Highly accurate water vapor measurements are indispensable for understanding
a variety of scientific questions as well as industrial processes. While in
metrology water vapor concentrations can be defined, generated, and measured
with relative uncertainties in the single percentage range, field-deployable
airborne instruments deviate even under quasistatic laboratory conditions up
to 10–20 %. The novel SEALDH-II hygrometer, a calibration-free, tuneable
diode laser spectrometer, bridges this gap by implementing a new holistic
concept to achieve higher accuracy levels in the field. We present in this paper the absolute validation of
SEALDH-II at a traceable humidity generator during 23 days of permanent
operation at 15 different H
Water vapor affects, like no other substance, nearly all atmospheric
processes (Ludlam, 1980; Möller et al., 2011;
Ravishankara, 2012). Water vapor not only represents a large direct feedback
to global warming when forming clouds but also plays a major role in
atmospheric chemistry (Held and Soden, 2000; Kiehl
and Trenberth, 1997). Changes in the water distribution, as vapor or in
condensed phases (e.g., in clouds), have a large impact on the radiation
balance of the atmosphere. This is the reason that water vapor is often
mentioned as the most important greenhouse gas and one of the most important
parameters in climate research (Maycock et al., 2011).
Water vapor measurements are often needed for other in situ atmospheric
analyzers to correct for their water vapor cross interference. The high
(spatial and temporal) variability of atmospheric water vapor, its large
dynamic range (typically 3–40 000 ppmv SEALDH-II native unit
for H
In particular for field weather stations, water vapor analyzers often are
seen as part of the standard instrumentation in atmospheric research. This
seems reasonable due to several reasons: slow H
As soon as hygrometers have to be deployed in harsh environments (e.g., on
airborne platforms), this situation changes entirely: the ambient gas
pressure (10–1000 hPa) and gas temperature (
Over the last decades, numerous hygrometers were developed and deployed on aircraft (Busen and Buck, 1995; Cerni, 1994; Desjardins et al., 1989; Diskin et al., 2002; Durry et al., 2008; Ebert et al., 2000; Gurlit et al., 2005; Hansford et al., 2006; Helten et al., 1998; Hunsmann et al., 2008; Karpechko et al., 2014; Kley and Stone, 1978; May, 1998; Meyer et al., 2015; Ohtaki and Matsui, 1982; Roths and Busen, 1996; Salasmaa and Kostamo, 1986; Schiff et al., 1994; Silver and Hovde, 1994b, a; Thornberry et al., 2015; Webster et al., 2004; Zöger et al., 1999a, b) (non-exhaustive list). While for some atmospheric questions the quality level of the data often is sufficient (e.g., typically climatologies), there are also a variety of questions, especially validation of atmospheric models, where the required absolute accuracy, precision, temporal resolution, long-term stability, comparability, etc. needs to be higher. These problems can be grouped into two major categories: accuracy-linked problems and time-response-linked problems. The latter is particularly important for investigations in heterogeneous regions in the lower troposphere as well as for investigations in clouds. In these regions, even two on average agreeing instruments with different response times yield local, large, relative deviations on the order of up to 30 % (Smit et al., 2014). It is important to keep in mind that the total time response of a system is a superposition of the time response components of the instrument itself as well as of the sampling inlet. These typically depend on numerous parameters like the type of inlet, inlet pipe length, pipe coating, pipe temperature, pipe heating, gas flow, and input air humidity level.
In contrast to time response studies, accuracy-linked problems in flight are
difficult to isolate since they are always covered by the spatial
variability (which leads to temporal variability for moving aircraft) of
atmospheric H
Therefore, in 2007 an international intercomparison exercise named
AquaVIT (Fahey et al., 2014) was carried out to compare airborne
hygrometers under quasistatic, laboratory-like conditions for upper-tropospheric and lower-stratospheric humidity levels. AquaVIT (Fahey et al.,
2014) encompassed 22 instruments from 17 international research groups. The
instruments were categorized in well-validated, oft-deployed “core”
instruments (APicT, FISH, FLASH, HWV, JLH, CFH) and “younger” non-core
instruments. AquaVIT revealed, in the important 1 to 150 ppmv H
The instruments by themselves might actually be more accurate than AquaVIT
showed, but deficiencies of the different calibration procedures (with their
different calibration sources etc.) might mask this. To summarize, AquaVIT
documented a span of up to 20 % relative deviation between the world's
best airborne hygrometers – but AquaVIT could neither assess absolute
deviations nor explain them, since a link to a metrological H
Therefore, we present in this paper the first comparison of an airborne
hygrometer (SEALDH-II) with a metrological standard for the atmospheric
relevant gas pressure (65–950 hPa) and H
This paper focuses on the metrological accuracy validation of the SEALDH-II. SEALDH-II is the airborne successor of the proof-of-concept spectrometer (SEALDH-I) study published in Buchholz et al. (2014), which showed the possibility and the achievable accuracy level for calibration-free dTDLAS hygrometry. The publication (Buchholz et al., 2014) demonstrates this for the 600 to 20 000 ppmv range at standard ambient pressure. The instruments SEALDH-I, SEALDH-II, and HAI (Buchholz et al., 2017) are all built with the design philosophy that every single reported value of the instrument should have a “related boundary/operation condition snap shot”, allowing us to exclude the possibility of any instrumental malfunction during the measurement. SEALDH-II is from this perspective the most extensive approach (capturing much more boundary condition data; Buchholz et al., 2016), while HAI can serve as a multichannel, multiphase hygrometer for a broader variety of scientific questions.
SEALDH-II integrates numerous different principles, concepts, modules, and
novel parts that contribute to or enable the results shown in this paper.
SEALDH-II is described in detail in Buchholz et al. (2016). The following
brief description covers the most important technical aspects of the
instrument from a user's point of view: SEALDH-II is a compact (width:
19 in. rack; height: 4 rack units
Approximately 80 different instrument parameters are controlled, measured, or
corrected by SEALDH-II at any time to provide an almost complete supervision
and detection of the spectrometer status – we termed this concept “holistic
dTDLAS spectroscopy”. This extensive set of monitoring data ensures reliable
and well-characterized measurement data at any time. The knowledge about the
instruments status strongly facilitates metrological uncertainties
calculations. SEALDH-II's calculated linear part of the measurement
uncertainty is 4.3 %, with an additional offset uncertainty of
SEALDH-II's measurement range covers 3–40 000 ppmv. The calculated mixture
fraction offset uncertainty of
SEALDH-II's data treatment works differently from nearly all other published
TDLAS spectrometers. Typically, instruments are set up in a way that they
measure the absorbance or a derivative measurand of absorbance and link it
to the H
In a very simplified way, our physical absorption model uses the
Equation (1) can be enhanced with the ideal gas law to calculate the
H
Equation (2) facilitates an evaluation of the measured spectra without any
instrument calibration at any kind of water vapor reference (Buchholz et al.,
2014; Ebert and Wolfrum, 1994; Schulz et al., 2007) purely based on first
principles. Our concept of a fully calibration-free data evaluation approach
(this excludes also any referencing of the instrument to a water standard in
order to correct for instrument drift, offsets, temperature dependence,
pressure dependence, etc.) is crucial for the assessment of the results
described in this publication. It should be noted that the term
“calibration-free” is frequently used in different communities with
dissimilar meanings. We understand this term according to the following quote
(JCGM, 2008): “calibration ... in a first step, establishes a relation
between the measured values of a quantity with measurement uncertainties
provided by a measurement standard ... [I]n a second step, [calibration] uses
this information to establish a relation for obtaining a measurement result
from an indication (of the device to be calibrated)”. Calibration-free in
this sense means that SEALDH-II does not use any information from
“calibration, comparison, test, adjustment” runs with respect to a
higher-accuracy “water vapor standard” to correct or improve any response
function of the instrument. SEALDH-II uses, as described in Buchholz et
al. (2016), only spectroscopic parameters and the 80 supplementary parameters
as measurement input to calculate the final H
Figure 1 right shows the validation setup. As a
well-defined and highly stable H The dew-point mirror
hygrometer used here
can measure far below 0
The humidity of the gas flow is set by the TSM generator but the absolute
H
Calibration of the D/FPH (dew- and frost-point mirror hygrometer,
MBW 373 LX, which is used as part of the THG) at the national primary water
vapor standards of Germany. The standard for the higher H
Overview showing all data recorded over 23 days of validation experiments. Measurements of the traceable humidity generator (THG) are shown in red, SEALDH-II data in black, and gas pressure and gas temperature in SEALDH-II's measurement cell in blue and green. Note: SEALDH-II operated the entire time without any malfunctions. The THG did not save data in the 35 ppmv section. The temperature increase during the 75 ppmv section was caused by a defect of the air conditioning in the laboratory.
One part of the validation was a permanent operation of SEALDH-II over a
timescale much longer than the usual air- or ground-based scientific
campaigns. In this paper, we present data from a permanent 23-day
(550 operation hours) operation in automatic mode. Despite a very rigorous
and extensive monitoring of SEALDH-II's internal status, no malfunctions of
SEALDH-II could be detected. One reason for this are the extensive internal
control and error handling mechanisms introduced in SEALDH-II, which are
mentioned above and described elsewhere (Buchholz et al., 2016). Figure 3
shows an overview of the entire validation. The multi-week validation
exercise comprises 15 different H
Detailed plot of the validation at 200 ppmv with six gas pressure
steps from 50 to 950 hPa. Each individual pressure level was maintained for
at least 60 min in order to avoid any dynamic or hysteresis effects and to
facilitate clear accuracy assessments. The
These well-understood, systematic pressure-dependent deviations will be visible in each further result plot of this paper. The impact and methods of compensation are already discussed in Buchholz et al. (2014). The interested reader is referred to this publication for a more detailed analysis and description.
SEALDH-II's primary target areas of operations are harsh field environments. Stability and predictability are to be balanced with extra levels of accuracy which might not be required or reliably achievable for the intended application. HOLS models are therefore deliberately traded for a stable, reliable, and unified fitting process under all atmospheric conditions. This approach leads to systematic, predictable deviations in the typical airborne accessible atmospheric gas pressure range (125–900 hPa) of less than 3 %. One has to compare these results for assessment of the non-systematic deviations of 20 % revealed during the mentioned AquaVIT comparison campaign (Fahey et al., 2014). Hence, for field and airborne purposes, the 3 % instrument uncertainty seems to be fully acceptable – especially in airborne environments where the water vapor content is locally very inhomogeneous (leads to rapid temporal variations) and therefore the sampling system enhances the instrument uncertainty significantly.
Short-term H
This comparison with AquaVIT should just provide a frame to embed the 3 %.
The H
Besides the pressure dependence discussed above, SEALDH-II's
accuracy assessment is exacerbated by the differences in the temporal
behavior between the THG's D/FPH and
SEALDH-II: Fig. 5 (left) shows an enlarged 45 min long section of measured
comparison data. SEALDH-II (black) shows a fairly large water vapor variation
compared to the THG (red). The precision of SEALDH-II (see
Sect. 2) is 0.056 ppmv at 0.4 Hz (which was
validated at a H
The results of this validation exercise are categorized in three sections
according to the following conditions in atmospheric regions:
mid-tropospheric range of 1200–600 ppmv (Fig. 6), upper-tropospheric range
of 600–20 ppmv (Fig. 7), and lower-stratospheric range of 20–5 ppmv
(Fig. 8). This categorization is also justified by the relative influence of
SEALDH-II's calculated offset uncertainty of
Figure 6 shows the summary of the pressure-dependent validations in the 1200–600 ppmv range. Each of the 48 data points represents the mean over one pressure measurement section of at least 60 min (see Fig. 4). A cubic polynomial curve fitted to the 600 ppmv results (blue) serves as an internal quasi-reference to connect with the following graphs. The 600 ppmv data (grey) are generated via a supplementary comparison at a different generator: the German national primary mid-humidity generator (PHG). These primary generator data at 600 ppmv indicate a deviation between PHG and THG of about 0.35 %, which is compatible with the uncertainties of the THG (see Sect. 3.1) and the PHG (0.4 %) (Buchholz et al., 2014). The PHG comparison data also allow a consistency check between the absolute values of (see Fig. 2) the PHG (calibration-free), the THG (D/FPH calibrated), and SEALDH-II (calibration-free).
Gas-pressure-dependent comparison between SEALDH-II and THG over a
H
In this range, the linear part of the uncertainty (4.3 %) and the offset
uncertainty (
The results in this range (Fig. 8) are dominated by the offset uncertainty.
It is important to mention at this point that the
Comparison results as in Fig. 6 but for the 200–600 ppmv range.
Comparison results as in Figs. 6 and 7 but for the 5–20 ppmv range. All spectra are determined with a calibration-free first-principles evaluation concept. The major contribution to the higher fluctuations at lower concentrations is the accuracy of the offset determination (for details see text).
Direct comparison of SEALDH-II versus THG for H
Figure 9 presents a summary of all 90 analyzed mole
fraction–pressure pairs during the 23 days of validation. The calculated
uncertainties (linear 4.3 % and offset
It should be noted that this result does not change the statement about SEALDH-II's uncertainties, since these are calculated and not based on any validation or calibration process. This is a significantly different approach between calibration-free instruments such as SEALDH-II and other classical spectroscopic instruments which rely on sensor calibration. SEALDH-II provides correctness of measurement values within its uncertainties because any effect which causes deviations has to be included in the evaluation model – otherwise it is not possible to correct for it.
As mentioned before, any calibration-free instrument can be calibrated too
(see, e.g., Buchholz et al., 2013). However, by doing so, one must accept to
a certain extent loss of control over the system, especially in environments
which are different from the calibration environment. For example, if a
calibration was used to remove an instrumental offset, one has to ensure that
this offset is stable long-term, which is usually quite difficult, as shown
by the example of parasitic water offsets in fiber-coupled diode laser
hygrometers (Buchholz and Ebert, 2014). Another option is to choose a
recalibration frequency that is high enough, i.e., minimizing the drift
amplitude by minimizing the time between two calibrations. This, however,
reduces the usable measurement time and leads to considerable investment of
time and money into the calibration process. For the case of SEALDH-II, a
calibration of the pressure dependence – tempting and easy to do – would
directly “improve” SEALDH-II's laboratory overall performance level from
The SEALDH-II instrument, a recently developed, compact, airborne,
calibration-free hygrometer (Buchholz et al., 2016) which implements a
holistic, first-principle dTDLAS approach, was stringently validated at a
traceable water vapor generator at the German National Metrology Institute
(PTB). The pressure-dependent validation covered a H
Due to its extensive internal monitoring and correction infrastructure,
SEALDH-II is very resilient against a broad range of external disturbances
and has an output signal temperature coefficient of only 0.026 % K
For future applications, the measurement path length of 1.5 m and hence
SEALDH-II's sensitivity could be relatively quickly enhanced by a factor of
5–10 by implementing a longer path absorption cell. A linear increase of the
absorption path yield a proportional scaling of the SEALDH-II's dynamic
range (currently at 1.5 m: 3–40 000 ppmv; lower limit defined by the
calculated offset uncertainty of
The underlying data for the results shown in this paper are raw spectra (time vs. photocurrent), which are compressed to be compatible with the instruments data storage. In the compressed state the total amount is approximately 6 GB of binary data. Uncompressed data size is approx. 60 GB. We are happy to share these data on request.
BB and VE conceived and designed the experiments. BB performed the experiments; BB and VE analyzed the data and wrote the paper.
The authors declare that they have no conflict of interest.
Parts of this work were embedded in the EMPIR (European Metrology Program for Innovation and Research) projects METEOMET-1 and METEOMET-2. The authors want to thank Norbert Böse and Sonja Pratzler (PTB Germany) for operating the German primary national water standard and the traceable humidity generator. Last but not least, the authors thank James McSpiritt (Princeton University) for the various discussions about reliable sensor designs and Mark Zondlo (Princeton University) for sharing his broad knowledge about atmospheric water vapor measurements. Edited by: Dietrich G. Feist Reviewed by: two anonymous referees