1.2 In situ hygrometers

In situ measurements of H₂O in the UT/LS have been reported for over five decades beginning with those of Brewer [1944] and Brewer et al. [1946]. Since then, a variety of experimental techniques have been developed and used. The most important of these are frost point detection, absorption and fluorescence techniques in the vacuum ultraviolet (VUV) spectral region (typically using Lyman-a radiation), and more recently, absorption measurements in the infrared using tuneable diode laser spectrometers. In the UT, techniques with lower sensitivity developed for meteorological soundings can also be applied and are briefly described below, but special maintenance and calibrations are required to make such measurements reliable up to tropopause altitudes. Due to the range of H₂O mixing ratios and the required calibration procedures, the development and operation of precise and accurate hygrometers for the UT/LS as well as their accurate calibration is still at the research level and is far from being routine.

One of the most crucial problems in making in situ measurements of H₂O is to ensure contamination-free probing. Proper selection of wall materials, optimisation of probe location, heating of the instrument and its inlet, large flow rates or open-cell designs have to be used in order to avoid or minimise measurement artefacts. The different approaches are discussed below for the individual hygrometers. The instruments presented here have proven that such contamination-free measurements are possible, while a number of other instruments described in the literature, though based on the same basic techniques, reported obvious contamination problems. The design of the inlet probe and its non-isokinetic sampling properties have been the subject of several studies, in particular when measurements in clouds were made.

The advantage of most in situ techniques is their higher precision and spatial resolution compared to remote sensing instruments; thus in situ instruments are well suited particularly for case studies on smaller scales. Their absolute calibration either in the laboratory or in flight, tracing to calibration standards can be repeated at regular intervals and possible instrument drifts are easier to detect. There is no need for additional algorithms for the measurement geometry. In situ instruments are therefore often used for the validation of space-borne remote sensing experiments. Since the operation of in situ instruments in the UT/LS requires the use of platforms such as balloons or aircraft, which are either expensive to operate, have limited availability, or both, the data sets cover only limited regions and time periods. This disadvantage is partly compensated by the large number of measurements made to date.

1.2.1 Frost point hygrometers

Most conventional frost point hygrometers are based on the chilled-mirror technique. In this technique, a layer of condensate (frost or dew) is formed on the surface of a small mirror. This layer of condensate is detected by an optical system and maintained at a constant extent by heating and cooling the mirror. Under these conditions, the amount of water vapour can be determined by measuring the temperature of the mirror, which represents the frost point or dew point temperature of the air passing over the mirror. The mirror temperature is controlled either by heating a resistor, which works against a cryogenic bath, or by a thermoelectric heat pump (Peltier device). The temperature of the mirror is measured using a thermistor or a platinum temperature sensor. The water vapour mixing ratio and relative humidity are derived from the frost-point temperature using ambient pressure and temperature.

On balloons, the measurements are made predominantly during descent to prevent contamination by the balloon wake. The altitude resolution depends on the time response and on the balloon descent rate. On aircraft, the spatial resolution is related to the speed of the aircraft.

The NOAA-CMDL and NRL frost point hygrometers

In the cryogenic frost point hygrometers used at the Climate Monitoring and Diagnostics Laboratory of the National Oceanic and Atmospheric Administration (NOAA/CMDL), the mirror is thermally connected to a cryogenic bath. The temperature of the mirror is controlled with a heating element, which heats the mirror against this cold sink. Solid state elements are used for frost-detection, and a simple controller is used to maintain a constant frost-layer.

The temperature of the mirror is measured using individually calibrated thermistors. The overall accuracy of the instruments is about 0.5ºC in temperature, which translates to roughly 10% in water vapour mixing ratio.

The hygrometers used at NOAA/CMDL were originally developed at the Naval Research Laboratory (NRL) in Washington D.C., U.S.A., [Mastenbrook and Dinger, 1961]. These instruments used a mirror, across which a temperature gradient of around 3ºC was maintained to concentrate the ice layer in the centre of the mirror. The extent of the ice layer was adjusted such that the ice edge would lie over the thermistor embedded in the mirror. It was assumed that the temperature at the ice edge is equal to the frost-point temperature. However, the accuracy to which the ice edge was maintained over the thermistor depended on the skill of the individual operator setting the instrument as well as on the accuracy of the controller to maintain an ice layer of constant size and shape. This mode of operation caused some variation in the data and was changed when the instruments were taken over by NOAA/CMDL.

The current version of the balloon-borne frost-point hygrometer, which has been used with only minor modifications since 1980, uses a mirror with a uniform temperature distribution allowing a more precise measurement of the frost-point temperature [Oltmans, 1985]. The instrument samples air in a 50 cm long and 2.5 cm diameter stainless steel tube. The flow through this tube is ram flow of about 5 m/s depending on the ascent/descent velocity of the balloon. Tests have shown that solar scatter within this tube does not affect frost detection. The largest uncertainties in the frost-point temperature measurement are the stability of the controller (<0.3ºC), uniformity of the mirror temperature (<0.1ºC), thermistor calibration (0.05ºC), and self-heating of the thermistor (<0.05ºC). The overall accuracy of this instrument is about 0.5ºC in frost point temperature, or about 10% in mixing ratio. The vertical resolution is mostly determined by the response time of the instrument, which is typically between 10 and 30 seconds for most of the region of interest.

The co-condensation of water and other trace gases (for example HNO₃) is not considered to be important, since water vapour is several orders of magnitude more abundant. Furthermore, at roughly 500 hPa the mirror is heated to about +80°C to remove any contaminants collected in the boundary layer. The time of measurement after this is not considered to be long enough to build up large amounts of contaminant tracers. Laboratory tests have been run using large amounts of CO₂ and CH₄ and have not shown any influences on the frost-point measurements.

A modified version of this instrument has been used in extensive laboratory intercomparisons. This version uses the same thermal, optical and electronic components, but has a special flow system, which allows its use in the laboratory. Intercomparisons with the Lyman-a instruments of the NOAA Aeronomy Laboratory (NOAA/AL) and Harvard University under controlled laboratory conditions are discussed in Chapter 2. Its measurement accuracy and response time is similar to the balloon instrument.

Roughly 200 instruments have been flown from Boulder, Colorado, U.S.A., and about 120 at other locations. On a few occasions multiple sondes have been flown at the same time and have always shown a good consistency amongst each other. The Mastenbrook data set until 1980 contains about 130 soundings from Washington D.C. and about 35 soundings from other locations. Table 1.1 summarises the NOAA-CMDL data sets obtained since 1964. All of the profiles before 1990 were obtained with an analog data recording system and data were extracted from a recorder strip chart at 30 s intervals. The instrument used the VIZ radiosonde for pressure and temperature information and data transmission to the ground station. Beginning in 1990, a digital interface (Tmax) to a Vaisala radiosonde was used, which provides 8 s resolution. All data are mapped onto a 250 m vertical resolution. Data from Washington D.C. are summarised in Mastenbrook and Oltmans [1983]. Those from 1964-1976 were obtained with an earlier version of the frost point hygrometer. Beginning in 1977 through the end of the program at Washington D.C. in 1980 the data were obtained using a redesigned instrument using solid state circuitry. These soundings were instrument development flights and may have incomplete data and, in some cases, somewhat suspect values. The data from 1964-76 were extracted from published data reports and not subjected to additional screening. Data for 1977-80 were also taken from previously processed data.

Data with the early version of the hygrometer were obtained since 1961 by NRL from balloon flights at different locations (Table 1.2). The hygrometers were also used onboard a NASA C141 aircraft for measurements up to 12.5 km altitude in 1974-1975 at mid-latitudes over the western United States and the eastern Pacific [Mastenbrook, 1976]. These historic data are available only from the published data reports.

The LMD-CNRS frost point hygrometers

Three kinds of instruments have been developed at the Laboratoire de Météorologie Dynamique (LMD), Centre National De Recherche Scientifique (CNRS), in France. In the first version of the hygrometer for balloons, the cooling and heating of the mirror was controlled by a thermoelectric module. Then a cryogen was used to cool the mirror for an instrument used on both aircraft and balloons. For all types of instruments, the temperature sensor of the mirror, which measures the frost-point temperature, is calibrated with respect to a platinum sensor which is traceable to the National Temperature Reference Standards. In addition, calibration and tests of the instruments are made with the LMD water vapour calibration system which allows the generation of frost-point temperatures down to ?90°C in a large air pressure range, from 1000 hPa to 20 hPa [Ovarlez, 1985]. Thus the hygrometers are tested in the same environmental conditions as those encountered during the balloon flights as well as during the aircraft flights. In order to verify the accuracy and proper operation of the hygrometers, comparisons between the frost point produced by the calibration device and the corresponding frost-point measured by the hygrometer (using the calibration of the mirror sensor) are performed and checked to fit within the precision of the hygrometer.

The thermoelectric hygrometer

This instrument was first developed for long-duration balloons (MIR = InfraRed Montgolfiere). In this version, the mirror temperature is controlled by a Peltier device, so that the hygrometer has the necessary autonomy for measurements for several days [Ovarlez, 1991, Ovarlez et al., 1996]. The uncertainty (2s ) on the frost-point determination is 0.5°C, the resolution is 0.1°C, and the time response is about 100 seconds in the middle stratosphere. The frost-point temperature range is -50 to -95°C. This instrument has also been used on open stratospheric balloons, for single profiles, before 1994 [Ovarlez and Ovarlez, 1994]. Since then, the performance of this instrument has been improved to be almost equivalent, in the stratosphere, to the cryogenic device described below.

Table 1.1 Listing of the NOAA/CMDL data set

Location	Station code	Lat.	Long.	Dates	No. of profiles	Remarks
Mid-latitudes
Boulder, Colorado	bl	40N	105W	1980-1998	194
Crows Landing, California	cl	37N	121W	7-11 May 1993	3	w/ NOAA & Harvard Ly-a on ER-2
Daggett, California	da	35N	121W	02 Feb 1992	2
Edwards, AFB, California	ew	32N	123W	25,26 Feb 1991	4	w/ NOAA Ly-a on ER-2, SAGE II
Ft. Sumner, New Mexico	fs	34N	104W	21 Sep 1996 18 May 1998	4	OMS
Laramie, Wyoming	lw	45N	105W	12 Feb 1983 30 Nov 1984, 29 Aug 1989	5	w/ SAGE II
Lauder, New Zealand	nz	45S	169E	14 Jan 1992 25,26 Nov 1992	2
Palestine, Texas	pl	32N	96W	07 May 1981 16,17 Apr 1983 11,13 Oct 1983 05 Jul 1985	8	BIC
Washington, D.C.	dw	39N	78W	1964-1980	129	NRL instrument
Tropics
American Samoa	sm	14S	171W	17/18 Jul 1986 19/20/21 Jan 1988	5	w/ SAGE II
CEPEX	cx	9S-2N	160E-157W	7-25 Mar 1993	13
Hilo, Hawai	hi	20N	155W	17 Jul 1991 22, 24 Mar 1992	3
Juazeiro do Norte, Brazil	jn	7S	64W	14 Feb 1997 11 Nov 1997	2	OMS
San Cristobal	sc	1S	90W	Mar, Sep 1998 Mar 1999	10	SOWER
Antarctica
McMurdo	mm	79S	167E	Sep, Oct 1990-92, Feb-Oct 1994	30	ASHOE/ MAESA
South Pole	as	90S		May 1990-Jan 1994	22
Arctic
Alert, NWT	al	82N	62W	Jan 1989,1990,1991	5
Fairbanks, Alaska	fb	65N	148W	7,10,11 Aug 1985	3	w/ SAGE II
Keflavik, Iceland	ic	64N	21W	24, 25 Mar 1994	2
Kiruna, Sweden	kr	68N	21E	Feb 1991,1992,1997 Jan-Mar 2000	6 5	SOLVE
Sodankyla, Finland	so	67N	27E	23 Jan, 17 Feb 1996	2

Table 1.2 Listing of the NRL balloon data set

Time	Location	Lat.	Long.	No. of profiles	Reference
April 1961	Hyderabad, India	17N	78E	3	Mastenbrook, 1962
Nov 1963	Kwajalein Atoll	9N	167E	5	Mastenbrook, 1965
Jan 1964-Dec 1965	Trinidad, West Indies	10N	61W	23	Mastenbrook, 1966
July-Aug 1965	Thule, Greenland	77N	69W	3	Mastenbrook, 1966

The cryogenic hygrometers for open stratospheric balloons (BSO)

In this instrument the mirror temperature is controlled cryogenically: the mirror is cooled through a copper rod immersed in liquid nitrogen enclosed in a Dewar container, and the heating is produced by a resistance wire wrapped around the copper rod [Ovarlez and Ovarlez, 1995]. The uncertainty on the frost point temperature determination is 0.3°C, and the resolution is 0.05°C. The response time depends on the frost point: the lower the frost point, the higher the response time, from about 20 seconds in the upper stratosphere to about 10 seconds in the upper troposphere. The frost point temperature range is -10 to -95°C. The range is quite large for stratospheric measurements, the extension to -10°C has been made for use on aircraft.

Table 1.3 LMD frost point hygrometer balloon flights

Project	Carrier	Date	Altitude range (km)
Mid-latitudes (44N, 0W)
French national program	BSO	12 Oct 1989	16 - 22.5
French national program	BSO	06 Oct 1990	14 - 25.5
TRAVERSE	BSO	07 Mar 1993	14 - 24
THESEO	BSO	05 May 1999	10 - 29.5
Arctic (68N, 21E)
CHEOPS	BSO	19 Jan 1990	14.5 - 24
	BSO	26 Jan 1990	11.5 - 22
	BSO	20 Nov 1991	11 - 24
	BSO	04 Dec 1991	12 - 25
	BSO	11 Dec 1991	13.5 - 24.5
EASOE	BSO	31 Jan 1992	13 - 23.5
	BSO	09 Feb 1992	13 - 28
	BSO	10 Mar 1992	13 - 23
	BSO	20 Mar 1992	11 - 27
SESAME	BSO	13 Feb 1995	11 - 26.5
	BSO	20 Mar 1995	11 - 27.5
ILAS validation campaign	BSO	14 Feb 1997	21.5 - 27.5
SABINE	BSO	10 Aug 1998	11 - 30
THESEO	BSO	18 Jan 1999	10.5 - 23.5
	BSO	30 Jan 1999	10 - 23
THESEO2000	BSO	19 Jan 2000	10 - 27
	BSO	25 Jan 2000
Tropics (0 to 15S)
AMETHYSTE	MIR	Jan - Feb 1994	17/22 - 26

The cryogenic hygrometer for aircraft

This instrument is the same as the open stratospheric balloon version; however, it has been adapted to aircraft installation and certification requirements [Ovarlez and van Velthoven, 1997]. During aircraft flights on the DLR Falcon (maximum altitude: 13 km), the ambient air flows into the instrument through an air inlet from the aircraft fuselage (a modified Rosemount temperature housing). Thus gas-phase measurements are accomplished without sampling ice particles. The airflow is generated by the difference between the dynamic pressure at the air inlet and the static pressure at the outlet. To prevent contamination from dense low-level clouds, the air inlet is opened after the first kilometres of the aircraft ascent. The air pressure is measured immediately at the sensor head of the hygrometer, so that the water vapour volume mixing ratio is given by the ratio between the water vapour saturation pressure at the measured frost-point and the measured air pressure.

Table 1.4 LMD aircraft missions

Project	Period	No. of flights	Latitudes	Longitudes	Comparisons
POLINAT*	Nov 1994	8	30-60N	0-30W
POLINAT*	May-Jun 1995	8	30-60N	0-30W
POLINAT2**	Sep-Oct 1997	8	30-60N	0-30W	Vay et al., 2000 Helten et al., 1999

* Ovarlez and van Velthoven [1997] and Ovarlez et al. [1999]; ** Ovarlez et al. [2000].

Other airborne frost point hygrometers

The Meteorological Research Flight (Farnborough, UK) has carried out humidity measurements since 1943 using frost point hygrometers. The first measurements were obtained from a B17 Flying Fortress up to 37700 ft [Brewer, 1944]. From 1944 to 1952, the hygrometers were employed onboard several Mosquito aircraft for measurements up to 44000 ft or 160 hPa [Bannon et al., 1952]. The instrument characteristics of this first hygrometer used in the UT/LS are described by Brewer et al. [1946]. Since 1954, a Canberra aircraft was used for regular measurements from the upper troposphere up to approximately 150 hPa. Data obtained from this aircraft are reported in several reports and publications, such as Helliwell [1960], Roach [1962], Cluley and Oliver [1978] and Foot [1984]. Measurements in the 1970s were made on a regular basis every 3-4 months between 45°N and 65°N. Cluley and Oliver [1978] estimate that the standard deviation of the frost point arising from random effects is 0.5°C; in the range 2 to 10 ppm this is equivalent to a standard deviation of about 7% in mass mixing ratio.

The Spyers-Duran [1991] NCAR instrument, a cryogenic frost point hygrometer, has been used on aircraft.

The hygrometers made by Buck Research, e.g. used on the German Falcon research aircraft by DLR [Busen and Buck, 1995] and on the NASA DC-8 aircraft, are cryogenic frost-point hygrometers, working in the range +30 to -95°C, with an uncertainty of 0.3°C, and a response time of 6 to 30 seconds depending on the environmental parameters.

Other commercial instruments are used on research aircraft, in the mid troposphere, such as the General Eastern instrument. They use a Peltier device to cool the mirror.

Chilled mirror hygrometers used for calibration standards

Advanced commercial chilled mirror hygrometers are used as calibration standards for some of the hygrometers discussed in this Assessment as well as by National Standard Institutes such as in the U.S., in Germany and in Spain.

The General Eastern Instruments thermoelectric sensor series 1311 DR (DRX) is designed for frost point measurements down to -70°C (-75°C). The accuracy of the instrument is quoted to ± 0.2°C (± 0.15°C). Lower frost points, i.e. -90°C (-95°C), can be determined with a hygrometer of MBW Elektronik AG (Switzerland), the K-1806/DP30-SHS (SHSX) system. This hygrometer system is accurate to ± 0.1°C. More technical details and other commercial chilled mirror hygrometers are reviewed by Wiederhold [1997].

1.2.2 Lyman-a hygrometers

The method used to measure H₂O by a fluorescence technique was developed by Kley and Stone [1978] and Bertaux and Delannoy [1978]. The photodissociation of H₂O molecules by radiation at wavelengths l < 137 nm produces electronically excited OH:

The electronically excited OH relaxes to the ground state by fluorescence or by collisions with other molecules M:

By measuring the intensity of the emitted fluorescence, the H₂O abundance can be determined. The number of fluorescence photons N_f is given by

[H₂O] and [air] denote respectively the concentration of H₂O and air molecules, J the photodissociation rate of reaction (1.1), F the quantum efficiency for excited state OH production via (1.1), A₀ the Einstein coefficient of reaction 1.2a, and kq the quenching coefficient of the OH radical in air (reaction 1.2b). In the UT/LS, that is, for altitudes below 20 km, [air]. kq/A₀ >> 1, and equation 1.3 can be approximated by

The factor C includes molecular coefficients from the literature as well as instrument specific quantities. If C is a constant, the number of detected fluorescence photons is proportional to the H₂O mixing ratio [H₂O]/[air] for measurements in the UT/LS. For measurements at higher altitudes, equation 1.3 has to be used to obtain correct water vapour mixing ratios.

In reality, C is a function of J and thus depends on the photon flux in the fluorescence volume, which in turn depends on variations of the lamp intensity and absorption by atmospheric gases. In the vacuum UV (VUV) spectral region, absorption by oxygen and water vapour has to be taken into account. The attenuation and variation of the VUV radiation have to be monitored when determining H₂O mixing ratios by measuring the fluorescence signal. The Lyman-a line at l = 121.6 nm coincides with a narrow deep minimum in the oxygen absorption cross section and thus enables measurements with the fluorescence technique down to the middle troposphere. Measurement of the absorption of Lyman-a radiation at higher concentrations can be used for quantification of H₂O abundances in the troposphere as well [e.g. Tillman, 1965]. Such instruments [e.g. Buck, 1976] are used on several research aircraft as part of the basic instrumentation.

The most advanced Lyman-a fluorescence hygrometers have been developed in the laboratories at NOAA in Boulder, at Harvard and in Jülich (Germany). Though they are based on the same technique, they differ in several experimental details, and also in their calibration procedure. The instruments are employed on different aircraft and balloons from the UT up to 35 km altitude. Rocket borne measurements in the mesosphere using this technique have been reported by Khaplanov et al. [1996].

One of the advantages of this technique is the large dynamic range for measurements from the UT at several hundred ppmv to dry stratospheric air masses where changes of the order of 0.1 ppmv can still be detected. Large flow rates through the hygrometers can be achieved for contamination-free measurement together with integration times on the order of 1 s for detection of small-scale features in the atmosphere.

The NOAA Lyman-a hygrometers

The NOAA balloon-borne Lyman-a fluorescence hygrometer was developed by Kley and Stone [1978]. The actual flight instrument for use in the stratosphere has been described and characterised by Kley et al. [1979]. This was an open-cell design with a radio-frequency discharge Lyman-a light source. The stray light intensity across the cell and the intensity in the fluorescence region were monitored by nitric oxide ionisation cells. The instrument was flown at night to eliminate solar scatter.

A characteristic of this and successive NOAA Lyman-a fluorescence hygrometers is their method of in situ calibration through the simultaneous measurement of water vapour concentration by absorption and OH fluorescence. The in situ calibration uses Beer’s law at the wavelength of Lyman-a radiation and has two constants, namely the water vapour absorption cross section at Lyman-a [Kley, 1984] and the distance between light source and nitric oxide intensity monitor. The balloon instrument makes use of the fact that, during ascent or descent, the absorption by water vapour across the cell changes from >90% in the middle troposphere to <10% in the upper troposphere, allowing the determination of absolute water vapour concentrations and, from the simultaneously-recorded OH fluorescence intensity, the determination of the fluorescence sensitivity constant C over a considerable range of water vapour absorption. Absorption by oxygen at Lyman-a varies between 35 and 20% between 500 and 250 hPa and is numerically considered using the measured cross section of O₂ [Kley, 1984]. Table 1 of Kley et al. [1979] gives an example for the determination of C for a typical balloon flight. The precision of C obtained from 48 in situ calibrations, over the altitude range of 6-10 km with water vapour absorption ranging from 90% to 5%, was 18%, and constituted the single largest error in the overall accuracy of the balloon-borne instrument. The 2s uncertainty of the water vapour absorption cross section at Lyman-a is 6%. Other error sources are negligible.

The first aircraft instrument was a closed cell design with a microwave discharge lamp. This instrument was operated in a pressurised container for thermal stability. Lyman-a intensity was monitored across the cell with a NO ionisation cell. This instrument was built for the Panama STEP U-2 campaign. The instrument design and calibration procedure are documented by Kley et al. [1983]. In short, the in situ calibration of the Panama instrument was achieved during level flight in the tropopause region whenever short duration events of large water vapour mixing ratios occurred, superimposed on much smaller background mixing ratios. Thus, the calibration is equivalent to the one described above for the balloon instrument except that the total pressure and, therefore, the oxygen absorption does not change during an absorption event. A disadvantage of the Panama instrument was the irregular and relatively infrequent occurrence of natural absorption events.

The U2 aircraft instrument was modified in 1981 for regular (every 15 minutes) calibration procedures. A small (200 cm³) cylinder is filled with 100-200 atm of air. The high pressure air is allowed to flow over a frit with a saturated cotton swab, then over a capillary into a second small cylinder, filling it with humidified air over a period of 15 minutes to a pressure of approximately 1000 hPa. A magnetically operated valve then opens and allows the second cylinder to empty its contents through a second capillary into the main air stream, upstream of the fluorescence chamber. The small pressure increase in the fluorescence chamber is negligible, but the mixing ratio of water is increased, initially to mixing ratios of approximately 500 ppmv. Set by the conductance of the second capillary, the mixing ratio decays exponentially to ambient levels with a time constant of approximately 1 min. From the measurement of transmitted Lyman-a light intensity it is possible to derive absolute water vapour mixing ratios (above ambient background). The calibration was performed every 15 minutes over a wide range of mixing ratios. Water vapour mixing ratios of approximately 500 ppmv at the high end, down to values around 20 ppmv at the low end were used to derive the calibration constant. Accounting for preabsorption of Lyman-a light before entering the field of view of the OH fluorescence detector, the response of the fluorescence signal was extremely linear. The slope of fluorescence intensity versus absolute water vapour mixing ratio yields the sensitivity constant. The precision of the in-flight calibrations (every 15 minutes during a 5 hour flight) was typically 1-2%.

Error sources for the U2 instrument include a negligible error from photon counting statistics from fluorescence and background signal and a contribution from water vapour outgassing. However, the latter contribution would decay to very small amounts after approximately 1.5 hours into the flight and played no role during the final descent profiles. Since the Lyman-a absorption cross section is accurately known with a 2s standard deviation of 6% [Kley, 1984] and the only other parameter (distance between the Lyman-a light source and the nitric oxide ionisation chamber) is measured, it follows that the calibration of the fluorescence instrument must be accurate to the uncertainty of the Lyman-a water vapour absorption cross section.

Since the calibration is always done at ambient conditions of pressure, flow rate, oxygen partial pressure and trace gas composition, errors in the calibration constant that might arise from systematic errors possibly connected to the above parameters are avoided.

The present NOAA Lyman-a instrument is also a closed cell design, optimised for a high flow rate to minimise the effect of trapped water. Heated inlet lines vaporise the ice before reaching the hygrometer. The Lyman-a source is a DC discharge lamp. The Lyman-a intensity is monitored with an iodine ionisation cell that is sensitive from 115 nm to 135 nm. A magnesium fluoride beam splitter samples the source with the ionisation cell placed equidistant from the source and the detected fluorescence region. This arrangement accounts for the preabsorption in the cell from oxygen and water. Water vapour is injected into the air stream and the absorption and fluorescence are measured as above to provide an in-flight calibration. A calcium fluoride filter is inserted during the calibration sequence to measure background radiation from the lamp at 309 nm (about 200 counts per second) and any radiation between 125 and 135 nm. Two hygrometers were built in an attempt to independently measure total water and water vapour. They were flown in the partially pressurised Q bay of the ER-2 [Kelly et al., 1989]. The instrument has since been modified to fly unpressurised on the WB-57.

The main sources of error of the present aircraft instrument are the accuracy of the water Lyman-a absorption cross section and the Poisson counting statistics of the signal and background. The 2s uncertainty of the water vapour absorption cross section uncertainty is 6%. The counting period for both instruments is 1 s. Count rates are in the range of 500-1000 counts per second per ppmv of water. For 4 ppmv the 1s counting error is about 4%. The total 2s error is 10% at 4 ppmv for 1 second data and 6.6% at 4 ppmv and 10 second data.

Areas of concern in the present NOAA instrument are spectral impurities in the DC lamp and trapped water. Other wavelengths are not as effective as Lyman-a in producing fluorescence. The spectral purity of the lamp is checked during the descents by using the oxygen and the water Lyman-a absorption cross-sections and the measured pressure and water to remove their absorption effects. If the lamp intensity is constant then the result should be independent of pressure unless there are spectral lines with different cross sections. The calcium fluoride cut-off filter gives a second measure of any Lyman-b and contamination, usually less than 0.1%.

Table 1.5 Missions of the NOAA-AL Lyman-a hygrometer

Mission	Location	Lat.	Long.	Dates	No. of flights	Carrier
-	Brazil	5S	39W	27 Sep 78	1	balloon
-	Wyoming	41N	105W	19 Sep 79 - 12 Feb 83	5	balloon
STEP	Panama	6-11N	77-82W	30 Aug 80 - 16 Sep 80	8	U-2
-	Texas	32N	96W	07 May 81, 10 Oct 83	2	balloon
ACE	California	38N	122W	02 Dec 80 - 21 Jan 83	23	U-2
-	Kansas	38N	97W	29 Apr 81 - 09 May 81	4	U-2
-	Texas	32N	96W	05 May 81 - 12 Oct 83	7	U-2
-	Wyoming	41N	105W	14 Jul 81	1	U-2
STEP	California			20 Apr 84 - 06 May 84	4	U-2
STEP	California			04 Apr 86 - 24 Apr 86		ER-2
STEP	Darwin + transits	12S	131E	08 Jan 87 - 19 Feb 87	11 6	ER-2
AAOE	Chile + transits Chile + transits	53S 44-90S	71W 71W	12 Aug 87 - 03 Oct 87 22 Aug 87 - 29 Sep 87	12 6 13 1	ER-2 DC-8
AASE	California Norway + transits Norway + transits	38N 59N 59-90N	122W 6E 6E	06 Oct 88 - 05 Dec 88 29 Dec 88 - 21 Feb 89 02 Jan 89 - 15 Feb 89	5 13 4 14 2	ER-2   DC-8
-				25 Feb 91, 26 Feb 91	2	ER-2
PEM west				01 Sep 91 - 21 Oct 91		DC-8
AASE II	Alaska, Maine, + transits	65N 45N	147W 69W	04 Oct 91 - 14 Oct 91 02 Nov 91 - 26 Mar 92	4 15 4	ER-2
SPADE				09 Nov 92 - 22 Oct 93	14	ER-2
ASHOE MAESA	California Hawaii New Zealand + transits	38N 20N	122W 160W	02 Feb 94 - 04 Nov 94	6 5 24 6	ER-2
WAM	Texas	25N	95W	09 Apr 98-07 May 98	6	WB-57
				99		WB-57

The Harvard Lyman-a hygrometers

The Harvard balloon-borne Lyman-a has been described by Weinstock et al. [1990]. Briefly, Lyman-a radiation from a RF discharge lamp photodissociates water vapour in a 6 inch duct. The OH fluorescence is collected at right angles to the lamp and airflow through an interference filter centred at 315 nm with a 10 nm bandwidth and detected with a photomultiplier tube. A photodiode sensitive to vacuum ultraviolet radiation was used to both monitor lamp intensity and, with the aid of an actuator to change absorption path length, to measure water vapour by absorption. The walls are maintained at about ?100°C to prevent outgassing while data is taken during a valve-down descent at velocities of 3-6 m/sec. A large fan is used to aid in the airflow through the duct.

This instrument performed successfully in 4 flights launched from Palestine, Texas, during the summers of 1987-1989, as described in Schwab et al. [1990]. The data were taken during balloon valve-down descent as tabulated below. The quoted accuracy of these 2-s data is 40% in 1987 for a 2-minute average and 30% in 1988 and 1989 for a 30-second average.

Table 1.6 The Harvard balloon data set

Date	Altitude range (km)	Latitude	Longitude
15 Jul 87	15-32	32 N	98 W
06 Jul 88	16-34	32 N	98 W
28 Jul 89	17-37	32 N	98 W
25 Aug 89	18-37	32 N	98 W

The Harvard Lyman-a instrument for the NASA ER-2 aircraft was described in detail by Weinstock et al. [1994], with an update in Hintsa et al. [1999]. Briefly, Lyman-a (121.6 nm) radiation from a RF discharge lamp photodissociates water vapour in a 3 inch square duct. The OH fluorescence is collected at right angles to the Lyman-a beam through a filter and detected with a photomultiplier tube. At ER-2 altitudes (5-21 km) the observed detector signal is proportional to the water vapour volume mixing ratio. Solar and lamp scatter near 315 nm is measured by using a quartz window to periodically block the Lyman-a beam. The fluorescence signal is normalised to lamp intensity measured by a vacuum photodiode opposite the lamp. A rear-surface MgF₂ mirror surrounding the diode reflects some of the radiation back across the duct to a second diode, allowing water measurements by direct (Beer’s Law) absorption in the mid-to-upper troposphere. Flow velocities in the duct are typically 30-70 m/sec for fast time response and to avoid (and directly test for) contamination from walls. Data are typically reported every 4 seconds, corresponding to 1 km horizontal resolution on the ER-2, but 8 Hz data are available for high resolution work (aircraft wake crossings, etc.). The precision is typically ±0.1 ppmv for a 4 second measurement in the stratosphere.

The instrument is calibrated in the laboratory by flowing known amounts of water vapour in air through the detection axis. Saturated air-water mixtures are prepared by passing air from 50 to 500 ccm at standard temperature and pressure through a two-stage bubbler apparatus, then diluted with dry air at approximately 5000 to 40000 ccm at standard temperature and pressure before being sent to the detection axis. Brooks flow controllers, individually calibrated to an accuracy of 1%, are used to control the respective flows. The bubbler apparatus is held in a well-insulated water bath with temperature measured to ±0.1 degree, providing a partial pressure of water accurate to better than 1% at the bubbler. The instrument response is linear up to concentrations of ~10¹⁶ H₂O/cm³ (typical of the mid-troposphere). The accuracy of the calibration system is 5% based on uncertainties in temperature, pressure, and gas flow, and has been verified by direct absorption both along and across the detection axis.

The calibration is checked in-flight by comparing Lyman-a photofragment fluorescence with direct absorption measurements of water vapour, similar to the method of Kley et al. [1979], using the atmosphere to provide a wide range of H₂O concentrations. Data at ER-2 cruise altitudes (where concentrations are ~10¹³/cm³) are used to provide a reference signal (I_o) for absorption on ascent and descent. At lower altitudes, concentrations are high enough (10¹⁴-10¹⁶/cm³) for measurements of I, and hence the water vapour concentration by direct absorption: ln(I/I_o)=s ? [H₂O]. l, where s is the absorption cross-section of H₂O at 121.6 nm and l (9.2 cm) is the path length. The first diode measures the lamp intensity at the mirror surface (to account for changes in lamp flux), and the second diode detects the intensity of the reflected beam after it has passed through the absorption path, so the ratio (diode 1 signal/diode 2 signal) is used for I and I_o. Absorption by O₂ is also accounted for. During nearly 100 flights from 1992 to 1997, the abundance of water vapour determined by absorption typically agreed to within ±10% of fluorescence measurements.

Level flight tracks of the ER-2 in the mid-to-upper troposphere provide near-ideal conditions to compare water vapour measured by fluorescence and absorption. Water vapour can vary by 10-100 ppmv at one altitude with other atmospheric conditions (temperature, pressure, etc.) nearly constant. This approach has very high stability, and in addition, O₂ number density changes very little on level flight tracks, so even large errors in the O₂ cross-section have only a minor effect on these measurements. Since 1995, these in-flight comparisons have been made on seven flights, with agreement between absorption and fluorescence averaging 1%. Figure 1.3 shows an example of the agreement between delta water vapour measured by absorption and fluorescence taken from a level segment of the July 25, 1996 flight. The slope is very close to unity, better than can be expected from estimated measurement uncertainties. The slight offset is related to the choice of initial and final segments and can be positive or negative on different flights. Forcing the best fit to go through the origin has at most a ~2% effect.

Based on laboratory calibrations and in-flight calibration checks, the instrument is accurate to ±5%, with an additional systematic uncertainty of 0.1 ppmv. A number of possible sources of error in the measurements have been examined. Contamination (desorption of water) from walls is checked for by periodically varying the duct velocity during each flight, and is found to be absent. To investigate the temperature dependence of quenching, the Lyman-a hygrometer was calibrated from room temperature to -30°C, and no change in sensitivity was found. Radiation from the Lyman-a lamp other than at 121.6 nm is another potential source of error. Lamp spectra from 115 to 150 nm usually show less than 1% intensity at wavelengths other than Lyman-a , and these are blocked by a flowing cell containing O₂ and N₂ in front of the lamp. The lamps do produce some radiation near 315 nm, but the use of an anti-reflection coated quartz window to block the Lyman-a beam (for measurements of background) reduces the required correction to the data to <2%.

Figure 1.3 Regression of delta water vapour by absorption and fluorescence on a level flight track on July 25, 1996. An average water vapour mixing ratio of 124 ppmv, corresponding to the start and end of the flight track, has been subtracted from the fluorescence measurement to give the changes over the flight segment. Because there is no way to get a true I₀ for absorption, only changes in water vapour can be compared. The slope is very close to unity, indicating excellent agreement between the two methods for this flight.

Table 1.7 The Harvard ER-2 data set

Campaign	Location	Latitude	Longitude	Altitude (km)	No. of flights	Dates
SPADE	Ames	21N - 41N	117-125W	8-20	3	09 - 20 Nov 92
	Ames	15N - 60N	117-125W	8-20	11	23 Apr - 18 May 93
CEPEX	Fij	17S - 2N	176W-178E	8-20	3	18 Mar - 06 Apr 93
STRAT	Ames	15N - 41N	117-123W	8-20	4	08 - 17 May 95
	Ames	20N - 60N	117-123W	8-20	6	20 Oct - 02 Nov 95
	Hawaii	2N - 35N	120-160W	8-20	3	05 - 09 Nov 95
	Ames	15N - 41N	117-123W	8-20	5	26 Jan - 02 Feb 96
	Hawaii	2N - 35N	120-160W	8-20	4	05 - 15 Feb 96
	Ames	15N - 41N	117-123W	8-20	4	18 - 30 Jul 96
	Hawaii	2N - 35N	120-160W	8-20	7	01 - 10 Aug 96
	Ames	15N - 41N	117-123W	8-20	4	13 - 21 Sep 96
	Hawaii	2N - 35N	120-160W	8-20	4	06 - 11 Dec 96
	Ames	15N - 41N	117-123W	8-20	5	04 - 13 Dec 96
POLARIS	Ames	15N - 60N	117-147W	8-20	4	16 - 24 Apr 97
	Alaska	64N - 90N	110-148W	8-20	7	26 Apr - 13 May 97
	Alaska	64N - 90N	110-148W	8-20	6	26 Jun - 10 Jul 97
	Alaska	20N - 90N	110-160W	8-20	8	08 - 21 Sep 97
	Hawaii	2N - 41N	117-160W	8-20	2	23 - 25 Sep 97
SOLVE	Sweden	68N	21E	8-20	6	20 Jan - 03 Feb 00
	Sweden	68N	21E	8-20	5	03- 15 Mar 00

The Jülich Lyman-a hygrometers (FISH)

The Fast In situ Stratospheric Hygrometer (FISH) developed at the Forschungszentrum Jülich (Germany) is described in detail by Zöger et al. [1999a]. Today, three different hygrometers exist, one for employment on large stratospheric balloons, and two for use on different aircraft (German Falcon of DLR, Dutch Cessna Citation of TU Delft, German Lear Jet of GFD, Russian M55 Geophysika of MDB).

FISH consists of a closed, vacuum-tight fluorescence cell, a Lyman-a radiation source, a PMT in photon-counting mode, detectors to monitor the VUV radiation output of the Lyman-a lamp, and a mirror drive that controls the measuring cycle: determination of the fluorescence and background count rate (N₁ and N₃) and of the lamp intensity U_B. Thus the relative count rate

with an instrumental parameter f_u, is used in equation 1.3, and C and f_u are determined by calibration in the laboratory.

The FISH measurement frequency is 1 Hz. At this frequency the noise equivalent mixing ratio at 3 ppmv is 0.2-0.15 ppmv, and the detection limit is 0.18-0.13 ppmv, depending on the quality of the lamp. Applying an autocovariance analysis on field data obtained with FISH confirms this precision.

On aircraft, FISH is usually connected to a forward-directed heated inlet that allows for measurement of total water. The enhanced sampling efficiency of larger particles has been quantified in Schiller et al. [1999]. The ram pressure maintains flow rates that allow an exchange of the air in the whole system within approximately 1 s. For balloon measurements, also using a heated inlet line, measurements are carried out only during descent, with a pump used to enhance the flow.

FISH is calibrated between flights in the laboratory using a calibration bench [Zöger et al., 1999a] under realistic conditions, that is varying the H₂O mixing ratio of the test air from a few ppmv to several hundred ppmv and the pressure from 1000 to 10 hPa. A frost point hygrometer (General Eastern 1311 DRX, recently also MBW K-1806/DP30-SHSX) is used as a reference instrument. The quoted accuracy of the frost point measurement is ± 0.15 K (± 0.1 K for the MBW instrument), corresponding to an uncertainty of 4% (3%) in the H₂O partial pressure at 210 K. Including an error of the pressure measurement in the calibration bench, the H₂O mixing ratio can be determined with an accuracy better than 5%. As a result of several field campaigns, the factors C and f_u determined in this calibration do not change significantly from flight to flight except when a detector has to be exchanged [Eicke, 1999].

FISH can also be calibrated during in-flight operation by measuring the absorption of Lyman-a radiation whenever the optical depth of H₂O is sufficient [Zöger et al., 1999a]. This method is occasionally applied as a cross check of the aforementioned laboratory calibration. Figure 1.4 shows an example using cross sections determined by Kley et al. [1984]. Both methods yield a discrepancy of the H₂O mixing ratios of 15%, which however is not significant within the combined errors (including that of O₂ absorption to be considered in the UT). As discussed in Zöger et al. [1999a], the discrepancy might be caused by differences of the effective cross sections used, which depend on the line shape of the individual Lyman-a radiation sources.

Figure 1.4 Relative count rate N_f* as a function of the H₂O mixing ratio for the FISH calibration for the aircraft mission on the 10th March 1997. Solid symbols and the corresponding regression line show the laboratory calibration before the flight using the frost point hygrometer as reference. Open symbols denote calibration points obtained in the same calibration cycle, but the H₂O mixing ratio is calculated from the simultaneous absorption measurements of FISH.

FISH measurements focus on the tropopause region at mid and high northern latitudes in different seasons. A couple of balloon flights, again at mid and high latitudes, extend the data set into the middle stratosphere. Recently, measurements from the Russian M55 Geophysika aircraft up to 20 km altitudes have been made at mid-latitude and in the tropics.

Table 1.8 Aircraft and balloon experiments with FISH

Project	Platform	Period	No. of flights	Location	Lat.	Long.	Altitudes
- 1	balloon	20 Sep 93	1	France	44N	0E	8-33 km
- 2	Falcon	Mar 95	2	Germany	47-54N	9-12E	5-13 km
STREAM IIa 3	Citation	May-Jun 96	8	Ireland	52-57N	14W-5E	5-13 km
POLSTAR I 3,4	Falcon	Jan-Feb 97	6 1	Sweden transit	66-82N 68-48N	10W-23E 10-21E	5-13 km
ILAS	balloon	11 Feb 97	1	Sweden	68N	21E	8-26 km
STREAM IIb 3	Citation	Mar 97	4 2	Sweden transit	63-73N 52-68N	10-22W 5-21E	5-13 km
-	Falcon	Aug 97		Germany	46-53N	2-18E	5-13 km
POLSTAR II 3	Falcon + Lear Jet	Jan-Feb 98	6 2	Sweden transit	62-82N 48-68N	10W-21E 10-21E	5-13 km
STREAM III	Citation	Jun-Jul 98	8 2 2	Canada transits Netherlands	46-56N 48-64N 52N	90-74W 80W-5E 5E	5-13 km
APE ETC	M55	Dec 98	5	Italy	44N	12E	8-20 km
THESEO	balloon	06 Feb 99 03 May 99	1 1	Sweden France	68N 44N	21E 0E	7-28 km 7-31 km
APE THESEO	M55	Feb-Mar 99	7 2	Seychelles transit	3N-10S 44N-5S	48-62E 12-55E	8-20 km 8-20 km
APE-GAIA	M55	Oct 99	3	transit	54S-37N	68-6W	8-20 km
THESEO 2000	balloon	27 Jan 2000 01 Mar 2000	1 1	Sweden Sweden	68N 68N	21E 21E	7-26 km 7-27 km

¹ Zöger et al. [1999b], ² Zöger et al. [1997], ³ Eicke [1999] and Schiller et al. [1998],

⁴ Schiller et al. [1999]

Others

A Lyman-a fluorescence hygrometer has been developed and operated by the Central Aerological Observatory (CAO) in Moscow, Russia [Yushkov et al., 1995]. The instrument is designed with an open layout for contamination-free measurements, therefore its use is restricted to nighttime observations (solar zenith angle <94°). Since 1991, a series of balloon flights up to 35 km using these hygrometers has been made at high northern latitudes [Khattatov et al., 1994, Yushkov et al., 1998].

1.2.3 Tuneable Diode Laser (TDL) spectrometers

JPL open-path tuneable diode laser spectrometers

Measurement of water vapour from aircraft platforms using external, open-path, optical absorption spectrometers provides a simple and direct in situ approach to the problem of obtaining water vapour abundances in the troposphere and lower stratosphere. With no containing walls or sample inlets, measurements can be made free of concerns related to wall reactions, flow rate restrictions, or "sticking", and with a high sampling rate limited only by the capability of the signal processing electronics and software. In addition, high-resolution detection of individual rotation-vibration lines in the near- and mid-infrared regions allows complete discrimination against liquid or solid H₂O (i.e. rain or ice particles) allowing accurate measurements of water vapour in clouds.

Two near-infrared aircraft instruments have been developed recently at the Jet Propulsion Laboratory (JPL) [May, 1998] and have been flown on the NASA ER-2 and DC-8 research aircraft since 1997. Both operate with similar electronics and software, but differ in their external optical arrangements.

The JPL laser hygrometers utilise second harmonic absorption spectroscopy [Webster et al., 1988]. While this technique alters the observed spectroscopic line shape and requires a quantitative transfer function for the signal processing electronics, gas concentration is ultimately derived from the Beer-Lambert law which relates observed line centre absorption to absorber number density,

r is the absorber number density, t (n ) is the observed transmission at wave number n , and L is the optical absorption path length. k(n ) is the absorption coefficient and requires knowledge of the line strength and molecular line shape for the transition of interest.

In the second harmonic detection technique a small-amplitude sinusoidal modulation is added to the base sawtooth laser current scan ramp. The detector signal is demodulated at twice the sine wave frequency to produce second harmonic line shapes. To quantify an observed 2f spectral line the amplitude of the sine wave modulation must also be known precisely. Techniques for determining the modulation amplitude are described by May and Webster [1993].

The ER-2 instrument contains a multipass Herriott cell [Herriott et al., 1964; Altmann et al., 1981] and was designed primarily for measurements in the stratosphere where water vapour concentrations typically fluctuate around a mean value near 5 ppmv. The optical path length is 11.13 m enabling measurements in the range from 400 to 0.1 ppmv. For the DC-8, which could not reach the stratosphere during CAMEX-3, measurements were desired from the ground to the aircraft ceiling of 42000 feet and an optical path length of 50 cm was chosen. In this case a multipass system is not required and the laser beam travels a simple "there and back" path between the laser/detector head, and a small return mirror. With this system, measurements from 30000 to 8 ppmv can be made.

The optical absorption paths are located outside the aircraft boundary layer on both aircraft. On the ER-2 this required locating the optical path >5 cm from the superpod skin. For the DC-8 the location required a much larger distance from the aircraft skin, and the optical path was positioned 15.3 cm away. The total optical path within the evacuated optical housings is <1 cm, which eliminates any residual H₂O absorption which could add uncertainty to the measurement.

The base spectrum scan rate for the ER-2 instrument is 10 Hz, meaning that measurements can, in principle, be made at this sampling rate. However, spectrum processing times dictate slower sampling rates and multiple spectra were averaged for the POLARIS and WAM missions to produce a net sampling period of 1-2 seconds. Recently, a new laser controller was designed which increased the signal detection frequency by more than an order of magnitude (from 20 kHz to 256 kHz). The base scan rate dropped to 8 Hz and the number of averaged scans adjusted accordingly to maintain an approximate 2 s sampling period. The higher frequency controller was used for the DC-8 instrument (where beam jitter was not a problem due to the double-pass arrangement) and is expected to result in significantly improved SNR for future ER-2 and WB57F flights. Figure 1.5 shows ER-2 data from a POLARIS flight (May 2, 1997) of the ER-2 where a measurement precision of +0.05 ppmv for a two second integration time is demonstrated.

Figure 1.5 ER-2 data from a POLARIS flight on 02 May 1997. The noise level in the data is +0.05 ppmv for a 2

If all spectral line parameters and the electronics response are known with high precision, the data processing matrix (with appropriate Beer's law corrections) would be sufficient to produce accurate water vapour volume mixing ratios at all pressures and temperatures. However, uncertainties in several independent parameters can cause pressure- and temperature-dependent errors that are difficult to quantify at the few percent level in a second-harmonic spectrometer. The most important of these parameters are the air-broadening coefficient and its dependence on temperature, the exact value for the laser modulation amplitude, and the detailed behaviour of the electronic signal demodulator with respect to residual amplitude (RAM) noise generated by the laser. Analytical laboratory calibrations are therefore used to derive correction curves to the data after initial processing.

The laboratory calibration procedure is to flow a standard gas mixture through the optical cell and then immediately through a NIST-traceable chilled mirror hygrometer (General Eastern model 1311DR, see Section 1.1.1). Two methods have been used to make a leak-tight chamber: either the entire H₂O instrument is installed in a thermal-vacuum chamber, or a telescoping tube is installed between the two mirrors of the optical cell. Calibration measurements are made over wide ranges in mixing ratio, temperature and pressure (50-200 ppmv, 50-500 hPa, 250-300 K) to map out differences between the actual instrument response, and the predicted response [May, 1998].

Number density is the fundamental quantity measured by the laser hygrometers, as defined by Equation 1.5. This can be converted to volume mixing ratio, dew point, etc. using measurements of pressure and temperature. For the POLARIS ER-2 flights the final accuracy in the measurements was a function of pressure and has been estimated at 5% for pressures <100 hPa, 8% for pressures 100-200 hPa, and 10% for pressures >200 hPa [May, 1998]. Using analytical correction curves derived from laboratory calibrations the accuracy for the CAMEX-3 DC-8 measurements is expected to be reduced to approximately 5% at all pressures.

A summary of the missions of the JPL laser hygrometers is given in Table 1.9 (flight count does not include engineering and test flights). For the ER-2 and WB57F, measurements were made from approximately 5 km altitude up to the aircraft ceilings in the 18-20 km range. The DC-8 measurements were made from the surface to the aircraft ceiling of 13 km.

Table 1.9 JPL laser hygrometer missions.

Mission	Platform	Dates	No. of flights	Deployment site	Latitude	Longitude
POLARIS	ER-2	Apr-Sep 1997	24	Fairbanks, Alaska	3S-90N	73-171W
WAM	WB57F	Apr 1998	3	Houston, Texas	9-46N	73-108W
CAMEX-3	DC-8	Jul-Sep 1998	20	Cocoa Beach, Florida	14-40N	63-86W
SOLVE	DC-8 ER-2	Dec 1999 - Mar 2000	24 11	Kiruna, Sweden	68N	21E

The LaRC/ARC diode laser hygrometer (DLH)

The Diode Laser Hygrometer (DLH), developed by NASA’s Langley and Ames Research Centers (LaRC and ARC), has flown on the NASA DC-8 aircraft in several field missions including those given in Table 1.10 [Vay et al., 1998; Vay et al., 2000; Cho et al., 2000]. The sensor consists of a compact laser transceiver mounted to a DC-8 window port and the sheet of retro-reflecting "road sign" material applied to the DC-8 engine enclosure that completes the optical path. The retro-reflecting sheet returns the beam to the laser transceiver thereby completing a 28.5 meter round trip in the free airstream. This sensor approach has a number of advantages including compactness, simple installation, fast response time, no wall or inlet effects, and wide dynamic measurement range (several orders of magnitude).

Using differential absorption detection techniques similar to those described in Sachse et al. [1987 and 1991], gas-phase water (H₂O(v)) is sensed along the external path. For dry conditions (generally altitudes > 6 km), the diode laser wavelength is swept across a strong, isolated line at 7139.1 cm^-1 while for altitudes typically < 6 km the laser wavelength sweep is locked onto a weaker line at 7133.9 cm^-1. By normalising the laser differential absorption signal with the laser power signal, the H₂O(v) measurement is unaffected by clouds, haze, plumes, etc. thereby enabling high spatial resolution measurements in and around clouds. The H₂O(v) mixing ratio is computed by an algorithm from the differential absorption magnitude, ambient pressure and temperature, and coefficients derived from laboratory calibration of the sensor. The calibration is conducted prior to each field mission and involves measuring the sensor response to humidified air flowing through each of two chambers (of lengths 1 and 3 meters). A matrix of optical depth and pressure values is generated in these chambers by varying: (1) moisture content in the flow (ultra dry air and humidified air typically ranging from 0° to 10° C dew point), (2) path length (1 or 3 m), and (3) chamber pressure from 100 to 1000 hPa. The moisture content in the calibration flow is monitored by an Edgetech model 2001 dew point hygrometer with accuracy of ± 0.2° C traceable to NIST. Sensor response to varying magnitudes of laser wavelength sweep is also recorded. From this collection of calibration data and temperature-dependent line parameter information taken from HITRAN 96 [Rothman et al., 1998], separate sets of calibration coefficients for the strong and weak H₂O(v) lines are calculated and provided to the algorithm which reduces raw sensor data to H₂O(v) mixing ratio.

Table 1.10 Missions using the LARC/ARC Diode Laser Hygrometer.

Mission	Dates	No. of flights	Deployment site	Latitude	Longitude
SUCCESS	Apr - May 1996	6	Salina, Kansas
PEM-Tropics A	Aug - Oct 1996	22	Honolulu, Hawai Tahiti Fiji New Zealand Easter Island	72S - 45N	150E-110W
SONEX	Oct - Nov 1997	12	Bangor, Maine Ireland Azores	10N - 70N	130W - 15E
PEM-Tropics B	Mar - Apr 1999	17	Hilo, Hawai Tahiti Fiji Easter Island Costa Rica	36S - 36N	87W - 163E
SOLVE	Nov 1999 - Mar 2000	24	Kiruna, Sweden	68N	21E

1.2.4 Radiosondes

Radiosondes have for several decades been the primary means of obtaining atmospheric vertical profile data from the surface to the lower stratosphere, for use in operational meteorological forecasts and, more recently, for assessment of climate trends from the archived data. The instrument packages are carried aloft by balloons (or dropped from aircraft or rockets, in the case of dropsondes) and are generally equipped with temperature, humidity, and pressure sensors. An overview of radiosonde instruments is given in WMO [1996], and Pratt [1985] reviews the quality of temperature and humidity data.

Currently, the global radiosonde network includes about 900 upper-air stations, and about two-thirds make observations twice daily (at 0000 and 1200 UTC). The network is predominantly land-based and favours the Northern Hemisphere. Radiosondes can achieve heights of about 35 km, although many soundings terminate below 20 km.

Measurements are made and transmitted during ascent, however the data archives may not contain the complete sounding. Radiosonde data are commonly acquired either every 2 s or every 10 s during flights. Given balloon ascent rates of about 5 m/s, the sounding data offer vertical resolution of 10-50 m. Operational data transmissions are required by the WMO at about 20 mandatory reporting levels up to 1 hPa (although soundings rarely attain that pressure level), and significant level data are required when the profiles deviate substantially from linearity between mandatory levels, thus detailed structure may not be fully captured in the archived operational reports. However, since 1995, the U.S. National Weather Service has archived data every 6 s for greatly enhanced vertical resolution.

Of the three basic radiosonde measurements, humidity observations are the most difficult to make because of the wide dynamic range needed for an instrument to measure water vapour concentrations varying by four or five orders of magnitude throughout the sounding. Humidity sensor performance depends on rapid exchange of water molecules with the air, and at cold temperatures, low water vapour mixing ratios make the measurement difficult. For this reason, the quality of humidity data from radiosondes is generally thought to decrease with decreasing water vapour content, temperature, and pressure [Elliott and Gaffen, 1991]. Stratospheric humidity data from radiosondes are essentially useless [WMO, 1996, Schmidlin and Ivanov, 1998], and much of the archived upper-tropospheric humidity data for the past several decades is unsuitable for the purposes of this Assessment.

The data come from radiosondes by various manufacturers, some producing multiple models, based on a variety of humidity sensor types. In each case, the fundamental quantity measured is relative humidity. The most commonly used are: goldbeater's skin sensors (used mainly in China, Russia, and the FSU); lithium chloride sensors (used in U.S. radiosondes before 1965, currently only in India); carbon hygristors; and thin-film capacitors. Goldbeater’s skin (beef peritoneum) sensors, more commonly used in the early decades of upper-air observations, give reliable humidity observations only at temperatures above 0°C [Schmidlin and Ivanov, 1998], making the upper-tropospheric and stratospheric observations of no value. Lithium chloride sensors, perform poorly at very high or very low humidity; have long time constants of response; are severely affected by wetting by rain [Mathews, 1965, Showalter, 1965]; and are deemed unreliable at pressures lower than 600 hPa [WMO, 1996].

Because of the widespread use of radiosondes carrying carbon hygristors and thin-film capacitors, much of the existing data archive is based on these sensors. Here, the recent findings on the performance of these two sensors in the upper-troposphere and the possibility of correcting some upper-tropospheric humidity data from thin-film capacitors are summarised.

Carbon hygristors are polystyrene strips, coated with a hygroscopic film containing carbon particles, whose electrical resistance increases with relative humidity. Carbon hygristors respond very slowly at temperatures below ?40°C (typical of the upper troposphere and lower stratosphere) and do not function at all at ?60°C. These data are considered so unreliable that many countries have special data reporting practices for low temperatures [Gaffen, 1993]. For example, humidity measurements below -40°C were not reported by the U.S. National Weather Service until 1993 (when they were introduced in response to requests from data users), and significant caveats are warranted with the post-1993 data [Wade, 1994, 1995]. Recent tests suggest that carbon hygristors have a moist bias at relative humidities above about 60% [Schmidlin and Ivanov, 1998]. They can also experience hysteresis after exiting clouds, resulting in significant humidity errors.

Thin-film capacitors (marketed as Humicap) are used on Vaisala and Meisei radiosondes. The capacitors have one electrode treated with a polymer film whose dielectric constant varies with ambient water vapour pressure. Though thin-film capacitors also respond more slowly at low temperatures, they are faster than carbon hygristors. Therefore, the upper-tropospheric humidity data from thin-film capacitors are potentially more valuable. There are three types of Vaisala humidity sensors:

a) Humicap A (RS80-A). This is the original RS80 humidity sensor and the only one in use until the mid-90s. It has a time constant of 100 s at -50° C and 400 s at -70° C, thus it will respond to 63% of a step change in humidity over a vertical distance of 0.5 and 2 km respectively. Experience has shown that there is a dry bias in the A type sensor at cold temperatures (see below).

b) Humicap H type (RS80-H). This sensor is smaller than the A type and responds more quickly. It appears not to suffer the same biases as the A type [Kley et al., 1997] and is capable of measurement up to the tropopause. The sensor’s reproducibility in the lower troposphere is quoted as ± 3% in relative humidity [http://www.vaisala.com]. Unfortunately, there is at present no reliable estimate of the accuracy of this sensor at the higher levels.

c) RS90 sensor. The new Vaisala radiosonde has an improved humidity sensor, which is designed to counter the problem of icing in clouds (see below). Other than this change the sensor is similar to the Humicap-H, with a modest improvement in accuracy in the upper troposphere. To date few countries have adopted the RS90 operationally, and its performance in the upper troposphere has not been determined.

The sensors are subject to several well-understood sources of measurement error. Corrections for some of these are now becoming available. However, before applying corrections to archived data, it is recommended to consult the operating weather services because the information on which type of sensor was used for a particular sounding is not archived with the data. The known error sources are:

1. An approximation for the temperature dependence of the calibration causes an error that increases with decreasing temperature, and is significant at temperatures below -30°C for RS80-A measurements and below -50°C for RS80-H measurements. Observations by Heymsfield and Miloshevich [1995] and Kley et al. [1997] showed significant low biases of up to 30% relative humidity at - 65°C in measurements from the RS80-A sensor within ice clouds (where the relative humidity should be at least saturated with respect to ice, about 55%). This error is merely a data processing error, not an inherent limitation of the sensor, and a temperature-dependent correction is available (see Chapter 2.2.4, figure 2.12).
2. A time-lag error is caused by an increase in the sensor's time constant (63% response time) with decreasing temperature. The time constant for both sensors exceeds one minute at temperatures below about -50°C and two minutes below about -60°C. Given typical balloon rise rates, the sensors will respond to 63% of a step change in humidity over vertical distances of about 0.6 or 1.2 km at these temperatures [Paukkunen, 1995]. The time constant for the sensors has been measured as a function of temperature, so it may be possible to correct these errors if the sounding data possess sufficient time resolution.
3. The sensors are susceptible to dry-bias errors that result from chemical contamination of the sensor polymer and from long-term sensor instability (sensor ageing). Corrections for these errors are being developed; however, the contamination issue is being addressed by Vaisala with changes in the radiosonde packaging, thus the applicability of the corrections will often be questionable for archived data.
4. Icing of the sensors when passing through clouds renders the remainder of the humidity profile invalid. The new Vaisala radiosonde, model RS90, addresses this problem by alternating observations between two sensors, heated to remove ice.

Although radiosonde sensors respond to relative humidity, sounding data can be reported in terms of either relative humidity or dew point depression, depending on the data archive. Differences in data reduction methods lead to small differences in reported dew point depression that are largest in cold dry conditions [Elliott and Gaffen, 1993]. In addition, errors in temperature and pressure observations will influence humidity variables computed using those data. Temperature and pressure errors tend to be larger in the upper-troposphere and stratosphere than in the lower troposphere. Many, but not all, radiosonde stations adjust the temperature data for solar and infrared radiation errors, and lag errors. However, these adjustments are empirical and not tailored to detailed atmospheric conditions (e.g. cloudiness) that influence the actual errors.

The simultaneous use of different radiosonde types within the global network can lead to differences in upper-tropospheric humidity across geopolitical boundaries, as seen by Soden and Lanzante [1996]. For long-term climate studies, changes in radiosonde instrumentation and methods of observation can introduce artificial signals into the data record [Gaffen, 1993,1994, Ross and Gaffen, 1998]. Therefore, care must be taken to verify both the spatial and temporal homogeneity of radiosonde data used to determine climatological features as well as interannual and longer changes in atmospheric water vapour.

In conclusion, of the many different radiosonde humidity sensors, only the thin film capacitor sensors are capable of measuring in the upper troposphere. The Humicap sensors are and have been widely used all over the world, and so offer, potentially, a valuable climatological archive. However, the existence of several uncorrected and poorly documented sources of measurement error at cold temperatures, as well as changes in radiosonde instrumentation and reporting practices, render most archived humidity data unsuitable for upper-tropospheric climate studies without extreme care in assessing the accuracy of individual radiosonde types and reporting practices and their changes with time. The availability of corrections to known errors and newer humidity sensor technologies may offer improved data in the upper troposphere. Global deployment of more reliable radiosondes on a routine basis would provide a very valuable source of in situ water vapour profiles.

1.2.5 MOZAIC sensors

MOZAIC (Measurement of Ozone by Airbus In-Service Aircraft), operated since August 1994, is a project for the automated measurement of ozone and water vapour onboard five Airbus A340 aircraft on scheduled flights of commercial European airlines [Marenco et al., 1998]. The flights cover a large extent of both hemispheres and will go on for several years on a regular time schedule. The percentage of different flight routes is shown in Figure 1.6. For the period between August 1994 and December 1998, data from more than 10,000 flights with an average of more than 8 hours length are available in the MOZAIC database at Centre National de Recherche Météorologique, (CNRM, Toulouse, France). Also, for nearly each flight altitude profiles of ascent and descent at the respective airports are available, giving a total of more than 19,000 profiles.

Figure 1.6 Overview of flight routes with MOZAIC water vapour measurements for the period August 1994 to December 1998. The numbers give the percentage of flight routes.

Relative humidity and temperature are measured in situ with an airborne sensing device AD-PS2 (Aerodata, Braunschweig, Germany) mounted in an appropriate housing (Model 102 BX, Rosemount Inc., Aerospace Division, United States) [Helten et al., 1998]. The humidity sensor (Humicap-H, similar to the Vaisala RS80) is a capacitive sensor with a hydroactive polymer film as dielectric whose capacitance depends on the relative humidity (see section 1.2.4). The temperature sensor, a platinum resistance sensor (PT100), is mounted parallel to the humidity sensor near its sensing surface for the direct measurement of the temperature.

The housing with both sensors is positioned outside the fuselage, 7 m backward from the aircraft nose on the left side just below the cockpit. The air is sampled at 7 cm distance from the aircraft skin, well outside the aircraft boundary layer. Inside the housing the air is divided into two parts. The main flow traverses straight through the housing. The minor flow makes a right angle turn to a smaller channel, perpendicular to the main channel, passing over the sensor elements before leaving the housing through a small outlet, located at the lower back side of the housing. The right angle turn of the secondary airflow protects the sensors against dust, water droplets, and particles. The internal boundary layer air is sucked off through small holes in the sidewalls of the housing, minimising internal boundary layer effects.

The air entering the housing is subject to adiabatic compression by the strong speed reduction in the inlet part of the housing. The conversion of the kinetic energy of the sampled air into heat leads to a significant temperature increase of the air sampled by the sensor. The temperature of the ambient air, called static air temperature (SAT), can be computed from the air temperature measured inside the housing, called total recovery temperature (TRT), if the air speed of the aircraft is known. TRT deviates by less than 1°C from the total air temperature (TAT) which is to be expected by an exact 100% conversion of the kinetic energy of the sampled air into heat inside the housing. The correction of the TRT is performed with an empirical correction factor provided by the housing manufacturer, determined from a series of wind channel experiments. The relationship between SAT and TAT is a function of the Mach number.

Due to adiabatic heating, the relative humidity measured inside the sensor housing, called dynamic relative humidity (RH_D), is much lower than the ambient relative humidity, called static relative humidity (RH_S), which has to be determined. Their ratio is given by

where E_W(SAT) and E_W(TAT) are the water vapour saturation pressures of pure liquid water at static air and total air temperature respectively. C_p (= 1005 J kg^-1 K^-1) and C_v (= 717 J kg^-1 K^-1) are the specific heats of dry air at constant pressure and volume respectively.

The ratio RH_S/RH_D is the product of two factors. The first compensates for adiabatic compression, while the second accounts for the different water vapour saturation pressures at SAT and TAT, respectively. The dynamic relative humidity, RH_D, measured by the sensing element is appreciably lower than the static relative humidity, RH_S, due to the adiabatic temperature increase in the housing. Therefore, at cruise altitude, typically above 8 km, the sensor is working in the lowest 10% of its full dynamic range. In this region the ratio RH_S/RH_D increases to values of 13, and it is obvious that individual calibrations of each sensor are necessary.

The temperature in the housing is typically 27°C higher than the ambient air temperature, so that the sensor temperature is always operated above -40°C. This increases the accuracy of the sensor, and makes its response faster in comparison with Humicap-H sensors operated on radiosondes.

Each MOZAIC humidity sensor is normally changed each month and calibrated in the laboratory before installation and again after 500 hours of flight operation. The calibrations are executed in an environmental simulation chamber in the laboratory [Smit et al., 1994]. Pressure, temperature, and frost point temperature are computer-controlled to simulate typical tropospheric conditions, including tropopause and lower stratosphere up to 15 km altitude. Frost point temperatures down to -70°C can be reached. The water vapour mixing ratio is determined with a Lyman-a fluorescence hygrometer [Kley and Stone, 1978; Kley et al., 1979; Helten et al., 1998], installed in the simulation chamber as a reference instrument for the measurement of the water vapour mixing ratio. The accuracy of the H₂O mixing ratio measurements in the fluorescence mode of the instrument is ±6%, including random and systematic errors with inclusion of the accuracy of the absorption cross sections of H₂O and O₂ at 121.6 nm wavelength [Kley, 1984], which determines the calibration of the fluorescence measurement.

These calibrations revealed that the relative humidity of a calibrated sensor (RH_C) for a constant temperature can be expressed by a linear relation

where RH_UC is the uncalibrated output from an individual sensor, and a and b are coefficients that result from the calibration procedure. Each calibration is executed at three temperatures:
-20°C, -30°C, and -40°C, resulting in three pairs of calibration coefficients a and b.

Figure 1.7 Mean uncertainty in percent RH of all 1995 MOZAIC relative humidity measurements (solid curve) as a function of static air temperature (bottom x axis). The corresponding altitude is indicated on the top x axis. The standard deviation of the mean is marked by the dashed curves. The region not covered by pre-flight and post-flight calibrations (lower troposphere, see text) is indicated with an estimated mean uncertainty (dashed/dotted line).

The mean of the pre- and post-flight calibration coefficients of each flight period are used to evaluate the measurements. The differences between both sets of these calibration coefficients give the main contribution to the uncertainty of the measurement [Helten et al., 1998]. The variations of the uncertainties of the RH measurements were determined as the mean of all individual total uncertainties over all MOZAIC data from 1995 as a function of SAT (Figure 1.7). The standard deviation is also shown. The uncertainty ranges from ±7% RH at -55°C (Å13 km) down to ±4%RH at -40°C (Å10 km). At a lower altitude, for SAT ranging between -40°C (Å9 km) and -20°C (Å6 km) the uncertainty is within ±(4-6) % RH, increasing above 0°C (near ground level) to ±8% RH. For the region below 5 km altitude, only an interpolation between our sensor calibrations below -20°C and the nominal calibration of the sensor manufacturer is used, indicated as a dashed dotted line in Figure 1.7. A standard deviation is also not quoted for this altitude range. For the data in 1999, the sensor calibration procedure in the environmental calibration chamber will also include this temperature (=altitude) range.

1.2.6 Cryogenic collection

Compared to other in situ techniques, cryogenic collection of water vapour is a somewhat cumbersome technique and it also offers a lower spatial resolution. Its use is restricted to some studies in the 60s and 70s. However, it allows one to measure also the concentration of the isotopic water molecules HDO and HTO. Historically, the first simultaneous in situ measurements of the vertical distribution of the major hydrogen compounds H₂O, CH₄, and H₂ in the stratosphere have been made using this technique [Pollock et al., 1980].

The NCAR cryogenic whole air sampler [Lueb et al., 1975] and its operation have been modified for a number of balloon flights for measurement of the water content of the air samples collected [Pollock et al., 1980]. The various sampling systems consist of up to 16 stainless steel cylinders designed as a cryopump and which are filled during balloon descent. In the laboratory, aliquots of the sample are drawn into separate containers for gas chromatographic analysis of CH₄, H₂, and other long-lived tracers. The amount of water vapour is measured by separating the water from air in a cold trap. This is followed by reduction of H₂O to H₂ whose amount is measured volumetrically.

Although the principle of this technique is very simple, it is complicated by contamination. There are several steps during sample collection and transfer at which the original amount of water in the samples can be disturbed. Pollock et al. [1980] determined correction factors in separate experiments to compensate these artefacts. For the measurement of the H₂O amount, the factors are generally below 20%. The accuracy of stratospheric H₂O measurement is 0.3-1.0 ppmv for most of the samples analysed.

Between 1975 and 1978, the NCAR whole air sampler was used on eight balloon flights at different latitudes and analysed for water (Table 1.11). Data of H₂O, the water isotopes, CH₄ and H₂ obtained during these flights including the individual uncertainties are listed in Pollock et al. [1980].

Table 1.11 Balloon-borne measurements using the NCAR cryogenic whole air sampler.

Date	Latitude	Longitude	Altitude [km]	No. of samples
02 Jun 1975	32N	96W	17-34	14
14 Aug 1975	52N	102W	20-43	8
24 Sep 1975	32N	106W	33-39	4
27 Jan 1976	32N	96W	16-34	13
27 Feb 1976	32N	96W	36-41	4
10 Aug 1976	52N	102W	32-40	12
14 Aug 1977	52N	102W	21-40	7
23 Apr 1978	65N	148W	21-40	7

Measurements of tropospheric H₂O and HTO vertical profiles were made aboard two aircraft between 1965 and 1973 [Ehhalt, 1971; Ehhalt, 1974]. Here, the water was separated from the air aboard the aircraft by passing the air through liquid N₂ or dry ice cooled traps and freezing out water. Sampling time varied from 15 to 60 min, depending on altitude. In the laboratory, the water amount was measured by weighing.

On the NCAR Queen Air flights, samples were collected at eight altitudes from 1.5 to 9 km. These flights were carried out in regular intervals in different periods given in Table 1.12. The flights on the Sabreliner extended the measurements to the upper troposphere, i.e. samples were taken at seven flight levels from 6 to 13 km. All the aircraft data are listed in Ehhalt [1974].

Table 1.12 NCAR aircraft measurements of H₂O and HTO in the 60s and 70s.

Period	Latitude	Longitude	Frequency / No. of flights
	Queen Air
Nov 1965	41N	102W	1
Feb-Jun 1966	41N	102W	3 per month
Jul 1966 - Jan 1967	41N	102W	monthly
Jul 1966 - Aug 1967	34N	125W	monthly
Jul 1966 - Aug 1967	36N	117W	monthly
Mar 1971 - Jan 1972	41N	102W	nearly monthly
Nov 1972	41N	102W	1
	Sabreliner
Mar 1971 - Sep 1973	32N	96W	irregular
Feb 1972	Florida		1
Sep 1972	37N	122W	1