Introduction
The stable isotopologues of atmospheric water (H2O) vapor
contain information about its condensation history, because the isotopologue
ratios are altered whenever H2O is subjected to phase changes
. Evaporation and condensation leave a fingerprint in
the isotopologue ratios due to differences in vapor pressure and kinetic
diffusion properties of the isotopologues
. This fingerprint is mainly
depending on temperature (vapor pressure), but it also contains information
of relative humidity (kinetic effect) during evaporation. Mixing of different
air masses induces further changes of the isotopic composition of air masses
probed .
In this paper we discuss changes in the ratio RD=D/H in
water vapor, which are expressed with respect to the Vienna Standard Mean
Ocean Water (VSMOW) reference standard as
δD(H2O)=RD(Sample)/RD(VSMOW)-1
(typically given in ‰).
On a global scale, remote sensing measurements of δD(H2O),
abbreviated herein as δD, can help to better constrain the
condensation history and transport pathways from the ground to the sampling
location. Tropospheric profile measurements have already been reported from
ground and space by means of Fourier Transform Infrared (FTIR) spectroscopy
or other techniques . Since
changes of δD are usually small, uncertainties in its
determination (accuracy) may lead to misinterpretation of the physical
processes involved in the H2O transport.
Quality assessment of remotely sensed δD has been performed
theoretically . So far empirical validation studies of
tropospheric δD remote sensing products have been made with
very few indirect references or have used a
δD reference that itself was not comprehensively validated
. Recently presented
reference data for space-based remote sensing observations of the
Tropospheric Emission Spectrometer (TES), which is important progress.
However, their profile measurements were limited to the lower ∼ 4500 m
of the troposphere and did not fully cover the altitude range of the remote
sensor's peak sensitivity.
Vertical profiles of calibrated in situ δD measurements with
well-known accuracy are highly desirable in order to prove the quality of
remotely sensed δD. Furthermore, highly resolved in situ
profiles can serve to study small-scale (tens to hundreds of meters) processes
related to tropospheric humidity, e.g., cloud formation and
air mass mixing .
In situ airborne measurements have been performed with different techniques
and on various platforms
. With the exception
of , these measurements were performed without what we
believe is a critical component especially for a validation experiment:
inflight instrument-performance analysis by measuring a calibration-gas
standard. Only if one can prove that airborne measurements of a
calibration-gas standard are reproducible within the stated instrument
uncertainty are these measurements an adequate source of data for remote
sensing validation purposes.
In this paper we present high-resolution tropospheric vertical profiles of
δD obtained in situ by the ISOWAT airborne tunable diode
laser spectrometer during the MUlti-platform remote Sensing of
Isotopologues for investigating the Cycle of Atmospheric water (MUSICA) remote sensing validation campaign. We
first describe our in situ instrument which has been adapted from a previous
prototype to meet the requirements for the profile measurements. We discuss
our calibration procedure and derive the uncertainties of our airborne data.
Then the MUSICA airborne campaign is introduced, and the vertical profiles
are presented. In the last part we introduce how measurements of water
isotopologues trace tropospheric transport processes.
ISOWAT II
Instrument setup
The ISOWAT II tunable diode laser spectrometer is based on the first
prototype version , which is presently being operated
onboard the IAGOS-CARIBIC passenger aircraft. ISOWAT II features a number of
modifications that were implemented for the instrument operation during the
MUSICA aircraft campaign. The basic instrument concept is reviewed briefly,
and the modifications are discussed in the following section.
The ISOWAT instrument is a tunable diode laser absorption spectrometer to
measure rotational–vibrational resolved absorption lines of the three
main isotopologues of water: H216O, H218O, and HDO at a
wavelength of λ= 2.65 µm (ν=3765 cm-1. In the
spectrometer (Fig. ), the beam of a distributed feedback
diode laser (Nanoplus GmbH) is collected by an anti-reflection coated
f=25.4 mm bi-convex CaF2 lens. The beam is then focused and steered
into an astigmatic-type multipass absorption cell (MPC, model AMAC-76-LW,
Aerodyne Research, Inc.) by means of a spherical-mirror telescope comprising
a f=250 mm concave and a f=-100 mm convex spherical mirror. The MPC
is aligned to N=110 passes to achieve a total absorption pathlength of l=36.1 m within a volume of V=0.5 L. The beam exiting the MPC is
collected and focused by a f=100 mm spherical mirror onto the
thermoelectrically (TE) cooled HgCdTe sample detector (SD, model J19-TE4,
Judson Technologies).
A fraction of the laser beam is picked off by a pellicle beam splitter (P-BS,
45 % / 55 %) upstream of the MPC. This beam is split again by a second
P-BS and it serves two purposes. One part is guided through a 10 cm long
reference-gas cell filled with a high mixing ratio of H2O. This beam is
subsequently focused onto the TE-cooled InSb reference detector (RD, model
J12-TE2, Judson Technologies). Spectra from the RD are used for active
locking of the laser wavelength to the absorption lines of interest by
applying a feedback to the laser-injection current ILD. The
second part of the beam is simply passing through the optical compartment of
the instrument and is focused onto the TE-cooled HgCdTe background detector
(BD, model J19-TE4, Judson Technologies). The BD is used to record spectra of
residual water vapor present in the optical compartment. This open-path
signal component is also accounted for in the SD spectra by including it in
the fit. The optical paths in the SD and BD beams are closely matched.
In addition to recording the residual humidity using the BD, for ISOWAT II
the optics compartment was sealed and actively dried to a relative
humidity below 5 %. To this end, a miniature pump (1410 Series, Gardner
Denver Thomas GmbH) constantly circulated air from the compartment
(Fig. ). This air was dried in a desiccant cartridge
containing molecular sieve (0.3 nm pore width) before it was re-directed
into the optics compartment. Furthermore, the optics compartment had an open
port to ambient pressure (instrument surrounding) via a second desiccant
cartridge. This was used as ventilation for the instrument as the aircraft
cabin pressure during the campaign was varying between 1000 hPa at ground
level and around 450 hPa at the highest flight level.
Optical system of the ISOWAT II spectrometer. LD: laser diode; P-BS:
pellicle-beam splitter; MPC: multipass cell; SD: sample detector; RD:
reference detector; BD: background detector; MP: miniature pump.
For ISOWAT II an entirely new digital data acquisition system has been
developed. This system has two main components. A compact embedded PC
(KONTRON, model pITX-SP 2.5" SBC) provides a user interface to adjust all
parameters for the laser control and data acquisition. These parameters are
transferred to the second component, i.e., a digital data processing unit
based on field-programmable gate array (FPGA, ALTERA, model Arria GX)
technology.
In ISOWAT II we employ wavelength-modulation spectroscopy (WMS) with second
harmonic detection. The laser wavelength is scanned via the laser-injection
current at a frequency of fscan=10 Hz across four absorption
lines of interest using a triangular waveform (Fig. ). In
addition, the laser wavelength is modulated with a sinusoidal waveform at a
frequency of fmod=37.5 kHz. All waveforms are generated by the
FPGA and are converted to analog voltages using a 16 bit digital-to-analog
converter.
The detector signals of SD, RD, and BD are first filtered using individual
analog bandpass filters (Sallen–Key fourth-order filter,
fc1=41 kHz, fc2=121 kHz) with 10 dB dec-1
efficiency. The individual signals are then amplified using computer-controlled low-noise amplifiers in order to utilize the full range of the
16 bit analog-to-digital converters (ADC). After digitization, three
individual software lock-in amplifiers realized in the FPGA demodulate the
detector signals at 2×fmod. The demodulated signals are
then fed through software low-pass finite impulse response filters with
a cut-off frequency of fc=500 Hz. The second-harmonic spectra
are then transferred to the embedded PC via USB connection for storage and
further processing.
In addition to the second-harmonic spectra we record direct absorption
spectra of the SD for power normalization purposes. To this end, the
modulated analog SD signal used for the WMS is amplified and digitized by a
16 bit ADC. A software low-pass filter with fc=500 Hz in the
FPGA equivalent to the one used for WMS signals further conditions the
signal. The spectral baseline is fit with a fourth-order polynomial (thick
green line in Fig. ). Furthermore the SD dark signal was
measured every 10 min by switching off the laser diode for 10 s. The direct
absorption signal was used to identify changes in the laser-power
transmission through the MPC.
Example spectra of ISOWAT II showing the up and down scan across the
absorption region of interest. Polynomial baseline (green trace) of the
direct absorption spectra (black trace) as well as detector zero signal (blue
trace) are used to normalize the second-harmonic absorption spectra (red
trace). AH216O is the peak-to-peak amplitude of the strong
H216O absorption line. The weak H216O line used in humid
conditions is depicted 20× enlarged.
Calibration
Instrument calibration was performed using a unit for ground-based pre- and
post-flight calibrations. Dry synthetic air (Air Liquide, ALPHAGAZ 1) from a
compressed cylinder was humidified in a 2 L volume bubbler filled with
500 mL of a liquid isotope working standard
(δD=-59.22± 0.7 ‰) at a temperature of
21 ∘C. This humidified air was subsequently diluted into a main flow
of dry synthetic air. By adjusting the flow through the bubbler as well as
the main dry air flow, the calibration humidity could be adjusted from a few
hundred ppmv to more than 30 000 ppmv at
δDCAL∼-137 ‰.
The relatively large amount of liquid isotope standard used in the
ground-based calibration unit was replaced daily. This was effectively
reducing drift in the isotopic composition due to evaporation of the bulk
water source to <0.5 ‰ in 4 h of calibration. This drift was
accounted for in the data analysis. The ground-based calibration unit was
independently examined prior to the campaign using a laboratory-based Picarro
water isotope analyzer in combination with a standard delivery module.
The repeatability was found to be better than 0.5 ‰ for
δD, and the uncertainty in the measurement of the liquid
standards was 0.7 ‰.
A dew-point hygrometer (Edge Tech, model DewMaster) was occasionally
connected in parallel to ISOWAT on the ground in order to perform absolute
humidity measurements that served as calibration of the H216O and HDO
absorption-line fit.
Similar to the first version of ISOWAT , during MUSICA the
instrument was equipped with a calibration-gas source for inflight
instrument-performance analysis. To provide a calibration-gas mixture of
known isotopic composition, a flow of 3.5 slpm
(1slpm=1000cm3 air at 1000 hPa) of ambient air was
first dried to H2O<5 ppmv in a desiccant cartridge containing
molecular sieve (0.3 nm pore width). A part (0.1 slpm) of this dry air flow
was humidified by passing it through a small temperature-controlled bubbler
(T=45± 0.05 ∘C) that contained a liquid-water isotope-working
standard (δD=-59.22± 0.7 ‰). The well-known
fractionation factor for evaporation of the liquid water standard is
α(T=45 ∘) ∼ 1.06, which leads to water vapor with
δD∼-116 ‰. The pressure in the bubbler thereby
corresponded to the ambient air pressure. The humidified flow was then mixed
into the main dry air flow to produce a calibration gas with a (pressure-dependent) humidity of around 3000 and 5000 ppmv at ground level and 6000 m
flight altitude. The inflight calibration-gas source was
cross-calibrated by three measurements before and after each flight using
this primary standard.
The calibration gas was fed into ISOWAT II two to five times per flight in a
variety of situations, including relatively rapid ascends or descends and
high turbulence at lower flight levels in order to ensure instrument
performance in these conditions. The MPC was first flushed with the
calibration gas for 7 min to allow for complete isotope exchange. Then a
calibration-gas spectrum was recorded for 2 min, after which the MPC was
flushed with ambient air for 7 min before continuation of ambient-air
measurements.
Data handling
Ten second-harmonic (2f) spectra are averaged in 1 s and normalized by
optical power at the sample detector (PLD). We therefore use the
spectral baseline of the simultaneously recorded direct absorption channel
together with the detector zero signal. The latter is linearly interpolated
between measurements taken every 10 min where the laser was switched off for
10 s. The normalized 2f spectra are then divided into windows around the
individual absorption lines.
For the strong H216O line we determine the peak-to-peak amplitude
AH216O based on the signal at line center and the minima of
the two negative lobes of the 2f absorption line. This is then compared to
AH216O from a previously recorded calibration spectrum in
order to derive the H2O mixing ratio. The shape and amplitude of the
strong H216O line are considerably affected by absorption-line self-broadening .
The other absorption lines are much less affected by line broadening, and in
each fit window a linear least squares fit is performed using singular value
decomposition (SVD) . The SVD fit includes constant and
linear baseline terms, as well as the line component from the calibration
spectrum, and it yields the mixing ratio of the individual isotopologues.
From the mixing ratios the isotopologue ratios are determined.
First we transferred the peak-to-peak absorption-line signal
AH216O onto a humidity scale based on the humidity
calibrations performed using the dew-point hygrometer. Then we determined the
humidity-dependent correction functions CF(H2O) for the raw
isotopologue ratio RD*=HDO/H216O as
depicted in Fig. a. For the CF(H2O) of
the strong H216O line we use an empirical sixth-order polynomial and for
the CF(H2O) of the weak line a first-order polynomial (black
traces in Fig. ).
In order to transfer the raw RD* to δD on the
VSMOW scale, we first divide the individual RD* measurements by
the daily average CF(H2O). We then multiply by the isotopologue
ratio of the calibration gas
RCAL=RVSMOW(δDCAL+1), where
RVSMOW=3.10693×10-4, and obtain δD by
solving
δD=RD*CFH2ORVSMOWδDCAL+1-1.
Figure b depicts this procedure for the calibration
measurements of 25 July 2013. The black dashed lines indicate the uncertainty
range, which is defined as the range of data in which 1σ of the data
fall.
Note that we have defined a transition interval between 10 000 and
12 000 ppmv in which we make use of both the strong and the weak
H216O absorption line. In this interval, the individual fit results are
used in a humidity-weighted manner in order to achieve a smooth transition
between the two fit regimes.
(a) Individual calibration measurements on 25 July 2013
pre- and post-flight together with the daily average calibration functions
for strong (red) and weak (blue) H216O line fit. (b)
Calibration measurements on VSMOW scale corrected for nonlinear response.
Dashed lines indicate the 1σ confidence band.
Uncertainty estimate
The uncertainty of our δD measurements was derived from the
calibration measurements that were performed on every flight day before and
after each flight as described in Sect. . We define the
uncertainty as the standard deviation of all daily calibrations to the daily
mean humidity correction function for a given humidity interval h:
ΔδDh=1n-1∑i=1nδDi-δDCAL2h.
The uncertainty was determined in humidity intervals (h) of 10 ppmv
between 100 and 25 000 ppmv, and it is depicted for all flight days as
black traces in Fig. .
Any differences in the instrument response before and after the flight was
assigned to instrumental drift. Even though the instrument temperature was
greatly stabilized, the drift was most likely caused by changes in the
temperature of components that are not well temperature-stabilized(see
Sect. for detailed discussion).
In addition we have used the inflight calibration-gas measurements to verify
that our uncertainty estimate was justified. The difference of these
calibration measurements to the expected δD of the inflight
calibration-gas standard is depicted as blue diamonds in
Fig. . These differences are generally better or
equal to the uncertainty estimation. We are thus confident that the
instrumental uncertainty of our flight measurements is equal to or better
than the stated uncertainty estimate.
Uncertainty analysis for all flight days. Grey symbols: deviation of
individual 1 Hz calibration measurements from calibration standard; black
lines: daily fit of Δ(δD) (Eq. );
green line: campaign average of Δ(δD). Inflight
calibration-gas measurements are depicted as the difference to the daily mean
(absolute value) by the blue symbols.
Sources of uncertainty
We have identified two main factors that define the measurement uncertainty
of ISOWAT II. Their importance varies throughout the investigated humidity
range.
At humidities below ∼ 1000 ppmv, optical interference fringes
superimposed onto the measured absorption spectra become relevant as the
measured absorption signal becomes weak. The fringes in our system during the
MUSICA campaign were with OD ∼ 1...2× 10-4 about 1.5–3
times higher than currently in ISOWAT-I , and their free
spectral range were such that they could not easily be distinguished from the
absorption lines by our absorption-line fit. The phase of the fringes was
also changing due to slight changes in instrument temperature, which
prohibited the recording of the fringes once for later subtraction while
sampling dry air (background spectrum). The fringes affected mostly the HDO
line at dry conditions as well as the weak H216O line below
15 000 ppmv. For example, the fractional absorption of HDO at 200 ppmv and
δD=-400 ‰ is around 4.2× 10-4, i.e.,
a signal to fringe ratio of ∼ 2–4. The fringe amplitude can be
translated to an absorption signal equivalent to around 1 ppmv (strong
H216O line).
An array of residual spectra is depicted in the lower panel of
Fig. . By repeatedly modulating the temperature of the
anti-reflection coated f=25.4 mm CaF2 collection lens (top panel in
Fig. ) we were able to identify this lens as the source
of the most dominant fringe. The fringe free-spectral range (FSR) is
estimated to be around 3.75 GHz, which translates into an optical resonator
length of around 2.8 cm assuming the refractive index of CaF2 of 1.43.
The center thickness of the lens is 11 mm. We conclude that multiple
internal reflections at a thinner part of the lens caused the fringe.
In a laboratory test after the campaign, replacement of this lens by a
reflective mirror objective has reduced the overall fringe level by a factor
of 2. Importantly, the FSR has become shorter by
a factor of 3 to 4. This will potentially make fitting of the fringe
structure easier for future measurements. We plan to carefully re-design the
optical setup of ISOWAT based on the results using the mirror objective.
(a) Temperature recording of the mount of the f=25.4 mm
collection lens. (b) Array of spectral residuals recorded in the
laboratory while repeatedly modulating the temperature of the f=25.4 mm
collection lens. The color code represents the residual amplitude scaled to
optical density. The black trace depicts a second-harmonic spectrum recorded
at 1 s integration time.
The second factor affecting the instrumental uncertainty was related to
slight changes in the temperature of the laser-gain medium. In the present
instrument, this temperature is set to a constant value
(∼ 28 ∘C). Then the laser emission wavelength
(λLD) is tuned by sweeping the laser-injection current
(ILD). Despite setting TLD to a constant value,
TLD was drifting during instrument operation. We actively control
the DC component of ILD in order to maintain the wavelength scan
in the spectral window around the absorption lines of interest (line lock).
Adjusting the DC component of ILD changes the operating point of
the laser. This affects the nonlinear response of laser power
(PLD) and λLD to a sweep of ILD.
Changes in PLD are accounted for by normalizing the
second-derivative spectra by the direct-absorption spectra. Changes in the
wavelength tuning of the laser lead to a distortion of the recorded spectra.
As we determine mixing ratios and δD by fitting of ambient
spectra to a previously recorded calibration spectrum, this distortion of the
spectra adds uncertainty to the fit.
It was found after the campaign that the major source of temperature drift
was a temperature-dependent potentiometer (100 ppm K-1) on the
commercial laser-driver unit. The resistance of this potentiometer changes
due to changes in ambient temperature, which in turn changes the temperature
set point of the laser-temperature control loop. Replacing this potentiometer
with a fixed resistor with low-temperature coefficient (2 ppm K-1)
will minimize the change of the temperature set point.
Meanwhile, we have developed an improved stabilization of the laser-operating
point by implementing a second control loop. In this approach we additionally
control the temperature of the aluminum mount of the laser diode using
resistive heaters. The temperature of the mount is adjusted such that
λLD is maintained stable, which is determined by analyzing
the position of the absorption lines within the reference-detector spectra.
This in turn allows us to maintain the DC component of ILD constant.
Consequently the laser operating point is stabilized. First tests in the
laboratory have shown that this approach minimizes the drift of the
laser-temperature set point by a factor of 11 even when using the
temperature-sensitive potentiometer.
MUSICA aircraft campaign
Objective and realization
The primary goal of the MUSICA airborne campaign was to perform validation
measurements for two remote sensing instruments: (i) the satellite-borne
Infrared Atmospheric Sounding Interferometer (IASI) instrument on both the
MetOp-A and B satellites and (ii) a ground-based FTIR spectrometer at the Izaña meteorological observatory (IZO). The
second goal was to investigate transport and processing of water vapor in the
North Atlantic free troposphere (FT).
Seven research flights were conducted from Las Palmas Airport (LPA), Canary
Islands, Spain. The Canary Islands are located in the subtropical North
Atlantic, approximately 200 km west off the coast of Morocco at
28∘ N latitude (Fig. ). The flights were
coordinated in time and location with the morning overpasses of the MetOp-A
and B polar-orbiting satellites . Both satellites are
equipped with an IASI instrument as scientific payload .
IASI measures nadir radiance spectra from which tropospheric H216O,
HDO,
and δD vertical profiles are retrieved .
MetOp overpasses took place on a trajectory from northeast to southwest,
where IASI scanned transversal to its flight direction. The individual
spectra were filtered based on data quality and cloudiness as described in
. The ground location of all quality-filtered vertical
profile measurements obtained during the campaign is depicted as circles in
Fig. .
The aircraft flights were performed in the line of sight of the ground-based
FTIR instrument stationed at IZO. The FTIR measures solar absorption spectra
to retrieve vertical profiles of H216O, HDO, and δD in
the troposphere . The blue lines in
Fig. indicate the line of sight for the beginning and the
end of the profile flight on 24 July 2013. The aircraft flight track of this
day is shown in red in Fig. .
The flights started with a relatively fast ascent to the ceiling altitude of
the aircraft (∼ 6800 m). The aircraft crew then waited for the
overpass of the satellites before performing a slow cascaded descent with up
to 10 selected levels. The levels were held for 5 min. Each flight took
around 3.5 h.
Aircraft flight track (red) of 24 July 2013 departing from Las
Palmas Airport (LPA). Also shown are all MetOp-A and MetOp-B measurements
(symbols) obtained during the campaign after a data-quality filter was
applied. The FTIR line of sight (blue lines) for 10:30 and 13:00 UTC (start
and end of profile measurements) at Izaña (IZO) is indicated. Map
source:
http://en.wikipedia.org/wiki/File:Map_of_the_Canary_Islands.svg.
Aircraft and instrumentation
For the campaign, several research instruments were mounted in a CASA
C-212-200 (Fig. ) of the Instituto Nacional de
Técnica Aeroespacial (INTA). The C-212 is a medium-sized transport
aircraft powered by two turboprop engines. It has an unpressurized cabin and
can reach a maximum ceiling altitude (depending on payload) of around
23 000 ft (7000 m) with the requirement for oxygen use by the pilots and
operators.
ISOWAT II was mounted in the foremost position of the aircraft cabin. Ambient
air was sampled by a rearward facing stainless steel inlet with 0.25 inch
(6.35 mm) outer diameter mounted on top of the aircraft fuselage
(Fig. ). A flow through the inlet line of around
15 slpm (1slpm=1000cm3 air at 1000 hPa) at
ground level and around 6 slpm at the aircraft ceiling level was established
by a membrane pump (MD1 VARIO SP, Vacuubrand GmbH) downstream of the ISOWAT
II instrument. Inside the aircraft, heated (T=40 ∘C)
electropolished stainless-steel tubing (ULTRON, Dockweiler AG) was used to
connect the inlet system to the instrument. A flow of 1.5 slpm was branched
off the main flow and fed into the ISOWAT II instrument for isotope ratio
measurements. All tubes inside the instrument as well as the MPC were also heated to 40 ∘C.
CASA C-212 aircraft of Instituto Nacional de Técnica Aerospacial
(INTA). The rear-facing ambient air inlet was located on top of the
fuselage. The inlet line was flushed at a high flow using a membrane pump.
ISOWAT was mounted in the foremost position of the aircraft and was equipped
with a second pump to establish the gas flow through the instrument.
Meteorological conditions
Two different meteorological conditions were encountered during the MUSICA
campaign which also mark the predominant conditions at the Canary Islands in
summer.
On 21 July 2013, 25 July 2013, and 30 July 2013 the marine
boundary layer (MBL) was well separated from the FT by a
temperature-inversion layer at an altitude of around 1100 m. This inversion
is a consequence of large-scale subsidence in the downward branch of the
Hadley circulation . The subsiding air is adiabatically
warmed as it reaches lower altitudes, thus yielding a well-developed
temperature-inversion layer. In the MBL, winds from northeast were
encountered, while in the FT the wind came from northwest. We refer to these
meteorological conditions as North Atlantic owing to the air mass origin in
the FT.
On 22 July 2013, 24 July 2013, 31 July 2013, and 1 August 2013
warm and mineral-dust laden air reached the Canary Islands from the east in a
layer of variable thickness at an altitude of approximately 2 to 5 km. This
Saharan air layer was positioned above a layer of MBL air, and it
presumably originated from the African boundary layer . We refer to these conditions as Sahara.
Figure shows satellite images of the Moderate
Resolution Imaging Spectroradiometer (MODIS) instrument of the days 21 July
2013 (a) and 31 July 2013 (b). In Fig. b one can
observe the dust-laden air as the hazy area in the right half of the image.
Note that the images are daily composites and may not reflect the conditions
during the actual hours of flight. The flight tracks are depicted in red.
Satellite images of the Canary Islands of 21 July 2013 (top) and 31
July 2013 (bottom), which are representative of “North Atlantic” and
“Sahara” conditions, respectively (for an explanation see
Sect. ). Images are daily composites of
data of the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the
NASA Terra and Aqua satellites (source:
https://earthdata.nasa.gov/labs/worldview/). The aircraft flight tracks
are depicted as red traces.
Results
Figure presents an overview of the vertical profiles of
H2O and δD as well as the estimated total uncertainty of
the isotope ratio Δ(δD) (see Eq. )
with a temporal resolution of 1 s. The black symbols show all inflight
calibration measurements that were performed during the campaign. We note
that the uncertainty of the inflight calibrations is in most cases smaller
than Δ(δD). This provides good evidence that our
uncertainty estimate can be understood as an upper limit.
For better clarity we only show in Fig. the data of the
descending branches of the flights. Ascent and descent took place at
different times and, more importantly, at slightly different longitudes. We
noticed strong longitudinal gradients in some of the flights, which prohibit
averaging of the ascent and descent profiles. Good agreement of ascent and
descent data is found where longitudinal gradients were low or sampling
locations coincided.
Vertical profiles of H2O (left), δD (center), and
instrumental absolute uncertainty (right) of all flight days evaluated at
1 Hz. For better clarity we show here only data with aircraft descending.
Black symbols denote the respective quantities of the inflight
calibration-gas measurements.
For a more detailed analysis we shall have a closer look at the measurements
during the descending branch of the flight on 31 July 2013.
Figure shows the corresponding vertical profiles of
temperature (T), H2O mixing ratio, and δD.
Measurements of temperature (T), H2O mixing ratio, and
δD taken with 1 s averaging on 31 July 2013. The profile is
divided into five layers for interpretation.
Studying the vertical profile in Fig. reveals large
spatial gradients in humidity and isotopic composition. We can divide the
profile in five different layers.
The lowest layer between sea level and ∼ 440 m is characterized by
a rather constant T, H2O, and δD.
Above this layer we observed a well-defined temperature inversion which
came along with a sharp negative gradient in both H2O and
δD. At its minimum at an altitude of ∼ 900 m, H2O
decreased to less than 2000 ppmv (relative humidity ∼ 8 %) at
δD<-300 ‰.
A third layer starts at ∼ 1025 m with a sharp positive gradient of
H2O and δD, which is then followed by relatively constant
humidity and slightly increasing δD until an altitude of
around 1900 m.
The fourth and thickest layer is characterized by a column of high
humidity and high δD, which stretches nearly 3.5 km between
around 1900 m and 5350 m altitude. It is a weakly stable layer with a
temperature laps rate of 8.6Kkm-1.
At an altitude of around 5350 m we encountered an abrupt decrease in
both H2O and δD as the aircraft flew at constant altitude
in easterly direction. Above we measured the lowest humidity at around
430 ppmv and δD≈-370 ‰.
In order to investigate the processes that may have led to the observed
humidity distribution, we depict δD as a function of H2O
(Fig. ). By doing so it becomes immediately evident that
δD measurements provide additional information, because we
observed a rather large range of δD for similar humidity,
especially when between around 1000 and 8000 ppmv. We may now be able to
identify individual hydration processes by comparing our measurements with
model calculations. The Rayleigh model describes an atmosphere in which water
vapor condenses at a relative humidity of 100 % and the condensate is
immediately removed from the atmosphere by precipitation. The isotopic
composition of the condensed phase is defined by the temperature-dependent
vapor-pressure isotope fractionation . The heavy
isotopologue HDO condenses preferentially over H216O due to its lower
vapor pressure. The remaining vapor is becoming more depleted in HDO as the
air becomes dryer.
Distribution of H2O and δD measurements (1 Hz)
during the descending branch of the flight on 31 July 2013. The color code
represents the sampling altitude. The grey shaded area depicts a range of
possible Rayleigh models assuming SST=22.5 ∘C and
RH=80 % for ocean water with δD=0 ‰
(VSMOW, lower boundary) and δD=+7 ‰ (upper
boundary). The solid lines represent air mass mixing of (4) MBL and UT air
and (5) air at intermediate altitude and UT air (numbers according to levels
in Fig. ).
Here we consider ocean water as the major source of atmospheric vapor. Around
the Canary Islands, the ocean water has
δDow≈+7 ‰ w.r.t. VSMOW
. The sea surface temperature (SST), which is used to
derive the initial isotope fractionation during evaporation, was around
22.5 ∘C (NOAA_OI_SST_V2 data,
http://www.esrl.noaa.gov/psd/). Assuming a relative humidity w.r.t. SST
of RHSST=80%, the initial vapor has
H2O≈ 23 000 ppmv and
δDev≈-70 ‰ with respect to the
liquid source reservoir. The grey shaded area in Fig.
depicts a range of Rayleigh models for
δDow=0 ‰ (lower boundary) and
δDow=+7 ‰ (upper boundary).
First consider the most humid measurements (layer 1) marked with (1) in
Fig. . These measurements were taken at an altitude of
less then 440 m above the ocean, and they are around 20 ‰ above the
Rayleigh model. The elevated δD may be due to the evaporation
of a fraction of sea spray into the lower atmosphere, a process
suggested, e.g., by .
We note that on this particular day the wind above the ocean was strong, and
the lowest atmospheric layer was very hazy. This was confirmed by particle
concentration measurements (not shown), which showed elevated concentrations
of small particles (size < 3 µm) below 500 m altitude.
If the droplets evaporate (even partially), δD of the vapor
becomes less negative . We can estimate the
fraction F of sea spray that evaporated without fractionation and mixed
with evaporated ocean water by solving
δD=FδDow+(1-F)δDev.
Here δD=-50 ‰ was measured, and
δDow=+7 ‰ as well as
δDev=-70 ‰ are assumed. Consequently
F=0.26, i.e., 26 % of the total water vapor may have originated from the
evaporation of sea spray without significant fractionation. Analyzing H2O
vapor samples for their δD, concluded that in
this particular study up to 50 % of the H2O vapor may have originated
from evaporated sea spray.
Above this humid layer we observed a very sharp humidity gradient, where the
humidity decreased by more than 1 order of magnitude. These data (layer 2)
remain close to the theoretical Rayleigh model (blue data in
Fig. ). This suggests that the dehydration of this air
occurred predominantly close to isotopic equilibrium as the air mass was
lofted and cooled. These observations were taken in the (locally) warmest
layer, where the lowest humidity suggests a temperature of last condensation
of around -12 ∘C at the sampling pressure. This in turn leads to
the conclusion that this air mass must have originated from much higher than
the sampling altitude. Descent and adiabatic warming lead to a strong
decrease of relative humidity versus ice to
∼8 % (not shown), a process termed subsidence drying
. The altitude of origin of this air mass was above
5000 m.
At an altitude between 1000 and 1900 m (layer 3), we observed a steady
increase of δD, while the humidity remained relatively
constant at around 3000 ppmv. This transition is well defined in
Fig. and shows a very tightly correlated group of data
(medium blue).
In layer 4, extending between around 1900 and 5400 m, we observed a
relatively compact distribution of δD and humidity. There,
δD is significantly elevated with respect to the Rayleigh
model. This suggests that a considerable fraction of the water content is
originating from processes other than slow ascent and dehydration of boundary
layer air. A model based on the method described by that
describes mixing of moist air typical of the MBL
(H2O= 22 900 ppmv, δD=-75.6 ‰) and dry
air from the upper troposphere (UT; H2O=1100 ppmv,
δD=-385 ‰) excellently fits our observations (solid
line (4) in Fig. ).
Layer 4 also showed elevated concentrations of dust (not shown) which
originated from the Sahara (see Fig. ).
This suggests that the observed air mass is likely a mixture of MBL air,
continental Saharan boundary layer air, and dry air of the UT. The range
of humidity and δD suggest that we observed different
fractions of moist and dry air. At the top of this layer we observed an
abrupt change towards dry UT air with δD close to the
Rayleigh model.
Above 5400 m (layer 5) we observed a dryer layer spanning a large range of
δD between -200 ‰ and -380 ‰. The
measurements could be fit to a mixing model line between a low-altitude
reservoir of H2O=4900 ppmv and δD=-133 ‰and a higher-altitude reservoir with H2O = 500 ppmv and
δD = -373 ‰ (solid line (5) in
Fig. ).
Conclusions
In this paper we present details of an instrument and methods to perform
measurements of δD. The instrument and methods were
successfully applied in vertical-profile flights during the MUSICA airborne
research campaign at the Canary Islands in the summer of 2013
. The campaign mainly focused on the validation of
remote sensing measurements from ground and space by performing simultaneous
in situ measurements. The airborne sampling measurements were performed with
the ISOWAT II tunable diode laser spectrometer. Modifications to ISOWAT
were implemented prior to the campaign aiming to provide
measurements with better accuracy and well-defined uncertainty.
During the MUSICA campaign seven flights were performed during which vertical
profiles of δD in the troposphere in the subtropical North
Atlantic near Tenerife were measured. The profiles were ranging between sea
level and around 7 km altitude. The flights were coordinated in time and
space with δD remote sensing measurements from ground by the
FTIR spectrometer at Izaña, Tenerife, as well as from space by the IASI
instruments onboard the MetOp-A and B polar orbiting satellites.
With ISOWAT II we demonstrated δD measurements in situ with a
temporal resolution of 1 s. This translated to a horizontal resolution of
around 75–95 m and vertical resolution of around 3 m. This allowed
resolving even very sharp vertical gradients found both in humidity and
δD. A rather extensive calibration protocol allowed
determining a robust uncertainty estimate of our δD
measurements, i.e., a mandatory requirement for validation measurements. The
uncertainty depends on humidity. For most conditions during the campaign
the uncertainty was around 10 ‰. Only in the arid upper troposphere was
the uncertainty somewhat higher.
Using one flight as an example, five different air mass layers could be
identified. Their humidity could be analyzed based on the H2O and
δD measurements as well as using relatively simple models for
dehydration and air mass mixing. Further data analysis is presently being done
to better describe both large- and small-scale atmospheric processes based on
our measurements.
In a recent paper by , the ISOWAT II in situ
measurements presented here were used for a first validation of space- and
ground-based remote sensing measurements of δD. We aim to
enlarge the data set for this purpose by performing regular profiling of the
troposphere with a new version of ISOWAT onboard the IAGOS-CARIBIC passenger
aircraft .