AMTAtmospheric Measurement TechniquesAMTAtmos. Meas. Tech.1867-8548Copernicus PublicationsGöttingen, Germany10.5194/amt-12-873-2019Tropospheric water vapor profiles obtained with FTIR: comparison with balloon-borne frost point hygrometers and influence on trace gas retrievalsTropospheric water vapor profiles obtained with FTIROrtegaIvaniortega@ucar.eduhttps://orcid.org/0000-0002-0067-617XBuchholzRebecca R.https://orcid.org/0000-0001-8124-2455HallEmrys G.https://orcid.org/0000-0001-5137-2902HurstDale F.https://orcid.org/0000-0002-6315-2322JordanAllen F.https://orcid.org/0000-0002-6178-4502HanniganJames W.jamesw@ucar.eduhttps://orcid.org/0000-0002-4269-1677Atmospheric Chemistry Observations & Modeling, National Center for Atmospheric Research, Boulder, Colorado, USACooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USANOAA Earth System Research Laboratory, Global Monitoring Division, Boulder, Colorado, USAIvan Ortega (iortega@ucar.edu) and James W. Hannigan (jamesw@ucar.edu)8February201912287389027August201817October201810January201924January2019This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/4.0/This article is available from https://amt.copernicus.org/articles/12/873/2019/amt-12-873-2019.htmlThe full text article is available as a PDF file from https://amt.copernicus.org/articles/12/873/2019/amt-12-873-2019.pdf
Retrievals of vertical profiles of key atmospheric gases provide a critical
long-term record from ground-based Fourier transform infrared (FTIR) solar
absorption measurements. However, the characterization of the retrieved
vertical profile structure can be difficult to validate, especially for gases
with large vertical gradients and spatial–temporal variability such as water
vapor. In this work, we evaluate the accuracy of the most common water vapor
isotope (H216O, hereafter WV) FTIR retrievals in the lower and upper
troposphere–lower stratosphere. Coincident high-quality vertically resolved
WV profile measurements obtained from 2010 to 2016 with balloon-borne NOAA
frost point hygrometers (FPHs) are used as reference to evaluate the
performance of the retrieved profiles at two sites: Boulder (BLD), Colorado, and at
the mountaintop observatory of Mauna Loa (MLO), Hawaii. For a meaningful
comparison, the spatial–temporal variability has been investigated. We
present results of comparisons among FTIR retrievals with unsmoothed and
smoothed FPH profiles to assess WV vertical gradients. Additionally, we
evaluate the quantitative impact of different a priori profiles in the
retrieval of WV. An orthogonal linear regression analysis shows the best
correlation among tropospheric layers using ERA-Interim (ERA-I) a priori
profiles and biases are lower for unsmoothed comparisons. In Boulder, we
found a negative bias of 0.02±1.9 % (r=0.95) for the 1.5–3 km
layer. A larger negative bias of 11.1±3.5 % (r=0.97) was found in
the lower free troposphere layer of 3–5 km attributed to rapid vertical
change of WV, which is not always captured by the retrievals. The bias
improves in the 5–7.5 km layer (1.0±5.3 %, r=0.94). The bias
remains at about 13 % for layers above 7.5 km but below 13.5 km. At MLO
the spatial mismatch is significantly larger due to the launch of the sonde
being farther from the FTIR location. Nevertheless, we estimate a negative
bias of 5.9±4.6 % (r=0.93) for the 3.5–5.5 km layer and 9.9±3.7 % (r=0.93) for the 5.5–7.5 km layer, and we measure positive biases of
6.2±3.6 % (r=0.95) for the 7.5–10 km layer and 12.6 % and
greater values above 10 km. The agreement for the first layer is
significantly better at BLD because the air masses are similar for both FTIR
and FPH. Furthermore, for the first time we study the influence of different
WV a priori profiles in the retrieval of selected gas profiles. Using NDACC
standard retrievals we present results for hydrogen cyanide (HCN), carbon
monoxide (CO), and ethane (C2H6) by taking NOAA FPH profiles as the
ground truth and evaluating the impact of other WV profiles. We show that the
effect is minor for C2H6 (bias <0.5 % for all WV sources) among all
vertical layers. However, for HCN we found significant biases between 6 %
for layers close to the surface and 2 % for the upper troposphere depending
on the
WV profile source. The best results (reduced bias and precision and r values
closer to unity) are always found for pre-retrieved WV. Therefore, we
recommend first retrieving WV to use in subsequent retrieval of gases.
Introduction
Water vapor is a ubiquitous atmospheric constituent with an extremely
important role in the lower and middle troposphere and stratosphere: it is
the most variable and critical greenhouse gas ; it plays a
key role in atmospheric chemistry, e.g., heterogeneous chemistry, aerosol
formation, and wet deposition ; it affects global
radiation through cloud formation ; and it acts as the main
source for precipitation in the lower atmosphere .
Middle and upper tropospheric and lower stratosphere stable water vapor
isotopes are key to understanding the water cycle feedbacks such as mixing of
air masses, dehydration pathways, and free-tropospheric moisture
.
Obtaining consistent long-term observations of vertical distributions of
water vapor is challenging but highly desirable in order to understand
climate evolution and feedback effects . There is a need to
measure water vapor vertical distribution for long-term monitoring but there
are only a few datasets, e.g., in situ balloon observations in Boulder,
Colorado, USA, are the longest dataset of water vapor with information from
the
lower to middle stratosphere . It has been
shown that ground-based Fourier transform infrared (FTIR) measurements
provide reliable long-term and continuous observations of the most common
water vapor isotope (H216O, hereafter H2O or WV)
. FTIR measurements have focused mostly
on integrated WV analysis among the Network for Detection of Atmospheric
Composition Change (NDACC, see http://ndacc.org, last access: 20 January 2019). For integrated WV
(IWV, i.e., total columns) FTIR measurements have been shown to be very precise with about
2.2 % using FTIR side-by-side intercomparisons .
MUSICA (Multi-platform remote sensing of isotopologues for investigating the
cycle of atmospheric water) is a project within the NDACC FTIR that uses
standard spectra from a subset of NDACC sites in order to generate a
long-term dataset of tropospheric water vapor profiles with degrees of
freedom (DOF) of about 2.8 and of about 1.6 for the ratio between the most
abundant isotopologue H216O and the heavy isotopologue
HD16O. Comparisons of FTIR
and operational radiosondes have been used to validate optimized WV profile
retrieval strategies .
studied the spatial–temporal variability in WV in the
free troposphere (Zugspitze, Germany) by exploiting the geometry of
measurements of differential absorption lidar (DIAL) and FTIR. In particular,
they assessed the variability under small space scales and timescales, i.e., a few
kilometers and minutes.
In this work, we evaluate the accuracy and precision of WV profiles using a
standard retrieval inversion with ground-based FTIR measurements. For the
first time, the retrieval validation uses coincident and well-characterized
balloon-borne in situ NOAA frost point hygrometer (FPH) measurements
. The FPH measurement technique has been used as a reference
to assess the accuracy of radiosonde relative humidity measurements due to
their high vertical time resolution and low uncertainties
. With the goal of assessing WV vertical
gradients, we studied both the influence of different WV a priori profiles and
the smoothing of highly resolved FPH profiles. Finally, ubiquitous strong WV
absorption signatures interfere in the retrieval of other gases. However,
there is a lack of knowledge of the quantitative effects of WV at different
altitudes. A second major part of this work seeks to use FPH profiles as the
ground-truth WV and quantitatively assess the impacts of other typical WV
profiles in the retrieval of selected tropospheric gases, hydrogen cyanide
(HCN), carbon monoxide (CO), and ethane (C2H6), using NDACC standard
retrievals.
MeasurementsFree tropospheric and boundary layer FTIR sites
FTIR direct solar IR absorption spectra are measured under clear-sky
conditions in two different locations: (1) Boulder, Colorado (hereafter BLD;
40.40∘ N, 105.24∘ W, 1600 m a.s.l.) and (2) Mauna Loa,
Hawaii (hereafter MLO; 19.40∘ N, 155.57∘ W, 3400 m a.s.l.).
The spectra at BLD have been recorded using a Bruker 120 HR spectrometer
operated since 2010 following standard measurement protocols of the Infrared
Working Group (IRWG)/NDACC (http://ndacc.org). The instrument is
located in the foothills laboratory of the National Center for Atmospheric
Research (NCAR) situated in the front range of the Rocky Mountains. Previous
studies have used the BLD dataset for satellite validation of NH3, mobile low-resolution FTIR validation of NH3 and
C2H6, and analysis of gases emitted by oil and
natural gas development . The MLO instrument
has been part of the long-term activities of the IRWG/NDACC. First IR solar
absorption spectra were recorded at MLO in 1991 using a Bomem DA02. In 1995 a
Bruker 120 HR began operating, which was upgraded in 2011 to a Bruker 125 HR.
The high-altitude site at MLO is normally above the boundary layer and the
measurements are sensitive mainly to free tropospheric and stratospheric air
masses. At both sites the spectra are recorded using optical band pass
filters maximizing the signal-to-noise ratio (SNR) over the near- and
mid-infrared spectral domain with a nominal spectral resolution of 0.004 cm-1
(optical path difference of 250 cm) using liquid-nitrogen-cooled
InSb and mercury cadmium telluride (MCT) detectors and a KBr beam splitter .
Balloon-borne NOAA frost point hygrometer
Highly precise and accurate in situ measurements of tropospheric and
stratospheric WV over Boulder, Colorado, and Hilo, Hawaii, are performed with
balloon-borne FPHs by the Global Monitoring Division of NOAA's Earth System
Research Laboratory (ESRL). These measurements are also part of the GCOS
Reference Upper Air Network (GUAN) and the NDACC. At both sites,
balloon-borne FPHs are launched once per month, preferably during conditions
of low winds and clear skies. The Boulder measurements started in 1980 and
are launched at the Marshall Field Site (1743 m a.s.l.), 10.5 km south of the BLD
FTIR measurement site . Monthly
NOAA FPH soundings at Hilo started in 2010 and the balloons are launched from
the National Weather Service facility at Hilo International Airport (10 m a.s.l.),
58.0 km east of MLO. In this paper we emphasize the comparisons at
BLD due to the shorter distance between the FTIR and balloon launch site,
although we perform identical comparisons and present results from MLO as
well.
A thorough description of the FPH measurement technique is available in
and . Briefly, the principle is to
condense WV from a stream of air onto a small, gold-plated mirror using a
cryogenic liquid to continually cool the mirror. Once a thin condensed layer
is deposited on the mirror, pulses of heat are applied as needed to maintain
a stable layer of condensate. Changes in frost (ice) coverage are detected by
measuring the mirror reflectivity using a small LED-based infrared beam and a
photodiode. The amount of heat applied is rapidly adjusted to produce a
stable frost layer, at which point the temperature of the mirror (frost point
temperature) is a direct measure of the partial pressure of WV in the air
stream above it via the Goff–Gratch equation . The water
vapor mixing ratio is calculated by dividing the WV partial pressure by the
dry atmospheric pressure. Since a FPH fundamentally makes temperature
measurements, only the thermistor embedded in each mirror requires
calibration. Each thermistor is calibrated using NIST traceable standards
(see ). A recent detailed analysis of WV mixing ratios
measured by the NOAA FPH shows the uncertainties (2σ) are <12 % for
the 0–5 km altitude layer, <8 % for 5–13 km, and <6 % for 13–28 km. The NOAA FPH vertical profile data employed here are
0.25 km vertical averages and their standard deviations are calculated from
the measurements made at 5–10 m vertical resolution during balloon ascent.
Retrieval of water vapor from FTIR
Prior to the retrieval of WV from the solar absorption spectra, a quality
control of each measurement is carried out, i.e., visual inspection of
spectra and assessment of the SNR. As mentioned in Sect. , we
only use spectra taken during cloud-free conditions. The spectra are analyzed
using the retrieval code SFIT4 0.9.4, which has been improved from its
predecessor SFIT2 . SFIT4
derives vertical profiles and the corresponding total vertical columns by
exploiting pressure broadening and temperature dependency of specific
absorption lines. The overall retrieval follows the optimal estimation method
applied to several micro-windows. The inverse problem is ill-posed and the
solution is constrained by an a priori profile (xa) and its
covariance matrix (Sa), which ideally should represent the
natural variability in the WV profile from climatological records
. Section describes, in
more detail, the different a priori profiles used in this study. In many cases
Sa is not well-known and an ad hoc constraint is used (e.g.,
). Constraining is important to select the solution
which, among the possible solutions of the ill-posed inversion, is the most
likely given prior knowledge. The forward model is nonlinear and the
following Gauss–Newton iteration is applied:
xi+i=xa+SaKiTKiSaKiT+Se-1⋅y-F(xi)+Kixi-xa,
where xi+1 is the retrieved state vector for the
(i+1)th iteration, K is the weighting function or
Jacobian of the forward model (F) calculated at each iteration,
Se is the measurement noise covariance matrix, and y is
the measurement state vector .
Many of the spectral windows used to retrieve NDACC standard gases contain WV
absorption signatures. Accurate WV profiles are required for the retrieval of
other gases because accurate quantification of the interfering WV reduces
retrieval uncertainty. WV can be retrieved using a range of absorption
features since it absorbs from the near- to far-infrared wavelengths. With the
goal of best characterizing this WV, we use retrieval settings that are commonly
used among NDACC sites. We use the 2600–2840 cm-1 spectral region to
simultaneously retrieve H2O and the isotopologue HDO. In this study, we
focus only on H2O. We use spectral micro-windows that are different to
those of the current MUSICA version and perform the
inversion on a linear scale (instead of a logarithmic scale used by MUSICA).
A short summary of the four micro-windows and interfering species included in
the analysis is given in Table . These micro-windows have
been chosen to maximize the information content and minimize total error. The
spectroscopic data used here are based on the line-by-line portion of the
HITRAN 2008 . The errors in the reported line parameters
are described in Sect. and are used to estimate the
systematic uncertainty in the retrieval. Most of the interfering species are
fitted as a scaling of the a priori vertical profile (CO2, N2O, and HCl)
with the exception of CH4, which is fit as a profile in micro-windows two,
three, and four. The Sa matrix is specified at each layer as a
fraction of the a priori profile, which allows for a linear scaled retrieval.
We adopted a maximum variability of 50 % in the diagonal covariance, which
exponentially decreased with increasing altitude. In order to prevent
sporadic vertical profile oscillations, we include a Gaussian correlation
length of 25 km in the off-diagonal elements of Sa. This
Sa has been optimized in order to obtain similar information
content for all a priori profiles presented in Sect. , a
requirement for efficient processing of decades of NDACC spectra. The
instrumental line shape (ILS) has been fixed with a unity modulation
efficiency and no phase error. The ILS does not play an important role in the
WV error budget and is of lower importance for tropospheric WV retrievals
.
Micro-windows for H2O retrieval including
interfering gases retrieved within those micro-windows. Column gases are
those retrieved by profile scaling of the initial profile while profile retrieval
is performed for the profile gases column.
Inputs into SFIT4 include vertical profiles of pressure, temperature, and the
volume mixing ratios (VMRs) of the atmospheric gases included in the fit.
Preceding the retrieval, SFIT4 employs the Air Mass Computer Program for
Atmospheric Transmittance/Radiance Calculation (FSCATM) ray tracing module to
calculate the atmospheric path . The input pressure and
temperature vertical profiles are obtained from the National Center for
Environmental Prediction (NCEP) reanalysis based on the NCEP/NCAR
analysis and forecast system to perform data assimilation using past data from
1948 to the present . These
profiles are obtained directly from NDACC (http://ndacc.org). These are
daily average profiles that extend to up to 0.4 mb (approximately 50 km).
Above 0.4 mb we use the monthly mean pressure and temperature profile from an
average of a 40-year simulation (1980–2020) of the Whole Atmosphere Community
Climate Model (WACCM) . These profiles are merged using a
cubic spline interpolation for pressure and a quadratic spline interpolation
for temperature.
We examined the effect of using more temporally refined temperature profiles.
In general, the 6-hourly temperature profile from the ERA-I reanalysis
model, produced by the European Center for Medium-Range Weather Forecasts
(ECMWF) , follows the daily average temperature profile shape
very well for both sites. The root-mean-square error (RMSE) between the 6-hourly data of ERA-I and daily average temperature profiles is less than
0.5 % using 2013 data for both BLD and MLO and the biases are less than
0.25 % for BLD and less than 0.1 % for MLO. These results suggest daily
mean temperature should be adequate for retrievals but we further
investigated the sensitivity of water vapor to this variability and found
that water vapor agrees within 1 % if using the daily average profile. The
temperature profile uncertainty is considered in the error analysis in
Sect. . With the exception of WV (see Sect. ), VMR input mean profiles of all other gases are taken
from the mean of a 40-year run of WACCM.
Characterization and error budget
The mean retrieval fit of the four micro-windows between 2010 and 2016 at BLD is
shown in Fig. . The small systematic residual structures
(black lines) are likely caused by spectroscopic parameter error but in
general the magnitude of residuals is low and within noise level (<0.1 %).
Mean retrieval fit between 2010 and 2016 for the spectral intervals of
WV. The observed and fitted lines are blue and green, respectively. The
absorption contribution for the different species is also shown in each
micro-window. The bottom black lines represent the mean residual and the gray
shadow is the standard deviation. Note that for visibility the residuals
have been multiplied by 10.
The information content of the retrieved WV vertical profile is characterized
within the averaging kernel matrix, A:
A=KTSe-1K+Sa-1-1KTSe-1K.
The rows of the mean A, known as averaging kernels (AKs), obtained
between 2010 and 2016 and color coded by altitude below 20 km are shown in Fig. a
for BLD. The maximum values are located at the surface; then
they decrease and remain steady to about 8 km and eventually decrease to zero
above 12 km. This indicates that most of the information content is derived
from the lower troposphere. The mean total column averaging kernel (TAK) is
shown in Fig. b. Typically, a unity TAK indicates that the
retrieval is not biased, while values of the TAK lower than unity indicate
underestimation and larger values than unity indicate overestimation with
respect to the a priori state vector. Hence, below 3 km the retrieval may
underestimate, between 3 and 8 km overestimate, and between 8 and 12 km underestimate
the real WV magnitude. The mean number of DOFs, given by the trace of the
A, is 2.4 and indicates the total number of independent pieces of
information in the retrieval. The vertical profile of the cumulative sum of
DOF is shown in Fig. c and shows that the first DOF is given in
the layers below 3 km, the second DOF is given between 3 and 6 km, and the
rest are given
above. Further optimization of the retrieval strategy might improve the
A but as explained before, one of the goals is to assess the
current retrieval strategy; therefore we do not investigate retrieval
constraints further. At MLO the vertical sensitivity is similar but starting
at 3.5 km.
(a) FTIR mean row averaging kernels, (b) mean total column averaging
kernel, and (c) cumulative sum of DOF of WV obtained in BLD from 2010 to 2016.
SFIT4 estimates an uncertainty budget that combines random, systematic, and
smoothing sources following the formalism given in . The
most important random error is normally the retrieval noise characterized
by the SNR in the spectral region of interest. The error covariance matrix
(Sn) is calculated with the following equation:
Sn=GySeGyT,
where the gain matrix Gy represents the sensitivity of the
retrieval to the measurement and is related with the averaging kernel as
A=GyK. Currently, the diagonals of the
Se matrix are constructed using the square of the inverse of the
SNR obtained from the noise in the spectra of interest, and off-diagonal
elements are not considered. The retrieval of WV is actually an estimate of a
state smoothed by the averaging kernel. The difference between these two
states is given by the smoothing error (Ss):
Ss=(I-A)Sa(I-A)T,
where I is a unit matrix. The smoothing error is treated separately
and not included in the total error analysis because Sa is
normally not well-known and consequently is often simplified. The model
parameter error represents the errors in the forward model parameters such as
temperature, solar zenith angle (SZA), and spectroscopic parameters. These
errors can contain both systematic and random components. We obtain the model
parameter covariance matrix as
Sb=(GyKb)Sb(GyKb)T,
where Sb is the error covariance and Kb
the weighting function matrices of the forward model parameters. The largest
contributors considered here are the absorption line parameters, temperature
profiles, and SZA. The uncertainty of the absorption line parameters, i.e.,
line intensity (S), air-broadened half width (γ), and temperature
dependence of γ (n), are taken from the lower limit reported in
HITRAN 2008 . These uncertainties are only considered
systematic and the errors reported in HITRAN for WV are 5 %, 1 %, and 10 % for
S, γ, and n, respectively. Furthermore, uncertainties due to the
retrieved interfering species are also considered. The error in the
temperature profile is considered to have both systematic and random
components.
These errors have been quantified with the mean (systematic) and standard
deviation (random) of the difference of long-term comparisons among NCEP
profiles with radiosondes launched near the sites and/or ERA-I reanalysis.
The measurement noise error is estimated with the square of the inverse of
the SNR as diagonal elements in the covariance matrix. The pointing accuracy
in the SZA is considered random and has been characterized with an error of
0.15∘. Figure shows the random and systematic
vertical profile uncertainties as percentages with respect to the mean mixing
ratio. The major systematic components in the lower troposphere are the
absorption line parameters S and γ but in the upper troposphere the
temperature contributes equally. The temperature and measurement noise are
the main components of the random uncertainty. The final uncertainty is
estimated from the error propagation of all components and is lower than 10 %
below 4 km and about 10 % above. The instrumental line shape uncertainty
plays a minor role in the total error budget.
Mean vertical profiles of the most important random (a) and
systematic (b) uncertainty components for the retrieval of WV in BLD from
2010 to 2016.
Comparison of water vapor vertical profiles
The total number of sonde observations is 90 at Boulder and 70 at Hilo from
2010 to 2016. The overall number of coincident dates of measurements under
ideal conditions is 56 and 36 for BLD and MLO, respectively. Figure presents a rough qualitative comparison of selected
WV profiles obtained with NOAA FPH measurements and FTIR retrievals in BLD.
To retain high vertical variability the FPH profiles are shown in 0.25 km
vertical averages of the sonde's ascent measurements (continuous
black lines). The FTIR profiles (in blue) represent the average profile weighted by
the error and the blue shading depicts the uncertainties propagated using the
individual profiles within 2 h of the FPH launch. The daily mean ERA-I
(henceforth ERA-d) a priori profiles used in the retrievals are also shown in
gray.
To quantitatively compare both measurements the high-vertical-resolution
balloon-borne profiles are re-gridded onto the altitude grid of the FTIR
retrieval by means of a linear interpolation. For BLD the nearest FPH point
to the surface is typically a few hundred meters above the first grid point of
the FTIR. In this case, we assume homogeneous WV close to the surface and use
the nearest-neighbor point. A proper comparison between FTIR and in situ
sonde profiles requires smoothing the in situ measurements using the FTIR AKs
and a priori profiles to account for its lower-vertical-resolution capability
(see Eq. 4 in ). Red profiles in Fig. represent smoothed FPH profiles. As pointed out by
the information of the WV AK is limited due to its
high variability through the troposphere. A goal of the present study is to
determine the extent of vertical structure gradients of retrieved WV
profiles; hence the comparison with in situ sonde measurements is carried
out mainly without smoothing. However, results are also presented for
smoothed comparisons following the formalism given in .
The temporal variability and its effect are studied in Sect. . To some extent the retrieved WV profiles
capture the vertical structure gradients identified with the in situ NOAA FPH
even though the a priori profile may be biased and smooth (see for example
14 September 2010 and 5 November 2010). Figure shows the same
but for selected vertical profiles at MLO. The near-surface mixing ratios at
this high-altitude site are significantly lower and the profiles show steeper
vertical gradients than at BLD. Note that the FTIR (MLO) and FPH (Hilo) are
about 60 km apart and, on some days, may have measured different air masses,
especially at the lowest FTIR retrieval levels. In BLD the launch site of the
FPH is only 10 km south of the ground-based FTIR.
Due to the limited number of DOF we combine grid points to assess several
layers, maximizing the number of points characterizing the boundary layer,
free troposphere, and upper troposphere–lower stratosphere. The following
layers have been chosen for BLD: 1.5–3.0, 3.0–5.0, 5.0–7.5, 7.5–10, 10–13,
and 13–17 km above sea level (a.s.l.) and 3–5.5, 5.5–7.5, 7.5–10,
10–13, 13–16, and 16–20 km a.s.l. for MLO. These layers have been chosen so that they
include three standard IRWG FTIR grid points. Comparison of ground-based
remote sensing with balloon-borne in situ measurements is challenging due to
spatial–temporal variability. The temporal and spatial variability are
characterized in the next two sections followed by the quantitative
comparison between FTIR and NOAA FPH.
WV vertical profiles for selected dates obtained with unsmoothed
in situ NOAA FPH measurements (black) and FTIR retrievals (blue) in BLD. The
ERA-d WV used as the a priori profile is shown in gray. The dates are shown at the top of
each plot. The FTIR profiles represent weighted mean profiles using
retrievals within 2 h of the radiosonde launch. The filled blue shadow
area represents the standard error propagation using the uncertainty in
individual retrievals. The gray shaded areas are FPH profiles of the 1-sigma
standard deviation of each mixing ratio. The number of retrieved profiles
within 2 h is shown in the upper left of each panel.
Same as Fig. but for MLO.
Panels (a, b) show the number of dates (black) and profiles (blue)
measured by the FTIR at BLD (a, c) and MLO (b, d) as a function of the
length of the time interval in minutes. The bottom panels show the temporal
variability in percent estimated with the ratio of the standard deviation to
the mean values for several layers as a function of the length of the time
interval. The length of the time intervals are defined as an increasing temporal
window, e.g., 0–30, 0–60, 0–120 min, and the number of retrievals
in each window is used to calculate the variability.
Temporal variability
Due to the lack of independent time-resolved WV vertical profiles we use
daily FTIR observations to assess the temporal variability. Figure
shows the number of dates and profiles and the
variability in WV as a percentage for several layers as a function of the length of
time interval starting from 0 to 3 min and gradually increasing, 0
to 10, 0 to 30, 0 to 60 min, etc. The retrievals produced during these
time intervals are used to calculate the temporal variability using the ratio
of the standard deviation to the mean values at several altitude layers. This
approach is sensitive only to the variability observed by the FTIR; however
the real variability might be greater because of potential lost variability
during retrieval smoothing. This proxy for variability has been estimated
using dates during coincident measurements between sondes and FTIR. The
number of dates and profiles is roughly the same below 10 min, indicating
the time that the FTIR takes to start a new measurement using the same band-pass filter for a standard set of observations. The variability in BLD among
different layers does not vary substantially and they remain within 1 %–2 %
of each other, indicating similar relative variability within all the
different tropospheric layers. In BLD the variability starts to increase from
about 1 % in 30 min to 6 % in 240 min. In contrast, at MLO the
variability is different among layers. A variability of up to 9 % is found
for the layer close to the instrument altitude (3–5.5 km); however the
variability is below 5 % for the layer between 5.5 and 7.5 km and about
3 % for the 13–16 km layer, indicating vigorous fluctuations and strong
convection near the MLO site. In general, these findings suggest that the
coincidence time interval to avoid variability larger than 2 % is
30 min
at BLD and 60 min at MLO. The air mass probed by the FTIR is changing during
the day due to the line of sight to the sun moving constantly such that after
some time the spatial variability may play an important role.
estimated that the spatial mismatch may play a role
for intervals longer than 30 min. The spatial mismatch is described in the
next section.
Spatial mismatch
If the spatial mismatch between the FTIR and sonde is considerably large, each
might probe distinctive air masses. Hence, natural WV variability would
affect a meaningful comparison . A
thorough assessment of the error component due to spatial difference between
the sonde and FTIR would require measurements of an extensive area
simultaneously and at different altitudes. However, this is hard to derive
due to lack of such observations. In this section, we aim to estimate the
spatial mismatch between the sonde location at various altitudes and FTIR
maximum sensitivity. We calculate the horizontal distance between the sonde
location and the line of sight of the FTIR. The effective horizontal position
sensitivity of the FTIR depends on the sun-pointing geometry and the vertical
WV profile distribution. We adopted a methodology applied by
to estimate this effective horizontal position. This
method assumes that the FTIR sensitivity is located at the point at which the
viewing direction of the instrument meets the altitude level of the mass-weighted WV profile. Using the mass-weighted WV of all sonde profiles we
roughly estimate an altitude of 3.8±0.9km in BLD. Using this altitude
and the SZA the horizontal distance from the ground-based site is calculated
for every measurement. Then, using the solar azimuth angle the latitude and
longitude are calculated after having traveled the given distance on the
given bearing. Once the location is found the haversine formula is applied to
determine the great-circle distance between two locations .
At BLD the mean distance with respect to the FTIR site location is 6.0±4.0km south, making the initial spatial mismatch with the sonde launch about
6.5 km. At MLO the mass-weighted WV profile is 6.0±0.6km and the
initial spatial horizontal mismatch is 47.0 km (see Fig. S1 in the
Supplement). Consequently, even co-located sonde launches may not exactly
probe the same air mass.
The spatial mismatch at different altitudes depends on the sonde trajectory
and the location of the FTIR sensitivity. At BLD the GPS location of the
sonde at every altitude is available for almost all profiles; hence the
distance between the FTIR and the sonde location can be calculated. Figure
shows the mean spatial mismatch between the FTIR and
the sonde profiles for the coincident time intervals of 0–30 and 90–120 min.
As mentioned above, the initial spatial difference close to the
surface is about 6 km. For the 0–30 min interval the horizontal difference
is below 10 km below 4.5 km in altitude, similarly for the 90–120 min coincident time interval, except
for one altitude, which is greater than 15 km. Above 5 km in altitude the
spatial mismatch starts to increase. A rapid significant increase in the
spatial mismatch is identified above 5 km for both 0–30 and
90–120 min
coincident time intervals. Interestingly, the greatest horizontal difference
is found for the 0–30 min interval with maximum values of about 70 km. This
analysis shows that the spatial mismatch depends on the complex convective
dynamics and not only in the coincidence time interval. Nevertheless, only
short temporal coincidence differences are encouraged to avoid temporal WV
fluctuations as shown above.
Vertical profile of the horizontal spatial mismatch between FTIR and
sonde profiles in BLD. As an example two coincident time intervals are used.
Influence of a priori profiles
The optimal estimation method is influenced by the a priori profile because
it may bias the solution of Eq. (). Since WV is highly
variable, even on the timescale of hours, using the most accurate a priori
profile
might improve the retrieval results. In general, the retrieval of WV can be
seen as an update of the a priori information. In order to study the effect
of the a priori, four different a priori profiles are used to retrieve WV,
which are then compared with balloon-borne NOAA FPH measurements: (1) a 40-year
simulation (1980–2020) of the WACCM mean profiles (WACCM is a global
model with 66 vertical levels from the ground to approximately 140 km
in geometric height, and the horizontal resolution is 1.9∘ by 2.5∘
(latitude by longitude) and is part of the NCAR Community Earth System Model
(for further details see ));
(2) daily varying (ERA-d) profiles; and (3) 6-hourly varying WV vertical profiles (00:00, 06:00,
12:00, and 18:00 UTC) obtained from ERA-I (ERA-6). In this case, the closest in
time to the measurements is used. ERA-I profiles extend to 1 mb and then are
merged with WACCM monthly mean profiles of WV using a spline interpolation.
We take the closest ERA-I grid point to represent the a priori at each
station, and we use (4) daily varying NCEP/NCAR (NCEP-d) reanalysis WV profiles
. Since the spatial resolution of NCEP is lower than
ERA-I, about 2.5∘×2.5∘, we interpolate WV spatially to obtain the
best WV profile. We have chosen the above four a priori profiles since they
are readily available and commonly used. With the aim being to capture
vertical gradients, the comparisons are carried out with unsmoothed and
smoothed in situ profiles.
An optimization of the dataset is carried out before the quantitative
assessment of vertical profiles. The difference between WV retrievals and
sonde profiles (Δx=xr-xs) shows a normal
distribution centered around zero for the layers defined in Sect. . Figure S2
in the Supplement shows an example of the
Δx distribution using ERA-d for the different layers. Extreme
outliers are identified for each distribution using the 95th percentile and
values above that are filtered out in order to avoid skewed results. Figure S3
shows the 95th percentile of the Δx as a function of the different
a priori sources and for different layers. The lowest values are found for
both ERA-d and ERA-6, and about 25 % larger values are found for both NCEP
and WACCM. Additionally, the difference between WV retrievals and a priori
profiles (xr-xa) provides further evidence in the measured
signal and to some extent the variability prescribed by the a priori profile
. For example, this difference is about 11±38 %
using ERA-6 while for WACCM it is about 29±32 % for the first layer. As
we expected, from these observations it can be seen that the 40-year WACCM
climatology as an a priori profile results in greater deviations compared to ERA-6.
A quantitative impact of the different a priori profiles in the retrieval of WV
vertical profiles is characterized by means of linear regression and
statistical analyses using the layers defined earlier. Since both NOAA FPH
and FTIR have altitude-dependent uncertainties, we adopted a weighted
orthogonal distance regression (ODR) analysis. For a thorough description of
weighted ODR applied in atmospheric sciences see . In order to
avoid temporal variability larger than 2 % according to conclusions in
Sect. , a mean WV profile (x¯r)
is obtained within a coincidence time interval of 0–30 min at BLD and 0–60 min
for MLO. The NOAA FPH WV mixing ratios are used in the abscissa axis
and the ODR accounts for uncertainties in both sets of measurements. In this
case we use the standard deviation of the NOAA FPH and FTIR uncertainty
propagated using the individual profiles within the coincident time interval.
The final number of vertical profiles used in the comparison is 31 and 30 in
BLD and MLO, respectively. Figure shows the slope,
intercept, and correlation coefficient (r value) obtained with the comparison
of retrievals using each of the a priori profiles with the unsmoothed NOAA FPH at
different layers at both sites. The error bars in the estimated parameters
are the standard errors. For layers below 10 km the best results are seen
with both ERA-I a priori profiles. In particular, we found that ERA-6 yields the best
comparison with a slope close to unity, the lowest intercept, and a
correlation coefficient of 0.95 for the layer of 1.5–3 km in BLD. For both
sites, the second layer, i.e., 3–5 and 5.5–7.5 km for BLD and MLO,
respectively, shows lower slopes likely due to gradients between the top of
the planetary boundary layer and free troposphere that are not captured by
the retrievals due to coarse vertical resolution and lower sensitivity (e.g.,
see Figs. and ).
Results of the ODR analysis between the NOAA FPH and FTIR using
different a priori profiles at different altitude layers. Error bars
represent the standard errors of the estimated parameters. Note that for
visibility the intercept obtained in the upper three layers has been
multiplied by a factor of 10.
For each coincidence profile the bias is characterized with the sum of
differences between x¯r and the sonde (xs) profiles
divided by the number of points (N) in each layer. As described before the
number of points in each layer is three. This definition indicates whether
the retrievals under- or overestimate the sonde values. The precision is
calculated as 2×σ/N, where σ is the standard
deviation. The bar plot in Fig. shows the median bias and
precision in parts per million and percentage with respect to the mean values of the NOAA
FPH for the different layers and a priori profile. The error bars in the bias are
estimated using the ±1⋅ standard error of the distribution. The bias
shows the dependency on the a priori profile. At both sites the first two layers
show a
negative bias for all a priori profile. At BLD the smallest bias is found for the 1.5–3 km
layer with -0.001±0.105×103ppm (-0.02±1.86 %) for
ERA-6 and the highest bias of -0.27±0.11×103ppm (-4.82±1.94 %) for WACCM climatology. The layer between 3 and 5 km shows a negative
bias of
between 5.56 % and 11.14 %. Interestingly, NCEP-d yields less biased
results in this layer. The layer of 13–17 km shows significantly larger
values for almost all a priori profiles (>15 %). The precision does not change
significantly among different a priori profiles. The best precision result
as a
percentage is below 5 %, found in the lowest layer of 1.5–3 km, and the
highest values of up to 15 % are found for layers between 5 and 10 km. As expected
based on the ODR analysis higher biases are found at MLO. Negative biases of
about 5 % for the 3.5–5 km layer and 10 % for the 5.5–7.5 layer are
found and a positive bias of 5 % is found for the 7.5–10 km layer.
Surprisingly, at both sites WACCM yields a lower bias for the layers above 13 km.
In general among all layers, the lowest biases are found using ERA-6 and
ERA-d for both sites.
The approach described above has been applied in the comparison of FTIR with
smoothed FPH profiles. Table presents a summary of the ODR
and statistical analysis using ERA-6 for unsmoothed and smoothed FPH
profiles at BLD where the spatial mismatch is known and the launch of the
sonde is in close proximity to the FTIR location. Among all layers the ODR
analysis shows similar results between unsmoothed and smoothed FPH
comparisons; however biases are significantly lower for unsmoothed
comparisons, indicating the limitation of the AK WV.
Statistical analysis results (bias and precision) of the FTIR WV
retrieved at different altitudes and using different a priori profiles for
BLD (a) and MLO (b). Bias and precision are given in mixing ratios
and as
percentages with respect to the mean values at each layer. The error bars in the
bias represent the standard error of the distribution. Note that for
visibility the bias and precision in mixing ratio from the two upper layers
have been multiplied by a factor of 10.
Summary of the ODR and statistical analysis using ERA-6 at BLD. Results for
unsmoothed (upper level) and smoothed (lower level) FPH comparisons are shown.
Retrieval settings of gases to study the influence of WV.
All interfering species are fitted with a scaling factor, except O3 in the
retrieval of CO and C2H6, and are fitted as vertical profiles.
GasMicro-windows (cm-1)Interfering speciesCO2057.7–2058.0; 2069.56–2069.76; 2157.50–2159.15O3, CO2, OCS, H2O, N2OHCN3268.04–3268.40; 3287.10–3287.35; 3299.40–3299.60H2O, C2H2, CO2, O3C2H62976.66–2977.059; 2983.20–2983.50; 2986.45–2986.85O3, H2O, CH4, CH3ClInfluence of WV on gas profile retrievals
Absorption of WV is normally present in the analysis of gases using FTIR
measurements. Even optimized micro-windows of gases include the WV and/or
isotopologue absorption lines in order to minimize its interference. In this
context, WV profiles are included in the retrieval process of other
atmospheric gases. Usually, the most accurate WV profile is recommended.
However, highly accurate and co-located WV profile measurements are rare and
typically reanalysis based or pre-retrieved WV profiles are used as reference
in the retrieval of other gases. In the latest case, WV is retrieved in
dedicated micro-windows and then the retrieved WV profile is used in the
retrieval of other gases .
studied the impact of WV interference in the retrieval
of carbon monoxide (CO) and further apply a joint retrieval strategy to
remove interference errors. There are few published data on the
quantitative impact of the WV profile using independent co-located WV
profiles. Findings from previous sections provide important insights into how
well the retrieved WV, and other WV priors, compare with the real WV profile,
in this case the NOAA FPH. In this section, we further exploit the FPH
measurements in order to study the influences of different WV profiles
typically used in the retrieval of selected tropospheric gases, i.e.,
hydrogen cyanide (HCN), carbon monoxide (CO), and ethane (C2H6). The WV
sources tested are ERA-6, ERA-d, NCEP, WACCM, and the retrieved WV profiles. Note
that we do not aim to study retrieval strategies of gases or the validation
of profile retrievals but rather to show the relative difference with respect
to the higher-precision WV profile (FPH measurements). Table
presents the interfering species with strong and/or weak
absorption signatures within each micro-window for all target gases. In all
cases, the selected settings have been chosen in order to maximize the
information content and minimize the total error in the retrieval. The
settings we follow are IRWG/NDACC standard operational retrieval parameters
with respect to micro-windows and interfering species. The WACCM climatology
is used for a priori profiles of interfering species. Spectroscopic line
parameters are adopted from HITRAN 2008 .
For the retrieval of HCN we followed a similar approach to that applied in
, , and . The settings applied
in the CO retrieval are part of an ongoing project in the IRWG/NDACC (Bavo Langerock,
personal communication, 2017), and for C2H6 we applied an
improved version applied in (Emmanuel Mahieu, personal
communication, 2017). Pressure and temperature profiles are from NCEP. For
the retrieval of WV we use ERA-d to imitate our typical retrieval strategy.
As for WV, full error analysis is performed, i.e., mainly considering
measurement noise error and forward model parameter errors (see Sect. ).
Example on 22 July 2014 of retrieval profiles of HCN, CO, and
C2H6 using the different WV a priori sources shown in (a). The
retrieval profiles in (b) and (c) represent the relative
difference
as a percentage with respect to the retrieval, which uses NOAA FPH WV.
The retrieval of HCN, CO, and C2H6 was performed only during dates with
NOAA FPH sonde measurements. Since the FPH profiles are used as the reference
we have limited spectra taken only within 1 h of the sonde launch based on
findings presented earlier. In all cases, the standard settings remain the
same and only the WV profile reference is changed. An example of the effect
of the WV profile in the retrieval of the different gases is shown in Fig. . The different WV profiles used on this day
(22 July 2014) are shown on top. The retrieved WV (black) is the closest in shape and
magnitude to the NOAA FPH profile (purple). All the other WV profiles show
significant differences with respect to the FPH. The gas profile retrievals
are shown in the left panels using a color scheme similar to that in the WV profile
panel. The relative difference at every retrieval level, defined as (xi-xfph)/xfph×100, is shown in the right panels. The lowest
relative difference in all grid points and for all gases always occurs when
using the retrieved WV profile (black). All other WV sources present
significant differences. For example, for HCN differences of up to -20 %
are found at 6–10 km if using ERA-I. CO and C2H6 also show important
differences but always below 10 %. This example suggests that the current
retrieval strategy of WV is suitable to avoid WV interference in the
retrievals of other trace gases.
In order to determine the general impact of the different WV sources for all
spectra recorded within 1 h of sonde launch for 6 years, we have performed an
ODR and statistical analysis similar to the one presented in Sect. .
In this case, the retrieval using NOAA FPH WV is used
as the reference. Figure shows the main results of the
ODR analysis for the three gases using the different WV sources and at
different layers. The best correlations (r value) and the lowest intercepts
are found using the pre-retrieved WV profiles for all three gases, in
agreement with the example given in Fig. . The slope
values are close to unity and within the uncertainty values for CO (middle)
and C2H6 (right) using the pre-retrieved WV. However, HCN on the left
shows the most notable difference with respect to unity. The intercept is
normally negligible for pre-retrieved WV for all gases. The bias and
precision results are shown in Fig. . Biases larger than
6 % and 1 % are found for HCN and CO, respectively, using WACCM WV in the
layer closest to the surface. C2H6 does not show a significant bias
among different layers and WV sources. Overall, these results suggest that
incorporating the pre-retrieved WV in the forward model improves the quality
of other retrieved gases.
Results of the ODR analysis in which the mixing ratios using different
WV sources at different layers are compared with the “truth” retrieved values
using the NOAA FPH WV for HCN (a), CO (b), and C2H6(c).
Error bars represent the standard errors of the estimated parameters.
Statistical analysis results (bias and precision) for HCN (a), CO (b),
and C2H6(c) using different WV profiles at different
altitudes. The error bars in the bias represent the standard error of the
distribution.
Conclusions
The aim of the present research was to determine the limitations in
retrieving real WV structural variability from the boundary layer to the
upper troposphere using a standard FTIR inversion, i.e., the current
retrieval strategy is not modified to correlate well with reference vertical
profiles. Highly precise and accurate vertical profiles of WV from NOAA
balloon FPH in situ sondes are used for the first time as reference to
evaluate FTIR WV profiles in BLD and MLO, allowing the characterization of the
retrievals in midlatitude boundary layer and subtropical free troposphere
locations.
The spatial–temporal variability in WV is inferred prior to a quantitative
comparison. By using daily continuous FTIR measurements we derive a temporal
variability for different altitudes and find that at BLD the different layers
are highly correlated and show comparable variability. In contrast, at MLO
the variability among layers is quite different, indicating vigorous
inhomogeneity due to local convection or long-range transport. The ideal
coincidence time between sonde launch and FTIR measurements is
0–30
and 0–60 min in BLD and MLO, respectively, to avoid variability larger than
2 % for all altitudes. The horizontal position with maximal sensitivity of
WV distribution is derived for each FTIR measurement. Then, based on the
sonde location at each altitude the horizontal spatial mismatch is
characterized. The insight gained from this evaluation is that the boundary
layer (about 1.5 to 3 km in Boulder) is the only layer in which the air mass
probed by the FTIR and NOAA FPH in situ is likely unchanged since the
horizontal difference remains below 10 km. We show that above 5 km the
spatial mismatch increased significantly up to 60 km horizontal distance at
about 10 km in altitude. This feature does not depend on the coincidence time
between measurements but rather on the local to synoptic meteorological
scales. More broadly, even co-located FTIR and sonde launch measurements
would have significant horizontal mismatches at different altitudes. Further
work is needed to establish the best methodology to validate FTIR profile
retrievals while avoiding a difference in measurement geometries.
This work offers a new assessment of the accuracy and precision of FTIR
retrievals at different altitudes. The analysis consists of the comparison of
WV for several atmospheric layers using ODR and statistical analysis, i.e.,
estimation of accuracy and precision. Furthermore, we study the effect of
different WV a priori profiles commonly used among NDACC stations (ERA-I, NCEP, and
WACCM profiles) and the limitations of the FTIR WV averaging kernels by
comparing unsmoothed and smoothed FPH profiles with FTIR retrievals. The
following overall conclusions can be drawn from the unsmoothed comparison of
WV using several layers: (1) using 6-hourly and daily ERA-I a priori profiles shows
the best correlation and comparison at both sites; (2) the lowest bias and
precision are found in the closest layer to the instrument (1.5–3 km at BLD
and 3–5 km at MLO). At BLD, we report a negligible negative bias of -0.001±0.105×103ppm (-0.02±1.9 %) and precision of 0.21×103ppm (3.7 %) for the 1.5–3 km layer while at MLO the bias
is -0.10±0.08×103ppm (-5.8±4.6 %) and the precision
is
0.16×103ppm (9.2 %) for the 3–5.5 km layer, which are larger
likely due to the significant spatial mismatch difference between the
locations of measurements; (3) high vertical variability probed by the sonde
in the second layer is not fully captured by the retrievals, although it is
considerably better than a priori profiles; (4) and one significant finding
to emerge is that the retrievals show encouraging results in the
10.5–13.5 km layer
at BLD and at 13–16 km at MLO (roughly the UTLS layer) with 13.1±5.3 % (BLD) bias and a precision of 10.6 % (BLD) but the bias increases to
about 40 % above this layer. Table was constructed to
show a representative analysis when the spatial mismatch is known and when
the location of the FTIR and the launch of the sonde are near each other. In
this table results are shown for unsmoothed and smoothed FPH profiles.
According to these results we infer that the interpretation of the averaging
kernels and degrees of freedom are quite conservative and WV retrievals
contain more information than expected. Among all layers, the biases are
lower for unsmoothed FPH profiles, indicating limitations of WV averaging
kernels. The findings of this study show that FTIR profiles can be used to
evaluate long-term records of WV in several unique partial columns in the
troposphere. Further research would explore the additional WV absorption
features in order to improve the information content, e.g., micro-windows
employed in the latest MUSICA version. Also, as we show, the ERA-I WV
profiles yield lower biases; hence we would construct a priori covariance
matrices for these that maximize accuracy and vertical structure.
The second goal of this study was to investigate the influence of WV in the
retrieval of other tropospheric gas profiles with DOF larger than 2. Here
we present results for three important gases, i.e., HCN, CO, and
C2H6,
using the WV NOAA FPH profile as reference and comparing to other WV
profiles, including the retrieved WV, ERA-I, NCEP, and WACCM profiles. In general, our
results recommend retrieving WV profiles first then using them as input to
the retrievals of other gases in order to reduce bias due to an imperfect WV
vertical profile. As an example (Fig. ) we show relative
differences of up to 25 % at 8 km, 8 % at 4 km, and 10 % at 3 km for
HCN, CO, and C2H6 if WV is not retrieved beforehand and used as the
input WV profile. Overall, a statistical comparison of all profiles in the
1.5–3.0 km layer shows a significant impact on HCN (about 6 % bias),
moderate impact on CO (about 1.2 % bias), and low impact on C2H6 (<0.5 % bias). This sensitivity study is the first comprehensive
quantitative investigation in this topic and provides a basis for future
error budget assessment. In principle we hypothesize that the effect of WV
profiles might be larger in humid regions within the boundary layer but
further research should be carried out to establish its quantitative
importance.
The NCAR FTIR water vapor retrievals can be obtained from the authors
upon request. Vertical Profile of Water Vapor from Balloon flight NOAA can be accessed through the websites ftp://aftp.cmdl.noaa.gov/data/ozwv/WaterVapor/Boulder_New/
and ftp://aftp.cmdl.noaa.gov/data/ozwv/WaterVapor/Hilo_New/ (last access: 4 February 2019).
The supplement related to this article is available online at: https://doi.org/10.5194/amt-12-873-2019-supplement.
JWH installed the FTIRs. EGH, DFH, and AFJ implemented and evaluated the
FPH measurements. IO, RRB, and JWH performed FTIR measurements. IO and JWH evaluated the
FTIR measurements and designed the analysis. IO prepared the paper with contributions from all co-authors.
The authors declare that they have no conflict of
interest.
The National Center for Atmospheric Research is sponsored by the
National Science Foundation. Any opinions, findings, and conclusions or
recommendations expressed in this publication are those of the author(s) and
do not necessarily reflect the views of the National Science Foundation.
Acknowledgements
This study has been supported under contract by the National Aeronautics and
Space Administration (NASA). We are grateful to the NOAA staff at MLO for
technical support and maintenance of the NCAR FTIR. Especially, we wish to
thank Paul Fukumura. We would like to thank David Nardini and Darryl Kuniyuki
for diligently preparing and launching the NOAA FPH instruments monthly from
Hilo, Hawaii. We thank Helen Worden for her valuable suggestions during the
NCAR internal review.
Edited by: Martin Riese
Reviewed by: Matthias Schneider and one anonymous referee
ReferencesBarthlott, S., Schneider, M., Hase, F., Blumenstock, T., Kiel, M., Dubravica, D.,
García, O. E., Sepúlveda, E., Mengistu Tsidu, G., Takele Kenea, S.,
Grutter, M., Plaza-Medina, E. F., Stremme, W., Strong, K., Weaver, D., Palm, M.,
Warneke, T., Notholt, J., Mahieu, E., Servais, C., Jones, N., Griffith, D. W. T.,
Smale, D., and Robinson, J.: Tropospheric water vapour isotopologue data (H216O, H218O,
and HD16O) as obtained from NDACC/FTIR solar absorption spectra, Earth Syst. Sci. Data, 9, 15–29, 10.5194/essd-9-15-2017, 2017.Dammers, E., Shephard, M. W., Palm, M., Cady-Pereira, K., Capps, S., Lutsch, E.,
Strong, K., Hannigan, J. W., Ortega, I., Toon, G. C., Stremme, W., Grutter, M.,
Jones, N., Smale, D., Siemons, J., Hrpcek, K., Tremblay, D., Schaap, M.,
Notholt, J., and Erisman, J. W.: Validation of the CrIS fast physical NH3
retrieval with ground-based FTIR, Atmos. Meas. Tech., 10, 2645–2667, 10.5194/amt-10-2645-2017, 2017.Dee, D. P., Uppala, S. M., Simmons, A. J., Berrisford, P., Poli, P., Kobayashi,
S., Andrae, U., Balmaseda, M. A., Balsamo, G., Bauer, P., Bechtold, P.,
Beljaars, A. C. M., van de Berg, L., Bidlot, J., Bormann, N., Delsol, C.,
Dragani, R., Fuentes, M., Geer, A. J., Haimberger, L., Healy, S. B.,
Hersbach, H., Hólm, E. V., Isaksen, L., Kållberg, P., Köhler, M.,
Matricardi, M., McNally, A. P., Monge-Sanz, B. M., Morcrette, J.-J., Park,
B.-K., Peubey, C., de Rosnay, P., Tavolato, C., Thépaut, J.-N., and Vitart,
F.: The ERA-Interim reanalysis: configuration and performance of the data
assimilation system, Q. J. Roy. Meteor. Soc.,
137, 553–597, 10.1002/qj.828, 2011.Dessler, A. E.: Cloud variations and the Earth's energy budget, Geophys.
Res. Lett., 38, l19701, 10.1029/2011GL049236, 2011.Finger, F. G., Gelman, M. E., Wild, J. D., Chanin, M. L., Hauchecorne, A., and
Miller, A. J.: Evaluation of NMC Upper-Stratospheric Temperature Analyses
Using Rocketsonde and Lidar Data, B. Am. Meteorol.
Soc., 74, 789–800, 10.1175/1520-0477(1993)074<0789:EONUST>2.0.CO;2,
1993.Franco, B., Bader, W., Toon, G., Bray, C., Perrin, A., Fischer, E., Sudo, K.,
Boone, C., Bovy, B., Lejeune, B., Servais, C., and Mahieu, E.: Retrieval of
ethane from ground-based FTIR solar spectra using improved spectroscopy:
Recent burden increase above Jungfraujoch, J. Quant.
Spectros. Ra., 160, 36–49,
10.1016/j.jqsrt.2015.03.017, 2015.Franco, B., Mahieu, E., Emmons, L. K., Tzompa-Sosa, Z. A., Fischer, E. V.,
Sudo, K., Bovy, B., Conway, S., Griffin, D., Hannigan, J. W., Strong, K., and
Walker, K. A.: Evaluating ethane and methane emissions associated with the
development of oil and natural gas extraction in North America,
Environ. Res. Let., 11, 044010,
10.1088/1748-9326/11/4/044010, 2016.Galewsky, J. and Rabanus, D.: A Stochastic Model for Diagnosing Subtropical
Humidity Dynamics with Stable Isotopologues of Water Vapor, J.
Atmos. Sci., 73, 1741–1753, 10.1175/JAS-D-15-0160.1, 2016.García, O. E., Schneider, M., Hase, F., Blumenstock, T., Sepúlveda, E.,
and González, Y.: Quality assessment of ozone total column amounts as
monitored by ground-based solar absorption spectrometry in the near infrared
(>3000cm-1), Atmos. Meas. Tech., 7, 3071–3084, 10.5194/amt-7-3071-2014, 2014.Garcia, R. R., Marsh, D. R., Kinnison, D. E., Boville, B. A., and Sassi, F.:
Simulation of secular trends in the middle atmosphere, 1950–2003, J.f Geophys. Res.-Atmos., 112,
d09301, 10.1029/2006JD007485, 2007.
Goff, J.: Saturation pressure of water on the new Kelvin temperature scale,
Transactions of the American society of heating and ventilating engineers,
63, 347–354, 1957.Hall, E. G., Jordan, A. F., Hurst, D. F., Oltmans, S. J., Vömel, H., Kühnreich, B.,
and Ebert, V.: Advancements, measurement uncertainties, and recent comparisons
of the NOAA frost point hygrometer, Atmos. Meas. Tech., 9, 4295–4310, 10.5194/amt-9-4295-2016, 2016.Hannigan, J. W., Coffey, M. T., and Goldman, A.: Semiautonomous FTS
observation system for remote sensing of stratospheric and tropospheric
gases, J. Atmos. Ocean. Tech., 26, 1814–1828,
10.1175/2009JTECHA1230.1, 2009.Hase, F., Hannigan, J. W., Coffey, M. T., Goldman, A., Höpfner, M.,
Jones, N. B., Rinsland, C. P., and Wood, S. W.: Intercomparison of retrieval
codes used for the analysis of high-resolution, ground-based FTIR
measurements, J. Quant. Spectrosc. Ra.,
87, 25–52, 10.1016/j.jqsrt.2003.12.008, 2004.Held, I. M. and Soden, B. J.: Water vapor feedback and global warming, Annu.
Rev. Energ. Env., 25, 441–475,
10.1146/annurev.energy.25.1.441, 2000.Hurst, D. F., Hall, E. G., Jordan, A. F., Miloshevich, L. M., Whiteman, D. N.,
Leblanc, T., Walsh, D., Vömel, H., and Oltmans, S. J.: Comparisons of temperature,
pressure and humidity measurements by balloon-borne radiosondes and frost point
hygrometers during MOHAVE-2009, Atmos. Meas. Tech., 4, 2777–2793, 10.5194/amt-4-2777-2011, 2011a.Hurst, D. F., Oltmans, S. J., Vömel, H., Rosenlof, K. H., Davis, S. M.,
Ray, E. A., Hall, E. G., and Jordan, A. F.: Stratospheric water vapor trends
over Boulder, Colorado: Analysis of the 30 year Boulder record, J.
Geophys. Res.-Atmos., 116, 1–12, 10.1029/2010JD015065,
2011b.Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L.,
Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Y., Chelliah, M.,
Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K. C., Ropelewski, C., Wang,
J., Leetmaa, A., Reynolds, R., Jenne, R., and Joseph, D.: The NCEP/NCAR
40-Year Reanalysis Project, B. Am. Meteorol. Soc.,
77, 437–472, 10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2,
1996.Kiehl, J. T. and Trenberth, K. E.: Earth's Annual Global Mean Energy Budget,
B. Am. Meteorol. Soc., 78, 197–208,
10.1175/1520-0477(1997)078<0197:EAGMEB>2.0.CO;2,
1997.Kille, N., Baidar, S., Handley, P., Ortega, I., Sinreich, R., Cooper, O. R.,
Hase, F., Hannigan, J. W., Pfister, G., and Volkamer, R.: The CU mobile Solar
Occultation Flux instrument: structure functions and emission rates
of NH3, NO2 and C2H6, Atmos. Meas. Tech., 10, 373–392, 10.5194/amt-10-373-2017, 2017.Kinnison, D. E., Brasseur, G. P., Walters, S., Garcia, R. R., Marsh, D. R.,
Sassi, F., Harvey, V. L., Randall, C. E., Emmons, L., Lamarque, J. F., Hess,
P., Orlando, J. J., Tie, X. X., Randel, W., Pan, L. L., Gettelman, A.,
Granier, C., Diehl, T., Niemeier, U., and Simmons, A. J.: Sensitivity of
chemical tracers to meteorological parameters in the MOZART-3 chemical
transport model, J. Geophys. Res.-Atmos., 112, d20302,
10.1029/2006JD007879, 2007.Korn, G. and Korn, T.: Mathematical Handbook for Scientists and Engineers:
Definitions, Theorems, and Formulas for Reference and Review, Dover Civil and
Mechanical Engineering Series, Dover Publications,
available at: https://books.google.com/books?id=xHNd5zCXt-EC (last access: 25 January 2019), 2000.Marsh, D. R., Mills, M. J., Kinnison, D. E., Lamarque, J.-F., Calvo, N., and
Polvani, L. M.: Climate Change from 1850 to 2005 Simulated in CESM1(WACCM),
J. Climate, 26, 7372–7391, 10.1175/JCLI-D-12-00558.1, 2013.Noone, D.: Pairing Measurements of the Water Vapor Isotope Ratio with Humidity
to Deduce Atmospheric Moistening and Dehydration in the Tropical
Midtroposphere, J. Climate, 25, 4476–4494,
10.1175/JCLI-D-11-00582.1, 2012.Oltmans, S. J., Vomel, H., Hofmann, D. J., Rosenlof, K. H., and Kley, D.: The
increase in stratospheric water vapor from balloonborne, frostpoint
hygrometer measurements at Washington, D.C., and Boulder, Colorado,
Geophys. Res. Lett., 27, 3453–3456, 10.1029/2000GL012133,
2000.Paton-Walsh, C., Deutscher, N. M., Griffith, D. W. T., Forgan, B. W., Wilson,
S. R., Jones, N. B., and Edwards, D. P.: Trace gas emissions from savanna
fires in northern Australia, J. Geophys. Res.-Atmos.,
115, d16314, 10.1029/2009JD013309, 2010.Pougatchev, N. S., Connor, B. J., and Rinsland, C. P.: Infrared measurements of
the ozone vertical distribution above Kitt Peak, J. Geophys. Res.-Atmos., 100, 16689–16697, 10.1029/95JD01296, 1995.Rinsland, C. P., Jones, N. B., Connor, B. J., Logan, J. A., Pougatchev, N. S.,
Goldman, A., Murcray, F. J., Stephen, T. M., Pine, A. S., Zander, R., Mahieu,
E., and Demoulin, P.: Northern and southern hemisphere ground-based infrared
spectroscopic measurements of tropospheric carbon monoxide and ethane,
J. Geophys. Res.-Atmos., 103, 28197–28217,
10.1029/98JD02515, 1998.
Rodgers, C. D.: Inverse Methods for Atmospheric Sounding: Theory and Practice,
World Scientific, Singapore, 2000.Rodgers, C. D. and Connor, B. J.: Intercomparison of remote sounding
instruments, J. Geophys. Res.-Atmos., 108, 4116,
10.1029/2002JD002299, 2003.Rothman, L., Gordon, I., Barbe, A., Benner, D., Bernath, P., Birk, M., Boudon,
V., Brown, L., Campargue, A., Champion, J.-P., Chance, K., Coudert, L., Dana,
V., Devi, V., Fally, S., Flaud, J.-M., Gamache, R., Goldman, A., Jacquemart,
D., Kleiner, I., Lacome, N., Lafferty, W., Mandin, J.-Y., Massie, S.,
Mikhailenko, S., Miller, C., Moazzen-Ahmadi, N., Naumenko, O., Nikitin, A.,
Orphal, J., Perevalov, V., Perrin, A., Predoi-Cross, A., Rinsland, C.,
Rotger, M., Šimečková, M., Smith, M., Sung, K., Tashkun, S., Tennyson, J.,
Toth, R., Vandaele, A., and Auwera, J. V.: The HITRAN 2008 molecular
spectroscopic database, J. Quant. Spectrosc. Ra., 110, 533–572, 10.1016/j.jqsrt.2009.02.013,
hITRAN, 2009.Rothman, L. S., Gordon, I. E., Babikov, Y., Barbe, A., Benner, D. C., Bernath,
P. F., Birk, M., Bizzocchi, L., Boudon, V., Brown, L. R., Campargue, A.,
Chance, K., Cohen, E. A., Coudert, L. H., Devi, V. M., Drouin, B. J., Fayt,
A., Flaud, J., Gamache, R. R., Harrison, J. J., Hartmann, J., Hill, C.,
Hodges, J. T., Jacquemart, D., Jolly, A., Lamouroux, J., Le Roy, R. J., Li,
G., Long, D. A., Lyulin, O. M., Mackie, C. J., Massie, S. T., Mikhailenko,
S., Mueller, H. S. P., Naumenko, O. V., Nikitin, A. V., Orphal, J.,
Perevalov, V., Perrin, A., Polovtseva, E. R., Richard, C., Smith, M. A. H.,
Starikova, E., Sung, K., Tashkun, S., Tennyson, J., Toon, G. C., Tyuterev,
V. G., and Wagner, G.: The HITRAN 2012 molecular spectroscopic database,
J. Quant. Spectrosc. Ra., 130, 4–50,
10.1016/j.jqsrt.2013.07.002, 2013.Scherer, M., Vömel, H., Fueglistaler, S., Oltmans, S. J., and Staehelin, J.:
Trends and variability of midlatitude stratospheric water vapour deduced from
the re-evaluated Boulder balloon series and HALOE, Atmos. Chem. Phys., 8, 1391–1402, 10.5194/acp-8-1391-2008, 2008.Schneider, M. and Hase, F.: Ground-based FTIR water vapour profile
analyses, Atmos. Meas. Tech., 2, 609–619, 10.5194/amt-2-609-2009, 2009.Schneider, M., Hase, F., and Blumenstock, T.: Water vapour profiles by ground-based
FTIR spectroscopy: study for an optimised retrieval and its
validation, Atmos. Chem. Phys., 6, 811–830, 10.5194/acp-6-811-2006, 2006.Schneider, M., Yoshimura, K., Hase, F., and Blumenstock, T.: The ground-based FTIR network's
potential for investigating the atmospheric water cycle, Atmos. Chem. Phys., 10, 3427–3442, 10.5194/acp-10-3427-2010, 2010.Schneider, M., Barthlott, S., Hase, F., González, Y., Yoshimura, K., García, O. E.,
Sepúlveda, E., Gomez-Pelaez, A., Gisi, M., Kohlhepp, R., Dohe, S.,
Blumenstock, T., Wiegele, A., Christner, E., Strong, K., Weaver, D., Palm, M.,
Deutscher, N. M., Warneke, T., Notholt, J., Lejeune, B., Demoulin, P., Jones, N.,
Griffith, D. W. T., Smale, D., and Robinson, J.: Ground-based remote sensing of
tropospheric water vapour isotopologues within the project MUSICA, Atmos. Meas. Tech., 5, 3007–3027, 10.5194/amt-5-3007-2012, 2012.Schneider, M., Wiegele, A., Barthlott, S., González, Y., Christner, E., Dyroff, C.,
García, O. E., Hase, F., Blumenstock, T., Sepúlveda, E., Mengistu Tsidu, G.,
Takele Kenea, S., Rodríguez, S., and Andrey, J.: Accomplishments of the MUSICA
project to provide accurate, long-term, global and high-resolution observations
of tropospheric H2O,dD pairs – a review, Atmos. Meas. Tech., 9, 2845–2875, 10.5194/amt-9-2845-2016, 2016.
Seinfeld, J. H. and Pandis, S. N.: Atmospheric Chemisty and Physics: from air pollution to climate change, Wiley-Interscience, Hoboken, New Jersey, 2 edn.,
2006.Sepúlveda, E., Schneider, M., Hase, F., Barthlott, S., Dubravica, D., García, O. E.,
Gomez-Pelaez, A., González, Y., Guerra, J. C., Gisi, M., Kohlhepp, R.,
Dohe, S., Blumenstock, T., Strong, K., Weaver, D., Palm, M., Sadeghi, A.,
Deutscher, N. M., Warneke, T., Notholt, J., Jones, N., Griffith, D. W. T., Smale, D.,
Brailsford, G. W., Robinson, J., Meinhardt, F., Steinbacher, M., Aalto, T.,
and Worthy, D.: Tropospheric CH4 signals as observed by NDACC FTIR
at globally distributed sites and comparison to GAW surface in situ
measurements, Atmos. Meas. Tech., 7, 2337–2360, 10.5194/amt-7-2337-2014, 2014.Suortti, T. M., Kats, A., Rivi, R., Kämpfer, N., Leiterer, U.,
Miloshevich, L. M., Neuber, R., Paukkunen, A., Ruppert, P., Vömel, H.,
and Yushkov, V.: Tropospheric comparisons of Vaisala radiosondes and
balloon-borne frost-Point and Lyman-α hygrometers during the
LAUTLOS-WAVVAP experiment, J. Atmos. Ocean. Tech.,
25, 149–166, 10.1175/2007JTECHA887.1, 2008.Sussmann, R. and Borsdorff, T.: Technical Note: Interference errors in infrared remote
sounding of the atmosphere, Atmos. Chem. Phys., 7, 3537–3557, 10.5194/acp-7-3537-2007, 2007.Sussmann, R., Borsdorff, T., Rettinger, M., Camy-Peyret, C., Demoulin, P., Duchatelet, P.,
Mahieu, E., and Servais, C.: Technical Note: Harmonized retrieval of column-integrated
atmospheric water vapor from the FTIR network – first examples for long-term records
and station trends, Atmos. Chem. Phys., 9, 8987–8999, 10.5194/acp-9-8987-2009, 2009.Trenberth, K. E. and Asrar, G. R.: Challenges and Opportunities in Water Cycle
Research: WCRP Contributions, Surv. Geophys., 35, 515–532,
10.1007/s10712-012-9214-y, 2014.Tzompa-Sosa, Z. A., Mahieu, E., Franco, B., Keller, C. A., Turner, A. J.,
Helmig, D., Fried, A., Richter, D., Weibring, P., Walega, J., Yacovitch,
T. I., Herndon, S. C., Blake, D. R., Hase, F., Hannigan, J. W., Conway, S.,
Strong, K., Schneider, M., and Fischer, E. V.: Revisiting global fossil fuel
and biofuel emissions of ethane, J. Geophys. Res.-Atmos., 122, 2493–2512, 10.1002/2016JD025767, 2016.Viatte, C., Strong, K., Walker, K. A., and Drummond, J. R.: Five years of CO,
HCN, C2H6, C2H2, CH3OH, HCOOH and H2CO total
columns measured in the Canadian high Arctic, Atmos. Meas. Tech., 7, 1547–1570, 10.5194/amt-7-1547-2014, 2014.Vigouroux, C., Stavrakou, T., Whaley, C., Dils, B., Duflot, V., Hermans, C., Kumps, N.,
Metzger, J.-M., Scolas, F., Vanhaelewyn, G., Müller, J.-F., Jones, D. B. A.,
Li, Q., and De Mazière, M.: FTIR time-series of biomass burning products
(HCN, C2H6, C2H2, CH3OH, and HCOOH) at Reunion Island
(21∘ S, 55∘ E) and comparisons with model
data, Atmos. Chem. Phys., 12, 10367–10385, 10.5194/acp-12-10367-2012, 2012.Vigouroux, C., Blumenstock, T., Coffey, M., Errera, Q., García, O., Jones, N. B.,
Hannigan, J. W., Hase, F., Liley, B., Mahieu, E., Mellqvist, J., Notholt, J.,
Palm, M., Persson, G., Schneider, M., Servais, C., Smale, D., Thölix, L., and
De Mazière, M.: Trends of ozone total columns and vertical distribution
from FTIR observations at eight NDACC stations around the globe, Atmos. Chem. Phys., 15, 2915–2933, 10.5194/acp-15-2915-2015, 2015.Vogelmann, H., Sussmann, R., Trickl, T., and Reichert, A.: Spatiotemporal variability of
water vapor investigated using lidar and FTIR vertical soundings above the
Zugspitze, Atmos. Chem. Phys., 15, 3135–3148, 10.5194/acp-15-3135-2015, 2015.
Wild, J. D., Gelman, M. E., Miller, A. J., Chanin, M. L., Hauchecorne, A.,
Keckhut, P., Farley, R., Dao, P. D., Meriwether, J. W., Gobbi, G. P.,
Congeduti, F., Adriani, A., McDermid, I. S., McGee, T. J., and Fishbein,
E. F.: Comparison of stratospheric temperatures from several lidars, using
National Meteorological Center and microwave limb sounder data as transfer
references, J. Geophys. Res.-Atmos., 100,
11105–11111, 10.1029/95JD00631,
1995.Wu, C. and Yu, J. Z.: Evaluation of linear regression techniques for atmospheric
applications: the importance of appropriate weighting, Atmos. Meas. Tech., 11, 1233–1250, 10.5194/amt-11-1233-2018, 2018.