remote sensing
Comment
The Difficulty of Measuring the Absorption of
Scattered Sunlight by H2O and CO2 in Volcanic
Plumes: A Comment on Pering et al. “A Novel
and Inexpensive Method for Measuring Volcanic
Plume Water Fluxes at High Temporal Resolution,”
Remote Sens. 2017, 9, 146
Christoph Kern
U.S. Geological Survey Cascades Volcano Observatory, 1300 SE Cardinal Ct, Vancouver, WA 98683, USA;
[email protected] or [email protected]; Tel.: +1-360-993-8922
Academic Editor: Prasad S. Thenkabail
Received: 25 April 2017; Accepted: 19 May 2017; Published: 27 May 2017
Abstract: In their recent study, Pering et al. (2017) presented a novel method for measuring volcanic
water vapor fluxes. Their method is based on imaging volcanic gas and aerosol plumes using a camera
sensitive to the near-infrared (NIR) absorption of water vapor. The imaging data are empirically
calibrated by comparison with in situ water measurements made within the plumes. Though the
presented method may give reasonable results over short time scales, the authors fail to recognize
the sensitivity of the technique to light scattering on aerosols within the plume. In fact, the signals
measured by Pering et al. are not related to the absorption of NIR radiation by water vapor within the
plume. Instead, the measured signals are most likely caused by a change in the effective light path of
the detected radiation through the atmospheric background water vapor column. Therefore, their
method is actually based on establishing an empirical relationship between in-plume scattering
efficiency and plume water content. Since this relationship is sensitive to plume aerosol abundance
and numerous environmental factors, the method will only yield accurate results if it is calibrated
very frequently using other measurement techniques.
Keywords: volcanic gases; water vapor; carbon dioxide; remote sensing; plume imaging; passive
degassing; infrared cameras; sulfur dioxide; UV cameras
1. Introduction
Water vapor (H2 O) and carbon dioxide (CO2 ) are the two most abundant constituents of
high-temperature volcanic gases. Measuring their emission rates from volcanic vents is a high priority for
volcano observatories and researchers around the world, as changes in their emission can be indicative of
processes occurring at depth. Unfortunately, both species are considerably more difficult to measure than
the less-abundant sulfur dioxide (SO2 ) due to the significant atmospheric H2 O and CO2 background
concentrations. The concentration of H2 O and CO2 in volcanic plumes can be measured with MultiGAS
systems [1,2] by positioning them in the plume or by Fourier Transform Infrared Spectrometers (FTIR) by
using a hot surface or an artificial heat lamp as a source of infrared radiation [3,4], but deriving emission
rates from these measurements is difficult.
Only recently, Butz et al. [5] were able to quantitatively measure CO2 in the Mount Etna plume
using solar occultation FTIR spectroscopy for the first time. This method shows great promise, as it
allows CO2 detection from a safe distance from the vent and without subjecting any instrumentation to
Remote Sens. 2017, 9, 534; doi:10.3390/rs9060534
www.mdpi.com/journal/remotesensing
Remote Sens. 2017, 9, 534
2 of 11
the harsh environment inside the plume. It could also potentially be automated. Instruments installed
in the field could track the sun as it moves behind the volcanic plume and measure the CO2 /SO2 ratio
whenever appropriate meteorological conditions exist. The main limitation of the method is that it
requires direct solar radiation to pass through the plume to the instrument without being scattered.
Thus, measurements are only possible along the direct path between the sun and the instrument. This
prohibits the integrative measurement of the entire plume required for deriving the emission rate.
However, emission rates can be determined by scaling the CO2 /SO2 ratio obtained by FTIR or other
methods with the SO2 emission rate measured by the established Differential Optical Absorption
Spectroscopy (DOAS) or SO2 camera techniques.
The recent success of solar occultation FTIR measurements in measuring volcanic CO2 raises the
question of whether the absorption of CO2 and H2 O can also be measured in scattered solar radiation
passing through the plume. This would allow for the direct determination of the emission rate, as the
volcanic plume could be scanned or imaged to determine the trace gas abundance in a cross-section.
Multiplied by the wind speed, this value then gives the emission rate. Volcanic SO2 is measured in this
way with scanning DOAS and SO2 cameras [6,7] but even though the absorption cross-sections of both
CO2 and H2 O exhibit characteristic features in the near-infrared (NIR) spectral region, measuring the
absorption of scattered solar radiation by these volcanic gases remains extremely challenging.
Following the general principle behind SO2 imaging in the ultraviolet spectral range,
Pering et al. [8] now present a novel method for imaging volcanic H2 O emissions and determining
emission rates. By introducing two bandpass filters in front of a NIR-sensitive camera system, they
measure the incident radiance of scattered solar radiation on and off of an H2 O absorption band.
By comparing the on-band radiance coming from the clear sky next to the plume with that coming
from the plume, they derive an optical depth. As is customary for SO2 camera measurements [7,9], this
optical depth is then normalized by the optical depth obtained from the off-band filter in an attempt to
correct for scattering of light on aerosols and water droplets in the plume.
The authors measured a clear signal at the location of the volcanic plume in their images.
However, a few calculations show that, contrary to their interpretation, the measured signal is unlikely
to be associated with the absorption of NIR radiation by water vapor in the volcanic plume. Instead, the
most probable explanation for their findings is that the effective atmospheric light path of the detected
radiation has been modified by aerosol scattering in the plume, and that their first-order scattering
correction is not a valid method for removing the associated signal. The calculations leading to these
conclusions are described below.
2. Simple Simulation of Expected Absorption
Light passing through a medium is attenuated according to the Beer-Lambert-Bouguer law
of absorption:
L(λ) = L0 (λ)·e−σ(λ)·S
(1)
Here, L(λ) describes the spectral radiance of light after passing through the medium, L0 (λ)
represents the radiance before entering the medium, λ is the wavelength, σ is the absorption cross-section
of the absorber, and S is the column density. The column density is the integral of the absorbing species
along the light path, and is generally the parameter that spectroscopic measurements are trying to
determine [10].
Using Equation (1), the expected absorption signal can be easily estimated for a given measurement
geometry if the absorption cross-section of the species of interest is known. The cross-sections of both
CO2 and H2 O are relatively well-known and available for download from the HITRAN database [11].
For the sake of brevity, and because this was the topic of the original study by Pering et al. [8], the
following sections will focus on H2 O, but the discussed concepts apply to CO2 as well.
H2 O has an absorption band at around 940 nm. The absorption cross-section of this band was
calculated for a line width of 0.01 cm−1 from the line strengths listed in the HITRAN database, and is
plotted in Figure 1. Note that the maximum cross-section is around 2 × 10−21 cm2 /molecule. This value
Remote
Sens.
2017,
9, 9,
534
Remote
Sens.
2017,
534
3 of
1111
3 of
differs from the 5.7 × 10−22 cm2/molecule (=19 cm2/g) cited by Pering et al. [8], but this could simply be
differs from the 5.7 × 10−22 cm2 /molecule (=19 cm2 /g) cited by Pering et al. [8], but this could simply
a result of using a greater line width for the calculation of the cross-section. In the end, it is the area
be a result of using a greater line width for the calculation of the cross-section. In the end, it is the area
under each line that will determine the absorbed energy. Since individual lines cannot be resolved,
under each line that will determine the absorbed energy. Since individual lines cannot be resolved,
the maximum amplitude of the cross-section is not of great importance for these measurements or
the maximum amplitude of the cross-section is not of great importance for these measurements or the
the below calculations.
below calculations.
Figure1.1.Absorption
Absorptioncross-section
cross-sectionofofwater
watervapor
vaporbetween
between880
880and
and1000
1000nm.
nm.Line
Linestrength
strengthdata
dataisis
Figure
fromthe
theHITRAN2012
HITRAN2012
database
[11].
from
database
[11].
The spectrum of a light-emitting diode (LED) is similar to the transmittance curve of a bandpass
The spectrum of a light-emitting diode (LED) is similar to the transmittance curve of a bandpass
filter. Both can be approximated by a Gaussian curve. A Gaussian with a 10 nm Full Width at Half
filter. Both can be approximated by a Gaussian curve. A Gaussian with a 10 nm Full Width at Half
Maximum (FWHM) is plotted in Figure 2a. Assuming this initial spectrum as the background
Maximum (FWHM) is plotted in Figure 2a. Assuming this initial spectrum as the background radiance,
radiance, the remaining spectral radiance after passing through a known amount of water vapor can
the remaining spectral radiance after passing through a known amount of water vapor can be calculated
be calculated according to Equation (1). The August-Roche-Magnus formula provides a tool for
according to Equation (1). The August-Roche-Magnus formula provides a tool for estimating the H2 O
estimating the H2O saturation vapor pressure ps as a function of temperature [12,13]:
saturation vapor pressure ps as a function of temperature [12,13]:
17.625 (2)
( ) = 6.1094 ∙ exp
17.625T
243.04 +
ps ( T ) = 6.1094· exp
(2)
243.04 + T
In this form, the temperature T must be inserted in units of degrees Celsius and the equation
yields
the vapor
pressure
ps in hPa.
can
thenofbedegrees
obtained
by comparing
the actual
In this
form, the
temperature
T Relative
must be humidity
inserted in
units
Celsius
and the equation
water
vapor
partial
pressure
(or
concentration)
to
the
saturation
pressure.
This
equation,
or
variant
yields the vapor pressure ps in hPa. Relative humidity can then be obtained by comparing thea actual
thereof,
is
probably
what
Pering
et
al.
[8]
meant
to
refer
to
in
their
Equation
(2),
though
their
equation
water vapor partial pressure (or concentration) to the saturation pressure. This equation, or a variant
appears
be reproduced
incorrectly.
thereof,
is to
probably
what Pering
et al. [8] meant to refer to in their Equation (2), though their equation
Applying
Equation
(2),
one
finds that 100% relative humidity corresponds to approximately 23
appears to be reproduced incorrectly.
hPaApplying
at 20 °C. Inserting
this
value
into the
gasrelative
law yields
a watercorresponds
vapor concentration
of about 6
Equation (2), one finds
thatideal
100%
humidity
to approximately
17
3
10 atmolecules/cm
. With
this information,
absorption
various water
vapor
23× hPa
20 ◦ C. Inserting
this value
into the idealthe
gasapproximate
law yields a water
vapor of
concentration
of about
17
3
column
amounts
can
be
simulated.
For
example,
a
heated
beaker
filled
with
water
might
induce
100%
6 × 10 molecules/cm . With this information, the approximate absorption of various water vapor
relativeamounts
humiditycan
over
20-cm pathFor
length
aboveait.heated
This corresponds
a column
of about
column
be asimulated.
example,
beaker filledtowith
waterdensity
might induce
19 H2O molecules/cm2. If this path were illuminated by an LED with a Gaussian emission
1.2
×
10
100% relative humidity over a 20-cm path length above it. This corresponds to a column density
19 H O molecules/cm
2 . If
remaining
light intensity
after
through
the water
column
can be
calculated
ofspectrum,
about 1.2 the
× 10
thispassing
path were
illuminated
by an
LED with
a Gaussian
2
according
to Equation
(1). Though
barely
different
thethrough
initial Gaussian,
spectrum
emission
spectrum,
the remaining
light
intensity
after from
passing
the water the
column
can beof
transmitted
light
is
plotted
in
Figure
2b.
Integration
over
the
spectral
radiance
gives
the
total
incident
calculated according to Equation (1). Though barely different from the initial Gaussian, the spectrum of
radiance, which
the parameter
thatIntegration
would beover
measured
by aradiance
camera gives
or other
photodetector
transmitted
light is is
plotted
in Figure 2b.
the spectral
the total
incident
(denoted
by
IA
s and IAb by Pering et al. [8]). The optical depth a’ is then obtained by taking the
radiance, which is the parameter that would be measured by a camera or other photodetector (denoted
analogous
the The
authors’
Equation
the obtained
absence ofbyin-plume
scattering,
IBs
bynegative
IAs andlogarithm,
IAb by Pering
et al. to
[8]).
optical
depth (1).
a’ is(In
then
taking the
negative
and IBb are
assumed to
to the
be identical
therefore
cancel
each other
out.)
logarithm,
analogous
authors’ and
Equation
(1). (In
the absence
of in-plume
scattering, IBs and IB
b
are assumed to be identical and therefore cancel each other out.)
Remote Sens. 2017, 9, 534
Remote Sens. 2017, 9, 534
4 of 11
4 of 11
=log
−
a=−
I As
I Ab
IBs
IBb
!
= =a0 + +
log
IBs
IBb
Figure
2. Radiance
behind
a bandpass
filter
the measurement
measurement
Figure
2. Radiance
behind
a bandpass
filterwith
withvarious
variousamounts
amountsof
ofwater
water vapor
vapor in
in the
light
path.
These
plots
were
calculated
according
to
Equation
(1).
The
Gaussian
depicted
in (a)
(a)
light path. These plots were calculated according to Equation (1). The Gaussian depicted in
represents
the
radiance
emitted
from
a
light
emitting
diode
(LED)
or
the
intensity
measured
behind
represents the radiance emitted from a light emitting diode (LED) or the intensity measured behind
an illuminated
bandpass
filter
centered
due to
to
an illuminated
bandpass
filter
centeredatat940
940nm.
nm.InIn(b),
(b),the
thespectral
spectral radiance
radiance is
is decreased
decreased due
absorption
by
100%
relative
humidity
along
a
20-cm
path.
Inset
(c)
shows
the
radiance
obtained
after
absorption by 100% relative humidity along a 20-cm path. Inset (c) shows the radiance obtained after
absorption
by by
H2H
O2O
in in
a standard
atmospheric
100% relative
relative
a standard
atmosphericcolumn.
column.In
Ininset
inset(d),
(d), the
the absorption
absorption by 100%
absorption
humidity
along
a 20-m
plume
was
added
totothe
(f) show
show the
the
humidity
along
a 20-m
plume
was
added
theatmospheric
atmosphericabsorption.
absorption. Insets
Insets (e) and (f)
same
calculation,
this this
timetime
assuming
that the
light has
passed
throughthrough
the vertical
column
same
calculation,
assuming
thatbackground
the background
light
has passed
the vertical
of Hcolumn
times.
of each
plot of
gives
total
measured
intensities
respectively,
H2O The
fiveintegral
times. The
integral
eachthe
plot
gives
the total
measuredIAintensities
b and IAs,
2 O five of
b and IAs , IA
andrespectively,
the optical depth
a’
can
be
calculated
from
these
(Equation
(3)).
In
each
of
the
three
examined
cases,
and the optical depth a’ can be calculated from these (Equation (3)). In each of the three
absorption
bycases,
the plume
is almost
when negligible
comparedwhen
to the
atmosphere,
the expected
examined
absorption
by thenegligible
plume is almost
compared
to theand
atmosphere,
and
the expected
optical depths
the plume
are less than 1%.
optical
depths associated
with associated
the plumewith
are less
than 1%.
(3)(3)
In the example of the beaker, only 0.086% of the total radiance is absorbed, so an extremely
5 of 11
precise measurement would be required to detect any absorption at all. It is unlikely that a
measurement
with sufficient
precision
possibleofwith
webcam,
as attempted
by Pering
et al. [8]. It
In the example
of the beaker,
onlyis 0.086%
the atotal
radiance
is absorbed,
so an extremely
is unclear
exactly
their measured
signal
originated
but it may
be related
to an imperfect
Remotemeasurement
Sens.
2017, 9,where
534 would
5 that
of 11 a
precise
be required
to detect
anyfrom,
absorption
at all.
It is unlikely
correction of light
water droplets
and
steam
above the
beaker. These
scattering
measurement
with scattered
sufficient on
precision
is possible
with
a webcam,
as attempted
by Pering
et al.effects
[8]. It
are
generally
greater
in
magnitude
than
the
absorption
of
water
vapor
itself,
as
will
be
discussed
in
is unclear
exactly
where
their
measured
signal
originated
from,
but
it
may
be
related
to
an
imperfect
In the example of the beaker, only 0.086% of the total radiance is absorbed, so an extremely precise
more detail
correction
ofbelow.
light
scattered
on water
droplets
and steamatabove
beaker.
These
scattering with
effects
measurement
would
be required
to detect
any absorption
all. It isthe
unlikely
that
a measurement
On
the
other
an H2O plume
degassing
from
a volcanic
fumarole
might
exhibit
100%
are sufficient
generallyprecision
greaterhand,
in
magnitude
than
the
absorption
of
water
vapor
itself,
as
will
be
discussed
is possible with a webcam, as attempted by Pering et al. [8]. It is unclear exactly in
relative
humidity
over asignal
path originated
of 20 m. Infrom,
this case,
same
calculation
shows that
about 6.8%
of the
more
detail
below.
where
their
measured
but itthe
may
be related
to an imperfect
correction
of light
light
would
be
absorbed
by
water
vapor
along
the
20-m
path.
In
the
absence
of
volcanic
gas,
the
light
On the on
other
hand,
an H
2Osteam
plume
degassing
from
a volcanic
fumarole
might
exhibit
100%
scattered
water
droplets
and
above
the beaker.
These
scattering
effects are
generally
greater
would
travel
through
20
m ofofbackground
atmosphere,
assuming
50%
humidity
inthe
the
in magnitude
than
the
of water
asand
will
be discussed
in relative
more
below.
relative
humidity
over
a absorption
path
20 m.
In thisvapor
case,itself,
the same
calculation
shows
thatdetail
about
6.8% of
background
air,
only
3.8%
signal
would
attenuated
same
light
path.100%
The
measured
On the
other
hand,
anof
Hthe
a volcanic
might
exhibit
light
would
be
absorbed
by
water
vapordegassing
along be
thefrom
20-m
path.along
In fumarole
thethe
absence
of volcanic
gas,relative
the light
2 O plume
optical
depth
a’
associated
with
the
plume
itself
would
therefore
be
about
3%.
Such
a
signal
be
humidity
a path
In this case,atmosphere,
the same calculation
shows 50%
that about
6.8%
of thecould
light
would
travelover
through
20ofm20ofm.
background
and assuming
relative
humidity
in the
detected
a only
reasonably
perhaps
even
with
a high
quality
webcam.
Certainly,
would with
be air,
absorbed
by water
along
the
20-m
path.
In
the absence
oflight
volcanic
gas,
light
background
3.8%
ofsensitive
thevapor
signaldetector,
would
be
attenuated
along
the
same
path.
Thethe
measured
a
more
expansive
plume
could
be
detected
in
this
way.
However,
such
a
measurement
would
require
would
travel
through
20
m
of
background
atmosphere,
and
assuming
50%
relative
humidity
in
the be
optical depth a’ associated with the plume itself would therefore be about 3%. Such a signal could
only 3.8%
of theansignal
would
be
attenuated
same
light path.
The measured
an background
experimental
setup
in which
active
source
of
radiation
(e.g.,the
isquality
placed
behind
the
volcanic
detected
with aair,
reasonably
sensitive
detector,
perhaps
evenalong
with
aLED)
high
webcam.
Certainly,
optical
depth to
a’ associated
with theDOAS
plumeor
itself
would
therefore
be about
3%.class
Suchof
a signal
could be is
plume,
similar
active
long-path
active
FTIR
experiments.
This
measurements
a more expansive plume could be detected in this way. However, such a measurement would require
detected with
a reasonably
sensitive
detector,
perhaps
even with
a high
qualityiswebcam.
Certainly,
a
summarized
in Table
1(A).
Things
become
more
solar
radiation
employed
as volcanic
the light
an
experimental
setup in
which
an active
source
of difficult
radiationif(e.g.,
LED)
is placed
behind the
more as
expansive
plume could
detected
in this way. However, such a measurement would require an
source,
will to
be active
shown
in thebenext
section.
plume,
similar
long-path
DOAS
or active FTIR experiments. This class of measurements is
Remote Sens. 2017, 9, 534
experimental setup in which an active source of radiation (e.g., LED) is placed behind the volcanic
summarized in Table 1(A). Things become more difficult if solar radiation is employed as the light
plume,
similar
to active
DOAS or active
FTIRsensing
experiments.
This
class
measurements
Table
1. Overview
of long-path
discussed geometries
for remote
of water
vapor
in of
volcanic
plumes. is
source,
as will be shown in the next section.
summarized in Table 1(A). Things become more difficult if solar radiation is employed as the light
Active-Source
Geometry
source,(A)
as will
be shown in
the next section.
Table 1. Overview of discussed geometries for remote sensing of water vapor in volcanic plumes.
Table 1. Overview of discussed geometries for remote sensing of Column
water vapor
in volcanic plumes.
densities:
(A) Active-Source Geometry
(A) Active-Source Geometry
(B) Solar Occultation Geometry
(B) Solar Occultation Geometry
(B) Solar Occultation Geometry
(C) Scattered-Light Geometry
(C) Scattered-Light Geometry
Background column density
=
Plume measurement:
=( + )
Column densities:
Example
assumptions:
Column densities:
Background column density
=
Background
column
density
Sb m
= ca L
Plume
width
= 20
Plume
measurement:
=
(
+
Plume measurement:
Ss = (c a + c3 v() L )=50% RH) 1
=3 × 1017 molecules/cm
Atmosphere:
Example
assumptions:
assumptions:
3 (
Plume: 2
=3 ×Example
1017 molecules/cm
+ =100% RH) 1
Plumewidth
widthL = =
Plume
20 20
mm
3
Calculated total17absorption
(920–960
nm):
3 3 = 50%
1
Atmosphere:
× 10
RH)
× 1017molecules/cm
molecules/cm(c
(a =50%
RH) 1
Atmosphere:c a = 3=3
2
17
3
1
Atmospheric
optical
depth
=
0.038
Plume:
3 ××101017molecules/cm
cv ==100%
100% RH)
3(c(a ++
Plume: 2cv ==3
molecules/cm
RH) 1
3
=0.068
Plume
+
atmosphere
optical
depth
Calculated total absorption (920–960
nm):
Calculated total absorption 3 (920–960 nm):
Atmospheric
optical
depth ab =
= 0.038
.
Plume optical
depth
Atmospheric
optical
depth
Plume
+ atmosphere
optical
depth as==0.038
0.068
PlumePlume
+ atmosphere
optical
optical depth
a0 depth
= 0.030 =0.068
= .
Plume optical depth
Column densities:
Column densities:
Background column density
=
Background column density Sb = c a H
=
Plume
measurement
Plume measurement
Ss = c a H + cv+L
Column densities:
Additional
example
assumptions:
Additional example assumptions:
Background
column density 22 =
2 2
Atmospheric
column
=8
molecules/cm
Atmospheric
column
c a H = 8 × 10×=2210
molecules/cm
+
Plume
measurement
21
21
2
Plume
column
=1.2
molecules/cm2
Plume
column
cv L = 1.2
× 10× 10molecules/cm
Additional
example assumptions:
3
3
Calculated
total
absorption
nm):
Calculated
total
absorption(920–960
(920–960
nm):
Atmospheric
column
=8 × a1022
molecules/cm2
Atmospheric
depth
b = 1.051
Atmosphericoptical
optical
depth
=1.051
21 molecules/cm2
Plume+column
× 10
Plume
atmosphere =1.2
optical
depth
a = 1.060
=1.060
Plume + atmosphere optical
depths
0
3
Plume total
optical
depth a =(920–960
0.009
Calculated
absorption
nm):
= .
Plume optical depth
Atmospheric optical depth
=1.051
=1.060
Plume + atmosphere optical depth
= .
Plume optical depth
Remote Sens. 2017, 9, 534
Remote Sens. 2017, 9, 534
6 of 11
Table 1. Cont.
6 of 11
(C) Scattered-Light Geometry
Columndensities:
densities:
Column
Backgroundcolumn
columndensity
density
/ sin(α)
Background
Sb ≈ c≈a H/ sin
(α)
Plume
measurement
h
i
Plume
measurement
ca H
Ss ≈ (1 − f ) sin
+ c L + f [c a H + cv L]
≈ (1 − ) (α() ) + v
+
+
Additional example assumptions:
Additional example
assumptions:
Elevation angle 4 4α = 10 deg
Elevation
angle
10 degnm):
3=
Calculated
total absorption
(920-960
3 (920-960 nm):
Calculated
total
absorption
Contribution of plume-scattered light 5 f = 0, 0.1
5
Contribution
plume-scattered
= 0, 0.1
Atmosphericofoptical
depth 5 ab =light
2.667, 2.667
5
Atmospheric
optical
depth
2.667,
2.667
Plume
+ atmosphere
optical
depth 5 a=
=
2.672,
2.334
s
5 a0 = 0.005,
optical depth
2.672, 2.334
PlumePlume
+ atmosphere
optical
depth 5 −=0.333
5
2 c is the
Plume
optical
depth
= 0.005,
−0.333
Conversion to relative humidity (RH) assumes a temperature
of 20 ◦ C.
volcanogenic
excess water
vapor
v
concentration in the plume. The total water vapor concentration is c a +cv and cannot exceed 100% relative humidity.
3 Total absorption is calculated by integrating the transmitted spectral radiance from 920 to 960 nm. 4 Calculation is
1 Conversion to relative humidity (RH) assumes a temperature of 20 °C. 2
is athe
performed based on the assumption that, for an elevation angle of 10 degrees, light has taken
pathvolcanogenic
through the
5 Two cases are considered: f = 0
water-rich
lower
atmosphere
that
is
five
times
longer
than
the
vertical
column.
excess water vapor concentration in the plume. The total water vapor concentration is + and
represents an idealized case with no scattering
occurring in the plume. f = 0.1 represents a more realistic case in
3
is calculated by integrating the transmitted
cannot
100%
relative
whichexceed
10% of the
detected
lighthumidity.
was scatteredTotal
in theabsorption
plume.
1
spectral radiance from 920 to 960 nm. 4 Calculation is performed based on the assumption that, for an
angle of 10Measurements
degrees, light has taken a path through the water-rich lower atmosphere that is
3. elevation
Solar Occultation
five times longer than the vertical column. 5 Two cases are considered: f = 0 represents an idealized
In contrast to measuring SO , which has a negligible atmospheric background concentration,
case with no scattering occurring 2in the plume. f = 0.1 represents a more realistic case in which 10% of
the quantitative detection of volcanic H2 O and CO2 is complicated by the fact that these species
the detected light was scattered in the plume.
have significant atmospheric background abundances. Any ground-based remote sensing technique
that relies
on solar Measurements
radiation as a light source will measure light that has passed through the entire
3. Solar
Occultation
atmospheric column above the measurement site at least once. If the radiation also passes through a
In contrast
to the
measuring
SO2, which
a negligible
atmospheric
background
concentration,
volcanic
plume,
column density
will has
be enhanced
over
the atmospheric
column by
the additionalthe
quantitative
detection
of
volcanic
H
2
O
and
CO
2
is
complicated
by
the
fact
that
these
species
have
amount of the target gas in the plume. Measured at sea level, the global average vertical
column
22
2
significant
abundances.
Any ground-based
remotethe
sensing
that
density ofatmospheric
water vaporbackground
is approximately
8 × 10 molecules/cm
[14], though
actualtechnique
value varies
significantly
with
latitude
and
is
much
larger
near
the
equator.
relies on solar radiation as a light source will measure light that has passed through the entire
A “solar
occultation”
measurement
is performed
by once.
pointing
a spectroscopic
at the a
atmospheric
column
above the
measurement
site at least
If the
radiation alsoinstrument
passes through
sun
and
measuring
unscattered
(“direct”)
solar
radiation
passing
through
the
atmosphere
[15,16].
volcanic plume, the column density will be enhanced over the atmospheric column by the additional
If a volcanic
located
sun and the
instrument,
radiation
additionally
amount
of the plume
target isgas
in thebetween
plume.the
Measured
at sea
level, thethe
global
average
vertical passes
column
through
it
(see
Table
1(B)).
If
such
measurements
are
performed
at
low
latitudes
near
solar
noon,
the
22
2
density of water vapor is approximately 8 × 10 molecules/cm [14], though the actual value varies
sun
is
close
to
the
zenith
and
the
light
path
is
close
to
a
vertical
column
through
the
atmosphere.
significantly with latitude and is much larger near the equator.
This
measurement
geometry
is advantageous
for twoby
reasons:
(1)aThe
low solar zenith
angle and
A “solar
occultation”
measurement
is performed
pointing
spectroscopic
instrument
at the
nearly
vertical
measurement
geometry
reduce
the
absorption
of
light
by
the
background
atmosphere
sun and measuring unscattered (“direct”) solar radiation passing through the atmosphere [15,16]. If
to a minimum, thus maximizing the relative contribution of the plume. (2) Perhaps more importantly,
a volcanic plume is located between the sun and the instrument, the radiation additionally passes
the influence of scattering can be neglected because any light scattered out of the direct path between
through it (see Table 1(B)). If such measurements are performed at low latitudes near solar noon, the
the sun and the instrument is unlikely to ever be detected by the instrument. Therefore, the light path
sun is close to the zenith and the light path is close to a vertical column through the atmosphere. This
of the measured radiation is known to always be the direct path between the sun and the instrument.
measurement geometry is advantageous for two reasons: (1) The low solar zenith angle and nearly
The calculations performed in the previous section can be repeated for the vertical atmospheric
vertical measurement geometry22reduce the absorption
of light by the background atmosphere to a
H2 O column. Inserting 8 × 10 molecules/cm2 into the Beer-Lampert-Bouguer law (Equation (1)),
minimum,
thus shown
maximizing
the relative
contribution
of the plume.
Perhaps
more
importantly,inthe
the spectrum
in Figure
2c is obtained.
This spectrum
shows(2)
that
virtually
all wavelengths
influence
of
scattering
can
be
neglected
because
any
light
scattered
out
of
the
direct
path
between
the
the measurement window are now absorbed to some degree, with several lines essentially
reaching
sunnull,
andi.e.,
thethe
instrument
to dark
ever in
bespots.
detected
by the instrument.
Therefore,
the light
path of
spectrum is
is unlikely
completely
Integration
over all wavelengths
shows
that about
the2/3
measured
radiation
is
known
to
always
be
the
direct
path
between
the
sun
and
the
instrument.
of the radiance is absorbed on its way through the atmosphere.
The
calculations
performed
in the previous
section can
be repeated
the vertical
atmospheric
This
spectrum can
now be compared
to the spectrum
obtained
with anfor
additional
fumarolic
water
22 molecules/cm2 into the Beer-Lampert-Bouguer law (Equation (1)), the
H2O
column.
Inserting
8
×
10
vapor plume in the light path. Again using a 20-m plume of 100% relative humidity (RH), a volcanic
2 canspectrum
spectrum
shownofin1.2Figure
is obtained. This
that virtuallycolumn,
all wavelengths
the
H2 O column
× 10212cmolecules/cm
be addedshows
to the atmospheric
resulting in a
22
measurement
absorbed to2some
degree, with
severalof
lines
essentially
reaching
null,
total columnwindow
of 8.12 ×are
10now
molecules/cm
. The associated
spectrum
transmitted
light
is plotted
i.e., the spectrum is completely dark in spots. Integration over all wavelengths shows that about 2/3
of the radiance is absorbed on its way through the atmosphere.
This spectrum can now be compared to the spectrum obtained with an additional fumarolic
water vapor plume in the light path. Again using a 20-m plume of 100% relative humidity (RH), a
volcanic H2O column of 1.2 × 1021 molecules/cm2 can be added to the atmospheric column, resulting
Remote Sens. 2017, 9, 534
7 of 11
in Figure 2d, but does not differ discernably from the atmospheric spectrum. Using the atmospheric
spectrum as a reference (IAb ), the optical depth added by the volcanic plume is only 0.0092. This signal
is small, but might be detectable with a precise instrument. Certainly, it should be possible to detect a
somewhat larger plume from a higher elevation volcanic vent (where the overhead H2 O column is
reduced) with the solar occultation geometry. Recent solar occultation FTIR measurements for the first
time succeeded in measuring volcanic CO2 emitted from Mount Etna, albeit at much higher spectral
resolution and longer wavelengths [5]. However, even more challenges arise when scattered solar
radiation is used as a light source, as is needed for scanning or imaging applications. This is discussed
in the next section.
4. Measuring H2 O Absorption in Scattered Sunlight
Scanning and imaging applications in the ultraviolet, visible, and NIR wavelength ranges use
solar radiation scattered in the Earth’s atmosphere as a source of light (Table 1(C)), the main advantage
being that the entire plume is illuminated at once. Therefore, the scanners or cameras can measure
the trace gas abundance throughout the plume and eventually derive emission rates rather than just
measuring relative gas abundances.
When measuring a gas species that is present in both the plume and the background atmosphere, the
scattered light measurements can become very dependent on the measurement geometry. For example,
water vapor has a scale height of about 2 km in the atmosphere [17], which means its concentration drops
to 1/e times its ground-level value at this altitude. Light detected by a scanning or imaging instrument
pointed at 10 degrees elevation above the horizon has traveled more than 10 km in the lowermost 2 km
water vapor layer. Therefore, the spectral radiance of this light will typically exhibit H2 O absorption
associated with about five times the vertical atmospheric H2 O column.
Figure 2e shows the spectral radiance reaching a detector behind a bandpass filter after passing
through 3.2 × 1023 H2 O molecules/cm2 (five times the average vertical atmospheric column).
The atmospheric absorption is quite strong now, and the relative contribution of an additional 20-m
plume of 100% RH is now even smaller (Figure 2f). In this scenario, the expected plume optical depth is
less than half of one percent. Clearly, a very precise measurement would be required to detect this signal.
In practice, however, such a measurement will probably not be possible. Up to now, the presented
examples have assumed that the influence of scattering aerosols in the water vapor plume is negligible.
This holds true for solar occultation measurements, where scattered radiation is not detected, but as
will be discussed next, it is not valid for measurements of scattered sunlight.
5. The Effect of Scattering in the Plume
Due to the high abundance of H2 O in high-temperature volcanic gas and the tendency for emitted
SO2 to form sulfate aerosols, most volcanic plumes contain some amount of aerosols and/or liquid
water droplets, particularly in areas proximal to the vent. These aerosols and water droplets can
potentially scatter solar radiation towards an observing instrument (shown in Table 1(C), path B).
In fact, aerosol scattering effects are the process by which these plumes become visible to the human eye.
As mentioned before, the conventional approach for first order correction of scattering in ultraviolet
SO2 camera measurements is to divide the trace gas optical depth by the optical depth obtained in a
wavelength range outside the trace gas absorption band (Equation (3)) [7,9].
This very same correction was also applied by Pering et al. [8]. Unfortunately, the method is
not valid for correcting absorption measurements of gas species that are abundant in the background
atmosphere. This can be easily demonstrated with the help of an example. Assuming the same
measurement geometry as discussed in the previous section, but now allowing for in-plume scattering,
the plume may appear brighter in the off-band channel (Pering et al. use 850 nm) due to additional
radiation being scattered toward the instrument on aerosols in the plume (Table 1(C), path B). Even for
relatively transparent plumes, it is reasonable to assume that 10% of the detected radiation could have
been scattered in the plume. In this case, the plume would appear ~10% brighter than the background
Remote Sens. 2017, 9, 534
8 of 11
in the off-band channel. The standard correction (Equation (3)) would then call for the measured
on-band (940 nm) optical depth to be increased by 0.1. Note that this correction is orders of magnitude
larger than the expected signal from H2 O absorption in the plume (~0.0046, Figure 2f), so the applied
correction would have to be extremely precise.
Unfortunately, it will not be possible to attain this level of precision. The problem lies in the
different atmospheric path that the plume-scattered radiation (Table 1(C), path B) has taken when
compared to the light coming through the plume from behind (Table 1(C), path A). In a single scattering
approximation, the plume-scattered radiation will have traveled straight through the atmospheric H2 O
column (again, assuming the sun is close to the zenith) before being scattered towards the detector. In
general, its atmospheric path will be significantly different than that of the light coming from behind
the plume. In our example, if we again assume a five times shorter path for the plume-scattered
radiation, light absorption by atmospheric H2 O will be significantly smaller, and the contribution of
this light to the total radiance will be much greater than the additional 10% measured in the off-band
channel. Inserting the numbers from the previous section for a standard atmosphere into Equation (1),
and assuming that 10% of the detected light traveled along path B in Table 1(C), we expect a detected
optical depth of −0.33 at 940 nm. The negative sign means that the plume actually appears 33%
brighter than the background. This is a direct effect of the shorter path through the atmospheric water
column taken by the plume-scattered radiation. Applying the correction from the off-band channel
(Equation (3)) would yield a corrected optical depth of −0.23. Since the off-band channel does not
account for atmospheric water vapor absorption, the derived differential optical depth is very far from
the true value associated with absorption of light by water vapor in the plume (only 0.0046).
While these numbers are valid only for the specific measurement geometry of the given example,
they illustrate the main difficulty of measuring the absorption of scattered solar radiation by volcanic
water vapor or CO2 . The problem lies in the fact that even small amounts of aerosols or condensed
water droplets in the plume cause changes in the effective atmospheric path length of the detected
radiation. Because H2 O and CO2 are abundant in the atmosphere, these path length changes
significantly affect the detected spectral radiance or integrated intensity. The instrument registers a
change in the amount of water vapor along the line of sight, but the majority of this change comes
from the modified effective path that radiation has taken on its way through the atmosphere. In the
discussed example, these atmospheric effects are about two orders of magnitude greater than the
signal expected from absorption of radiation by H2 O in the plume itself. Moreover, the example is
by no means an extreme case. Plume-scattered radiation will oftentimes account for more than 10%
of the incident light, particularly when imaging plumes close to their source vents. The discussion
also shows that the dual channel correction approach (Equation (3)) is not valid when measuring gas
species with significant atmospheric abundances, because it does not take potential changes of the
effective path length through the atmospheric background into account.
6. The Physics behind the Measurements Presented by Pering et al.
According to the above discussion, the dual channel camera employed by Pering et al. [8] was
probably not able to measure the extremely weak absorption of scattered sunlight by water vapor in
the volcanic plume. In both the Vulcano Island and Mount Etna measurements, scattering on plume
aerosols and water droplets most likely led to a change in the effective atmospheric path length of the
detected radiation. Since a change in atmospheric path brings a change in atmospheric water vapor
absorption with it, the instrument would have responded to this effect. This means that the parameter
derived from the imagery was not primarily associated with in-plume water vapor absorption but
instead was essentially a measure of the contribution of in-plume aerosol scattering to the total detected
radiation (parameter f in Table 1(C)).
While this does not appear to have been the intent of the measurements, others have shown that
monitoring the radiance arriving at the detector due to in-plume aerosol scattering may in some cases
give a first-order approximation of the plume’s water content [18,19]. However, factors other than
Remote Sens. 2017, 9, 534
9 of 11
the plume’s water abundance also impact the plume scattering efficiency. Environmental conditions
such as the temperature and relative humidity of the background air, the vent exit temperature, and
the concentration of sulfate aerosols and other condensation nuclei in the plume all affect the plume
aerosol optical depth and thus the expected scattering efficiency of the plume. In addition, even
if the plume and environmental conditions remained constant, geometric parameters such as the
solar zenith angle, the solar azimuth angle relative to the observation azimuth, and the measurement
elevation angle will all influence the expected contribution from in-plume aerosol scattering to the
total detected radiance.
Given the sensitivity of the measured signal to such a large number of potentially varying
parameters, the method can only be expected to achieve quantitative results if very frequently calibrated
with an independent measure of plume water content. Pering et al. chose to calibrate their measurements
using simultaneously recorded measurements of relative humidity within the plume. This approach
could work, though measuring the total water content in both the vapor and liquid phase would require
a heated instrument. Otherwise, the presence of condensed liquid water in the instrument or gas inlet
would simply lead to a saturation of the relative humidity at 100%.
Pering et al. also reported coincident emission rates of SO2 and H2 O from the Northeast Crater of
Mount Etna. Figures 3 and 4 in [8] show what appears to be a positive correlation between H2 O and SO2
fluxes, but the above discussion introduces some ambivalence into these results. Since the measured
NIR water vapor absorption signal appears to be mainly associated with the scattering of light on
aerosols in the plume, the signal is expected to vary with the aerosol abundance. The abundance of
aerosols in the plume could be related to the water abundance, but it is certainly also related to the
abundance of sulfate aerosols and other condensation nuclei within the plume. Therefore, it seems
plausible that the efficiency of scattering within the plume could be modulated by the sulfate aerosol
concentration (which in turn is likely related to the SO2 output), even if the total water abundance
remained constant over time. Thus, the performed measurements are probably not well-suited to
provide quantitative information on the contemporaneous output fluxes of SO2 and H2 O.
7. Conclusions
Measuring the absorption of scattered sunlight by CO2 or H2 O in volcanic plumes is challenging.
The main difficulty in taking such measurements lies in the potential for a change in the effective
atmospheric light path of the detected radiation caused by scattering on plume aerosols. Since both
CO2 and H2 O have significant atmospheric background concentrations, any change in the atmospheric
light path will also affect the measured absorption signal, regardless of the CO2 or H2 O abundance
inside the plume. This difficulty is not limited to ground-based measurements. Earth-observing satellite
instruments face the very same challenges, which must be overcome before reliable measurements of
volcanic CO2 or H2 O can be made.
Pering et al. [8] reported on NIR-imaging measurements of volcanic plumes at Vulcano Island
and Mount Enta. Their novel method measures the absorption of NIR light by water vapor. However,
the examples discussed here show that the method is extremely sensitive to the scattering of light on
aerosols within the plume. The signal caused by changes in the effective atmospheric light path due to
in-plume scattering is about two orders of magnitude larger than the expected absorption signal from
water vapor in the plume itself. The proposed method for correcting aerosol scattering by normalizing
the measured optical depth with an optical depth obtained outside the water absorption wavelength
region (Equation (3)) does not work, as it does not take absorption of light by atmospheric water vapor
into account. Therefore, similar to other methods that attempt to relate a plume’s visible opacity to
its water content [18,19], the technique can only produce quantitative information if very frequently
calibrated with an independent measurement method. Also, since the aerosol abundance within the
plume can influence its scattering efficiency, the NIR-imaging technique cannot easily distinguish
between a change in the abundance of sulfate aerosols or water vapor in a volcanic plume.
Remote Sens. 2017, 9, 534
10 of 11
Despite the challenges associated with the quantitative remote sensing of volcanic CO2 and H2 O,
a few approaches do show promise. For one, any type of active-source measurement (e.g., active-source
FTIR or active long-path DOAS) can be successful, since this geometry eliminates the need for detected
radiation to pass through the entire atmosphere before reaching the sensor. When attempting to
use sunlight, solar occultation measurements have the great advantage that by far the majority
of the detected radiation has passed through the atmosphere (and plume) in a straight line from
the sun to the sensor [5]. Scattered-light measurements are much more difficult because they are
susceptible to variations in atmospheric radiative transfer. One starting point could be the application
of sensors to high-elevation volcanoes, thereby reducing the overhead atmospheric CO2 and H2 O
columns. For example, water vapor absorption was recently detected in the plume of Sabancaya
Volcano, Peru, at 6000 m above sea level using the DOAS technique [20]. Scattered-light measurements
should also focus on completely transparent volcanic plumes to minimize in-plume scattering effects.
This typically means measuring a plume at a distance from the vent, rather than directly at the point
of emission. Through a combination of these approaches, continuous monitoring of volcanic CO2
and H2 O emissions may become possible without the need to install instrumentation within the gas
plumes themselves.
Acknowledgments: The author would like to thank the Remote Sensing editorial board for welcoming this
comment, and Peter Kelly for the helpful discussions during its preparation. In addition, the author would like
to thank Jake Lowenstern, Nicole Bobrowski, Ulrich Platt and three anonymous reviewers for their thoughtful
comments which greatly improved the manuscript.
Conflicts of Interest: The author declares no conflicts of interest.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Aiuppa, A.; Federicao, C.; Giudice, G.; Gurrieri, S. Chemical mapping of a fumarolic field: La Fossa Crater,
Vulcano Island (Aeolian Islands, Italy). Geophys. Res. Lett. 2005, 32, L13309. [CrossRef]
Shinohara, H. A new technique to estimate volcanic gas composition: plume measurements with a portable
multi-sensor system. J. Volcanol. Geotherm. Res. 2005, 143, 319–333. [CrossRef]
Burton, M.R.; Oppenheimer, C.M.; Horrocks, L.A.; Francis, P.W. Remote sensing of CO2 and H2 O emission
rates from Masaya volcano, Nicaragua. Geology 2000, 28, 915–918. [CrossRef]
Oppenheimer, C.; Kyle, P.R. Probing the magma plumbing of Erebus volcano, Antarctica, by open-path FTIR
spectroscopy of gas emissions. J. Volcanol. Geotherm. Res. 2008, 177, 743–754. [CrossRef]
Butz, A.; Dinger, A.S.; Bobrowski, N.; Kostinek, J.; Fieber, L.; Fischerkeller, C.; Giuffrida, G.B.; Hase, F.;
Klappenbach, F.; Kuhn, J. Remote sensing of volcanic CO2 , HF, HCl, SO2 and BrO in the downwind plume
of Mt. Etna. Atmos. Meas. Tech. 2017, 10, 1–14. [CrossRef]
Galle, B.; Oppenheimer, C.; Geyer, A.; Mcgonigle, A.J.S.; Edmonds, M.; Horrocks, L. A miniaturised
ultraviolet spectrometer for remote sensing of SO2 fluxes: a new tool for volcano surveillance. J. Volcanol.
Geotherm. Res. 2002, 119, 241–254. [CrossRef]
Mori, T.; Burton, M. The SO2 camera: A simple, fast and cheap method for ground-based imaging of SO2 in
volcanic plumes. Geophys. Res. Lett. 2006, 33, L24804. [CrossRef]
Pering, T.D.; Mcgonigle, A.J.S.; Tamburello, G.; Aiuppa, A.; Bitetto, M.; Rubino, C.; Wilkes, T.C. A Novel and
Inexpensive Method for Measuring Volcanic Plume Water Fluxes at High Temporal Resolution. Remote Sens.
2017, 9, 1–13. [CrossRef]
Kern, C.; Kick, F.; Lübcke, P.; Vogel, L.; Wöhrbach, M.; Platt, U. Theoretical description of functionality,
applications, and limitations of SO2 cameras for the remote sensing of volcanic plumes. Atmos. Meas. Tech.
2010, 3, 733–749. [CrossRef]
Platt, U.; Stutz, J. Differential Optical Absorption Spectroscopy—Principles and Applications; Springer: Berlin,
Heidelberg, 2008.
Rothman, L.S.; Gordon, I.E.; Babikov, Y.; Barbe, A.; Chris Benner, D.; Bernath, P.F.; Birk, M.; Bizzocchi, L.;
Boudon, V.; Brown, L.R. The HITRAN2012 molecular spectroscopic database. J. Quant. Spectrosc. Radiat. Transf.
2013, 130, 4–50. [CrossRef]
Remote Sens. 2017, 9, 534
12.
13.
14.
15.
16.
17.
18.
19.
20.
11 of 11
August, E. F. Ueber die Berechnung der Expansivkraft des Wasserdunstes. Ann. Phys. 1828, 89, 122–137.
[CrossRef]
Alduchov, O.A.; Eskridge, R. E. Improved Magnus Form Approximation of Saturation Vapor Pressure.
J. Appl. Meteorol. 1996, 35, 601–609. [CrossRef]
Randel, D.L.; Vonder Harr, T.H.; Ringerud, M.A.; Stephens, G.L.; Greenwald, T.J.; Combs, C.L. A New Global
Water Vapor Dataset. Bull. Am. Meteorol. Soc. 1996, 77, 1233–1246. [CrossRef]
Francis, P.; Burton, M.R.; Oppenheimer, C. Remote measurements of volcanic gas compositions by solar
occultation spectroscopy. Nature 1998, 396, 567–570. [CrossRef]
Duffell, H.; Oppenheimer, C.; Burton, M. Volcanic gas emission rates measured by solar occultation
spectroscopy. Geophys. Res. Lett. 2001, 28, 3131–3134. [CrossRef]
Tomasi, C. Vertical Distribution Features of Atmospheric Water Vapor in the Mediterranean, Red Sea and
Indian Ocean. J. Geophys. Res. 1984, 89, 2563–2566. [CrossRef]
Matsushima, N.; Shinohara, H. Visible and invisible volcanic plumes. Geophys. Res. Lett. 2006, 33, 2–5.
[CrossRef]
Girona, T.; Costa, F.; Taisne, B.; Aggangan, B.; Sorvigenaleon, I. Fractal degassing from Erebus and Mayon
volcanoes revealed by a new method to monitor H2 O emission cycles. J. Geophys. Res. Solid Earth 2015, 120,
2988–3002. [CrossRef]
Kern, C.; Masias, P.; Apaza, F.; Reath, K. A.; Platt, U. Remote measurement of high preeruptive water vapor
emissions at Sabancaya volcano by passive differential optical absorption spectroscopy. J. Geophys. Res.
Solid Earth 2017, 122. [CrossRef]
© 2017 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).