1.5 Summary and conclusions

In this Chapter, a variety of different techniques to measure water vapour have been presented with the description of the most important instruments providing data for this Assessment. The instruments are designed and optimised for specific measurement tasks and therefore their characteristics differ as well.

Table 1.29 summarises the most important technical parameters of in situ and balloon-borne, airborne and ground-based remote sensing instruments for the following discussion, i.e. precision and accuracy, along with information on the approximate range (mixing ratio or altitude) and the time and/or vertical resolution. Radiosonde characteristics are not listed here due to their limited capabilities for UT/LS altitudes.

Table 1.29 echoes the known advantage of in situ techniques for measurements whenever small-scale structures have to be investigated. The fastest instruments certainly are open-cell tuneable diode laser spectrometers and Lyman-a fluorescence hygrometers, the latter depending on purge rates. However, in order to achieve precision on the order of 0.1 ppmv in the stratosphere, integration times of at least 1 s are typically required for all instruments. Modern frost point hygrometers especially designed for measurements in the UT/LS can be as fast as 10-30 s, also permitting good spatial resolution both from balloons and aircraft. Remote sensing instruments operated from balloons, aircraft or from the ground provide vertical profile measurements with a resolution of several 100 m in the case of LIDARs, a few kilometres for the infrared and far infrared spectrometers and approximately 10 km for the microwave instruments.

The accuracies given in Table 1.29 are based on known or estimated systematic uncertainties and errors in the calibration procedures and/or the retrieval algorithms including parameters such as absorption cross sections and line strengths. They do not consider intercalibration between different instruments that can result in either smaller or larger discrepancies (see discussion in the Chapter 2). The accuracies given in Table 1.29 range from approximately 5 to 10%. In fact, the quoted accuracy for the different instruments is remarkably similar. As with the precision, it depends strongly on the method used to determine this quantity by each group and by the number of individual errors that have been added. We recommend that the Table be interpreted as saying that almost all the techniques, if properly applied and calibrated, can claim an accuracy of approximately 5-8%.

Reversing this statement, none of known techniques can currently provide a measurement of water vapour that is more accurate than 5%. The reason for this is, that producing standards for low H2O mixing ratios with high accuracy and subsequently the determination of more accurate cross sections and line parameters is limited by available techniques.

The time spans and altitude ranges for the satellite data were given in Table 1.22. Precision and accuracy may be judged from the summary estimates of the random and systematic errors in Table 1.30 for single representative profiles of each of the stratospheric satellite data sets. Precision is based on the random error, while accuracy is normally a combination of both the random and systematic error components of Table 1.30. Those estimates are based on the information in Section 1.4 or its references. Although some error estimates have been cited at an altitude (e.g., SAGE II), an approximate conversion to pressure-altitude has been performed here. For some data sets the errors were cited at specific altitudes or pressures, while for others they have been given for a range of pressure-altitudes. The errors in Table 1.30 have been estimated for ranges of vertical levels to achieve better consistency across altitude or pressure-layer boundaries. It is important to note that some systematic error components are often quasi-random from profile to profile (e.g., profile registration errors due to pointing uncertainties for limb measurements). Those error components become much less significant for daily or seasonal zonal means. Therefore, a zonal-average stratospheric error profile is dominated by the truly systematic error components, whose signs are often unknown. Some data sets extend well into the mesosphere; the reader must refer to the individual sections and the cited references for those error estimates.

 

Table 1.29 Summary of calibrations and uncertainties of the non-satellite borne instruments discussed in this Assessment: (1) systematic uncertainties; (2) reproducibility of calibration or flight-to-flight variability of uncertainties, or additional random calibration uncertainties; (3) precision or random uncertainty (given for typical integration interval); (4) total accuracy (i.e. sum of all systematic and random errors), given in percentage of mixing ratio for typical measurement range of instrument. Unless otherwise stated, the uncertainties do not contain potential sampling artefacts. The numbers of uncertainties are for typical measurement conditions and for the most advanced versions of the instruments; for details the reader is referred to the individual instrument descriptions and references therein. Secondary calibration procedures, applied irregularly for a number of instruments for crosscheck, are given in brackets.

laboratory
technique or acronym
range
resolution time or vertical

calibration procedure

(secondary calibration procedure)

range for calibration
standard, reference instrument or spectroscopic parameters used
uncertainties
source
magnitude
NOAA-CMDL FPH -15 to -95°C 10-30 s thermistor calibration   thermistor, NIST traceable

(1) thermistor calibration

(2) instrumental

(3)

(4)

0.05°

0.4°

0.5° / 10%

LMD-CNRS FPH

-10 to

-95°C

10-20 s

temperature sensor calibration

(humidity generation in a calibration bench)

  platinum sensor, traceable to national T reference standard

(1) sensor calibration

(2)

(3) at 10 s

(4)

0.1° (2s )

0.1°

0.05°

0.3° / 3-6%

NOAA-AL Lyman-a PFF 600 - 0.15 ppmv 1s in-flight absorption measurement by injection of H2O

500-20 ppmv

ambient p

s H2O(121.6 nm) [Kley, 1984]

(1) s H2O

(2)

(3) at 1 s

(4) at 10 s

6% (2s )

0.15 ppmv

6.6% (2s )

Harvard Lyman-a PFF 0.2 - 500 ppmv 0.125 s

generation of humidified air

(absorption measurement in the laboratory and in-flight)

10-300 ppmv

15-400 hPa

T, p measurement at saturation

(1)

(2)

(3) at 4 s

(4)

5%

3%

2%/0.1ppmv

6%

Jülich FISH 500 - 0.2 ppmv 1 s

generation of humidified air via a laboratory bench

(in-flight absorption measurement)

2-500 ppmv

50-200 hPa

Frost Point Hygrometer GE 1311 DRX, calibration traceable to NIST and national H2O standard

(1) calibration of FPH

(2)

(3) at 1 s

(4)

0.15° / 4%

3%

0.15 ppmv

6%

JPL TDL

ER-2:

400 - 0.1 ppmv.

DC-8: 30000 -
8 ppmv

0.1 s generation of humidified air in a calibration chamber

50-200 ppmv

50-500 hPa

250-300 K

Frost Point Hygrometer GE 1311 DR, calibration NIST traceable

(1) calibration of FPH

(2) FPH reproducibility

(3) at 2s

(4)

0.2° / 5%

3%

1%

5-10%

LaRC/ ARC DLH 1 - 20000 ppmv 0.05 s generation of flowing humidified air in 1 or 3 m long chamber equivalent of 0-1000 ppmv 100-1000 hPa

DPH Edgetech 2001, NIST traceable calibration

spectral parameters [Rothman et al., 1996]

(1) Calibration

(2) electronics, thermal

(3) at 50 msec

(4)

3%

8%or1ppmv

2%

10%or1ppm

Jülich MOZAIC > 20 ppmv 0.5-180s generation of humidified air (calibration chamber)   Lyman-a absorption: s H2O(121.6 nm) [Kley, 1984]

(1) s H2O

(2) flight-to-flight

(3)

(4)

6% (2s )

30%

8% (2s )

4-7% RH

NRL

WVMS

(ground-based)

40-80 km

1-7 d

10 km

blackbody calibration   line parameters [Pickett et al., 1998; Liebe et al. 1993]

(1) baseline, instrumental

+ profile retrieval errors

(3) 7d

(4) 40-80 km

0.5-0.1 ppmv

0.15 ppmv

0.6-0.2 ppmv

MPAE ground-based micro-wave 40-80 km

1-7 d

10 km

blackbody calibration   line parameters [Pickett et al., 1998; Liebe et al. 1993]

(1) baseline, instrumental

+ profile retrieval errors

(3) 7d

(4) 40-80 km

0.5-0.1 ppmv

0.15 ppmv

0.6-0.2 ppmv

Bern airborne micro-wave 15-75 km

30 min

8-15 km

blackbody calibration   line parameters [Pickett et al., 1998; Liebe et al. 1993]

(1)

(2)

(3)

(4)

0.6 ppmv
LaRC LASE tropo-sphere

2 min

300-500 m

self-calibrating (DIAL technique) 0.01-20 g/kg s H2O(813 nm) [Ponsardin and Browell, 1997]

(1) s H2O

(2)

(3)

(4)

6%
DLR DIAL

UT/LS

5-100 ppmv

1-2 min

300-500 m

self-calibrating (DIAL technique)  

s H2O(935.43, 935.35, 942.82 nm) [Poberaj and Weiss, priv. comm.];

[Rothman et al., 1996]

(1) s H2O

(2) laser transmitter

(3) profile retrieval

(4)

3%

2%

1-5%

5-7%

Karlsruhe MIPAS 5-40 km 2-3 km calibration based on "deep space" (+20° elevation angle) and internal blackbody spectra recorded during flight  

HITRAN data base [Rothman et al., 1998],

spectral intervals within 808-825 cm-1, 1210-1245 cm-1, 1589-1610 cm-1

(1) line parameters

(2) profile retrieval

(3)

(4)

5%

1-5%

4-8%

6-11%

JPL MkIV 20000-1 ppmv 2 km self-calibrating (solar occultation) - HITRAN data base [Rothman et al., 1998], 12 intervals at 1500-4630 cm-1

(1) line parameters

(2) profile retrieval

(3)

(4)

5%

5-12%

7-13%

SAO FIRS-2 10-40 km 2-4 km; 60 min per profile at 4 km vertical reso
lution
ambient-temperature blackbody + cold space view (intensity calibration)   HITRAN data base [Rothman et al., 1998], with strengths and positions from Coudert [1999] and Toth [1991]; 18 intervals at 80-480 cm-1

(1) line parameters

(2) profile retrieval

(3)

(4)

3%

4%

5%

 

Table 1.30 Estimates of random and systematic error of satellite stratospheric H2O profiles

Instrument and data set
Random error
Systematic error
LIMS (version 5)

20-15% from 1 to 5 hPa

15-10% from 5 to 10 hPa

10% from 10 to 50 hPa

31-24% from 1 to 5 hPa

24-20% from 5 to 10 hPa

20-37% from 10 to 50 hPa

SAGE II (version 5.9)

10-5% from 3 to 10 hPa

5-14% from 10 to 25 hPa

14% from 25 to 300 hPa

6-13% from 3 to 7 hPa

13% from 7 to 25 hPa

13-27% from 25 to 100 hPa

27% from 100 to 300 hPa

ATMOS (version 3) 9-11% from 1 to 300 hPa 6% from 1 to 300 hPa
HALOE (version 19)

9-7% from 1 to 10 hPa

7-13% from 10 to 40 hPa

13% from 40 to 100 hPa

10-14% from 1 to 10 hPa

14-19% from 10 to 40 hPa

19-24% from 40 to 100 hPa

MLS (version 0104)

4% from 1 to 10 hPa

3% from 10 to 50 hPa

3-8% from 50 to 100 hPa

6-9% from 1 to 10 hPa

9-16% from 10 to 50 hPa

16-50% from 50 to 100 hPa

MAS 5-10% from 1 to 50 hPa 10-15% from 1 to 50 hPa
ILAS (version 4.20)

More than 10% above 2 hPa

10-5% from 2 to 300 hPa

30% from 1 to 2 hPa

30-10% from 2 to 7 hPa

10% from 7 to 300 hPa

POAM III (version 2) 5% from 3 to 100 hPa 15% from 3 to 100 hPa

It should be appreciated that there are instances when comparisons between data sets consistently indicate very good agreement, such that one might conclude that the total error for the least accurate data set is too conservative. In other words, some of its systematic error components may be overestimated or have unknown but opposing signs. Conversely, it should also be apparent that only known error mechanisms have been characterised for each data set, and it is often the intercomparison with other data sets that reveals any remaining problems.

Similar estimates of random and, especially, systematic errors for the MLS and TOVS upper tropospheric humidity data sets can not be characterised as easily. These products are derived as layer-averaged results for a region of the atmosphere where the vertical gradient of water vapour (and temperature for TOVS) is changing significantly. Estimates of systematic error are dependent on the atmospheric state for the measurement, and it has been more appropriate to report the sensitivity of the retrieved upper tropospheric humidity to those atmospheric state variables. Those sensitivities are provided in the MLS and TOVS subsections and their cited references. The reader should consult them for the details. Even so, the precision of the MLS and TOVS upper tropospheric humidity data sets is generally good, which is why the spatial water vapour patterns that they report are useful. SAGE II data are reasonably accurate in the upper troposphere, but they do have a "clear-air" bias.

The TOVS upper tropospheric humidity measurements and two of the stratospheric data sets (SAGE II and HALOE) have relatively long records, and they are being used to look for changes that may be related to increases in greenhouse gases and/or surface warmings. The reader is referred back to their individual sections for the more complete estimates of measurement and/or retrieval errors that could affect such trend analyses. The primary concern for the long record of TOVS data is the intercalibration of the series of instruments that have been flown on the operational satellites. The Version 5.9 SAGE II H2O time series is affected by uncertainties in its correction for aerosol extinction between the tropopause and about 27 km (for volcanically perturbed periods). SAGE II H2O data may also be affected by changes in the reflectivity of the scan mirror and in the characteristics of the filter for the NO2 (453 nm) channel that is part of the aerosol correction algorithm. The HALOE time series seems to be free of long-term instrument and/or retrieval effects. However, its near-global trends may be affected by changes in HALOE sampling frequency and pattern, particularly for the late 1990s.

In summary, all of the satellite data sets are well characterised and can be used to define zonal-mean distributions of water vapour and its large-scale, zonal variations. Because their precision is generally good, they can provide the seasonal and even interannual cycles of water vapour. Longer-term changes can also be addressed with HALOE data at those latitude zones where the sampling has been consistent over the life of the experiment. It is now the purpose of Chapters 2 and 3 of this report to evaluate whether the distributions and long-term changes from each of these data sets are consistent with each other and meaningful.

The data sets of the individual instruments presented in this Chapter provide a wide coverage in space (different altitude range and regions) and in time. First measurements in the UT/LS date back to the 1940s. The longest continuous and well-documented record of measurements in the UT/LS is that of the balloon-borne NOAA-CMDL frost point hygrometers beginning in 1964 up to present days. Since the late 1970s, the Lyman-a technique provides complementary measurements both from balloon and aircraft. In the 1980s and 1990s, the number of in situ and remote sensing instruments for balloon-borne and airborne measurements has increased rapidly (see Figure 1.2) involving established and new techniques. The only ground-based technique ranging into the stratosphere is microwave spectroscopy, which has been used for continuous measurements at different places since 1993 (Figure 1.9).

Satellite instruments of stratospheric water vapour are available since the late 1970s (Table 1.22). Today, they are the most important data source for most applications due to their nearly global coverage and the continuity of the measurements.

A large number of techniques is available for probing the upper troposphere and tropopause region, i.e. radiosonde and MOZAIC sensors, LIDARs, satellites (operational upper tropospheric humidity and more recently, also extended retrievals of other satellite experiments) for regular measurements, as well as airborne in situ hygrometers on campaign-basis. However, the large variability and strong gradients in the H2O distribution at these altitudes plus the sometimes limited accuracy and precision, the characterisation of the instruments or their low sampling frequency, are still limiting the factors for determining the global distribution of H2O and its possible long-term changes.