Operational weather observations often serve as primary observational data in atmospheric research. Within the scope of SPARC, many areas of inquiry require profiles of atmospheric temperature, pressure, humidity, and winds, and radiosonde observations are an important source of this information. The high vertical resolution of radiosonde data offers a distinct advantage over some satellite data and reanalysis products. This article presents some basic, but perhaps not widely known, information about radiosonde data (particularly temperature) and derived products, that may be useful in SPARC-related investigations of such topics as: the climatology, variability, and trends of temperature in the lower stratosphere and upper troposphere; upper-tropospheric water vapour; the quasi-biennial oscillation; volcanic impacts on stratospheric temperature; and tropopause characteristics.
Several organisations provide basic radiosonde data for research purposes. A few groups have created specialised upper-air temperature datasets, some with a focus on the stratosphere. Each has somewhat different approaches to station selection, data processing, treatment of missing data, etc. It is important to understand these differences both in selecting products for particular applications, and in comparisons of radiosonde datasets with those from other observing systems (e.g., in the WMO-UNEP-SPARC Temperature Trends Assessment, Chanin and Ramaswamy, 1999). Seven products are briefly summarised here.
- Historical Arctic (and Antarctic) Rawinsonde Archive: Individual soundings for the Arctic and Antarctic, beginning in the 1950's (Kahl et al., 1992). Monthly-mean gridded soundings are currently being prepared by the University of Wisconsin-Milwaukee.
- Freie Universitat Berlin Northern Hemisphere stratospheric analyses: Hand-drawn synoptic analyses of radiosonde data at 100, 50, 30, and, in some months, 10 hPa, starting in 1964 (Labitzke and van Loon, 1995).
- Central Aerological Observatory Arctic monthly mean stratospheric temperatures: Data for 58 stations at 100 and 50 hPa, starting in 1959 and 1961, respectively (Koshelkov and Zakharov, 1998).
- NOAA Air Resources Laboratory (ARL) global and regional seasonal anomalies: Two data products, including 1) temperature anomalies at 50, 30, 20, and 10 hPa based on 5 stations in the equatorial zone, 5 in the north subtropics, and 3 in the northern United States, beginning in 1958 (Angell, 1991); and 2) layer-mean virtual temperature in five layers: 850-300, 300-100, 100-50 and 100-30 hPa, and surface-100 hPa, computed from geopotential heights of the layer endpoints, for seven zonal bands, two hemispheres, and the globe, based on a 63-station network (Angell, 1988).
- NOAA Geophysical Fluid Dynamics Laboratory global temperature analyses: Gridded global monthly data from objective analysis and interpolation at eleven pressure levels up to 50 hPa, for 1958-89 (Oort and Liu, 1993).
- NOAA National Climatic Data Center Comprehensive Aerological Research Dataset (CARDS): An ongoing effort to create a complete historical radiosonde data archive (Eskridge et al., 1995). Currently, monthly data are available for a "core" set of about 200 stations (Wallis, 1998) through 1990.
- UKMO Hadley Centre for Climate Prediction and Research monthly station data and gridded monthly temperature anomalies: Several versions of monthly data at nine pressure levels up to 30 hPa, beginning in 1958. Data are interpolated to three grids of different resolution, missing data are reconstructed using principal component analysis, and some artificial data inhomogeneities are adjusted using Microwave Sounding Unit data as a reference (Parker et al., 1997).
The earliest radio soundings were made in the late 1920's and 1930's by V. Vaisala in Finland (Rossi, 1973), P. Molchanov in Russia (Zaitseva, 1993), and P. Idrac and R. Bureau in France (Gaffen, 1993). However, archived data generally begin in the 1940's, with significant expansion of the network at the time of the International Geophysical Year (1957-58). Currently, the 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.
For the stratosphere, however, the sampling by radiosondes is considerably worse than for the troposphere. Individual soundings terminate in the lower stratosphere at balloon burst height or when the balloon is lost by the tracking system, which, during the early decades, was by optical theodolite. We have recently examined radiosonde data from stations in two well-known global climate monitoring networks: the 63-station NOAA/ARL network used by J.K. Angell, (1988) and the approximately 150-station Global Climate Observing System (GCOS) Upper-Air Network, which together total 180 stations.
Figure 1 shows the number of soundings available in the CARDS database for this network for three sample years: 1960, 1975, and 1990. If data were available for all 180 stations, twice daily, the totals would be 131,400. The dataset expands with time at all levels, and the fraction of soundings reaching 50 hPa (~21 km) increases from 56 to 67 to 78%. However, the fraction reaching 10 hPa (~31 km) is only about 18% in each of the three years, severely limiting stratospheric sampling. Furthermore, stratospheric analyses involving multiple pressure levels may be affected by the decreased sampling with height within the lower stratosphere.
Figure 1. Number of archived soundings for a 180-station network in the CARDS database for three sample years. If data were available twice daily for all station, the totals would be 131,400/year. Note the poor sampling in the stratosphere compared with the troposphere.
In recent years, the number of radiosonde stations and soundings has decreased. The closure of the OMEGA radio navigation system threatened wind-finding at stations relying on this method of tracking radiosondes, although many affected stations have implemented alternate systems. Cost-saving measures in many countries have included closing stations and reducing observing schedules.
Numerous factors influence the accuracy of radiosonde temperature measurements, and an excellent review is given in the WMO Guide to Meteorological Instruments and Methods of Observation (WMO, 1996). Solar and infrared radiation impinging on the sensor can cause errors which depend in complex ways on the configuration of the radiosonde and environmental factors. The shape, radiative properties, and exposure of the temperature sensor affect heat transfer between the sensor and the environment, and these factors vary from one radiosonde model to another (Luers and Eskridge, 1998). Solar effects vary with elevation angle (McMillin et al., 1992). Infrared effects are both from the environment (air, surface, clouds, aerosols) and from radiosonde components (instrument housing, balloon). Generally the daytime temperature errors exceed those at night, and the errors increase with height (Luers and Eskridge, 1998), making daytime stratospheric temperature data most prone to bias errors. In fact, day-night differences in stratospheric temperatures have long been used in schemes to adjust daytime observations for inclusion in stratospheric analyses (Teweles and Finger, 1960). Radiation errors in the stratosphere can approach several degrees (McMillin et al., 1992, Luers and Eskridge, 1998), and the situation near the tropopause is particularly complex because of the ambient temperature gradients and the resulting long-wave radiative fluxes (McMillin et al., 1992).
Radiosonde data at some stations, and for some periods of time, have been adjusted for radiation errors (Gaffen, 1994). In some cases (e.g., Vaisala radiosondes), adjustments are made at the station by data-processing software and are therefore incorporated into archived data. Numerical weather prediction centres apply their own corrections to operational data, and these affect the model-analysed fields but not the archived sounding data.
Meteorological balloon size influences the rise rate of radiosondes which, in turn, influences sensor lag errors. Moreover, the final height attained by the balloon before burst depends both on balloon type and environmental factors. In the stratosphere, meteorological control of burst height could influence the representativeness of climatological data, with better sampling in warm conditions (Parker and Cox, 1995).
Balloons also influence the temperature of air in their wakes, with differential effects at night and day and errors increasing with decreasing air density (Tiefenau and Gebbeken, 1989). Balloon wake effects depend on the distance between the balloon and the sensor. One documented example of this effect is the lengthening of the cord from 7 to 15 m in Japanese radiosondes in 1968, which resulted in decreases of 50 hPa day-night temperature differences of 2-3 K (Suzuki and Asahi, 1978). The effect on climatological data was to lower temperatures by up to 1.3 K (at 100 hPa) in the decade following the change compared with the preceding decade (Gaffen, 1994). In general, however, balloon wake effects on climatological upper-air data are not well documented.
All of the sources of error discussed above can change over time, with important repercussions for the temporal homogeneity of climatological datasets (Gaffen, 1994, Parker and Cox, 1995, Lanzante, 1996, Gaffen et al. in preparation). Because radiosondes are operational data sources, instruments and observing practices are continually changing as technology improves and national meteorological services' requirements change. Station history information can be useful in identifying artificial signals in radiosonde data caused by observing system changes.
Under the auspices of the WMO Commission on Instruments and Methods of Observation, an international survey in 1990 gathered radiosonde station history information (Gaffen, 1993), which has subsequently been digitised and is available electronically (Gaffen, 1996). This "metadata" product includes information about changes in station location, instrumentation, observing practices, data processing methods, observing schedule, etc., which may affect the spatial and temporal homogeneity of station data. However, the station histories are not complete and contain considerable uncertainty, and so do they not provide definitive identification of all artificial signals in radiosonde data (Gaffen et al. in preparation). The metadata include a list of more than 175 radiosonde models, which highlights the complexity of the complete global data archive.
Because of their long period of record and high vertical resolution, radiosonde data are appealing for temperature trend studies. Until recently, most estimates of upper-air temperature trends based on radiosonde data (e.g., Angell, 1988, Oort and Liu, 1993, Labitzke and van Loon, 1995, Pawson and Naujokat, 1997) have not taken into account the possibility of time-dependent biases due to instrument changes, except perhaps to note their existence. However, discrepancies between temperature trend estimates from different observing systems have prompted closer examination of the reliability of radiosonde-based trends estimates.
Parker et al., (1997) have attempted to adjust radiosonde temperature data for known changes in instrumentation and to estimate the effect of the adjustments on calculated trends. Their adjustments were limited to data from stations in Australia and New Zealand where instrument changes were documented during the period 1979-95. The adjustments reduced the estimated zonal-mean temperature change between the periods 1965-74 and 1987-96 at about 30°S and 30 hPa from -2.5 K to about -1.25 K. These results indicate the potentially large sensitivity of estimated temperature trends to the identification and adjustment of artificial signals in time series.
A recent examination of the sensitivity of temperature trends to several radiosonde data quality issues showed that trends are very sensitive to attempts to adjust the data for possible artificial discontinuities (Gaffen et al., in preparation). Adjustments based solely on statistical identification of discontinuities effectively eliminate most statistically significant trends, because they reduce both artificial step-like changes in time series as well as real abrupt temperature changes which contribute to trends (Pawson et al., 1998). A second approach, incorporating station history information, results in adjusted time series whose trends are statistically significantly different from the original trends at the 99% confidence level. For example, adjustments were made to 50 hPa temperatures at three Japanese radiosonde stations to reduce a step-like increase of about 0.7 K associated with a 1981 change from Meisei RSII-56 to Meisei RSII-80 radiosondes. At one station, Minamitorishima, the 1970-95 trends, and their 99% confidence intervals, before and after the adjustment were 0.30± 0.14 and -0.12± 0.13 K/decade, respectively, which illustrates the large impact of the adjustment.
Santer et al. (submitted) recently compared lower stratospheric temperature trends from two versions of the UKMO Hadley Centre data set described above. They find that global and hemispheric trends for the two versions differ due to the differing spatial coverage associated with interpolation to two different grids. In the Southern Hemisphere, where sampling is poorest, the gridding effects are larger than in the Northern Hemisphere. Southern Hemisphere trends for 1973-93 in the two versions of the dataset are -0.235 and -0.268 K/decade, differing by about 13%.
As another example of the effect of dataset choice on trends, Figure 2 compares 1959-91 temperature trends at Kagoshima, Japan, based on the CARDS and Hadley Centre data sets. The CARDS data show cooling, statistically significant at the 99% confidence level, at all levels from 300 to 30 hPa. The Hadley Centre data show significant cooling only between 200 and 100 hPa. These differences appear to be related to differences in the observation times (0000 or 1200 UTC or both) of soundings used to create monthly means. In Figure 3, the difference in 30 hPa monthly mean temperatures from the two datasets is shown. The downward shift in the differences in 1981 accounts for most of the difference in trends.
Figure 2. Temperature trends at Kagoshima, Japan, based on Hadley Centre and CARDS data. Trends at each level are for 1959-91 and 99% confidence levels are shown. The differences are likely due to inclusion of soundings for different observing times during different periods in the two datasets. The trends at each level are slightly shifted to avoid overlap.
Figure 3. Differences (CARDS minus Hadley Centre) in monthly 30 hPa temperatures at Kagoshima in the CARDS and Hadley Centre datasets.
These few examples demonstrate that temperature trends from radiosonde data are sensitive to several factors that diminish the quality of the data for climate monitoring purposes. Some of these factors, for example historical changes in instrumentation that affect the temporal homogeneity of the data, are beyond the control of an individual researcher. Others, such as choice of radiosonde data product, are individual decisions, but information that may be useful in making such decisions in not always readily available. Overall, it is important to remember that these are operational meteorological observations, and their use for trend detection must be made with care.
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. However, in the upper troposphere, it is not entirely clear how credible the observations may be. Peixoto and Oort, (1996) presented global climatological humidity and moisture flux variables derived from radiosonde data up to 300 hPa, and Gutzler, (1996) presented tropical specific humidity data up to 300 hPa as well. Ross and Elliott, (1996) limited their analysis of humidity over North America to levels up to 500 hPa, as did Gaffen et al., (1992).
Recent WMO-sponsored comparisons of the performance of contemporary radiosonde humidity sensors, both in test chambers and during test flights, concluded that "no sensor appeared to respond to humidity changes at temperatures colder than -30°C" (Schmidlin and Ivanov, 1998). In recognition of the poor humidity sensor response in cold environments, many countries have special data reporting practices for low temperatures (Gaffen, 1993). On global average, temperatures below -30°C occur at pressures below about 400 hPa (Peixoto and Oort, 1992), which suggests that upper-tropospheric humidity data be used with caution, if at all.
Radiosonde sensors respond to relative humidity, but sounding data can include 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 also largest in cold dry conditions (Elliott and Gaffen, 1993), which further reduces the quality of humidity observations in the upper troposphere compared with the lower troposphere.
Radiosonde data have long been the basis for observational studies of the tropopause, particularly in the tropics (e.g. Angell and Korshover, 1964, Reid and Gage, 1981). Because of their high vertical resolution, and the requirement to include tropopause level information in operational soundings, radiosonde archives contain a wealth of information about the tropopause. However, the same caveats already noted apply to radiosonde tropopause data.
Humidity data at tropopause levels are highly uncertain, so the best tropopause humidity information from radiosondes is likely to be calculated saturation vapour pressures based on observed tropopause temperature and pressure, bearing in mind the potential uncertainties in temperature data noted above. Because the geopotential height of the tropopause is calculated from the hypsometric equation based on all the tropospheric temperature data, it integrates any biases in the temperature and humidity data. Day-night differences in the tropopause level may, in part, reflect day-night differences in temperature errors.
Long-term tropopause data records are potentially influenced by changes in observing methods. Figure 4 shows monthly anomalies of tropopause temperature and pressure from radiosonde data at Papeete, Tahiti for 1971-90. The largest signal in these data is the period of anomalously high values before 1976, the year in which the station switched from using Mesural 1943A sondes, with bimetal temperature sensors, to Mesural 1944C sondes, carrying thermistors (Gaffen, 1996). The slow response of the bimetal sensors probably accounts for the early high biases in tropopause temperature, because of the decrease of temperature with height in the troposphere. The high bias in tropopause pressures are more difficult to explain. Both sondes carried aneroid capsule pressure sensors, but the 1944C sensors were pre-calibrated, whereas the earlier model required calibration at launch time. Why this could lead to biases is unclear, but the clearly spurious signal in the tropopause temperature data suggests the pressure data could also be problematic.
Figure 4. Monthly anomalies of tropopause temperature (blue) and pressure (black) at Papeete, Tahiti (150°W, 17°S). Triangle indicates the date of a radiosonde instrument change that may account for the anomalously high values before 1976.
Radiosonde data archives are a rich source of information about the troposphere and lower stratosphere and figure prominently in many SPARC-related investigations. However, the observations have limitations that may be important in some applications. Sampling by radiosondes is limited spatially and favours Northern Hemisphere, mid-latitude land areas. The sampling of the lower stratosphere is substantially worse than for the troposphere. Temperature data, particularly in daytime and particularly early in the record, may have biases of up to several degrees. The nature of the temperature biases in a long data record may change over time due to changes in instrumentation and observing methods, with important implications for trend studies. Humidity data in the upper troposphere and stratosphere are poor and should be used with caution, if at all. Radiosondes provide fine vertical resolution near the tropopause level; however, tropopause data may incorporate errors in temperature, humidity, and pressure observations in complex fashion.
Angell, J.K., 1988: Variations and trends in tropospheric and stratospheric global temperatures. J. Climate, 1, 1296-1313.
Angell, J.K, 1991: Stratospheric temperature change as a function of height and sunspot number during 1972-89 based on rocketsonde and radiosonde data. J. Climate, 4, 1170-1180.
Angell, J.K., and J. Korshover, 1964: Quasi-biennial variations in temperature, total ozone, and tropopause height. J. Atmos. Sci., 21, 479-492.
Chanin, M.-L., and V. Ramaswamy, and co-authors, 1999: Trends in stratospheric temperatures. WMO Scientific Assessment of Stratospheric Ozone Change: 1998, World Meteorological Organization, Global Ozone Research and Monitoring Project Report in press, Chapter 5, Geneva.
Elliott, W.P., and D.J. Gaffen, 1991: On the utility of radiosonde humidity archives for climate studies. Bull. Amer. Meteor. Soc., 72, 1507-1520.
Elliott, W.P., and D.J. Gaffen, 1993: Effects of conversion algorithms on reported upper air dewpoint depression. Bull. Amer. Meteor. Soc., 74, 1323-1325.
Eskridge, R.E., O.A. Alduchov, I.V. Chernykh, Z. Panmao, A.C. Polansky, and S.R. Doty, 1995: A Comprehensive Aerological Research Data Set (CARDS): Rough and systematic errors. Bull. Amer. Meteor. Soc., 76, 1759-1775.
Gaffen, D.J., 1993: Historical changes in radiosonde instruments and practices, WMO/TD-No. 541, Instruments and Observing Methods Report No. 50, World Meteorological Organization, Geneva, 123 pp.
Gaffen, D.J., 1994: Temporal inhomogeneities in radiosonde temperature records. J. Geophys. Res., 99, 3667-3676.
Gaffen, D.J., 1996: A digitized metadata set of global upper-air station histories, NOAA Technical Memorandum ERL-ARL 211, Silver Spring, MD, 38 pp.
Gaffen, D.J., A. Robock, and W.P. Elliott, 1992: Annual cycles of tropospheric water vapor. J. Geophys. Res., 97, 18185-18193.
Gaffen, D.J., M.A. Sargent, R.E. Habermann, and J.R. Lanzante, Sensitivity of tropospheric and stratospheric temperature trends to radiosonde data quality, in preparation for J. Climate.
Gutzler, D., 1996: Low-frequency ocean-atmosphere variability across the tropical western Pacific. J. Atmos. Sci., 53, 2773-2785.
Kahl, J.D., M.C. Serreze, S. Shiotani, S.M. Skony, and R.C. Schnell, 1992. In-situ meteorological sounding archives for Arctic studies. Bull. Amer. Meteor. Soc., 73, 1824-1830.
Koshelkov, Yu. P., and G.R. Zakharov, 1998: On temperature trends in the Arctic lower stratosphere, Meteorologia i Gidrologia, 5, 45-54.
Labitzke, K., and H. van Loon, 1995: A note on the distribution of trends below 10 hPa: The extratropical northern hemisphere. J. Meteor. Soc. Japan, 73, 883-889.
Lanzante, J.R., 1996: Resistant, robust and non-parametric techniques for the analysis of climate data: Theory and examples, including applications to historical radiosonde station data. Intl. J. Climatol., 16, 1197-1226.
Luers, J.K., and R.E. Eskridge, 1998: Use of radiosonde temperature data in climate studies. J. Climate, 11, 1002-1019.
McMillin, L.M., M. Uddstrom, A. Coletti, 1992: A procedure for correcting radiosonde reports for radiation errors. J. Atmos. Oceanic Tech., 9, 801-811.
Oort, A.H., and H. Liu, 1993: Upper-air temperature trends over the globe, 1956-1989. J. Climate, 6, 292-307.
Parker, D.E., and D.I. Cox, 1995: Towards a consistent global climatological rawinsonde data-base, Intl. J. Climatology, 15, 473-496.
Parker, D.E., M. Gordon, D.P.N. Cullum, D.M.H. Sexton, C.K. Folland, and N. Rayner, 1997: A new global gridded radiosonde temperature data base and recent temperature trends. Geophys. Res. Lett., 24, 1499-1502.
Pawson, S., and B. Naujokat, 1997: Trends in daily wintertime temperatures in the northern stratosphere, Geophys. Res. Lett., 24, 575-578.
Pawson, S., K. Labitzke, and S. Leder, 1998: Stepwise changes in stratospheric temperature. Geophys. Res. Lett., 25, 2157-2160.
Peixoto, J.P., and A.H. Oort, 1992: Physics of Climate, American Institute of Physics, New York, 520 pp.
Peixoto, J.P., and A.H. Oort, 1996: The climatology of relative humidity in the atmosphere. J. Climate, 9, 3443-3463.
Reid, G.C., and K.S. Gage, 1981: On the annual variation in height of the tropical tropopause. J. Atmos. Sci., 38, 1928-1938.
Ross, R.J., and W.P. Elliott, 1996: Tropospheric water vapor climatology and trends over North America: 1973-93. J. Climate, 9, 3561-3574.
Rossi, V., 1973: On the development of experimental aerology in Finland. Vaisala News, 58, Ref. No. B 227.
Santer, B.D., J.J. Hnilo, T.M.L. Wigley, J.S. Boyle, C. Doutriaux, M. Fiorino, D.E. Parker, and K.E. Taylor, 1998: Uncertainties in 'observational' estimates of temperature change in the free atmosphere. J. Geophys. Res., submitted.
Schmidlin, F.J., and A. Ivanov, 1998: Radiosonde relative humidity sensor performance: The WMO intercomparison-Sept. 1995. Preprint volume. 10th Symposium on Meteorological Observations and Instrumentation, American Meteor. Soc., 68-71.
Suzuki, S., and M. Asahi, 1978: Influence of solar radiation on the temperature measurement before and after the change of the length of suspension used for the Japanese radiosonde observation, J. Met. Soc. Japan, 56, 61-64.
Teweles, S., and F.G. Finger, 1960: Reduction of diurnal variation in the reported temperatures and heights of stratospheric constant pressure surfaces, J. Meteorol., 17, 177-194.
Tiefenau, H.K.E., and A. Gebbeken, 1989: Influence of meteorological balloons on temperature measurements with radiosondes: Nighttime cooling and daylight heating. J. Atmos. Oceanic Tech., 6, 36-42.
Wallis, T.W.R., 1998: A subset of core stations from the Comprehensive Aerological Reference Dataset (CARDS), J. Climate, 11, 272-282.
WMO, 1996: Measurements of upper air pressure, temperature, and humidity. Guide to Meteorological Instruments and Methods of Observation, Chapter 12, WMO-No. 8, Sixth edition, WMO, Geneva.
Zaitseva, N., 1993: Historical developments in radiosonde systems in the former Soviet Union. Bull. Amer. Meteor. Soc., 74, 1893-1900.