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

Подождите немного. Документ загружается.

252 Analytical Techniques for Atmospheric Measurement

ions. Normally the acetone ions, mass 59, are transmitted around 2.4 times better than

the H

and this must be accounted for in the quantification. Since the transmission

of the ions measured can change markedly as the detector ages it is advisable to ascertain

the transmission regularly when operating such instrumentation in the field.

In measurement mode, the resulting signal is the number of counts detected at each

mass per second, which can be converted to a mixing ratio if the rate constant for the

protonation reaction and the reaction time are known (see Section 5.4.1). All exothermic

proton transfer reactions proceed at, or close to, the collisional rate, and where specific

reactions rates have not been measured they can be calculated (Su & Chesnavich, 1982).

In practice the PTR-MS is calibrated in the laboratory against commercially available gas

standards but derived rate coefficients are checked for consistency with theoretical ones.

In addition to the uncertainty in the gas standards there is statistical error associated with

the count rates that must be considered in the overall error assessment. This noise signal

(NS) is determined by Equation (5.21) (Hayward et al., 2002), where mean signals are in

counts per second and dwell time is the time spent counting at a given mass.

NS =

mean signal (cps)



(mean signal (cps) ×dwell time (s)

(5.22)

The statistical error of PTR-MS data with a 2-second integration time is typically ±10%

at volume mixing ratios around 5 nmol/mol and ±30% at 1 nmol/mol. The errors in the

absolute values are between 10 and 30%, being largely determined by uncertainty in the

gas-phase standard. Optimisation of the PTR-MS technique for airborne measurement

resulted in detection limits for most species between 10 and 100 pmol/mol, which are

determined from the background (air through catalyst) values at each mass. The accuracy,

precision and detection limit of this method are currently worse than gas chromatographic

methods (see Section 5.3.4). However, the measurement frequency, and hence data

coverage, is considerably better. The high frequency of measurement is ideally suited to

fast-moving platforms such as jet aircraft that travel at typical speeds of 150 ms

−1

. The

PTR-MS has been operated on the ground (Salisbury et al., 2003), on planes (Crutzen

et al., 2000) and on ships (Williams et al., 2004). In the study conducted on ships two

PTR-MS systems were used to measure species such as DMS, methanol, acetone and

acetonitrile in the air and the surface seawater simultaneously.

The main disadvantage of the technique for atmospheric research is that several

compounds can contribute to the signal at a given mass. For example, mass 59 corre-

sponds to protonated acetone, propanal and glyoxal, all of which may be present in

ambient air (Williams et al., 2001). In some cases it is possible, by examining the potential

contributing compounds or the variation of the mass signal, to deduce a most likely

candidate for a given mass. However, in many cases the concentration determined for a

given mass cannot be assigned to a single compound unambiguously. Making the PTR-

MS technique more compound-specific is a key goal in the future, see Section 5.5. Some

groups have replaced the commonly used quadrupole detector with a TOF MS (Blake

et al., 2004). While the PTR-TOF system has demonstrated improved mass resolution

over a quadrupole, there are considerable drawbacks. First, improved resolution will

never resolve some of the key overlapping species in the atmosphere (e.g. acetone and

propanal) since they have identical masses. Secondly the PTR-TOF system is necessarily

Mass Spectrometric Methods for Atmospheric Trace Gases 253

much larger and heavier than the quadrupole, and the sensitivity at this stage in the

development is some 2 orders of magnitude less. At the time of writing, it would seem

that a PTR-ion trap would be a prudent line of instrument development.

An example of a vertical profile of acetone, methanol and acetonitrile mixing ratios up

to 13 km is shown in Figure 5.13. These data were obtained on board a German research

aircraft flying over the Mediterranean. It is clear that there is a high degree of structure

in the profiles. By studying the back trajectories at different altitudes it was possible to

determine the origins of the airmasses as: 8–13 km from Southeast Asia; 4–8 km from

the Atlantic Ocean region; and 0–4 km from Eastern Europe. The chemical signatures of

these airmasses are also distinct and reflect the sources in the aforementioned regions.

From previous studies we know that acetonitrile is a good marker for biomass burning,

methanol is primarily released from vegetation and that acetone can be produced from

both anthropogenic and biogenic sources. Thus we can see that in July, biomass burning

is important in Eastern Europe and Southeast Asia but not in the airmasses from the

West. By careful analysis of such datasets we may make estimates of the global sources for

these compunds. These data are described in detail by Lelieveld et al. (2002). The clear

vertical structure in this plot is obtained because of the key advantage of the PTR-MS

(and API-CIMS) techniques, namely high measurement frequency.

5.4.3 The use of CIMS for radical measurement (OH)

The hydroxyl radical is the most important oxidant in the lower atmosphere, and therefore

the primary removal mechanism for most trace gas species in ambient air. As a result there

has been considerable interest in measuring the concentration of this radical. However,

as OH is found at very low concentrations (ca. 1 ×10

molecules cm

−3

 and the radical is

inherently very reactive (chemical lifetime 0.1–1 seconds), so its measurement represents

500

120 140 160

Acetonitrile pptv

180 200 220

Altitude (km)

1000 1500

Methanol pptv Acetone pptv

2000 2500

Methanol

Acetone

Acetonitrile

Figure 5.13 An example of PTR-MS data collected from an aircraft.

254 Analytical Techniques for Atmospheric Measurement

a considerable challenge. Three methods have emerged as capable of making routine

atmospheric measurements of OH, two optical methods (DOAS and LIF described in

Chapters 3 and 4, respectively) and one mass spectrometric. The mass spectrometric

method described here has an OH sensitivity similar to LIF but is about 10 times more

sensitive than DOAS. In the mass spectrometric method, the determination of OH is

indirect as the OH radicals are rapidly titrated by

prior to measurement. This

titration of OH is achieved via the reactions:

OH +

+M →H

+M (5.23)

→

+HO

(5.24)

O +M →H

+M (5.25)

A typical experimental arrangement is shown in Figure 5.14. Air is drawn into the

instrument at a rate of about 10L min

−1

. Sufficient levels of

are introduced into

the inlet to ensure that the OH is almost completely reacted within a few milliseconds.

This prevents significant loss of the OH to the walls of the inlet. H

is not present at

significant concentrations in the atmosphere and can be accurately measured by negative

ion chemical ionisation mass spectrometry. A number of important features distinguish

this radical measurement from the API-CIMS system described in Section 5.4.1, although

in both cases the ions are filtered with a quadrupole (Section 5.3.2.1). For the ionisation it

was found that direct ionisation of the air sample was unsuitable for OH measurements as

it generated a high background signal and interferences by other ion–molecule reactions

and other radicals or unstable species also generated in the air stream. Chemical ionisation

is achieved by NO

−

reagent ions which are generated separately from the sample air and

Figure 5.14 A schematic diagram of a mass spectrometer to measure OH. (Berresheim et al., 2000 with

permission from Elsevier.)

Mass Spectrometric Methods for Atmospheric Trace Gases 255

then drawn through the sample by an electric field. To generate these ions, a radioactive

emitter 

241

Am is placed in a sheath flow of purified air enriched with HNO

vapour

and high purity propane. The chemical ionisation then proceeds via the reaction:

+NO

−

→ H

−

+HNO

(5.26)

Although the majority of the NO

−

reactant ions are clustered with neutral HNO

and/or H

O molecules, they may be nominally considered to be NO

−

as all clusters have

been found to react at very similar rates. As described in Section 5.4.1, by measuring the

reagent ion NO

−

at m/z =63 and the product H

−

at m/z =99 a concentration of

(and hence OH) can be deduced, if the rate coefficient and the reaction time

are known. Nonetheless, because of uncertainties in the rate coefficient and possible inlet

effects, direct calibration is also performed by producing a known concentration of OH

in the sample air by online photolysis of atmospheric water vapour whose concentration

in turn is accurately determined by dewpoint hygrometry (see Chapter 4 Section 5.4.3.2).

A further complication of this measurement is that under certain atmospheric conditions

significant quantities of OH can be generated (or recycled) within the system through

the reactions of NO or ozone with HO

and RO

. To minimise this potential positive

measurement artefact, excess propane is added downstream from the

titration step

to remove OH that might be generated from HO

. For background signal measurements,

propane is added at the beginning of the titration zone to remove ambient OH.

OH measurement time resolution with this system is typically 30 seconds. The detection

limit is about 2 ×10

molecules cm

−3

based on 5-minute signal integration, while

precision and accuracy are estimated to be 48 and 54% respectively. Such a system

has been deployed on the ground for long-term measurements of OH at the German

Weather Service’s (DWD) Meteorological Observatory, Hohenpeissenberg (Berresheim

et al., 2000). A regularly updated evaluation of the OH monitoring data can be seen on

the website http://www.dwd.de/en/FundE/Observator/MOHP/. A similar system has also

been adapted for use on aircraft and flown in several field measurement campaigns (e.g.

Eisele et al., 2001).

A novel extension to this method has been described by Hanke et al. (2001). This has

enabled the on-line measurement of peroxy radical measurements, namely HO

and the

sum of the organic peroxy radicals, RO

. This method relies on the amplifying chemical

conversion of peroxy radicals via the chain reaction with NO and SO

, to make gaseous

sulphuric acid, which is detected by CIMS. The ratio of the OH reaction rates with SO

and NO is equal to the ratio of the H

and the total peroxy radical concentration

HO

+RO

. By reducing the O

ratio, the H

-forming reactions of RO

can

be so disfavoured that measured H

is almost entirely formed by HO

. In this way

it has proven possible to measure HO

and the sum of HO

and RO

radicals in the

atmosphere, with a time resolution of 1 minute and a detection limit of 0.5 pptv. These

radicals were accurately measured at mixing ratios between 1 and 30 pmol/mol (pptv). An

alternative but related method for measuring RO

+HO

has been developed which does

not use the amplification, and instead relies on the extreme sensitivity of the CIMS system

to determine OH (Edwards et al., 2003). In this case the peroxy radicals are converted

with NO to OH, in a pre-reactor held at constant low pressure. This instrument has

been employed to measure the evolution of peroxy radicals in the northern hemisphere

winter–spring transition (Cantrell et al., 2003).

256 Analytical Techniques for Atmospheric Measurement

5.4.4 Coupled GC-MS capabilities

The combination of a gas chromatograph and a mass spectrometer has become one

of the most powerful and widely used tools for the analysis of complex mixtures such

as air. In this application, spectra are collected for separated compounds as they exit

from a chromatographic column. By monitoring specific ions the mass spectrometer can

detect trace gas species more specifically and sensitively than many conventional GC

detectors (see Chapter 8). A variety of GC-MS quadrupole systems are commercially

available which can be reliably automated and deployed for long-term measurements.

Furthermore, the MS can be operated in either EI or CI mode, offering a broad range of

detectable species. While the GC-MS can detect many species specifically and sensitively,

analysis times of typically 1 hour limit the temporal data coverage. This is in contrast

with the CIMS methods described above that are capable of measuring on-line albeit with

less sensitivity and specificity. Since 2000, considerable effort has been invested to make

the GC-MS measurements more rapidly. Some current examples of these applications

are given below.

A global network of GC-MS systems (Agilent 5973) has been set up by Prinn et al. (2000)

in order to monitor the long-term changes of trace gas species. The sites have been selected

in remote locations to assess changes in the background concentrations of chemically and

radiatively important trace gas species. The sites include Cape Grim, Tasmania (41



S, 145



E),

Cape Matatula, American Samoa (14



S, 171



E), Ragged Point, Barbados (13



N, 59



W), Mace

Head, Ireland (53



N, 10



W), and Trinidad Head, California (41



N, 124



W).

Of particular interest are the man-made hydrochloroflourocarbons (HCFCs) 134a,

141b and 142b that have been introduced as replacements for the stratospheric ozone

depleting chlorofluorocarbons (e.g. CFC 11 and 12). At least four measurements are

made per day with approximately 1% precision and the calibrated data are available on

the network homepage http://cdiac.ornl.gov/ndps/alegage.html. Lifetimes of HCFCs are

significantly shorter than that of CFCs because they are capable of being scavenged by

OH in the troposphere. For this reason, HCFCs do not cause as much ozone depletion

per molecule on a 100-year timescale as compared to the CFCs (Montzka & Fraser, 2002).

However the mole fractions of these species are steadily increasing in both hemispheres.

If the release rate of such species is known, and global measurements are available,

then it is theoretically possible to determine the global average OH. In this way GC-

MS measurements can contribute to the long-term assessment of the global oxidising

capacity.

The high sensitivity of the GC-MS can be exploited to measure trace gas species at

extremely low mixing ratios. For example, the ambient concentrations of halon 1202

CF

 can be reliably measured in ambient air even though it is present at only

0.045 pmol/mol (pptv) (Fraser et al., 1999). These measurements were made using

a magnetic sector mass spectrometer with an EI source. The presence of previously

unknown but potentially important gas SF

was discovered using the same instrument

(Sturges et al., 2000). These compounds were first observed at mixing ratios of around

0.005 pmol/mol (pptv). Assuming this as a detection limit we may estimate that if 200

tonnes of such material was emitted anywhere in the world it would be detectable by this

instrument, provided the troposphere is well mixed. Therefore even when relatively small

emissions of long-lived species occur in the atmosphere they can be detected by GC-MS

Mass Spectrometric Methods for Atmospheric Trace Gases 257

Jan-65

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.0

0.5

Jan-70 Jan-75 Jan-80

Mean date of firn air

Concentration (ppt)

Jan-85 Jan-90 Jan-95 Jan-00

Concentration (ppt)

0.14

0.12

0.10

0.08

0.06

0.04

0.02

0.00

Figure 5.15 An example of GC-MS data collected from ﬁrn air. (Courtesy of Dr W. Sturges, UEA.)

all over the world. Figure 5.15 shows the atmospheric mixing ratios of SF

and SF

since January 1965. The air for this analysis was pumped up from where it was naturally

trapped in consolidated deep snow, in eastern Antarctica. Air from 1965 was determined

to be 100 m below the snow surface, where after freezing in the ice matrix it has remained

isolated from mixing with present-day air. SF

has a current growth rate of about

0.008 pptv per year (ca. 6% per year). The similarity of the growth curves for SF

and

has the researchers to speculate that the former is a breakdown product of the latter

in high-voltage equipment.

The examples given above all involve the application of GC-MS with EI ion sources. The

commercially available negative ion chemical ionisation source has also been applied to

atmospheric measurements. The results are very promising and worth mentioning here.

By feeding a small flow of methane into the ionisation chamber, thermalised electrons can

be generated that can interact with organohalogen and organonitrate species, to produce

measurable ions at m/z = 35 (Cl), 79 (Br), 127 (I) or 46 NO

−

.

+e 70eV →CH

+2e thermal (5.27)

RONO

+e thermal →RO +NO

−

(5.28)

RX +e thermal →R +X

−

m/z =

Cl

Br

127

I (5.29)

The results show simple spectra with little co-elution and low background signals. The

sensitivity is comparable to that of the electron capture detector (0.01 pmol/mol (pptv)

in a 1000 ml sample) and the combination of retention time and ion measurement means

compounds can be identified easily and with a high degree of certainty. Such measure-

ments can be of great interest over the open ocean where emissions of organohalogen

and alkyl nitrates have been identified.

In the ground-based examples given above, the GC-MS analysis times are typically

about 1 hour. Considerable progress has been made in reducing the analysis time of

GC-MS systems used in atmospheric analysis. By miniaturising the gas chromatograph

and cryogenic preconcentrator, a 2–5 minute duty cycle has shown to be possible for

258 Analytical Techniques for Atmospheric Measurement

measurements of the C

to C

carbonyl compounds and methanol (Apel et al., 2003).

This instrument has been flown on a jet aircraft and shown to be able to detect low

pmol/mol levels off the target analytes. While this is not as fast as the systems described

in Section 5.4.1 and 5.4.2, the GC-MS system is generally more sensitive, offers a broader

range of potentially detectable compounds and unlikely to be affected by interfering

compounds at the same mass.

5.4.5 The isotope ratio mass spectrometer

The isotope ratio of a gas phase compound is generally defined as the ratio of a less

abundant isotope to the most abundant isotope. The ‘natural’ abundance of the

Cto

C is 1.1%. However, there are subtle but significant variations in the ratios and these

can be determined using high precision isotope ratio mass spectrometry (IRMS). From

a knowledge of the isotopic composition for a trace gas species in the atmosphere, we

may learn much about its sources and chemistry. Gases emitted from certain sources

have characteristic isotopic ratios. For example it has been found that methane emitted

from rice paddies contains 2% less carbon 13 

C than atmospheric methane. Moreover,

removal of methane from the atmosphere by OH, O

D and Cl causes the isotope fraction

to change, and so the isotope ratio can also indicate the relative importance of the removal

mechanisms (Brenninkmeijer et al., 2003). This is information that is not accessible by

measuring the concentration of the trace gases alone. Indeed, our understanding of the

global cycles of CH

and CO

stems partly from isotopic analysis, and our knowledge of

past climate change is based on isotopic analysis of H

O in ice cores or carbonates in

sediments. Isotopic measurements of stratospheric carbon dioxide have shown how such

measurements can be used to examine carbon exchanges between the atmosphere and

biosphere on interannual to glacial–interglacial time scales (Boering et al., 2004). It is

interesting to note that one of the oldest atmospheric measurements, the determination of

oxygen abundance, is now most accurately performed with an isotope mass spectrometer

able to measure subtle variations in the approximately 20% oxygen content. The same

techniques are now being applied to volatile organic carbon compounds which are

present at much lower (nmol/mol-pmol/mol or ppbv-pptv) atmospheric concentrations

(Goldstein & Shaw, 2003).

In the case of C

, the largest deviation from the standard for naturally occurring

molecules is approximately 10%, but more usually deviations are observable in the fourth

significant digit. These small but important changes are normally expressed as the per

mil (‰) difference to an agreed standard. To measure these effects requires an analytical

set-up of extremely high precision, with standard deviations in the range 4–6 significant

figures. The precision of trace gas concentrations determined by the GC-quadrupole

systems described in Section 5.4.3 are typically between 1 and 5%. This is clearly not

sufficient for the isotopic analyses, which are performed exclusively on magnetic sector

instruments with electron impact sources and multiple collectors as detectors to ensure

the requisite stability and hence precision. A schematic diagram of suitable apparatus is

shown in Figure 5.16. Normally the source is operated at 70 eV and ionisation occurs at

high efficiency. Typically out of 1000 neutrals entering the source, one ion reaches the

detector. In IRMS, mass resolution is sacrificed to optimise precision. Hence an electric

Mass Spectrometric Methods for Atmospheric Trace Gases 259

sector that can degrade the later is uncommon. A uniform magnetic field is required to

produce multiple well-defined beams emerging towards the detectors. This is typically

provided by a permanent magnet of around 0.75 tesla. A series of fixed Faraday cups

(multicollector system) are used as the detector system with one cup being allocated for

each ion (see Figure 5.16 and Section 5.3.3). Although Faraday cups are less sensitive than

electron multipliers, they are superior for IRMS because: the high count rates needed

for high precision measurements are in their normal operating range whereas high ion

currents would damage an electron multiplier; the Faraday cup is able to detect very

large ion currents without gain adjustment; and the cup is stable and has a long working

lifetime (ca. 10 years).

J.J. Thompson and F.W. Aston were the first to establish the nominal terrestrial isotopic

abundances of the elements. Between 1940 and 1980 the straightforward method of dual

inlet analysis was used to determine isotopic ratios in many fields. For CO

analysis this

method involved a pre-separated and combusted sample, and a reference gas sample each

being attached to an IRMS inlet. The sample and reference are then alternately admitted

into the IRMS, allowed to stabilise and the response (e.g. for m/z 44, 45, 46) recorded

at a plateau level. The measurement cycle is then repeated many times. This method

was superseded in the 1980s by the continuous flow (CF) method (Barrie et al., 1984).

This approach involves separating the individual compounds by gas chromatography,

combusting or reducing the eluting compounds to CO

or H

O, and continuously

analysing each peak with the IRMS. When analysing air for trace gases at low concentra-

tions such as methane or volatile organic hydrocarbons, the sample is first cryogenically

concentrated (see Chapter 8), before separation by gas chromatography. Ambient air

contains approximately 18 mol/mol (ppmv) of methane and it is necessary to concen-

trate 25 mL to measure the C

. For other organic gases at lower ambient volume

mixing ratios <100pptv a correspondingly higher volume is required (10–20 L). The

organic gases exiting the GC enter a combustion chamber where analytes such as methane

or other hydrocarbons are combusted to CO

with 99.99% efficiency in a NiO filled

chamber heated to around 1000



C. Following combustion the CO

peaks corresponding

to the separated and combusted trace gases are introduced to the mass spectrometer

entrained in a stream of helium carrier gas. For VOCs, depending on the size of the

peaks the precision varies from 0.2 to 3.6 per mil ‰, 10

−4

 (Rudolph et al., 1997).

Figure 5.16 A schematic diagram of a continuous ﬂow isotope ratio mass spectrometer.

260 Analytical Techniques for Atmospheric Measurement

5.5 Future developments

The importance of mass spectrometry in atmospheric research will almost certainly grow

in the future. Continuing technological improvements are expected to reduce the size,

power, analysis times and cost of MS systems. Increased efforts are being made to develop

miniature mass spectrometers which are hand portable but retain the performance of

laboratory based instruments. For atmospheric research this would considerably simplify

the investigation of sources in the field and installation on balloon or aircraft platforms

where weight and space are critical. It is expected that as vacuum technology is improved

and pumps become smaller considerable progress can be made. This is because in many

cases it is not the mass analyser but the associated electronics and vacuum technology

that limit the miniaturisation (Badman & Cooks, 2000). Considerable funds have been

made available to develop this field with the aim of providing specific, portable detectors

capable of detecting toxic gases rapidly. Atmospheric research stands to gain from these

rapid developments, particular in the analysis of organic trace species which play a key

role in global air chemistry (Williams, 2004).

Presently both GC-MS and CIMS type instrumentation are used to measure trace gases

in the atmosphere. Extensive validation though intercomparison has taken place between

GC-MS and CIMS methods (Kato et al., 2004 and references therein); between PTR-MS

and API-CIMS methods (Sprung et al., 2001), and between two differently configured

PTR-MS systems (de Gouw et al., 2004). All the intercomprisons to date have reported

generally satisfactory results. The GC-MS system offers the atmospheric researcher high

selectivity and sensitivity but long analysis times. In contrast, CIMS systems provide

better spatial and temporal coverage through faster measurements, but are currently

less sensitive and selective. It can be expected that the capabilities of both systems will

converge in the coming years. By replacing the quadrupole used in the CIMS with an ion

trap, much higher selectivity (but not yet sensitivity) has been shown to be possible with a

PTR-MS type system on the ground (Prazeller et al., 2003). The more sensitive API-CIMS

system has also been coupled to an ion trap and even aircraft measurements conducted

successfully (Kiendler & Arnold, 2003). An alternative approach using a triple quadrupole

to sequentially select ions, fragment ions and then detect the fragmented ions has also

been shown to be viable (Reiner et al., 1998). For CIMS systems it seems that further ion

schemes will be employed to detect new ranges of molecules. For example, it was shown

that peroxy acetyl nitrate (PAN) type compounds can be detected by a CIMS type method

using a I

−

reagent ion (Slusher et al., 2004). GC-MS systems will continue to improve the

frequency of measurement and the number of quantifiable compounds. New narrow bore

columns and miniature GC ovens with low thermal mass enable faster chromatography

to be achieved. Fast in situ measurements by GC will become more common. In both

Europe and the United States new high altitude, long range aircraft will shortly become

available to researchers. As space will be limited, competition between these techniques

can be expected. It will be interesting to see which proves more successful.

The field of isotope ratio mass spectrometry is expected to develop rapidly in the

coming years. Advances in the analysis techniques for compound specific measurements

of carbon isotopes in small samples and the availability of reliable commercial mass

spectrometers open a new field to atmospheric researchers. Considerable progress in the

isotopic characterisation of sources and photochemical processes can be expected. The

Mass Spectrometric Methods for Atmospheric Trace Gases 261

aforementioned carbon isotope studies will be further expanded through the simultaneous

examination of N, Cl, Br or O isotopes.

Generally it can be expected that in the future mass spectrometers will become more

accessible to the atmospheric researcher, replacing less sensitive and selective systems

currently in use. Mass spectrometric determinations of trace gases in the atmosphere will

become more routine, both on the ground though global instrument networks, and in

the air on commercial and research aircraft. Experience gained in this way will aid the

investigation of extra terrestrial atmospheres by mass spectrometry for trace gas species,

which has already begun (Fenselau & Caprioli, 2003).

Acknowledgements

Information and helpful advice is gratefully acknowledged from my colleagues Prof.

F. Arnold (Max Planck Institute – Heidelberg), Dr C.A.M. Brenninkmeijer and Dr J.

Crowley (Max Planck Institute – Mainz), Dr H. Berresheim (Deutsche Wetter Dienst), Dr

W. Sturges (University of East Anglia), T. Klüpfel and G. Eerdekens (ORSUM Group –

Max Planck Institute – Mainz). Mrs Feyerherd from the Max Planck Institute graphics

department is thanked for her enormous help and patience in the preparation of figures.