Data Preparation

GLOBALVIEW uses discrete and quasi-continuous measurements from fixed surface and tower sites and moving ship and aircraft sites. Discrete samples are collected in situ at weekly to monthly intervals and returned to the collaborating measurement laboratory for analysis. Quasi-continuous samples are measured in situ using systems located at the sampling location.

Each measurement record used to derive GLOBALVIEW-CO2 has been carefully edited and selected by the organization or institution contributing the observations. Observations used to derive the data product have been selected for baseline conditions to exclude samples influenced by local source and sinks . Baseline selection is site-specific. In most instances, selection is done by the PIs before submission to GLOBALVIEW and based on their knowledge of local conditions. The measurement records are accumulated at NOAA ESRL along with documentation and references. Wherever possible, NOAA ESRL attempts to reproduce the selected data set based on descriptions in the literature. Users are encouraged to review the literature for further details on baseline selection strategies.

Quasi-continuous data used to derive GLOBALVIEW have been preprocessed to produce a single value per day. Often this averaging process is performed by the collaborating laboratories using well-established methods that have been published in the literature. In some instances, the averaging is done at ESRL and in cooperation with the contributing laboratories. Table 1 summarizes the different averaging strategies. Users are encouraged to review the literature and contact the measurement labs directly for details about and access to the actual observations.

Data Integration

Comparison Experiments

A primary challenge for the GLOBALVIEW project and for the atmospheric trace gas measurement community is to ensure that measurements made using independent techniques can be integrated into larger cooperative data sets without introducing significant biases. This is a difficult challenge (see Data Integration). Several strategies exist to make this assessment including ongoing comparisons of 1) measurements of air from the same high-pressure cylinders; 2) measurements from glass flasks filled from the same high-pressure cylinders; 3) measurements from low-pressure cylinders decanted from high-pressure cylinders; 4) measurements of air from the same ambient samples; and 5) measurements from the same location (co-located) using different methodologies. Ongoing direct comparison of co-located atmospheric measurements is one of the more effective strategies [See, for example, Masarie et al., 2001]. Where ongoing comparison experiments of atmospheric measurements do not exist, we must rely on other less direct methods. In all instances, selected measurements are compared to other measurement records that are nearby in latitude as an additional assessment of potential calibration or sampling problems. The first step in assessing measurement comparability between independent records is to ensure data are reported on a consistent standard scale.

Standard Scale

The majority of laboratories contributing to the GLOBALVIEW-CO2 data product are members of the World Meteorological Organization (WMO) Global Atmosphere Watch (GAW) network. Data from the GAW network are reported relative to the WMO CO2 Mole Fraction Scale, which is maintained and propagated by the Central CO2 Laboratory (CCL). GAW laboratories are required to maintain direct traceability of their internal calibration scale to the CCL. A few laboratories contributing to the data product are not part of the WMO GAW program and thus provide data referenced to some other scale. This section describes ongoing efforts to assess the comparability of calibration scales and atmospheric observations.

1. The WMO CO2 Mole Fraction Scale

   The WMO Scale is based on regular determinations of the mole fraction of CO2 in dry air from a set of 15 primary standards using a high precision manometric system with NIST-traceable measurements to temperature, pressure and volume [Tans et al., 2003; Zhao et al., 1997]. Uncertainty of the WMO Scale is estimated to be ~0.06 µmol mol-1 (one sigma). Reproducibility of the determinations is about 0.03 µmol mol-1 (one sigma), based on repeated manometric analyses [Zhao and Tans, 2006]. The scale as defined by the primary standards (projected 30 year average lifetime) is subsequently propagated to a set of 9 secondary (transfer) standards (3-4 year average lifetime) using relative nondispersive infrared (NDIR) measurement techniques. NDIR measurement reproducibilityp is ~0.01 µmol mol-1. Propagation of the WMO Scale from the Primary cylinders to working standards, via intermediate Secondary standards maintained by the CCL, has a reproducibility of ~0.02 µmol mol-1 (one sigma). Cylinders are calibrated for other laboratories against the transfer standards using the NDIR methodology. The use of a calibration hierarchy enables the CCL to occasionally re-assign, when justification for such a change is strong, the value of a primary or secondary standard and propagate the change, in a straightforward manner, to all dependent calibrations.

   Recent History

   In 1995, the WMO designated NOAA ESRL as the Central CO2 Laboratory (CCL) responsible for the maintenance of the absolute WMO Mole Fraction Scale for carbon dioxide. Before that time, the scale had been maintained by the Scripps Institution of Oceanography (SIO).

   In 1990, ESRL prepared 15 CO2-in-air reference gas mixtures in large aluminum high pressure cylinders for use as primary standards, ranging in CO2 mole fraction from approximately 250 to 520 µmol mol-1. These cylinders were calibrated four times at SIO by the NDIR method from mid-1991 to 1999. In 1996, ESRL began making absolute manometric determinations of its 15 "primary" standards. Between 1996 and 2001, values assigned to the 15 primaries were based on both SIO NDIR measurements and ESRL manometric determinations. Starting in 2002, the values assigned to the primaries have been based on manometric determinations by ESRL alone.

   Revisions to calibrations provided before 2005 by the CCL have been made. This is mainly due to revisions of the calibrations performed by Scripps between 1991 and 1999. Until 1996 the assigned values of the primary standards were based entirely on the infrared calibrations by Scripps. The average of all assigned values to the primaries increased by 0.16 µmol mol-1 from 1993 to 2002. Since then the average of all assigned values of the primaries has decreased by 0.01 µmol mol-1.

   The CO2 Mole Fraction Scale is defined by a polynomial curve fit to the Primary Standards. This is done to smooth out the uncertainty of assigned values to individual Primary Standards caused by the imprecision of the absolute calibrations, which is 0.03 ppm (see above). In September 2005, the WMO scale was revised, and a quadratic curve fit was used. After another set of calibrations of the Primaries in 2006-2007, the individual Primaries were revised by only minor amounts, up to 0.01 ppm. However, in defining the revised scale for 2007, we chose to use a cubic polynomial for the curve fit, which led to mole fraction-dependent differences between WMO-X2007 and WMO-X2005 between -0.03 and 0.03 ppm in the range of ambient air. All laboratories who have had cylinders calibrated by the CO2 CCL should have received revised calibration assignments based on the new scale. If you would like to receive revised values based on the new scale please contact Duane Kitzis (NOAA/ESRL).

2. Traceability to the WMO Scale

   Not all data contributed to the Cooperative Atmospheric Data Integration Project for CO2 are directly traceable to the WMO Mole Fraction Scale. A few laboratories have never had their standard gases calibrated by the CCL and report CO2 measurements relative to some other scale (see Table 2). Measurements from these laboratories are not directly traceable to the WMO Mole Fraction Scale. Several other laboratories have, at one time, had their standards calibrated by the CCL but have not maintained a routine recalibration schedule. Because the mole fraction of CO2 contained in high-pressure cylinders can potentially change with time due to CO2 adsorption or production within the cylinder or regulator, or through other effects, a laboratory's internal scale may potentially change with time relative to the WMO scale, which itself is anchored through absolute manometric determinations. Without routine recalibration by the CCL to reestablish direct traceability to the WMO-X2007 scale, laboratories contribute CO2 data that are no longer directly traceable to the WMO scale. Please note that recent calibration with the CCL does not necessarily imply measurements are on the most current WMO Mole Fraction Scale.

   Table 2. Traceability to the WMO scale based on calibration by the CCL.

   LABORATORY [lab #]	RECENT CAL. DATES1	# of CYLINDERS3	REPORTED SCALE
   ESRL [01]	2010 05 / 2009 12	20 (recal) / 20 (recal)	WMO
   CSIRO [02]	2003 11 / 2002 12	10 (recal) / 11 (cal)	WMO4
   NCAR [03]	2011 04 / 2006 04	6 (recal) / 6 (cal)	WMO4
   SIO [04]	--	--	Scripps 08A
   EC [06]	2009 06 / 2008 09	2 (cal) / 8 (recal)	WMO
   LSCE [11]	2008 01 / 2007 07	1 (cal) / 7 (recal)	WMO4
   NIWA [15]	2010 04 / 2008 10	2 (cal) / 2 (cal)	WMO
   IMS [17]	1998 10	5 (cal)	WMO4
   JMA [19]	2011 01/ 2008 10	14 (cal) / 14 (recal)	WMO
   NIES [20]	2006 09 / 2002 12	5 (recal) / 2 (cal)	NIES09
   UBA/UHEI-IUP [23]	2007 12 / 2006 04	6 (cal) / 6 (recal)	WMO4
   SNU [24]	2004 09 / 2003 06	4 (cal) / 5 (cal)	2008A SIO
   AEMET [27]	2008 09 / 2006 03	2 (cal) / 8 (cal)	WMO
   ENEA [28]	2008 12 / 2008 02	4 (recal) / 4 (cal)	WMO
   PNRA/DNA [29]	2001 07 / 2000 12	3 (cal) / 3 (cal)	WMO4
   FMI [30]	2007 05 / 2007 02	2 (cal) / 6 (cal)	WMO
   CAMS [33]	2009 04 / 2008 10	6 (cal) / 6 (cal)	WMO
   HMS [35]	2010 10 / 2007 11	3 (cal) / 6 (cal)	WMO
   SAWS [36]	2009 04 / 2007 05	2 (cal) / 2 (recal)	WMO
   MPI-BGC [45]	2009 04 / 2009 02	1 (cal) / 2 (cal)	WMO

   1Only the two most recent calibration events are shown.
   2Calibration made at ESRL relative to the ESRL secondary standards.
   3Initial (cal)ibration and subsequent (recal)ibration by the CCL (NOAA) are specified.
   4Traceability to the WMO Mole Fraction Scale has lapsed. A recalibration schedule of every 3 years is thought to be the minimum frequency for maintaining traceability to the WMO scale.
   5Insufficient number of cylinders calibrated to properly link laboratory internal scale to WMO Mole Fraction Scale. The minimum number of standards required to establish traceability to the WMO Mole Fraction Scale is three.
   6Internal scale is indirectly linked to WMO Mole Fraction Scale.

3. Comparisons of Standard Scales

   In an attempt to assess differences in standard scales among organizations making CO2 measurements, laboratories contributing to GLOBALVIEW-CO2 have participated in recent interlaboratory intercomparison or round robin (RR) experiments endorsed by the WMO and IAEA. Based on preliminary results from the 2002-2006 intercomparison of standard gases, the majority of participating laboratories agreed to within 0.1 µmol mol-1 [see Table 3]. The preliminary results were kindly provided by Dr. Lingxi Zhou (CAMS) [Zhou et al., in preparation]. Final results were not available at the time of this update. Results from previous RR experiments are also available [Peterson et al., 1999].

   Table 3. 2002-2006 Round Robin Results [per comm. Lingxi Zhou, 2008-08-20]: Differences from NOAA ESRL (Lab minus ESRL).

   LABORATORY [lab #]	MEASUREMENT DATE	DIFFERENCE from NOAA (µmol mol-1)
   Nominal Tank Concentration		355	365	385
   CSIRO [02]	2005 10	-0.01	-0.03	-0.08
   NCAR [03]	2006 06	0.07	-0.04	-0.04
   SIO [04]	2005 06	0.10	0.02	0.02
   EC [06]	2005 05	0.06	-0.05	-0.06
   MRI [10]	2003 07	-0.16	-0.16	-0.08
   LSCE [11]	2002 12	0.10	0.03	0.05
   NIWA [15]	2005 05	-0.08	-0.08	-0.09
   IMS [17]	2002 10	0.08	0.02	-0.03
   JMA [19]	2004 01	0.13	0.00	-0.02
   NIES [20]	2003 04	-0.10	-0.15	-0.14
   UBA/UHEI-IUP [23]	2002 10	-0.01	-0.06	-0.06
   SNU1 [24]	--	--	--	--
   AEMET1 [27]	--	--	--	--
   ENEA [28]	2002 11	0.05	-0.15	-0.26
   PNRA/DNA1 [28]	--	--	--	--
   FMI [30]	2003 01	-0.02	-0.04	-0.14
   CAMS [33]	2004 08	-0.03	-0.20	0.02
   HMS [35]	2003 02	0.06	-0.21	-0.06
   SAWS [36]	2005 12	-0.02	-0.09	-0.18
   MPI-BGC [45]	2003 12	0.04	0.02	-0.02

   1Did not participate in the 2002-2006 RR experiment.

   Based on available comparison information, we estimate that data used to derive GLOBALVIEW-CO2 are comparable to within 0.2 µmol mol-1. At present, the Cooperative Atmospheric Data Integration Project for Carbon Dioxide has made no standard scale adjustments to any of the measurement records integrated into GLOBALVIEW-CO2. Records that appear to be affected by a serious scale discrepancy have been omitted at this time.