Introduction
Formaldehyde (HCHO) is the most abundant aldehyde and one of the
highest reactive species in the Earth's atmosphere, with a mid-day
lifetime on the order of a few hours . Being
a product of the oxidation of most of the volatile organic compounds
(VOCs), HCHO plays a primary role in tropospheric chemistry and in
the control of air quality. Besides dry and wet deposition of
generally lesser importance , HCHO is predominantly
removed from the atmosphere via catalytic pathways that are
photochemically driven . Indeed, large losses occur
when HCHO reacts with the hydroxyl radicals (OH) available in the
atmosphere to yield water vapor (H2O) and the HCO radical.
By consuming OH, the main sink of methane (CH4) in the
troposphere, the HCHO abundance contributes to increasing the
lifetime of this major greenhouse gas. HCHO also decomposes by
photolysis either in H2 (molecular hydrogen) and CO (carbon
monoxide) or in H (hydrogen) and HCO. Since HCO reacts with oxygen,
all these catalytic pathways release CO and hydrogen oxides
(HOx), making HCHO a key component in the production of
CO by oxidation of hydrocarbons . As
HOx radicals are major oxidizers in the atmosphere, HCHO
also constitutes a useful tracer of the atmospheric oxidative
capacity . Moreover, in the presence of sufficient
amounts of nitrogen oxides (NOx), for instance in polluted
air masses over urban areas , the HOx
produced by the HCHO removal processes converts nitric oxide to
nitrogen dioxide, which results in a net production of tropospheric
ozone by photolysis O3;.
On the global scale, CH4 oxidation by OH constitutes the main
source throughout the background troposphere, accounting for more
than half of the overall production, while the remainder generally
results from the oxidation of most of the non-methane VOCs (NMVOCs).
However, where strong NMVOC emissions take place over continents,
the oxidation of these compounds can dominate the methane-originated
HCHO production, especially in the continental boundary layer
. Among the NMVOCs
emitted over continents, biogenic compounds are dominant during the
growing season of vegetation, providing ∼85 % of the
total emissions, with the largest contribution coming from isoprene
e.g.,. Global HCHO production from
anthropogenic NMVOCs is relatively reduced, but is significantly
enhanced over largely populated and industrialized areas. NMVOCs
from pyrogenic sources (mainly biomass burning) generally provide
small HCHO contributions on the global scale, although fire events
can enhance HCHO emissions in specific areas
see, e.g.,. In addition, only a negligible fraction
of HCHO (< 1 %) results from direct emissions to the
atmosphere by various sources such as biomass burning, vegetation
or incomplete fossil fuel combustion
e.g.,.
Formaldehyde has already been intensely observed, using measurements
obtained from in situ instruments
e.g.,, aircraft campaigns
e.g., and various satellite
sensors e.g.,,
as well as ground-based remote measurements derived from UV-Visible
passive Multi-AXis Differential Optical Absorption Spectroscopy
(MAX-DOAS) instruments e.g.,
and from high-resolution infrared solar spectra recorded with Fourier
transform infrared (FTIR) spectrometers
e.g.,. However,
few long-term trends of HCHO loadings exist that are suitable for
trend analysis, particularly due to the
lack of extended consistent data sets. Offering regular and quasi
global geographical sampling of the Earth's atmosphere, UV-Vis
satellite sensors such as SCIAMACHY (SCanning Imaging Absorption
spectroMeter for Atmospheric CHartographY), GOME, GOME-2 (Global
Ozone Monitoring Experiment) and OMI (Ozone Monitoring Instrument)
provide some sensitivity in the free troposphere and have been used
recently to produce regional decadal trends of HCHO columns at the
global scale . Nonetheless, most space-borne
observational campaigns are time-limited, added to the fact that
such measurements can be considerably affected by instrumental
ageing as well as by noise and error sources in the lowermost
atmospheric layers, where the bulk of HCHO lies.
Although the seasonal intra-day variation of HCHO has been studied
in field campaigns in different environments
or using ground-based MAX-DOAS and space-borne UV-Vis measurements at
various locations , consistent diurnal
observations of HCHO columns remain sparse and time-limited. Hence
the uncertainties on the intra-day cycle remain high, added to the
fact that the diurnal pattern of HCHO may vary considerably from site
to site according to many local factors, such as the emissions of
NMVOC precursors, the chemical regime and the
influence of the planetary boundary layer. More particularly, the
HCHO diurnal cycle may be responsible for significant short-term
variability that needs to be accounted for when comparing results
derived from space-borne instruments, according to their respective
overpass times. Furthermore, the HCHO intra-day modulation remains
incompletely captured by the chemistry transport models (CTMs),
especially for remote conditions . Therefore, the
characterization of the HCHO diurnal cycle using high-quality
ground-based observations is definitely required for
validation/calibration efforts of satellite sensors and models.
Ground-based instruments, such as the high-resolution FTIR
spectrometers distributed worldwide at strategic locations and part
of the Network for the Detection of Atmospheric Climate Change (NDACC;
see http://ndacc.org), are important contributors to the
monitoring of the Earth's atmosphere. An optimized retrieval strategy
has been recently developed to derive HCHO total columns from
ground-based FTIR solar spectra recorded at the high-altitude NDACC
station of Jungfraujoch (Swiss Alps, 46.5∘ N,
8.0∘ E, 3580 ma.s.l.). The results have been
successfully validated against MAX-DOAS measurements and simulation
of two CTMs, GEOS-Chem
Goddard Earth Observing System – chemical transport model;
and IMAGES v2
Intermediate Model of Annual and Global Evolution of Species;,
over the 2010–2012 time period . The Jungfraujoch FTIR
observational database now covers more than 30 years (back to 1988 in
the case of HCHO observations) and is unique worldwide in terms of the
quality and density of the measurements as well as of their temporal
coverage. Time series of high-quality geophysical data derived from this
database are particularly appropriate for multi-decadal studies of many
important constituents of the Earth's atmosphere, including HCHO and
its VOC precursors. Used as comparative and complementary data, they
are also of crucial importance for the calibration and validation of
models as well as of current and future satellite sensors.
In the present study, we use the observational database of
ground-based solar spectra recorded by two high-resolution FTIR
spectrometers operated at the Jungfraujoch station, in order to produce
a 27-year time series of the HCHO total column (from 1988 to mid-2015). To
the best of our knowledge, it represents the longest time series of remote
HCHO measurements. We first investigate the intra-day variation of
HCHO total columns in the remote troposphere, using a consistent subset
of observations spanning more than 20 years. As this 20-year subset
provides robust statistics without inducing errors and/or biases
resulting from the use of different sources of measurements, we also
characterize the HCHO diurnal cycle on a monthly basis by adjusting
a fitting parametric model to the observed intra-day variations. These
parameters being determined according to the observations, we employ
this model to scale all individual HCHO measurements of the entire
Jungfraujoch database at a given moment of the day with the aim of
removing the effect of intra-day variability in the HCHO time series.
Such a parametric model will be useful for carrying out comparisons
between ground-based FTIR and space-borne UV-Vis HCHO columns at
the overpass time specific to each satellite sensor (e.g., 09:30 LT
for GOME-2B and 13:30 LT for OMI).
In the second part of this study, we employ simulations made with
the state-of-the-art GEOS-Chem 3-D CTM to investigate the main
sources and drivers of HCHO above Jungfraujoch. First we compare
the ground-based FTIR observations with HCHO total columns simulated
by the CTM, taking into account the vertical resolution and specific
sensitivity of the FTIR retrievals. Then sensitivity runs are
performed with the aim of evaluating the contribution of different
precursor species or source category (from anthropogenic, biogenic
and biomass burning sources) to the HCHO loadings and seasonality.
Finally, we analyze the multi-decadal FTIR time series of the
Jungfraujoch station (1988–2015) in order to study the
inter-annual variability and deduce trends of HCHO columns in the
remote troposphere of the mid-latitudinal Northern Hemisphere.
This paper is organized as follows: we briefly introduce in
Sect. the FTIR instrumental setups and data sets, as well
as the GEOS-Chem model. In Sect. , we investigate the
HCHO diurnal variation and describe the fitting parametric model and
how it is adjusted to the observations. We report in Sect.
the results of the comparison between FTIR measurements and GEOS-Chem
simulations, as well as of the sensitivity runs. Section
is devoted to the analysis of the 1988–2015 time series of HCHO
total columns above the Jungfraujoch station, involving trends
determination. Section concludes this study with
discussions of the results and identifies avenues for future work.
Data sets
FTIR instrumentation and measurements
The long-term HCHO time series presented and investigated in this
study has been produced from the analysis of solar spectra recorded
between January 1988 and June 2015 under clear-sky conditions at
the high-altitude International Scientific Station of the
Jungfraujoch (hereafter ISSJ; Swiss Alps, 46.5∘ N,
8.0∘ E, 3580 ma.s.l.). These spectra were
recorded using two very high spectral resolution FTIR
spectrometers. A “home-made” instrument was primarily in operation
until the mid-1990s and then progressively replaced by a more
sensitive commercial Bruker-120 HR instrument. A thorough
description of the instrumentation is given by . The
consistency among the HCHO columns retrieved from the two subsets
is evaluated in Sect. using all available
coincident measurement days of 1995–1997.
The ISSJ is mainly located in the free troposphere during winter
and represents a unique site to study the atmospheric background
conditions over central Europe. During the rest of the year,
injections of air masses from the planetary boundary layer can
occur, bringing biogenic and anthropogenic pollutants from the
nearby valleys. Furthermore, the very high dryness due to the
altitude, combined with the presence of the Aletsch Glacier (adding
to the local dryness) in the immediate vicinity of the station,
limits significantly the interference by water vapor in the
infrared solar measurements. More details regarding the
measurement site can be found in and ,
as well as in references therein.
The overall database investigated here consists of 10 979 solar
spectra, of which 1436 were recorded by the “home-made”
spectrometer over the 1988–1997 period and 9542 were obtained
with the Bruker instrument between 1995 and June 2015 (referred
to below as the LIEGE and BRUKER data sets, respectively), both
equipped with indium antimonide (InSb) cooled detectors. The
spectra were recorded using optical filters maximizing the
signal-to-noise (S / N) ratio over the 2400–3310 cm-1
spectral domain, thus encompassing the ν1 infrared band
of HCHO centered at 2783 cm-1. The typical spectral
resolution (defined here as twice the maximum optical path
difference) alternates between 0.003 and 0.005 cm-1
for both instruments, with the highest resolution adopted for
minimum variation of the air-mass and observing geometry,
predominantly around midday. S / N ratios vary from 550 to
more than 3100 (average spectra resulting from several
successive individual scans).
Summary of the retrieval strategy for HCHO applied in this study
above Jungfraujoch. A thorough description is given by .
Retrieval code
SFIT-2 v3.91 algorithm
A priori profile
From WACCM v6 simulations
Regularization
Optimal estimation method
Covariance profile
From WACCM v6 simulations (multiplied by 1.5)
Spectroscopic database
HITRAN 2008
Microwindows (and interfering species)
2763.425–2763.600 cm-1 (HDO, CH4, O3, N2O, CO2)
2765.725–2765.975 cm-1 (HDO, CH4, O3, N2O, CO2)
2778.200–2778.590 cm-1 (HDO, CH4, O3, N2O, CO2)
2855.650–2856.400 cm-1 (HDO, CH4, O3, N2O, H2O)
The retrieval strategy applied to both spectral data sets is
the one developed and described by . A short
summary of this strategy is given in Table .
Characterization of the FTIR retrievals indicates a sensitivity
throughout the troposphere (up to 12 km altitude). The
mean degree of freedom for signal (DOFS) over the entire
data set is ∼1; hence, only total columns of HCHO may be
obtained. In addition, the individual observations characterized
by a DOFS value lower than 0.35 have been discarded. This DOFS
threshold was set arbitrarily to ensure sufficient statistics at
low zenith angles for representing the HCHO diurnal cycle around
mid-day (see more details in Sect. ).
also provide a complete error budget of the HCHO measurements,
quoting the total systematic and random components at ±14
and ±21 %, respectively.
GEOS-Chem simulations and processing
GEOS-Chem is a global 3-D chemical transport model capable of
simulating global trace gas (more than 100 tracers) and aerosol
distributions. In the present study, GEOS-Chem is driven by the
Goddard Earth Observing System v5 (GEOS-5) assimilated
meteorological fields from the NASA Global Modeling Assimilation
Office (GMAO). The GEOS-5 data are available at a native
horizontal resolution of 0.5∘ × 0.667∘ and
at a 6 h temporal frequency (3 h for surface
variables and mixing depths). These meteorological fields
provide a description of the atmosphere on the basis of 72
hybrid pressure σ levels from the surface up to
0.01 hPa. For the simulations used here, the GEOS-5
meteorological fields are degraded to
a 2∘ × 2.5∘ horizontal resolution and 47
vertical levels, lumping together levels above
∼80 hPa. We apply the standard full chemistry
GEOS-Chem simulation that includes detailed
O3–NOx–VOC–aerosol coupled chemistry
described by and , with updates
by .
Conversely to , who used GEOS-Chem version 9-01-03,
we employ here model version 9-02
(http://acmg.seas.harvard.edu/geos/doc/archive/man.v9-02/index.html)
that implements a new isoprene oxidation scheme as standard. This
chemistry is largely based on the work of
and has been proven to yield more HCHO from the isoprene oxidation
pathways for weakly polluted conditions under very
low NOx concentrations; see. Nonetheless,
results provided by version 9-01-03 of GEOS-Chem (according
to the same standard full chemistry simulation) are also
provided hereafter for comparison purposes. The isoprene
oxidation scheme applied in GEOS-Chem v9-01-03 is described in
.
In GEOS-Chem, the global biomass burning emissions are derived
from the Global Fire Emissions Database (GFED) v3
and the global biogenic emissions are obtained with the Model of
Emissions of Gases and Aerosols from Nature (MEGAN) v2.0
in GEOS-Chem v9-01-03 and v2.1 in
version 9-02. Over Europe, the anthropogenic emissions of CO,
NOx, SOx (sulfur oxides), ammonia and NMVOCs
are provided by the European Monitoring and Evaluation Programme
(EMEP; http://www.ceip.at/) regional inventory for the year
2010 , except for ethane and propane, which are
derived from an offline simulation , and acetone,
which is from the REanalysis of the TROpospheric chemical
composition (RETRO;
http://gcmd.gsfc.nasa.gov/records/GCMD_GEIA_RETRO.html)
global emission inventory for the base year 2000. The
global and regional inventories are scaled to the years of
interest using energy statistics . Annual average
CH4 concentrations are prescribed over four latitude
bands (0–30; 30–90∘) on the basis
of CH4 measurements from the NOAA Global Monitoring
Division flask measurements.
In addition to the standard full chemistry simulations of
GEOS-Chem v9-02 (hereafter called standard runs), three other
runs also implementing the standard full chemistry have been
performed with version 9-02, but in each of them either
the anthropogenic, biogenic or biomass burning emissions of
NMVOCs and NO (nitric oxide) have been turned off. These
additional simulations are referred to below as
non-anthropogenic, non-biogenic and non-biomass burning runs,
respectively. It is worth noting that CH4
concentrations in these sensitivity runs are still derived
from the NOAA measurements, as for the standard simulation.
The different GEOS-Chem data sets used in the present study
are obtained from July 2005 to May 2013 simulations, for which
the GEOS-5 meteorological fields are available. These
simulations were preceded by a 1-year run for chemical
initialization, restarted several times to remove the effect
of initial conditions.
The GEOS-Chem outputs consist of HCHO volume mixing ratio
(VMR) profiles simulated at the closest pixel to the
Jungfraujoch station and saved at a 3 h time step.
The comparisons between the GEOS-Chem simulations and the
FTIR total columns account for the vertical resolution and
sensitivity of the FTIR retrievals. To this end, the
individual VMR profiles simulated by GEOS-Chem have been
regridded onto the 39-layer vertical scheme adopted at the
ISSJ for the FTIR retrievals, according to a mass
conservative interpolation.
Then these profiles have been smoothed according to the
formalism of by convolving them with the FTIR
averaging kernels, seasonally averaged over March–May,
June–August, September–November and December–February,
as well as over successive years, on the basis of the July
2005–May 2013 FTIR data set. The GEOS-Chem total columns
have eventually been computed from these smoothed VMR
profiles by using the corresponding regridded air density
profiles simulated by the model. The comparisons between the
FTIR measurements and the GEOS-Chem simulations are performed
over the 919 days with observations available in the July 2005–May 2013 time range.
Intra-day variation of the HCHO abundance, represented by the
0.5 h time step average (as grey circles) and corresponding standard
error of the mean (as error bars) of the HCHO total columns (in
moleccm-2) derived from all individual observations made by the
Bruker instrument between January 1995 and June 2015 at the ISSJ. It is
expressed according to the hour of the day (UTC + 1) and for each month
of the year. No smoothing has been applied here. The solid color curves are
the fits of the monthly running averages of the individual observations by
the parametric model (Eq. ), associated with the 95 %
confidence and prediction intervals delimited by the dotted and dashed color
curves, respectively. The running average used here corresponds to
a 0.5 h time step and a 1.5 h wide integration length. The
coefficient of determination (R2) of the model fit is provided for each
month in the legend. All the monthly fits by the parametric model are
displayed on the same graph in Fig. a,
Sect. .
Formaldehyde diurnal variation
Observed intra-day modulation
In this section, we investigate the HCHO diurnal variation above
the ISSJ using the total column measurements derived from the
January 1995–June 2015 BRUKER data set, owing to their overall
improved temporal regularity and measurement density compared to
the LIEGE data set. The total columns have been normalized to
the mean annual pressure at the ISSJ (654 hPa) in order
to avoid the effects due to pressure variation between different
days/seasons in the retrieved HCHO columns.
Figure shows the intra-day modulation of the
HCHO abundance above Jungfraujoch averaged over each month of
the mean year (with the HCHO total columns averaged every
0.5 h as grey circles), according to the FTIR
observations made over the entire 1995–mid-2015 time period
(a global view of the observed and modeled monthly intra-day
cycles is available in Fig. a,
Sect. ).
At the global scale, the diurnal cycles of HCHO loadings depend
on local chemical regimes, which generally vary across the
seasons and determine the HCHO formation and loss, as well as
local emissions of NMVOC precursors and their diurnal
variability. For instance, it has been shown that daytime
photochemical production and anthropogenic NMVOC emissions drive
the HCHO diurnal modulation over highly populated and
industrialized areas, such as in Belgium and Holland
see. However, at a remote site such as the
ISSJ, located in the free troposphere during most of the
year, the overall sunrise to sunset modulation of the HCHO
total columns is predominantly driven by the atmospheric
photochemistry and the CH4 oxidation (see
Sect. ): enhanced insolation and higher
temperatures intensify the concentration in OH radicals and
hence the photochemical oxidation rate of VOC precursors of
HCHO. It is characterized by a.m. increases and
p.m. decreases of the HCHO columns, peaking around
mid-day and in the early afternoon. The amplitude of the
intra-day modulation varies largely from month to month: from
0.2×1015 moleccm-2 in December up
to 1.0×1015 moleccm-2 during
summertime. The weaker amplitude in winter is due to
the relatively high solar zenith angle (SZA) around noontime,
inducing less radiation, as well as to the weak moisture,
both hindering the formation of OH. A similar pattern of the HCHO
diurnal cycle was reported over the remote ocean by MAX-DOAS
measurements .
The FTIR measurements are irregularly distributed throughout
the day, with most of the observations performed before mid-day
because of frequent high cirrus cloud occurrences in the
afternoon as well as the mountainous summits around the station,
occulting the sun at SZA larger than ∼80∘
during wintertime see Fig. 4 in. As
a consequence, the relative uncertainty associated with the
p.m. observations increases (see the error bars in
Fig. ). Furthermore, the retrievals from
low-SZA spectra (around mid-day) are characterized by relatively
low DOFS values, such as illustrated in Fig. a,
due to the very weak solar absorptions by HCHO for low air masses.
This contributes to increasing the uncertainty in the retrieved
total columns and explains the fluctuations of the running
average observed around noontime during the summer months (see
Fig. ). The diurnal variation of the HCHO
abundance also shows for some months (e.g., August and September
in Fig. ) a small re-increase in the total columns at the
end of the day. This results from the fact that only
observations made during later (earlier) days of the month are
available at this moment for the first (last) 6 months of the
year (see Fig. b), due to the increasing
(shortening) day length. Given the typical seasonal cycle of
HCHO throughout the year, such measurements hence provide
somewhat larger (lower) HCHO columns.
Average of the mean DOFS values and the mean observational day
of the month (a and b, respectively) of all individual
observations made by the Bruker instrument between January 1995 and
June 2015 at the ISSJ, according to the hour of the day and for each
month of the year. The average has been calculated with a time step and
an integration length of 0.5 h (no smoothing has been applied
here).
Parametric model of the diurnal variation
The diurnal modulation of the HCHO abundance above Jungfraujoch
corresponds to a large short-term variability that should be
accounted for when comparing ground-based observations to, e.g.,
satellite measurements and model results. Moreover, it is worth
describing such a modulation in order to remove it before
investigating seasonal/inter-annual variability of HCHO in the
following parts of this study. Therefore, we have adjusted
a fitting parametric model to each monthly subset, the results
being presented in Fig. . To this end, the
intra-day modulation used to adjust the parametric model consists
of a running average (not shown in Fig. ) of
all individual FTIR measurements within each month, calculated
using a 0.5 h time step and a 1.5 h wide
integration length (compatible with the HCHO lifetime). The
smoothing associated with the running average helps to dampen
the contribution of very high HCHO loadings that correspond to
“polluting” events. The previous analysis has highlighted the fact that
modeling the HCHO diurnal cycle merely according to the seasons
would not capture the large differences observed from month to
month, especially during spring and fall. Hence we have also
adjusted the fitting parametric model while considering monthly
diurnal cycles, in order to keep enough p.m. observations
(i.e., statistics) for adjusting the model with high reliability.
The model used here (described in Eq. ) consists in
a re-parametrization of the standard statistical form of the
Weibull continuous probability distribution of
a random variable
x. In this study, it determines the HCHO total column (y) as
a density function of the hour of the day (x), according to the
amplitude (a), the scale parameter (b), the shape parameter
(c) and the location parameter (or threshold; x0)
of the distribution. The Weibull density function is a versatile
distribution capable of adopting the characteristics of other
types of distributions, according to the value of the shape
parameter (c), and is widely used to mimic peaking
distributions with asymmetric slopes.
y=ac-1c1-cc×x-x0b+c-1c1cc-1×exp-x-x0b+c-1c1cc+c-1c
The model has been adjusted to the observations and the
parameters calculated by using the iterative
Levenberg–Marquardt algorithm that minimizes
the sum of the squared differences between the observations
and the predicted values of the model until convergence occurs
(i.e., when the residuals no longer decrease significantly).
The model fit for each month is represented as solid color
curves in Fig. , along with the associated
95 % confidence and prediction bands. The coefficients
of determination (R2) calculated between the monthly
observations and model fits are high and range from 0.78 to
0.95 (see Fig. ). The parameters a, b,
c and x0 determined by the adjustments are given
for each month in Table S1 in the Supplement so that one may
reproduce the model fits using Eq. ().
The model is able to reproduce with reliability the diurnal
modulation of HCHO loadings above Jungfraujoch and allows for
its quantitative characterization for a typical day of each
month of the year, thanks to the large BRUKER statistical
database. Since this model is parameterized, we use it in this
study to scale individual FTIR measurements on a given moment
of the day before carrying out a comparison with GEOS-Chem
outputs (see Sect. ) and a long-term trend
study (see Sect. ). Nonetheless, such a model
cannot be used to extrapolate the HCHO total columns outside
the actual range of measurements. In addition, this model is
only valid if the condition in Eq. () is true; otherwise,
the Weibull distribution collapses and the results are null
(e.g., for the very first hours of the day).
x>x0-bc-1c1c
Given the lower DOFS values characterizing the retrievals performed
around mid-day, especially in summer (see Fig. a),
the a priori profile adopted for the retrievals will have an impact
on the HCHO total columns obtained from such observations. The a
priori profile used in for the HCHO
retrieval is derived from WACCM v6
Whole Atmosphere Community Climate Model; see, e.g.,
simulations above ISSJ over
the 1980–2020 period. According to sensitivity tests adopting either
a lower or higher realistic a priori profile (i.e., derived from
ACE-FTS occultation measurements and GEOS-Chem simulations; see
Fig. S1 in the Supplement), the diurnal cycle fitted by the
parametric model will show somewhat lower or larger amplitude,
respectively (see Fig. S2). We gauge at ±10 % the
maximum systematic uncertainty that can affect the fit of the HCHO
intra-day variations due to the choice of the a priori profile.
Such maximum uncertainty is reached around noontime for the summer
months only, and as such can be considered highly conservative.
Monthly diurnal cycles of HCHO total columns above Jungfraujoch
adjusted by the parametric model (Eq. ) to the 1995–2015 BRUKER
FTIR observations (a) and simulated by the standard run of GEOS-Chem
over the July 2005–May 2013 time period (b). The red circles in
frame (a) illustrate an example of the HCHO column derived from an
individual observation made in June at
13:02 (UTC + 1) and scaled to
09:00 using the parametric model of HCHO intra-day variation (see explanation
in Sect. ).
In order to remove the effect of the intra-day cycle, the
pressure-normalized total columns obtained from all individual
measurements have been scaled to 09:00 (UTC + 1) on the
basis of the parametric model described previously. Using
Eq. () that links the daytime to the HCHO columns, as
well as the constant parameters determined for each month, the
total column derived from a solar spectrum is scaled according to
the difference between the actual time of the corresponding
observation and 09:00 (UTC + 1). An example is illustrated
for June by the color circles in Fig. a (see
Sect. ): an initial total column of
2.356×1015 moleccm-2 derived from an
observation made at 13:02 (UTC + 1) is decreased to
2.072×1015 moleccm-2 when scaled at
09:00. Here we have chosen 09:00 as a reference
time because observations for every month are available at this
moment of the day, added to the fact that the gap between the
different monthly intra-day regimes in terms of HCHO columns is
minimal in the morning.
Formaldehyde simulated by GEOS-Chem
Comparison with the GEOS-Chem standard run
Figure presents the monthly diurnal variations
as adjusted by the parametric model on the basis of the FTIR
observations in Sect. (Fig. a)
and as derived from the 3 h outputs of the GEOS-Chem v9-02
standard run spanning the mid-2005–mid-2013 time period
(Fig. b). The diurnal modulation of HCHO
loadings above Jungfraujoch represents a large short-term
variability superimposed on seasonal and inter-annual variations
(Fig. a) that should be accounted for when
comparing ground-based observations to, e.g., model results.
However, Fig. b shows that GEOS-Chem does not
capture the observed monthly intra-day variation. That a CTM is
unable to simulate the observed HCHO diurnal modulation,
especially for remote conditions, has already been reported with
IMAGES v2 . This can be partially explained by the
fact that different oxidation pathways of VOC precursors leading
the HCHO production, which are numerous, might not be optimally
implemented (especially very short-lived VOCs) or merely not
considered in the model. In addition, large uncertainties remain
concerning the various sources of precursor emissions, their
geographical distribution and how this latter one can influence the
air masses over remote sites such as the ISSJ. Finally, the CTMs
dilute local enhancements in HCHO due to their relatively coarse
spatial resolution (2∘ × 2.5∘ here), which
also affects the comparison with observations.
For this reason, we compare in this section HCHO total columns
simulated by GEOS-Chem and obtained from the 09:00 model
outputs to the daily averaged FTIR total columns calculated on
the basis of the 09:00-scaled individual observations
(see Sect. ). This comparison is made over the
May 2005–June 2015 time period for the days with available
FTIR observations only. In this study, some outliers, likely due
to transport of air masses with high NMVOC precursor
concentrations up to ISSJ, have been discarded from the FTIR data
set. Indeed, the CTMs are generally unable to simulate such
“polluting” events for the reasons evoked above. These outliers
have been identified as the daily averaged HCHO total columns
with relative anomalies to the curve fitted by the method of
Sect. and Fig.
here below higher than the 95th percentile value of all
relative anomalies of the 2003–2015 data set.
Monthly averaged total columns of HCHO and associated 1σ bars displayed on a 1-year time base, from the individual
09:00 (UTC + 1) scaled FTIR measurements performed above the ISSJ between
July 2005 and May 2013. Note that the daily observation values with relative
anomalies to the curve fit calculated by
Sect. , higher than the 95th percentile
value of all relative anomalies of the data set, have been excluded from the
present data set. The green curve and shaded area show on a 1-year time
base the running mean fit to the daily averaged columns (with a 2-month wide
integration time and a 15-day time step) and the associated 1σ
standard deviation, respectively. The solid red line and shaded area
represent corresponding information, but deduced from the smoothed outputs of
the GEOS-Chem v9-02 standard run. The dashed red line corresponds to the same
1-year time base running mean, but obtained from the smoothed outputs of
the GEOS-Chem v9-01-03 standard run. Note that the 1σ standard
deviations around the running mean are calculated on the basis of the daily
averaged columns and hence include inter-annual fluctuations as well as
variability of the monthly mean.
Figure presents the good agreement
(R=0.72) in terms of seasonal cycles of HCHO loadings
above Jungfraujoch, obtained from the FTIR observations and such
as simulated by the GEOS-Chem v9-02 standard run. A similar
seasonal comparison over the mid-2010–2012 time range has been
carried out by , but involving results from the
GEOS-Chem v9-01-03 standard run (in dashed red line in
Fig. ). This comparison illustrates the higher
HCHO columns simulated by version 9-02 of GEOS-Chem compared
to version 9-01-03, due to the implementation of the new
isoprene oxidation scheme that enhances the
HCHO formation under very low NOx concentrations. We refer
to , Sect. 4.1, for the discussion regarding the
mean seasonal bias of the GEOS-Chem results to the FTIR HCHO total
columns, which is here -4.7 ± 31.3 %. As the
model does not reproduce the HCHO intra-day variations at the
ISSJ, this bias increases to -21.3 ± 26.4 % for the
comparison involving the 12:00 (UTC + 1) GEOS-Chem outputs and the
12:00-scaled individual FTIR observations (with
R=0.69).
(a) Monthly mean of the daily averaged HCHO total column
(in moleccm-2) above Jungfraujoch simulated by GEOS-Chem v9-02
over the July 2005–May 2013 time period, according to the standard and
sensitivity (i.e., non-anthropogenic, non-biomass burning and non-biogenic)
runs. In the sensitivity simulations, the anthropogenic, biomass burning and
biogenic emissions of NMVOCs and NO have been shut off, while the CH4
concentrations are still derived from NOAA measurements, as for the standard
simulation. The grey shaded area corresponds to the 1σ standard
deviation associated with the standard run. (b) HCHO total column
from the sensitivity runs, as relative to the HCHO amount simulated by the
standard run (in % of the latter). (c) The same
as (b) but for OH. (d) Monthly total carbon emissions (in
TgCmonth-1) by category, integrated over
30–60∘ N and -130–35∘ E.
GEOS-Chem sensitivity tests
In this part of the study, we investigate the influence of the
different categories of emission sources implemented in GEOS-Chem
v9-02 on the HCHO abundance simulated by the model above
Jungfraujoch. To this end, we compare the results from the
standard run and from the three sensitivity runs performed without
either anthropogenic, biomass burning or biogenic emissions of
NMVOCs and NO. Figure a shows the
monthly averaged HCHO total columns at the ISSJ, derived from
these simulations from July 2005 to May 2013.
Figure b presents the HCHO total columns from
the three sensitivity runs relative to the amount from the
standard run.
In Fig. a and b, it can be
seen that none of the missing emission sources significantly
impacts the simulated HCHO loadings in summer at the ISSJ, with
the HCHO columns derived from either the non-anthropogenic or
non-biogenic runs still accounting for ∼95 % of the
total columns from the standard run (Fig. b).
During this season, biogenic primary NMVOCs such as isoprene may
constitute a significant source of HCHO, especially in the
continental boundary layer. However, it is most likely that
a large part of these very short-lived NMVOCs are already oxidized
before being transported to the ISSJ. Hence these compounds do not
contribute directly to the HCHO loadings above Jungfraujoch,
although they release biogenic secondary products that can be
transported to the upper troposphere and in turn can be oxidized,
adding to the HCHO abundance in the upper tropospheric layers.
During wintertime, the absence of anthropogenic emissions
decreases the HCHO burden down to 75 % of the standard
run (Fig. b), with a December–February average
of 82 % over July 2005–May 2013. Due to their longer
lifetimes and more intensive anthropogenic combustion during this
season, more elevated concentrations in reactive anthropogenic
compounds can be measured in winter at the ISSJ
, which contributes to the HCHO
formation. Finally, the missing biomass burning emissions have
almost no effect on the simulated HCHO loadings above Jungfraujoch
(Fig. b). These tests suggest that the
contribution of anthropogenic, biogenic and pyrogenic NMVOCs to
the HCHO burden above Jungfraujoch is quite limited, and that the
oxidation of CH4 (not included in the emission sources
shut off here) by OH dominates the HCHO production and constitutes
the main driver of its seasonality.
It is worth noting that turning off the emission sources in the
GEOS-Chem sensitivity tests investigated here may eventually
result in slightly enhanced HCHO amounts (by 2–3 %)
produced by GEOS-Chem compared to the standard run, as shown in
Fig. b for the non-biomass burning run and, to
a lesser extent, for the non-biogenic run during winter. In these
particular cases, with part of the emission sources missing
globally, enhanced HCHO might be due to a lower concentration in
associated NMVOCs for which OH is the main sink, inducing enhanced
OH loadings above Jungfraujoch available for oxidation of other
VOC precursors of HCHO such as CH4. Indeed,
Fig. c illustrating the monthly averaged OH
amounts simulated by the different sensitivity runs, relative
to the standard run, shows OH total columns from the non-biomass
burning and non-biogenic runs increased by up to 10 %.
However, it is hard to evaluate the impact of the excluded NO
emissions, since this compound plays a key role in both HCHO
formation (through the NMVOC oxidation pathways) and destruction
(by contributing to the regeneration of OH). Investigating the
balance between all these reactions would require a specific
study that is beyond the scope of the present work.
According to the GEOS-Chem simulation performed without
anthropogenic emissions, the wintertime contribution of
anthropogenic NMVOC precursors to the HCHO total column above
Jungfraujoch varies from year to year (from 15 to 25 %;
Fig. b). However, the wintertime total carbon
emissions of anthropogenic origin as implemented in GEOS-Chem
from the inventories are approximately constant throughout the
entire July 2005–May 2013 time range
(∼5 TgCmonth-1; Fig. d)
when integrated over the source regions of emissions impacting
the ISSJ. According to , this domain encompasses
central and western Europe, as well as North American
mid-latitudes (i.e., between 30–60∘ N and
-130–35∘ E). Hence, this suggests that the HCHO
production from oxidized anthropogenic compounds and its
inter-annual variability at the ISSJ are mainly driven in
GEOS-Chem by the simulated OH burden and its year-to-year
fluctuation (Fig. c).
Formaldehyde time series
We investigate hereafter the long-term evolution of the HCHO
abundance at ISSJ, using both the LIEGE and BRUKER FTIR data sets.
We employ the HCHO total columns derived from all individual FTIR
observations made over the 1988–2015 time period, corrected to
the equivalent 09:00 values via the parametric model
described previously and eventually combined as daily means.
Consistency between the FTIR data sets
We first evaluate the consistency of the measurements derived from
both instruments. To this end, we compare the HCHO total columns
derived from solar spectra recorded on the same days, using the
1995–1997 overlap time period while accounting for the diurnal
modulation of the HCHO abundance. Figure shows
a scatterplot of the scaled (to 09:00) BRUKER vs. LIEGE
total column daily means, including the 25 days available
over the 1995–1997 years. The comparison demonstrates a very
good agreement between both data sets, with a high coefficient of
determination (R2=0.89) for both low and high HCHO columns
(corresponding globally to measurements performed during winter and
summer). Moreover, the mean difference between the BRUKER and LIEGE
daily means is -0.10 ± 0.36 × 1015 moleccm-2.
Given the good consistency and absence of significant bias, the
LIEGE and BRUKER data sets will be jointly used hereafter
to investigate the long-term variation of the HCHO abundance above
Jungfraujoch.
Scatterplot of the daily average (and the associated 1σ
standard deviation as error bars) HCHO column measurements derived from FTIR
observations made by the LIEGE and BRUKER instruments at the ISSJ, over the
1995–1997 time period. These daily means are compared for days with
coincident observations, after scaling to 09:00 (see text). The solid black
line is the linear regression between both data sets (R2=0.89), along
with the 95 % confidence and prediction intervals delimited by the
blue and red lines, respectively.
FTIR time series of daily mean HCHO total columns and associated
1σ standard deviation bars above Jungfraujoch, from January 1988 to
June 2015. All individual measurements have been re-scaled to 09:00 (see
text) and then averaged over the days. The blue curves correspond to the
functions fitted to all daily means (including trend component and seasonal
modulation) by the bootstrap method of , over the 1988–1995,
1996–2002 and 2003–2015/06 time periods, inclusively.
Formaldehyde multi-decadal trend
Combined together, the LIEGE and BRUKER data sets constitute
a unique database covering more than 27 years (from
January 1988 to June 2015), providing to the best of our knowledge the
longest consistent time series of remote ground-based observations
of HCHO worldwide. In this part of the study, we employ the HCHO
total columns derived from all individual FTIR observations made
over the 1988–2015 time period. The entire daily mean total
column time series is
illustrated in Fig. . We have applied to the whole
data set a running mean characterized by a 3-year integration
length and a 6-month time step, in order to minimize the influence
of the large intra-annual variability of HCHO. This has revealed
a significant maximum of HCHO columns between the end of 1995 and early
1996, as well as a minimum around the beginning of 2003. The trend
and associated uncertainty affecting each subset (i.e., the daily
mean total column subsets spanning the 1988–1995, 1996–2002 and
2003–2015 periods, respectively) have been determined using
a statistical bootstrap resampling tool combining
a linear function and a third-order Fourier series that accounts for
the strong seasonal modulation of HCHO (in blue curve in
Fig. ).
Trends of daily mean 09:00-scaled HCHO total columns derived from
the ground-based FTIR observations at the ISSJ, calculated over the
1988–1995, 1996–2002, 2003–2015 and 1996–2015 time periods by the
bootstrap resampling tool. These trends are calculated using either
year-round or seasonal (over DJF, MAM, JJA and SON, respectively)
observations. They are expressed in annual rate of change
(%year-1) relative to the 1988.0, 1996.0 and 2003.0 columns
modeled by the bootstrap tool. The trends significant at the 2σ
confidence level are in boldface.
Observations
1988–1995
1996–2002
2003–2015
1996–2015
Year-round
2.94±2.02
-3.68±1.00
0.81±0.62
-0.61±0.26
Winter
5.66±5.04
0.31±2.39
-0.08±1.16
-1.05±0.54
Spring
4.92±2.86
-3.79±2.05
1.90±1.22
-0.3±0.52
Summer
4.19±3.29
-3.91±1.63
0.29±1.16
-0.65±0.47
Fall
1.24±3.19
-2.96±2.53
0.70±1.26
-0.41±0.61
Analysis of the whole time series indicates
a statistically significant rate of change (at the 2σ
confidence level) of the HCHO abundance over each time period:
4.35 ± 2.98, -7.22 ± 1.97 and
1.20 ± 0.92 × 1013 moleccm-2year-1
for 1988–1995, 1996–2002 and 2003–2015, respectively. Using the
1988.0, 1996.0 and 2003.0 columns modeled by the bootstrap tool
as references, we obtain the following relative annual trends:
2.94 ± 2.02 %year-1 up to 1995,
-3.68 ± 1.00 %year-1 between 1996 and 2002,
and 0.81 ± 0.62 %year-1 from 2003 onwards. It
is worth noting that the choice of the reference hour for scaling
the individual HCHO columns has no significant impact on the
calculated rates of change. For example, the relative annual
trends obtained from HCHO total columns scaled at 12:00,
i.e., when the difference between the 12 monthly intra-day regimes
is near its maximum, are 2.55 ± 1.75 %year-1
(1988–1995), -3.26 ± 0.90 %year-1 (1996–2002)
and 0.70 ± 0.54 %year-1 (2003–2015). However,
these trends may differ when calculated over specific seasons
only. The corresponding results are summarized in
Table .
The HCHO increase observed above Jungfraujoch between 1988 and
1995 may be related to the sharp rise of the atmospheric
CH4 growth rate from the 1980s to the beginning of the
1990s , which is its main precursor in the
background troposphere (see Sect. ). Above the
ISSJ, calculated discrete annual changes in
CH4 total column derived from FTIR observations equal
to 0.72 and 0.31 %year-1 for 1987–1988 and
1995–1996, respectively. In addition, all seasons also present
a significant positive rate of change of HCHO loadings at the
2σ confidence level over this time period, excepting
fall (see Table ). Nonetheless, the data set
covering this time range is mainly constituted of FTIR
measurements recorded with the LIEGE instrument, which are
sparser than those obtained with the BRUKER spectrometer from
1995 onwards (as is obvious in Fig. ). This may
explain the relatively large 2σ confidence levels
associated with the trends determined over this period.
Conversely, the decreased HCHO loadings from 1996 to 2002 may
be due to the global stabilization of the CH4
concentrations during most of this period ,
which was also observed at the ISSJ , combined
with reduced emissions mainly from fossil fuel sources in the
Northern Hemisphere and short-term
variations of the atmospheric OH burden .
However, it should be noted that the wintertime minimum of
HCHO is not affected by this decrease (see
Table ).
The observed rise of the HCHO total columns at the ISSJ
during the last decade (over 2003–mid-2015), which mainly
occurs during spring (see Table ), may be
explained by the highly studied global renewed increase in
atmospheric CH4, starting in the mid-2000s
. This renewal is also observed in the
CH4 columns derived from ground-based FTIR
measurements at the ISSJ.
It is also worth noting that the decrease in many
anthropogenic precursors of HCHO as a result of pollution
abatements in the Northern Hemisphere has no apparent
influence on the HCHO evolution during the last decade,
probably due to the fact that the HCHO formation coming from
these oxidized anthropogenic compounds is dominated by the
methane-originated production, as pointed out in
Sect. . Globally over the 1996–mid-2015
time period, the high-rate depletion of the HCHO loadings at
the ISSJ over 1996–2002 still dominates the slow renewal from
2003 onwards: the rate of change of the HCHO burden is
-0.61 ± 0.26 %year-1 from January 1996 to
June 2015.
Summary and conclusions
In this study, we use the strategy developed by to
retrieve HCHO total columns from high-resolution ground-based FTIR
solar spectra recorded at the high-altitude station of Jungfraujoch.
Because of its localization, this site allows for the study of the
continental background conditions in the remote troposphere at
mid-latitudes of the Northern Hemisphere. Using the large statistics
that represent the January 1995–June 2015 data set of solar spectra
recorded with a Bruker instrument, we first investigate the HCHO
diurnal variations above the station. These variations, resulting in
a.m. increases and p.m. decreases peaking around mid-day and in the
early afternoon, are mainly driven by the atmospheric photochemistry,
the intra-day insolation modulation and the CH4 oxidation.
Then, we characterize quantitatively these monthly diurnal variations
by adjusting a parametric model to the observations, consisting in
a re-parametrization of the standard statistical form of the Weibull
continuous probability distribution of a random variable. The equation
of the model and its parameters determined on the basis of the
observations are provided. As this model is fully parameterized and
links the daytime to the HCHO columns, it is eventually used to scale
all the individual FTIR measurements on 09:00 (i.e., when the
difference between the monthly intra-day regimes is minimal) in order
to remove the effect of the intra-day modulation.
In the next part of the study, we perform a GEOS-Chem v9-02 simulation
of the HCHO loadings above Jungfraujoch over the July 2005–May 2013
time period. As the analysis of the model outputs revealed that
GEOS-Chem does not reproduce the observed diurnal variations of the
HCHO columns, we use the daily mean 09:00-scaled FTIR measurements
to compare with the simulated 09:00 total columns, accounting for the
vertical resolution and sensitivity of the FTIR retrievals. Over this
period, the enhanced HCHO burden simulated by GEOS-Chem v9-02 compared
to version 9-01-03 reduces the mean bias with the observations,
due to the implementation of the new isoprene oxidation scheme in
version 9-02. Results from GEOS-Chem sensitivity runs (turning off
successively either the anthropogenic, biogenic or biomass burning
emissions of NMVOCs and NO implemented in the model) are also
investigated and suggest that the HCHO loadings above Jungfraujoch,
as well as its seasonal and inter-annual variations, are predominantly
led by the atmospheric CH4 oxidation. The anthropogenic
precursors of HCHO are estimated to contribute up to 25 % to
the wintertime HCHO total columns, while the impact of each of the
other emission sources is limited to 5 %.
Finally, we exploit the large database of FTIR solar spectra recorded
at the Jungfraujoch station by two high-resolution spectrometers
spanning the 1988–1997 and 1995–2015/06 time periods, respectively.
After checking the consistency between both subsets in terms of
retrieved HCHO columns, we combine them in order to produce a 27-year
time series of HCHO total columns, which is to our best knowledge the
longest time series of remote HCHO observations worldwide. Employing
the parametric model, the intra-day variation is removed by scaling
all the individual measurements of the data set to 09:00. We
eventually use the so-scaled entire time series to study the long-term
evolution of the HCHO columns in the background troposphere. Trend
analysis reveals an increasing HCHO burden between 1988 and 1995
(2.9 %year-1), followed by a sharp depletion
over 1996–2002 (-3.7 %year-1) and a slow
renewal of the growth rate from 2003 onwards
(0.8 %year-1). This long-term evolution
above Jungfraujoch is likely to be related to the fluctuations of
the atmospheric CH4 as well as to the short-term variability
of the OH concentrations.
Regional decadal trends at the global scale of mid-morning and
early-afternoon HCHO columns have been recently derived from combined
SCIAMACHY–GOME-2A, B and OMI measurements, respectively, over the
2004–2014 time period . Over western Europe, these
trends show an overall significant decrease in the HCHO abundance
(between -1.5 and -3.0 %year-1), mainly attributed
to effective pollution regulation measures . According to
the ground-based FTIR observations, we observe by contrast a weak
significant increase (less than 1 %year-1) of the
HCHO total columns above the ISSJ, over approximately the same time
period (2003–2015). These opposite trends may be explained by the fact
that the space-borne measurements cover entire regions (more
specifically Germany, France and Spain) that are largely under
the influence of anthropogenic emissions of NMVOC precursors, while the
Jungfraujoch HCHO columns are generally characteristic of the remote
troposphere and mainly originate from the CH4 oxidation (see
results in Sect. ).
Due to its very short lifetime, the abundance and spatial distribution
of HCHO in the atmosphere can be closely related to the emissions of
its NMVOC precursors and resemble their distribution in the atmosphere
if the NMVOC lifetime is short enough to avoid the spatial relationship
being smeared by atmospheric transport. Conversely, emissions of
long-lived VOCs (such as CH4) will produce a global HCHO
background with no detectable localized signal. As the HCHO loading
above the ISSJ predominantly originates from the photochemical
oxidation regimes of CH4, such inverse modeling studies will
be difficult to carry out on the basis of the ground-based FTIR
measurements of HCHO. Nevertheless, identifying in the Jungfraujoch
time series the large HCHO columns that are due to the injection of
“polluted” air masses from the planetary boundary layer (e.g., from
the nearby industrialized valleys), by the use of backward trajectories
models, can help in the determination of significant trends of HCHO
according to the origin of the air masses.
As HCHO is a key component in the global catalytic cycle responsible
for generating or destroying tropospheric O3 (depending on the
NOx levels), monitoring and understanding of the HCHO evolution
for background conditions are of primary importance. Indeed, many
questions arise as regards the renewed increase in atmospheric
CH4, the main precursor of HCHO in the remote troposphere.
According to and , oxidized CH4
still represents an important source of HCHO production in the
uppermost tropospheric layers, with contributions that vary from 40 to
more than 50 %, depending on the air masses. More particularly,
a sharp increase in the ethane (C2H6) burden of close to
5 %year-1 since 2009, attributed to the massive
growth of shale gas exploitation in North America, has recently been
highlighted above Jungfraujoch . Therefore, as
C2H6 is a HCHO precursor and shares most of its sources with
CH4, there are some concerns as to the impact on the evolution
of the HCHO loadings. Ground-based FTIR measurements combined with model
simulations can undoubtedly help with these issues.
The parametric model implemented in this study and the quantitative
characterization of the monthly intra-day variations of HCHO may be
a very useful tool in future works dedicated to the comparison between
ground-based FTIR and space-borne HCHO measurements. Indeed, long-term
consistent data sets of regular HCHO observations are increasingly
required for calibration/validation efforts of present satellite
instruments such as OMI and GOME-2. Furthermore, from 2017 onwards,
the space-borne monitoring observations are planned to continue with
TROPOMI (TROPOspheric Monitoring Instrument) and a third GOME-2
instrument. By scaling the FTIR HCHO columns to the respective overpass
times of the satellite sensors, this parametric model applied to the
Jungfraujoch long-term time series may be of high value for future
validation/calibration tasks in remote conditions at mid-latitudes of
the Northern Hemisphere. In the future, it will be useful to evaluate
the uncertainty associated with the scaling of the FTIR columns via
the parametric model. However, we anticipate that it will be far below
the random error inherent to the retrieval of individual HCHO total
columns 21 %; see. The parameters
of the fitting model are made available as Table S1 in the Supplement.