Heard D.E. (editor) Analytical Techniques for Atmospheric Measurement

Подождите немного. Документ загружается.

162 Analytical Techniques for Atmospheric Measurement

the high levels observed >20 ppt were shown to be the cause of O

depletion events. In

addition, BrO measurements have also been performed at mid-latitude locations such as

the Dead Sea (Hebestreit et al., 1999), a salt lake (Stutz et al., 2002), and the mid-latitude

coastal marine boundary layer (Saiz-Lopez et al., 2004). ClO has also been detected by

DOAS in a salt lake mid-latitude boundary layer (Stutz et al., 2002).

A number of iodine species have been detected in the marine boundary layer by LP-

DOAS, including IO (Alicke et al., 1999; Allan et al., 2000a), OIO (Allan et al., 2001)

and I

(Saiz-Lopez & Plane, 2004). Figure 3.6(a) shows a spectral fit confirming the

presence of I

: the grey line is the atmospheric OD spectrum after subtracting several

absorbing species (e.g. H

O NO

and OIO, but not I

, compared to a reference spectrum

of I

(black line) (Saiz-Lopez & Plane, 2004). The molecule was detected at nighttime

concentrations of up to 90 ppt together with OIO levels of 10 ppt (averaged over the

4.2 km pathlength, illustrated in Figure 3.6b). Figure 3.7 shows measurements of the IO

radical off the north coast of Tenerife, Spain, which is probably more typical of open

ocean conditions. The measurements (solid points) are compared with model predictions

of IO from a constrained model that includes a full treatment of the processing of

iodine-containing species in sea-salt aerosol (solid lines) (McFiggans et al., 2000).

Finally, commercial DOAS systems have been also routinely used in air pollution–

monitoring networks (e.g. Kourtidis et al., 2000; Xie et al., 2004). The commercial

instruments usually consist of a transmitter and a receiver telescope attached to a

computer-controlled spectrometer (Section 3.3.1.1). Because of the higher concentrations

involved when studying urban pollution, optical pathlengths of typically a few hundred

× 10

–3

–3 × 10

–3

2 × 10

–3

–2 × 10

–3

1 × 10

–3

–1 × 10

–3

Differential optical density

540 580570560550

Wavelength (nm)

Time (GMT)

(a)

(b)

100

18.00 0.00 6.00 12.00 18.00

OIO / ppt

/ ppt

Tidal Height (m)

(i) (ii)

Figure 3.6 Spectral ﬁt and concentration proﬁles of I

. (a) The atmospheric OD spectrum (grey) is

matched to an OD reference spectrum for I

(black) corresponding to a retrieved concentration of

88±41 ppt. (b) The I

and OIO concentration proﬁles with the corresponding detection limits (thin black

lines) are plotted against tidal height (thick broken lines). The white and grey backgrounds indicate day

and nighttime periods, respectively. (Saiz-Lopez & Plane, 2004, reproduced with permission from the

American Geophysical Union.)

UV-Visible Differential Optical Absorption Spectroscopy 163

185 186 187 188 189 190 191 192

Julian days

193 194 195 196 197 198 199

Mixing ratio (ppt)

Predicted IO

Measured IO

Figure 3.7 Time series measurements of IO made by LP-DOAS off Tenerife, Spain. The measured

concentrations (solid points) are compared to model predictions of IO (solid lines) for a number of days

during the ﬁeld campaign. (McFiggans et al., 2000, reproduced with permission from the American

Geophysical Union.)

metres are used. Generally, DOAS monitoring networks emphasise measurements of

urban pollutants such as SO

, NO, NO

and aromatics.

3.3.2 Laser sources

3.3.2.1 OH radical measurements in the atmosphere

Hydroxyl radical has been detected in the atmosphere by optical absorption in the



−X

 electronic transition around 308 nm (Perner et al., 1987; Dorn et al., 1988;

Comes et al., 1992; Mount, 1992). Because the rotational constant of OH is very large,

rotational lines in the (0,0) vibrational band can be resolved even at atmospheric pressure.

Hence, a laser can be used as the light source in a DOAS instrument, such as the example

shown in Figure 3.8 (Hausmann et al., 1997). A picosecond laser at 616 nm is frequency-

doubled to 308 nm, and spectra are recorded with a high resolution Echelle spectrograph

coupled to a PDA detector. This set-up provides simultaneous monitoring of a spectral

width of 0.26 nm, which covers a number of OH absorption lines (Figure 3.9). An open

multi-reflection or White cell (38.5 m long) is used to achieve an absorption path of more

than 3 km. This is necessary because OH has a typical daytime peak concentration of only

about 5 ×10

molecules cm

−3

. The OD

min

of the instrument is about 2 ×10

−4

,sothe

detection limit of OH is about 15 ×10

molecules cm

−3

(Hausmann et al., 1997). There

have been other laser systems used for OH detection, for example a pulsed XeCl excimer

laser covering several OH lines (Mount, 1992) and a continuous wave narrow spectral

linewidth laser which scans very rapidly over several OH rotational lines (Armerding

et al., 1996).

164 Analytical Techniques for Atmospheric Measurement

Figure 3.8 Schematic diagram of a long-path DOAS instrument employing a laser light source for

tropospheric OH detection. The instrument consists of a high resolution Ebert spectrometer-photodiode

array (PDA) detector system in combination with an open path multiple-reﬂection cell. (Hausmann et al.,

1997, reproduced with permission from the American Geophysical Union.)

UV-Visible Differential Optical Absorption Spectroscopy 165

Figure 3.9 Wavelength interval used for OH tropospheric observations. The laser output spectrum,

the absorption spectrum around 308 nm of the A-X(0,0) band of OH (with individual rotational lines

identiﬁed), and the absorption spectra of other trace gases that potentially interfere in the LP-DOAS

detection of OH are plotted in panels (A) to (F), respectively. Note that the absorption band depicted in (F)

corresponds to an unidentiﬁed atmospheric component. (Dorn et al., 1996, reproduced with permission

from the American Geophysical Union.)

Note that in contrast to traditional LP-DOAS, where the retrieved concentration is

integrated over several kilometres, the multi-pass arrangement provides a measurement

integrated over a few tens of meters, which can be considered as in situ monitoring. Inter-

comparison of OH measurements by LP-DOAS and in situ laser-induced fluorescence

(LIF) systems has shown very satisfactory agreement (Hofzumahaus et al., 1998).

The spectra are analysed in a similar way to the broadband DOAS technique

(Section 3.3.1.2), as illustrated in Figure 3.9. This shows the laser output spectrum

(Panel (A)), and reference spectra for OH, SO

, HCHO and naphthalene (Panels (B)–(E),

respectively). Panel (F) records the absorption band of an unknown absorber. Very

precise calibrations of the reference spectra are needed because of the very narrow

wavelength interval used (307.96–308.19 nm) (Dorn et al., 1996; Mount et al., 1997;

Brauers et al., 2001). Figure 3.10(A) shows an atmospheric absorption spectrum at 308 nm,

and the corresponding superimposed fitted reference spectra together with a ninth-order

polynomial. The high-order polynomial is applied to correct the laser spectral baseline

and the broadband component of the atmospheric spectrum. Figure 3.10(B) depicts the

atmospheric spectrum after subtracting all the fitted reference spectra except for OH,

166 Analytical Techniques for Atmospheric Measurement

Figure 3.10 Spectral de-convolution of OH: (A) a ﬁt (dark line) to an atmospheric absorption spectrum

(grey line); (B) atmospheric spectrum with all species but OH ﬁtted out (grey line), and the ﬁtted OH

reference spectrum (dark line); (C) the residual obtained after all species are ﬁtted and subtracted from

the atmospheric spectrum. (Dorn et al., 1996, reprinted with permission from the American Geophysical

Union.)

and Figure 3.10(C) shows the residual after subtracting all reference spectra. The OH

absorption corresponds to an optical density of 2×10

−4

, demonstrating the ability of the

DOAS system to provide in situ measurements of OH radicals.

3.3.2.2 Cavity ring-down spectroscopy

Cavity ring-down spectroscopy (CRDS) was first developed in the late 1980s (O’Keefe

& Deacon, 1988), and since then its use for laboratory and atmospheric spectroscopy

has grown substantially (Paul & Saykally, 1997; Scherer et al., 1997; Ball & Jones, 2003;

Brown, 2003). The principle of the method is based on the gradual decrease of light

intensity as a light pulse undergoes multiple reflections between two highly reflective

mirrors in an optical cell. Figure 3.11 is an illustration of a typical CRDS set-up (Brown,

2003). The laser pulse is introduced into the CRD cavity through one of the mirrors. The

intensity of light transmitted out of the cavity by leaking through the mirrors will decay

exponentially with time (Paul & Saykally, 1997; Scherer et al., 1997). This intensity is

usually measured by placing a photomultiplier, with an appropriate optical filter, on the

far side of the optical cavity from the laser. The cavity length L is the distance between

the mirrors. The transit time between the mirrors should be large compared to the laser

pulsewidth in order to discriminate discrete pulses or cavity modes inside the cell, thus

ensuring the ring-down decay is exponential. If an absorbing species is now introduced

into the cavity, then the decay in light intensity will be caused both by reflection losses

and by molecular absorption, and as a result decay will be faster. The change in time

decay can be used to measure the OD of the absorbing species (Ball & Jones, 2003;

Brown, 2003).

The time decay depends ultimately on the mirror transmission T

=1−R, where R is

the reflectivity of the mirrors (see Equation 3.19)). The time for the beam to traverse the

cavity is given by L/c, where c is the speed of light. For a known absorption cross-section

UV-Visible Differential Optical Absorption Spectroscopy 167

Figure 3.11 Schematic diagram of a cavity ringdown spectrometer. (Brown, 2003, reproduced with

permission from American Chemical Society.)



, the absorption coefficient of species i (A

,incm

−1

 of the gas injected in the cavity

cell can then be written as:





−





= 

(3.18)

where 1/ and 1/

are the first-order decay rates for the empty CRD cell and in the

presence of the absorbing gas of concentration c

, respectively. The factor R

is the ratio

of L to the length in the cell over which the absorbing species is present. For atmospheric

applications, extinction due to Rayleigh and Mie scattering, and absorption by other

gas-phase species, must be considered together with the losses by mirror transmission in

the definition of 

in the absence of the target gas (Brown, 2003):



= c

⎛

⎝

+





⎞

⎠

(3.19)

where 

and 

are the wavelength-dependent extinction coefficients for Rayleigh and

Mie scattering, and A

are the absorption coefficients for the gas-phase species other than

the target compound. Hence, monitoring the CRD time decay with the empty cavity

and in the presence of the sample gas allows the measurement of the gas concentration

following the Beer–Lambert law.

Cavity ring-down spectroscopy has been used increasingly since 1990s for kinetic and

spectroscopic studies of a number of species of interest in atmospheric chemistry (Zhu

& Johnston, 1995; Atkinson & Hudgens, 1997; Newman et al., 1998, 1999; Schulz &

Simpson, 1998; Wheeler et al., 1998; Ashworth et al., 2002; Mazurenka et al., 2003; Nakano

et al., 2003). This proliferation of laboratory applications of CRDS has been due to the

high sensitivity against minimum detectable fractional absorbance per pass ∼10

−6

, and

a good time resolution 10

−6

–10

−4

s (Ball & Jones, 2003; Brown, 2003). In addition,

168 Analytical Techniques for Atmospheric Measurement

CRDS instruments have been used for field measurements of atmospheric species such

as NO

and N

(King et al., 2000; Ball et al., 2001; Brown et al., 2001; Simpson, 2003).

These nitrogen compounds are in thermal equilibrium NO

+NO

+M ↔ N

+M.

In order to achieve simultaneous measurements of both species, the CRDS instrument

contains two flow systems and ring-down cavities. The first flow system is kept at ambient

temperature for the detection of ambient NO

, whereas the second is heated up to

∼80



C to thermally decompose N

, therefore yielding the sum of the NO

and N

concentrations. An accurate measurement with CRDS requires a precise determination

of the empty cavity ring-down time 

. For instance, to measure the NO

radical 

includes other absorbers such as water vapour, NO

and the aerosol contribution. A

measurement of 

can therefore be obtained by titrating with NO NO

+NO →2NO

,

without affecting the concentrations of the other constituents (Brown et al., 2003).

An extension of CRDS is to use a broadband laser source (typically a dye laser with

a 20 nm bandwidth), and replace the photomultiplier with a spectrometer and a two-

dimensional CCD with fast time-clocking to record spectra with a better than 1 s time

resolution (Ball et al., 2001). Cavity ring-down absorption spectra are therefore obtained

simultaneously over a range of wavelengths, so in principle this technique combines the

advantages of CRDS with conventional DOAS. The technique has so far been employed

to detect the NO

radical by monitoring the spectral region between 650 and 670 nm

(Ball et al., 2001).

Finally, it should be noted that CRDS and LP-DOAS are complementary techniques

because they probe different spatial scales. The latter averages the trace gas concentration

over typically several kilometres and thus provides some degree of remote sensing, whereas

CRDS is an in situ measurement. The techniques have similar detection limits. Thus,

combining the two techniques at a field site where there is likely to be significant

heterogeneity because of local sources and sinks (e.g. in an urban environment or the

marine coastal zone) provides a very useful way of studying fast chemistry and transport.

Another new locked cavity method which offers absorbance detection limits as low

as 10

−13

has been developed. The noise-immune cavity-enhanced Optical Heterodyne

Modulation Spectrometry (NICE-OHMS) combines a high reflectivity cavity with signif-

icant noise reduction by modulation of the laser frequency (Ye et al., 1998). Although,

to date, the technique has not been applied to atmospheric measurements, its potential

capabilities have been demonstrated by detecting absorbances of ∼3 ×10

−11

from C

by using an ultra-stabilised Nd:YAG laser (Ma et al., 1999).

3.4 DOAS using scattered sunlight

The earliest spectroscopic remote sensing measurement was the observation of O

in the

atmosphere by Dobson and Harrison in 1926. In the mid-1970s, Noxon measured NO

in the upper troposphere and stratosphere by observing scattered sunlight at the zenith

(Noxon, 1975, 1976). This pioneering work was followed by a wide variety of studies

of many important atmospheric species (Noxon, 1981, 1983; McKenzie & Johnston,

1982; Solomon et al., 1987a,b; Mount et al., 1988; Pommereau & Goutail, 1988; Roscoe

et al., 1990; Smith & Solomon, 1990; Wahner et al., 1990a,b; Perner et al., 1991, 1994;

Fiedler et al., 1993; Sanders et al., 1993; Arpag et al., 1994; Pfeilsticker & Platt, 1994;

UV-Visible Differential Optical Absorption Spectroscopy 169

Pommereau & Piquard, 1994; Vanroozendael et al., 1994; Aliwell & Jones, 1996b; Aliwell

et al., 1997; Eisinger et al., 1997; Kreher et al., 1997; Platt et al., 1997; Richter et al.,

1999; Roscoe et al., 1999; Wittrock et al., 2000; Friess et al., 2001, 2004; Allan et al., 2002;

Honninger & Platt, 2002; von Friedeburg et al., 2002; Leser et al., 2003). Note that DOAS

measurements performed using lunar or stellar light are not the subject of this chapter,

but have been discussed in previous reviews of the DOAS technique (Platt, 1994; Plane &

Smith, 1995). This section contains a brief overview of the spectroscopic principles and

the main applications of scattered-sunlight DOAS, or passive DOAS.

3.4.1 Spectroscopic principles

As the source of light is now outside the earth’s atmosphere, the Beer–Lambert law

(Equation 3.1) is modified (Solomon et al., 1987a):

 = I

 exp



−SCD



(3.20)

where SCD is the slant column density of an atmospheric species, integrated along the

path followed by solar photons through the atmosphere. If n species are present, the total

OD is given by:

OD = ln









i=1

 =



i=1



SCD

(3.21)

Photons traversing the atmosphere are subject to Rayleigh and Mie scattering, and thus

wavelength-dependent extinction coefficients are introduced in the analysis in the same

manner as for the LP-DOAS technique (Section 3.2.1).

Passive DOAS spectra can therefore be analysed using the same analysis procedure as

for LP-DOAS, but there are several additional considerations when using scattered solar

radiation as the light source. The first is that the unattenuated I

 spectrum cannot

be measured (the same applies in LP-DOAS, but boundary layer spectra can at least be

normalised by dividing by the spectrum of the lamp in the transmitter – Section 3.3.1.2).

Instead, the I

 spectrum is usually taken as the solar spectrum when the solar zenith

angle (SZA) is at its minimum (i.e. at local noon), since the slant column through the

atmosphere is then a minimum (Noxon, 1975, 1976; Noxon et al., 1979). The objective

is to remove the Fraunhofer lines present in the solar spectrum, since these structures

correspond to up to 30% absorption, and are much larger than molecular absorption

features. Taking the ratio of the spectra at high and low SZAs largely removes the Fraun-

hofer structure, as shown in Figure 3.12. The two atmospheric scattered sunlight spectra

in Figure 3.12(a) are taken at 70



and 90



SZA, and the ratio compared to an O

reference

spectrum in Figure 3.12(b). This demonstrates how the strong absorption bands of O

become visible in the 317–335 nm region after the solar Fraunhofer structure is removed.

Another factor to consider when retrieving atmospheric species from scattered sunlight

spectra is the so-called ‘Ring effect’, named after Grainger and Ring (1962). This describes

the partial ‘filling-in’ of the Fraunhofer structures in the solar spectrum as sunlight

traverses through the atmosphere, resulting in a decrease of the ODs of Fraunhofer lines.

170 Analytical Techniques for Atmospheric Measurement

Intensity (arb. units)

Wavelength (nm)

320 330 340 350 360 370

0.0

0.5

1.0

reference

Atmospheric spectrum

Figure 3.12 Scattered sunlight atmospheric spectra: (a) atmospheric spectra taken at 70



(grey line) and



(solid line) SZA, showing the dominant Fraunhofer structure (the signal intensity is shown on the

left-hand axis); (b) the ratio of the atmospheric spectra in (a) (grey line) – the Fraunhofer lines largely

disappear to reveal the strong O

absorption bands in the region 317–335 nm, which are compared to

an O

reference spectrum (solid line). Note that both spectra have been normalised for comparison.

The Ring effect is most likely due to Raman scattering (Kattawar et al., 1981; Fish & Jones,

1995; Vountas et al., 1998; Sioris & Evans, 1999). The size of the Ring effect depends

strongly on SZA, with the result that at larger SZA the observed Fraunhofer lines are

weaker. Accurate removal of the Ring effect is necessary because it represents a change in

OD that can be an order of magnitude larger than that from absorption by atmospheric

trace gases. An artificially generated Ring spectrum is usually employed in the analysis

of scattered sunlight spectra (Chance & Spurr, 1997). An additional complication is that

cloudy conditions increase the optical path of scattered sunlight reaching the ground,

which can greatly enhance the Ring effect (de Beek et al., 2001).

Analysis of scattered sunlight spectra yields the SCD (Equation 3.21). However, the

vertical column density (VCD) is a more useful quantity. The conversion from SCD to

VCD is obtained by means of the so-called air mass factor (AMF) (Solomon et al., 1987a;

Perliski & Solomon, 1993):

VCD =

SCD

AMF

(3.22)

The AMF depends on the radiative transfer of the atmosphere and can be calculated

from radiative transfer models (Marquard et al., 2000; von Friedeburg et al., 2002).

UV-Visible Differential Optical Absorption Spectroscopy 171

These models must include parameters such as wavelength-dependent Rayleigh and Mie

scattering, vertical profiles of pressure, temperature and the species of interest, as well as

the SZA and azimuth angle of the receiving telescope. For SZA < 75



, the AMF is well

approximated by the simple geometrical relationship 1/ cosSZA.

3.4.2 Zenith-viewing geometry: Retrieval of NO

vertical

proﬁles

Zenith-sky spectroscopy is based on the collection of scattered photons using a zenith-

pointing telescope, from either a ground-based or airborne instrument. Although routine

monitoring of O

and NO

in the stratosphere has been conducted for many years

using zenith-pointing instruments (Noxon, 1975; Harrison, 1979; Noxon et al., 1979;

Syed & Harrison, 1980; Noxon, 1981; McKenzie & Johnston, 1982, 1983; Noxon, 1983;

McKenzie & Johnston, 1984; Mount et al., 1988; Pommereau & Goutail, 1988; Roscoe

et al., 1990; Goutail et al., 1994; Gil et al., 1996; Sarkissian et al., 1997; VanRoozendael

et al., 1997; Vaughan et al., 1997; Preston et al., 1998; Roscoe et al., 1999; Gil et al.,

2000; Liley et al., 2000), here we will focus on ground-based DOAS measurements of

. During the night, NO

has been monitored using direct observations of the moon

as a light source (Noxon et al., 1978, 1980; Noxon, 1983; Sanders et al., 1987; Solomon

et al., 1989a,b; Wagner et al., 2000). Since 1990s NO

columns have been measured for

an hour or so before sunrise, by collecting scattered sunlight using zenith and off-axis

DOAS when the SZA is less than about 95



(Weaver et al., 1996; Aliwell & Jones, 1998;

von Friedeburg et al., 2002). In fact, NO

is photolysed so rapidly that the decrease in

column as the solar terminator sweeps downward before sunrise can be used to

determine the vertical profile of the radical in the troposphere and lower stratosphere

(Smith & Solomon, 1990; Smith et al., 1993).

Figure 3.13(a) shows how NO

is sequentially removed by photolysis as the SZA decreases

to 90



(sunrise). The VCDs of NO

are derived in a similar way to the NO

concentrations

determined with an LP-DOAS instrument (Allan et al., 2002) (Section 3.3.1.2), where each

processed atmospheric spectrum is divided by a processed daytime reference spectrum. Note

that the lifetime of NO

against photolysis is only a few seconds in full sunlight, so the persis-

tence of NO

when SZA is less than 90



indicates a large quantity of N

is present. This is in

equilibrium with NO

NO

+NO

↔ N

, so the relatively slow thermal decomposition

of N

has the effect of delaying the disappearance of NO

after sunrise. This effect occurs

in semi-polluted environments with NO

mixing ratios above 500 ppt (Coe et al., 2002).

The generic optimal estimation method (OEM) developed by Rodgers (1976, 1990) for

solving atmospheric data inversion problems has been successfully developed to retrieve

vertical profile information of NO

from zenith-sky spectroscopic measurements of

column abundance made through sunrise (Coe et al., 2002). A forward model described by

the expression y =Kx predicts the column density time series, y, that would be observed

from a vertical concentration profile, x, connected through a weighting function, K. The

elements of the matrix K are the degree of photolysis of NO

as a function of height,

SZA and time (Coe et al., 2002). In fact, since x is unknown the forward model is

inverted to yield the backward model x = K

−1

y. The inversion process begins from an

a priori guess of the NO

profile x

; the backward model is then used to calculate an