIGHT_JOINTUNC (umolCO2 m-2 s-1) Joint uncertainty estimation for NEE_VUT_USTAR50_NIGHT, including random uncertainty and USTAR filtering uncertainty HH: not produced DD: [SQRT(NEE_VUT_USTAR50_NIGHT_RANDUNC^2 + ((NEE_VUT_84_NIGHT - NEE_VUT_16_NIGHT) / 2)^2)] for each day WW: [SQRT(NEE_VUT_USTAR50_NIGHT_RANDUNC^2 + ((NEE_VUT_84_NIGHT - NEE_VUT_16_NIGHT) / 2)^2)] for each week MM: [SQRT(NEE_VUT_USTAR50_NIGHT_RANDUNC^2 + ((NEE_VUT_84_NIGHT - NEE_VUT_16_NIGHT) / 2)^2)] for each month YY: [SQRT(NEE_VUT_USTAR50_NIGHT_RANDUNC^2 + ((NEE_VUT_84_NIGHT - NEE_VUT_16_NIGHT) / 2)^2)] for each year NEE_CUT_USTAR50_DAY (umolCO2 m-2 s-1) Average daytime NEE, from NEE_CUT_USTAR50 HH: not produced DD: average from half-hourly data (where NIGHT is 0) WW-YY: average from daily data NEE_VUT_USTAR50_DAY (umolCO2 m-2 s-1) Average daytime NEE, from NEE_VUT_USTAR50 HH: not produced DD: average from half-hourly data (where NIGHT is 0) WW-YY: average from daily data NEE_CUT_USTAR50_DAY_SD (umolCO2 m-2 s-1) Standard Deviation of the daytime NEE, from the NEE_CUT_USTAR50 HH: not produced DD: from half-hourly data (where NIGHT is 0) WW-YY: from daily data NEE_VUT_USTAR50_DAY_SD (umolCO2 m-2 s-1) Standard Deviation of the daytime NEE, from the NEE_VUT_USTAR50 HH: not produced DD: from half-hourly data (where NIGHT is 0) WW-YY: from daily data NEE_CUT_USTAR50_DAY_QC (adimensional) Quality flag for NEE_CUT_USTAR50_DAY HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_USTAR50_DAY_QC (adimensional) Quality flag for NEE_VUT_USTAR50_DAY HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_CUT_USTAR50_DAY_RANDUNC (umolCO2 m-2 s-1) Random uncertainty of NEE_CUT_USTAR50_DAY, from the random uncertainty of the single daytime half-hours HH: not produced DD-YY: from random uncertainty of individual half-hours where NIGHT is 0 (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used to calculate the daytime aggregation in the day. NEE_VUT_USTAR50_DAY_RANDUNC (umolCO2 m-2 s-1) Random uncertainty of NEE_VUT_USTAR50_DAY, from the random uncertainty of the single daytime half-hours HH: not produced DD-YY: from random uncertainty of individual half-hours where NIGHT is 0 (rand(i)) = [SQRT(SUM(rand(i)^2)) / n], where n is the number of half-hours used to calculate the daytime aggregation in the day. NEE_CUT_USTAR50_DAY_JOINTUNC (umolCO2 m-2 s-1) Joint uncertainty estimation for NEE_CUT_USTAR50_DAY, including random uncertainty and USTAR filtering uncertainty HH: not produced DD: [SQRT(NEE_CUT_USTAR50_DAY_RANDUNC^2 + ((NEE_CUT_84_DAY - NEE_CUT_16_DAY) / 2)^2)] for each day WW: [SQRT(NEE_CUT_USTAR50_DAY_RANDUNC^2 + ((NEE_CUT_84_DAY - NEE_CUT_16_DAY) / 2)^2)] for each week MM: [SQRT(NEE_CUT_USTAR50_DAY_RANDUNC^2 + ((NEE_CUT_84_DAY - NEE_CUT_16_DAY) / 2)^2)] for each month YY: [SQRT(NEE_CUT_USTAR50_DAY_RANDUNC^2 + ((NEE_CUT_84_DAY - NEE_CUT_16_DAY) / 2)^2)] for each year NEE_VUT_USTAR50_DAY_JOINTUNC (umolCO2 m-2 s-1) Joint uncertainty estimation for NEE_VUT_USTAR50_DAY, including random uncertainty and USTAR filtering uncertainty HH: not produced DD: SQRT(NEE_VUT_USTAR50_DAY_RANDUNC^2 + ((NEE_VUT_84_DAY - NEE_VUT_16_DAY) / 2)^2) for each day WW: SQRT(NEE_VUT_USTAR50_DAY_RANDUNC^2 + ((NEE_VUT_84_DAY - NEE_VUT_16_DAY) / 2)^2) for each week MM: SQRT(NEE_VUT_USTAR50_DAY_RANDUNC^2 + ((NEE_VUT_84_DAY - NEE_VUT_16_DAY) / 2)^2) for each month YY: SQRT(NEE_VUT_USTAR50_DAY_RANDUNC^2 + ((NEE_VUT_84_DAY - NEE_VUT_16_DAY) / 2)^2) for each year NEE_CUT_XX_NIGHT (umolCO2 m-2 s-1) NEE CUT nighttime percentiles (indicated by XX) calculated from the 40 estimates for each period -- XX = 05, 16, 25, 50, 75, 84, 95 HH: not produced DD: XXth nighttime percentile from 40 daily NEE_CUT_XX_NIGHT WW: XXth nighttime percentile from 40 weekly NEE_CUT_XX_NIGHT MM: XXth nighttime percentile from 40 monthly NEE_CUT_XX_NIGHT YY: XXth nighttime percentile from 40 yearly NEE_CUT_XX_NIGHT NEE_VUT_XX_NIGHT (umolCO2 m-2 s-1) NEE VUT nighttime percentiles (indicated by XX) calculated from the 40 estimates for each period -- XX = 05, 16, 25, 50, 75, 84, 95 HH: not produced DD: XXth nighttime percentile from 40 daily NEE_VUT_XX_NIGHT WW: XXth nighttime percentile from 40 weekly NEE_VUT_XX_NIGHT MM: XXth nighttime percentile from 40 monthly NEE_VUT_XX_NIGHT YY: XXth nighttime percentile from 40 yearly NEE_VUT_XX_NIGHT NEE_CUT_XX_NIGHT_QC (adimensional) Quality flag for NEE_CUT_XX_NIGHT -- XX = 05, 16, 25, 50, 75, 84, 95 HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_XX_NIGHT_QC (adimensional) Quality flag for NEE_VUT_XX_NIGHT -- XX = 05, 16, 25, 50, 75, 84, 95 HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_CUT_XX_DAY (umolCO2 m-2 s-1) NEE CUT daytime percentiles (indicated by XX) calculated from the 40 estimates for each period -- XX = 05, 16, 25, 50, 75, 84, 95 HH: not produced DD: XXth daytime percentile from 40 daily NEE_CUT_XX_DAY WW: XXth daytime percentile from 40 weekly NEE_CUT_XX_DAY MM: XXth daytime percentile from 40 monthly NEE_CUT_XX_DAY YY: XXth daytime percentile from 40 yearly NEE_CUT_XX_DAY NEE_VUT_XX_DAY (umolCO2 m-2 s-1) NEE VUT daytime percentiles (indicated by XX) calculated from the 40 estimates for each period -- XX = 05, 16, 25, 50, 75, 84, 95 HH: not produced DD: XXth daytime percentile from 40 daily NEE_VUT_XX_DAY WW: XXth daytime percentile from 40 weekly NEE_VUT_XX_DAY MM: XXth daytime percentile from 40 monthly NEE_VUT_XX_DAY YY: XXth daytime percentile from 40 yearly NEE_VUT_XX_DAY NEE_CUT_XX_DAY_QC (adimensional) Quality flag for NEE_CUT_XX_DAY -- XX = 05, 16, 25, 50, 75, 84, 95 HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) NEE_VUT_XX_DAY_QC (adimensional) Quality flag for NEE_VUT_XX_DAY -- XX = 05, 16, 25, 50, 75, 84, 95 HH: not produced DD: fraction between 0-1, indicating percentage of measured and good quality gapfill data WW-YY: fraction between 0-1, indicating percentage of measured and good quality gapfill data (average from daily data) USTAR THRESHOLD ESTIMATION METHODS There are cases where not enough data are present to calculate a USTAR threshold (for both the Change Point detection method - CP - described in Barr et al. and the Moving Point detection method - MP - Reichstein et al. 2005) or where it is not possible to identify a clear change point (CP method only). This has an impact in the uncertainty estimation that is underestimated (less or no USTAR threshold values available) but it should be considered as a general indication of difficulties in the application of the USTAR filtering for the specific sites or years. Here below are reported the years and which of methods could not be applied to estimate an USTAR threshold in the present dataset: YEAR, Method NOT applied 1999,CP 2001,CP 2003,CP 2004,CP 2007,MP+CP 2008,MP+CP 2009,MP+CP 2010,CP 2011,CP USTAR THRESHOLDS USED IN NEE_CUT_USTAR50 and NEE_VUT_CUT_USTAR50: - NEE_CUT_USTAR50: u* threshold 0.396086 - NEE_VUT_USTAR50, YEAR 1999: u* threshold 0.480061 - NEE_VUT_USTAR50, YEAR 2000: u* threshold 0.488428 - NEE_VUT_USTAR50, YEAR 2001: u* threshold 0.462651 - NEE_VUT_USTAR50, YEAR 2002: u* threshold 0.434054 - NEE_VUT_USTAR50, YEAR 2003: u* threshold 0.419929 - NEE_VUT_USTAR50, YEAR 2004: u* threshold 0.41095 - NEE_VUT_USTAR50, YEAR 2005: u* threshold 0.36225 - NEE_VUT_USTAR50, YEAR 2006: u* threshold 0.337466 - NEE_VUT_USTAR50, YEAR 2007: u* threshold 0.276875 - NEE_VUT_USTAR50, YEAR 2008: u* threshold 0 - NEE_VUT_USTAR50, YEAR 2009: u* threshold 1.38207e+267 - NEE_VUT_USTAR50, YEAR 2010: u* threshold 0.2612 - NEE_VUT_USTAR50, YEAR 2011: u* threshold 0.30058 - NEE_VUT_USTAR50, YEAR 2012: u* threshold 0.331333 - NEE_VUT_USTAR50, YEAR 2013: u* threshold 0.351261 MODEL EFFICIENCY SELECTION FOR NEE_CUT_REF and NEE_VUT_REF: The reference NEE has been selected on the basis of the Model Efficiency. Starting from the 40 different NEE estimations (obtained filtering the data with 40 different USTAR thresholds) it has been calculated the Model Efficiency between each version and the others 39. The reference NEE has been selected as the one with higher Model Efficiency sum (so the most similar to the others 39). In this dataset have been selected as reference: NEE_CUT_REF = filtered using the USTAR percentile 58.75 (USTAR value: 0.4233) NEE_VUT_REF filtered on YEAR 1999 using the USTAR percentile 48.75 (USTAR value: 0.477056) NEE_VUT_REF filtered on YEAR 2000 using the USTAR percentile 48.75 (USTAR value: 0.48675) NEE_VUT_REF filtered on YEAR 2001 using the USTAR percentile 48.75 (USTAR value: 0.458313) NEE_VUT_REF filtered on YEAR 2002 using the USTAR percentile 48.75 (USTAR value: 0.429923) NEE_VUT_REF filtered on YEAR 2003 using the USTAR percentile 48.75 (USTAR value: 0.418482) NEE_VUT_REF filtered on YEAR 2004 using the USTAR percentile 48.75 (USTAR value: 0.408643) NEE_VUT_REF filtered on YEAR 2005 using the USTAR percentile 48.75 (USTAR value: 0.359429) NEE_VUT_REF filtered on YEAR 2006 using the USTAR percentile 48.75 (USTAR value: 0.334167) NEE_VUT_REF filtered on YEAR 2007 using the USTAR percentile 48.75 (USTAR value: 0.275875) NEE_VUT_REF filtered on YEAR 2008 using the USTAR percentile 48.75 (USTAR value: 0) NEE_VUT_REF filtered on YEAR 2009 using the USTAR percentile 48.75 (USTAR value: 0) NEE_VUT_REF filtered on YEAR 2010 using the USTAR percentile 48.75 (USTAR value: 0.261125) NEE_VUT_REF filtered on YEAR 2011 using the USTAR percentile 48.75 (USTAR value: 0.299907) NEE_VUT_REF filtered on YEAR 2012 using the USTAR percentile 48.75 (USTAR value: 0.32865) NEE_VUT_REF filtered on YEAR 2013 using the USTAR percentile 48.75 (USTAR value: 0.35025)