Under the UK-focused Greenhouse gAs and Uk and Global
Emissions (GAUGE) project, two new tall tower greenhouse gas (GHG)
observation sites were established in the 2013/2014 Northern Hemispheric
winter. These sites, located at existing telecommunications towers, utilized
a combination of cavity ring-down spectroscopy (CRDS) and gas chromatography
(GC) to measure key GHGs (
The adverse effects of anthropogenically driven increases in greenhouse gas
concentrations on global temperatures and climate have been well established
(IPCC, 2013). Governmental efforts to curb these
emissions include the UK 2008 Climate Change Act, which will soon be amended
to require the UK to produce net-zero emissions by 2050 (Parliament of
the United Kingdom, 2008, Chapter 27). This in turn motivated the Greenhouse
gAs Uk and Global Emissions (GAUGE) project, which aimed to better quantify
the UK carbon dioxide (
This paper describes the establishment of two new UK GHG tall tower (TT)
sites funded under the GAUGE project. Here we provide an analysis of the
observations made at the sites and investigate the error associated with
empirical instrument-specific water correction algorithms and the
Nafion®-based sample drying approach used at these TT sites.
A further paper, currently in preparation, will discuss the integration of
these new sites with the existing UK Deriving Emissions linked to Climate
Change (DECC) network (Stanley et
al., 2018) funded by the UK Department of Business, Energy and Industrial
Strategy (BEIS) and provide a full uncertainty analysis for data collected
at all the DECC–GAUGE sites. A second companion paper, also in preparation,
will discuss the integration and inter-calibration of all the
Like the UK DECC network, the new sites, Bilsdale (BSD) and Heathfield (HFD), are equipped with a combination of cavity ring-down spectrometer (CRDS) and gas chromatograph (GC) instrumentation (Stanley et al., 2018). These instruments, along with the associated calibration gases (linked to WMO calibration scales) and automated sampling systems are located at the base of telecommunication towers within the UK. Further details of the sites and instruments used along with a description of the data collected to date are provided in the subsequent sections.
The precision, stability, relative autonomy and robustness of CRDS instrumentation has led to a rapid increase in it's deployment in global, continental and regional GHG monitoring networks including the GAUGE network, the European Integrated Carbon Observing System (ICOS) (Yver Kwok et al., 2015) and the Indianapolis Flux Experiment (INFLUX) (Turnbull et al., 2015). These instruments also claim the advantage of being able to measure undried (“wet”) air samples, which are then post-corrected to “dry” values using an inbuilt algorithm (Rella, 2010).
Initially, it was hoped that the inbuilt water correction would remove the need for sample drying, inherent in most other methods (e.g. FTIR or NDIR) but subsequent studies questioned its stability over time and between instruments (Yver Kwok et al., 2015; Chen et al., 2010; Winderlich et al., 2010). In response to this, researchers have typically developed their own water corrections or have returned to sample drying in order to minimize the effect (Welp et al., 2013; Winderlich et al., 2010; Schibig et al., 2015; Rella et al., 2013). As such the examination of any errors or biases induced by drying and water correction methods is essential for fully quantifying the uncertainty of CRDS measurements.
For ease of servicing, the CRDS instrumentation at GAUGE and UK DECC Network
sites was initially deployed using an identical drying method to that of the
co-located GC instrumentation. This method relied on drying the sample with
a Nafion® water-permeable membrane in combination with dry
zero air as a counter purge gas. Here, due to the moisture gradient between
the sample and the counter purge, the water passed from the wet sample
through the membrane to the dry counter purge. Drying in this manner has a
history of successful application for the measurements of halocarbons
(Foulger and Simmonds, 1979),
A study by Welp et al. (2013) examined this issue and
concluded that the leakage was small and well within the WMO compatibility
guidelines. However, the drying approach used by Welp et al. (2013) is not directly comparable to that of the
GAUGE sites as they used dry sample gas as the counter purge rather than
zero air. That study also only examined two water contents (0 % or 2 %
As such, this paper also aims to quantify the magnitude of
Nafion®
Two new tall tower sites, Heathfield (HFD; 50.977
The Heathfield tower is located in rural East Sussex, 20 km from the coast and 157.3 m above sea level. The closest large conurbation (Royal Tunbridge Wells) is located 17 km NNE from the tower. The area surrounding the tower is > 90 % woodland and agricultural area with some residential (0.7 %) and light industrial areas (0.3 %) (East Sussex in figures, 2006). Notable local industry includes a large horticultural nursery located only 200 m north of the tower.
Locations of the GAUGE Bilsdale (BSD) and Heathfield (HFD) sites, shown in black, and the UK DECC Mace Head (MHD), Ridge Hill (RGL), Tacolneston (TAC) and Angus (TTA) sites, shown in grey.
Bilsdale is a remote moorland plateau site within the North York Moors National Park. The base of the tower is located 379.1 m a.s.l. It is 25 km NNW of Middlesbrough (the closest large urban area) and 30 km from the coast. The tower is situated in a predominantly rural area, including moorland, woodland, forest and farmland (North York Moors National Park Authority, 2012; Chris Blandford Associates, 2011).
Inverted stainless steel intake cups were mounted at 42, 108 and 248 m a.g.l. (metres above ground level) on the BSD tower and 50 and 100 m a.g.l.
at HFD. Air was pulled through the intake cups via
Both sites are equipped with a CRDS (G2401 Picarro Inc., USA, CFKADS2094 and
CFKADS2075 deployed at Bilsdale and Heathfield respectively) taking high-frequency (0.4 Hz)
The sample lines and calibration and standard gas cylinders are linked to two
multiport valves (EUTA-CSD10MWEPH, VICI AG International, Valco Instruments Co. Inc.,
Switzerland), one for the CRDS and a second for the GC–ECD; the output of
each valve is connected to the intakes of the instruments. Filters (7
The automated switching of valves and control of GC–ECD temperatures and
flows, as well as logging the data and a range of other key parameters
(flows, pressures, temperatures), are achieved using custom Linux-based
software (GCWerks,
All samples measured on the GC–ECD (air, standards and calibration) are
dried using a Nafion® permeation dryer (MD-050-72S-1,
Perma Pure, USA) prior to analysis. The counter purge gas for the dryer is
generated from compressed room air. The counter purge is dried to < 0.005 %
A generalized schematic showing the initial Bilsdale and Heathfield site set-up of the cavity ring-down spectrometer (CRDS) and the gas chromatograph–electron capture detector (GC–ECD) including the dry gas generator (TOC) and back pressure regulator (BP). Note that Bilsdale has three inlets, while Heathfield has only two as shown here. The Nafion® drying system located downstream of the CRDS multiport valve was removed at both sites in 2015. Black arrows and lines show the direction of sample, standard and calibration gas flow. Grey dashed lines and arrows show the flow path of the Nafion® counter purge gas.
In an attempt to minimize the water correction required for dry mole
fraction CRDS measurements, CRDS samples were initially dried using a
Nafion® dryer in an identical manner to those of the GC–ECD. When
functioning correctly this drying method resulted in air samples with water
mole fractions between 0.05 % and 0.2 %
Due to concerns that the mole fraction gradient between the sample and the
Nafion® counter purge might lead to
The drying technique implemented in this study uses a
Nafion® dryer which relies on a dry counter purge air stream. Measurements of these air streams were made at BSD, HFD and the University of Bristol (UoB) laboratory using the respective sites' CRDS instrumentation. All counter purge streams showed mole fractions of
Motivated by the possibility of
Data collected in the first 5 min immediately following the
injection, the typical line equilibration period, were excluded from the
fit. This avoids using data adversely affected by the effect of rapid
changes in
A water correction was then determined from a fit between the mean
wet
Instrument-specific water corrections for the Bilsdale (BSD),
Heathfield (HFD) and University of Bristol (UoB) CRDS instruments. The
parameters shown are the mean
The fit was conducted using orthogonal distance regression weighted by both
the minute-mean standard deviation of the
As Picarro analysers are not calibrated for
The typical temporal stability and mole fraction dependence of the CRDS
water correction were examined using a laboratory-based CRDS (G2301, Picarro
Inc., USA;
The accuracy of the CRDS water correction determined through the water
droplet test, as described in Sect. 2.3.4, was assessed through a series
of simple dew point generator (DPG; Li-cor LI-610 portable dew point
generator, USA) experiments. Here, dry air from four cylinders with varying
In brief, the output of the cylinder regulator was plumbed to the input of
the DPG. A T-piece connected prior to the DPG input vented any excess gas
via a flowmeter (F1, Fig. 3a) ensuring that the DPG input remained at
close to ambient atmospheric pressure throughout the experiment. The output
of the DPG passed through a second T-piece with the overflow outlet also
connected to a flowmeter (F2) to ensure that the CRDS input pressure
remained near ambient. A third flowmeter (F3) was placed on the outflow of
the Nafion counter purge. Flowmeters F1 and F2 had a range of 0.1–1 L min
The cryogenic water trap consisted of a coil of
The cylinders used during the dew point generator CRDS water
correction, Nafion® counter purge and UoB instrument-specific
water tests. Most measurements were made in-house and only corrected for
linear drift against a standard calibrated at WCC-EMPA, Dübendorf,
Switzerland, and hence are simply indicative of the expected mole fractions.
Those marked
A schematic of the humidification system used in the
The experiment was conducted in a temperature-controlled laboratory at 19
Multiple measurement blocks of each sample treatment were conducted after a lengthy initial stabilization period. This period allowed for the establishment of equilibrium between the water in the condenser block of the DPG and the sample gas and lasted at least 2 h (sometimes up to 5 h). The treatment blocks varied in length depending on the time required for the concentration to stabilize. At least 15 min of data was collected after the concentration stabilized.
It is important to note that the DPG was not calibrated but the
Flow rates, cylinder pressure, chamber temperature and
An experiment was designed to observe gas exchange across the
Nafion® membrane by measuring the counter purge gas before
(CP
In this experiment, a system (Fig. 3b) was constructed allowing the
controlled humidification, using a DPG, of two high-pressure cylinders, one
of dry near-ambient and one above-ambient
Other differences include the placement of the Nafion®, water
trap and the addition of a multiport valve. In this experiment the
humidified cylinder air exiting the third T-piece is split with half passing
through the Nafion® before reaching a four-port two-position valve,
V1 (EUDA-2C6UWEPH, VICI AG International, Valco Instruments Co. Inc., Switzerland, actually a
six-port valve configured as a four-port valve). The other half bypassed the
Nafion® and connected directly to V1. The first output of V1
connected to a multiport valve (EUTA-CSD10MWEPH, VICI AG International, Valco Instruments Co. Inc., Switzerland) while the second connected to a pump (Picarro
vacuum pump S/N PB2K966-A) set to a flow rate matching that of the CRDS (0.3 L min
Counter purge air, both before (CP
As the reliability of CRDS water correction was also under investigation, it
was important to isolate the effect of the Nafion® from that
of the CRDS water correction. To do this the experiment was conducted in
three stages (see Fig. S5). Firstly, the
All
The counter purge measurements made during the humidification experiments
represent a combination of effects.
Hence the difference between the mean of CP
In order to remove any valve switching or line equilibration effects the
first 5 min of data of each sample period was discarded and the mean of the
final 15 min period of each sample type at each dew point was calculated.
The uncertainty of this mean was determined as the 95 % confidence
interval based on the larger of either the standard deviation of the minute
means or average of the standard deviations of the minute means. Examples of
the raw data collected during the experiment are given in Fig. S5. As the
experiment was subject to a small temporal drift, the mean CP
While experiments 2.3.5, 2.3.6 and 2.3.7 were designed to isolate key processes, other possible sources of error or bias may exist. These include adsorption and desorption effects within the regulator and walls of the tubing, gas solubility within the condenser of the dew point generator, and instrumental drift.
Regulator and tubing adsorption and desorption effects have been previously
examined by Christoph Zellweger and Martin Steinbacher (2017, personal communication).
They found that for Parker Veriflo regulators, as used in this
experiment, the effects can be quite large, up to 0.5
As discussed earlier, a lengthy equilibration period was used at the start
of each DPG run and following any change in DPG set point. This was to
account for the dissolution of sample gas, in particular
Although small, any time-dependent drifts were accounted for by temporally interpolating between each block of data. Also key to the design of this experiment is the examination of differences between two very similar mole fractions rather than absolute mole fractions. As such, any systematic errors that might drive a systematic offset cancel out and any mole-fraction-dependent biases are minimized.
CRDS calibration and standard cylinder mole fractions and usage
start dates for the Heathfield (HFD) and Bilsdale (BDS) sites. Where
available the mole fractions measured prior to and after deployment are
given. Reported mole fractions from the WCC-EMPA, Dübendorf, Switzerland,
are given as mean
Calibration procedures for both the CRDS and GC–ECD are as described in
detail in Stanley et al. (2018). In
brief, CRDS measurements are calibrated using a close-to-ambient standard
(“working tank”) and a set of three calibration cylinders, which span the
typical ambient range (Table 3). Only a small number of elevated
observations, < 0.4 % of the
Assigning mole fractions to values outside the range of the calibration
suite will increase the error. The magnitude of this error will depend on
the magnitude of the mole fraction difference between the closest
calibration cylinder and the sample. This error has been estimated using
measurements made at the Heathfield site of cylinders of known CO mole
fractions, 6 and 57 nmol mol
Daily 20 min long measurements of the ambient standard are used to
account for any linear drift, while monthly measurements of the calibration
suite are used to characterize the nonlinear instrumental response. This
calibration procedure is controlled by the GCWerks software and allows near-real-time examination of calibrated data. During the period that the
Nafion® drying system was used these standards were partially
humidified as they passed through the wet Nafion® dryer. The
level of humidification is dependent on that of the air samples measured
prior to the standard. The moisture content of the standard closely tracks
that of the air samples with variations in the humidity of the samples
clearly reproduced in the standard (Fig. S1). However, the moisture
content of the standard is generally slightly lower. On average the standard
has a mean moisture content 88 % that of the average of the 30 min of
air sample either side of the standard (on average 0.02 %
In contrast, due to the time taken to take replicate measurements of the
calibration cylinders (at least 240 min) only the first 20 min
measurement block of each calibration cylinder is significantly humidified.
with the water content of the calibration measurement dropping rapidly to
< 0.02 %
All CRDS standards and calibration gases are composed of natural air, some
spiked or diluted with scrubbed natural air (TOC gas generator, model no.
78-40-220, Parker Balston, USA), to achieve the required concentrations of
GC–ECD measurements are made relative to a natural air standard of known
GC–ECD standard cylinder mole fractions and usage start dates.
GC–ECD and CRDS standards and calibration cylinders were, where possible,
calibrated both before and after deployment at the sites. If these two
measurements agreed then a mean mole fraction was used, otherwise a linearly
drift-corrected mole fraction was used. The CRDS cylinders were calibrated
through WMO linked calibration centres (either WCC-EMPA, Dübendorf,
Switzerland, or GasLab MPI-BGC, MPI, Jena, Germany). This ties the ambient
measurements to the WMO
The short-term (1 min) precision of the CRDS data was determined as the
mean of the standard deviations of the 1 min mean data. This was
calculated from measurements of the standard cylinder and the calibration
suite allowing the relationship between
The mean absolute short-term precision for all cylinders was consistent
between the two sites across all three gases. At BSD the short-term
precision was 0.024
The long-term reproducibility of a 20 min mean was estimated as the mean
standard deviation of the daily 20 min measurements of the standard
cylinders used at each site. A total of eight standard cylinders have been used
in succession at the two sites with the usage periods and
Repeatability of individual injections on the GC instruments was calculated
as the standard deviation of the hourly standard injection. These were found
to be < 0.3 nmol mol
A three-stage data flagging and quality control system was used for the HFD
and BSD data. Initially, automated flags based on the stability of key
parameters including cell temperature and pressure and instrument cycle
time (the time taken to collect and process each measurement) were applied.
Here, data with a cycle time > 8 s were filtered out along
with any data with cell temperature outside the range of
The long-term trend in mole fraction at each site was estimated as the mean linear trend in the minute-mean data over the period 2014–2017, inclusive. Seasonal and diurnal trends in the data were assessed using monthly and hour-of-day means of trimmed detrended minute-mean data developed using the Python numpy package. Here the long-term trend was removed by using a least-squares fit between a quadratic and the minute-mean data. The data for each hour (or month for the seasonal plots) were trimmed following the approach of Satar et al. (2016), who removed the highest and lowest 5 % of all data points.
The minute-mean
Minute-mean
HFD is located in southern England, just south of London (Fig. 1). Here,
high
Both sites show a clear relationship between
Minute-mean
Interestingly, Winderlich et al. (2010) suggest that their ability to observe gradients on an hourly timeframe is only revealed due to their use of buffer volumes and fast switching (every 3 min). In contrast, the measurements presented here, made without buffer volumes and at a much lower switching rate, were also able to identify gradients between the heights (Fig. S7). This suggests that the use of buffer volumes and fast switching is not required in order to observe these trends.
The timings and magnitude of the HFD and BSD seasonal cycles are similar,
with
Mean diurnal cycle by season of detrended hourly mean values for
As with the seasonal cycle, the shape of the
Although there is a very large range in the minute-mean
Mean seasonal cycle of detrended hourly mean values for
Of the five gases measured at HFD and BSD, CO is the only gas to show a
decrease in mole fraction between 2013 and 2017, roughly
While the range of minute-mean CO mole fractions was significantly larger at
BSD, 63 to 9500 nmol mol
The range of
The
A previous study, Nevison et al. (2011)
examined the monthly mean
Clear diurnal cycles in
The long-term trend in the SF
The annual instrument-specific water corrections, determined through regular
droplet tests, are typically very similar at each site, often within the 95 % confidence interval of the triplicate runs (Table 1), suggesting that
the corrections are fairly stable between years and instruments. The
residuals of the corrections are generally quite small, with 25th and
75th quartiles of
While instrument-specific CO water corrections were calculated, the large
minute-mean variability inherent in the G2401 CO measurements (> 4 nmol mol
Plots of the residuals typically show a common pattern, with the residual of
zero at 0 %
Reum et al. (2019) previously identified this
pattern in water correction residuals and linked it to a sensitivity of the
cavity pressure sensor at low water vapour mole fractions. They proposed an
alternative fitting function incorporating the “pressure bend”, although
they do not recommend using this fit for data collected during the droplet
test due to the paucity of stable data typically obtained between 0.02 %
It is also important to note that the magnitude of the dip observed by
Reum et al. (2019) in their controlled water tests,
The poor performance of the CRDS pressure sensor at low
In contrast, without the humidification of the standard the error when
sampling with Nafion® drying may well be significant. It is
difficult to quantify this error, as it will vary with sample water content
and the sensitivity of the individual instrument's pressure sensor to low
The sample mole fraction dependence of the CRDS water correction was
examined by conducting water droplet tests using dry cylinders of above- and
below-ambient mole fractions (Sect. 2.3.5). Specific above- and below-ambient water corrections were calculated based on these data sets (Table 1
and Fig. S9). If the water correction was independent of sample mole
fraction, then the residuals should be identical for both correction types.
Although the above- and below-ambient residual plots are similar, they do
differ slightly with the residual of the above mole fraction sample, becoming
more positive at higher
The change with water in the difference in
The change in the difference between dry mole fractions calculated using the
earliest instrument-specific water correction and subsequent water
corrections for each instrument with water concentration is shown in Fig. 8a and b. For a typical air sample (1.5 %
The
A comparison of the individual daily and weekly tests, Figs. 8c and d and
10e and f, conducted using the UoB instrument, show the daily tests to be
far more similar than the weekly tests. That is, the variability over the
3-month period of the weekly test is much larger than that of the 5 d
period of the daily test. However, the variability of the weekly tests is
similar to that of the annual tests, Fig. 8a and b, suggesting that,
within the bounds of the data typically observed at the BSD and HFD sites,
the use of annually derived instrument-specific water corrections is
sufficient. This may not be the case at sites with higher levels of humidity
and
The change in the CRDS water correction with sample
Although the UoB CRDS was not deployed in the field, we expect the results of
the DPG tests to be typical of most Picarro G2401
Change in the counter purge in (CP
Unlike the ambient and below-ambient samples, the CRDS water correction
error of the above-ambient sample, UoB-04, exceeded the WMO internal
reproducibility guidelines for both
The full range of
Nafion® membranes, when combined with a dry counter purge gas
stream, can be used to effectively dry air samples. This drying process is
driven by the moisture gradient between the wet sample and the dry
counter purge. In a similar manner, as long as the membrane is permeable to
the gas, a sample-to-counter purge gradient in any other trace gas species
will also drive exchange. In an effort to quantify the magnitude of
The counter purge experiments conducted with both the ambient (UoB-15) and
above-ambient (UoB-16) mole fraction cylinders show identical changes in
Estimates of the maximum error associated with the measurement of
ambient
The
The
The newly established Bilsdale and Heathfield tall tower measurement
stations provide important new data sets of GHG observations. These
high-precision continuous in situ measurements show clear long-term
increases in baseline
The two drying methods implemented at Bilsdale and Heathfield –
Nafion® drying with an empirical water correction and an
annual empirical water correction without drying – have a number of
practical and scientific advantages and disadvantages. The
Nafion® drying method, once installed and running correctly
can provide reliable drying to between 0.05 %
As shown in Table 5, the systematic errors associated with the
Nafion® drying method, as applied at these two sites, were
small, < 0.01
By comparison the annual CRDS empirical water correction has a narrower
optimum range with minimal systematic errors only at water contents very
near 0 % and between 0.5 % and 2.5 %
This weakness in the CRDS water correction also has notable implications for
sample drying. Namely, while sample drying may not be an inherent source of
bias, the partial drying of the sample puts it within the range of peak
error in the CRDS water correction (0.05 % to 0.5 %
Considering the relatively narrow humidity range observed at Bilsdale and
Heathfield, with no observations > 2.4 %
While these errors are significant relative to the WMO internal
reproducibility goals, they are for the majority of observations smaller than
the extended WMO measurement compatibility goals (
Future improvements to the Bilsdale and Heathfield records include the addition of target tanks at the sites. Although the use of target tanks does not directly influence measurement uncertainty, they allow independent long-term monitoring of instrument performance and are a useful tool for assessing measurement uncertainty. The development of a full uncertainty analysis incorporating such target tank measurements along with an assessment of the calibration strategy, instrumental, water correction, and sampling errors and errors induced by the isotopic composition of the calibration gases is also planned. Further work to fully characterize the humidity-dependent error in the water correction of each instrument, like that of Reum et al. (2018), possibly using a piecewise post hoc correction, would also be beneficial in an effort to reduce the estimated error associated with the observations.
The BSD and HFD
The supplement related to this article is available online at:
AS led the paper with contributions from SO, KS, DY, AM, ML, CR and TA. AS designed and performed all drying related experiments and analysed the data. AS, SO, KS, DY, CR and TA shared responsibility for installing, maintaining and managing the data collected at the HFD and BSD sites. AM and ML contributed atmospheric transport modelling outputs and analysis.
The authors declare that they have no conflict of interest.
This article is part of the special issue “Greenhouse gAs Uk and Global Emissions (GAUGE) project (ACP/AMT inter-journal SI)”. It is not associated with a conference.
The National Physical Laboratory (NPL) took responsibility for the Heathfield site on 30 September 2017 through the UK's Department for Business, Energy and Industrial Strategy as part of the National Measurement System programme.
The authors would also like to acknowledge the support of Carole Helfter and Neil Mullinger from the NERC Centre for Ecology and Hydrology (CEH), Edinburgh, Scotland, who helped to maintain and run the Bilsdale site. Lastly the authors would like to thank Joseph Pitt from the University of Manchester for the use of the dew point generator used during a series of preliminary studies.
This study was funded under the NERC Greenhouse Gas Emissions and Feedbacks programme as part of the Greenhouse gAs UK and Global Emissions (GAUGE) area grant number NE/K002449/1NERC. This grant also covered the establishment and early running costs of the stations. Operating costs of the Bilsdale site after 17 September 2016 were funded by the UK Department of Business, Energy and Industrial Strategy (formerly the Department of Energy and Climate Change) through contract TRN1028/06/2015.
This paper was edited by Dominik Brunner and reviewed by two anonymous referees.