Abstract

Ten of the most common radiosondes used throughout the world since 1960 have been evaluated concerning potential use of their temperature data for climate studies. The VIZ; Space Data Corp.; Chinese GZZ; Japanese RS2-80; Russian RKZ, MARS, and A-22; and Vaisala RS80, RS 12/15, and RS18/21 radiosondes were evaluated by modeling the temperature of the sensing element relative to the temperature of the air in which it is immersed. The difference, designated as the temperature error, was calculated under various environmental conditions. Validation and sensitivity analysis studies were performed on each radiosonde model as a means of estimating the environmental parameters that influence the temperature error and the resulting accuracy of the day and nighttime temperature profiles. Environmental parameters to which some sondes were sensitive include cloud cover, surface temperature, solar angle, ambient temperature profile, blackbody temperature, and the ventilation velocity. The ventilation velocity was found to depend strongly on the position of the sensor in the balloon wake. It is believed that the results of these analyses provide the best guidelines available to anyone wishing to perform climate studies using radiosonde data.

The research work presented in this paper indicates that climate trends can currently be estimated with a subset of the worldwide upper-air data. Trends can be calculated for monthly averaged, nighttime soundings with some confidence for the Vaisala RS80 (models not using the RSN80 and RSN86 corrections), Vaisala RS 12/15, Vaisala RS 18/21, Chinese GZZ (below 25 km), Russian RKZ, Russian MARS, and Russian A-22 (below 20 km) radiosonde models. The analysis presented in this paper shows that all of the above radiosondes have small errors in individual radiosonde soundings at night (< ±1°C) and the errors of the monthly averaged data are estimated to be less than ±0.5°C, except for the A-22 (±0.8°C). In addition, temperature data from the Japanese RS-2-80, the Russian A-22 above 20 km, Vaisala RS80 (RSN80 and RSN86 corrections applied), and VIZ can be made suitable for climate analysis if the appropriate temperature correction models are used to correct the data.

1. Introduction

Upper-air measurements of temperature, pressure, humidity, and winds have been made by balloon borne radiosonde sensors for over 50 years. These data provide a unique set of observations from which long-term changes in the earth’s atmosphere, on the order of decades, may be established. Difficulties in establishing and interpreting this dataset arise as a result of the numerous models of radiosondes that have been used throughout the world to make these measurements. These instruments have different error and response characteristics, have undergone modifications and improvements, and have insufficient documentation that establishes when the changes occurred, though Gaffen (1993) and Oakley (1993) have provided much needed information. The measurements were made to fulfill a need for understanding and predicting present weather, with little foresight into a future possible use as a climatic trend indicator.

Stratospheric temperature data, measured from radiosonde instruments, are expected to be a first-order indicator of atmospheric heating and circulation changes. A long-term change on the order of 1°C, or less, can be an indicator of significant atmospheric changes. Thus the temperature data extracted from radiosonde instruments must be precise, at least in the mean, have a high level of accuracy, and be without significant bias. Furthermore, the data collected from different instruments must be compatible to avoid spurious discontinuities between areas utilizing different radiosondes.

On the brighter side, however, is the fact that only a few radiosonde instruments were used in making a very large percentage of the world’s total radiosonde measurements and that sites from which they were launched span all of the continents and some ocean areas. The United States’ VIZ, the Finnish Vaisala, and the Russian, Chinese, and Japanese instruments provide a potentially stable subset of observations. With these instruments a reasonable amount of information is available about sonde performance, documentation, and modification chronology, from which long-term global temperature trends may be deduced.

a. Need for temperature corrections

The measurement of atmospheric temperature by every radiosonde instrument is affected by heating from sources other than the air itself. Solar and infrared radiation, heat conduction to the temperature sensor from its attachment points, and infrared radiation emitted by the sensor are heat sources or sinks that make the temperature of the sensor different from that of the air in which it is embedded. Thus a correction is required on every radiosonde instrument to deduce the atmospheric temperature from the temperature of the radiosonde sensor. For most radiosondes this correction is small at altitudes below 15 km but can become quite significant (several degrees) at altitudes between 20 and 30 km. For many of the early radiosondes, the recommended temperature correction was on the order of 3°–10°C (McInturff and Finger 1968), while for most of today’s instruments the stipulated correction is usually less than 3°C. The temperature error, however, varies from day to day and with geographic region, depending on solar angle, cloud cover, ground temperature, balloon rise rate, and other parameters. Another potential source of error in many radiosonde systems is the length of the tether that attaches the radiosonde to the balloon. As the balloon rises in the atmosphere it expands in size reaching a diameter of 5–10 m at 30 km. If the tether is not of adequate length, the radiosonde sensor, at high altitudes, comes into the wake recirculation region. In this recirculation zone, the ventilation air velocity over the sensor may be considerably less than the balloon rise rate, thus decreasing the convective cooling (see the appendix). None of the correction tables used with today’s radiosondes take into account all of these possible sources of error. Thus, to produce a set of compatible stratospheric temperature data from radiosonde instruments requires the development of a correction technique for each radiosonde that takes into account those parameters that significantly affect the temperature of the radiosonde sensor.

Differences between air temperature and the temperature of the sensor are not the only source of discrepancies. As instrument manufacturers have constantly improved, modified, and automated their instruments through the years, changes in hardware and software have been made that affect the temperature measurements. The noise level in the transmitted temperature signal, data processing used to convert the sensor’s mechanical or electronic response to a temperature, and upgrading of temperature correction tables as a result of international radiosonde intercomparisons are additional sources of temperature discrepancies that must be addressed.

b. Worldwide radiosondes

The Gaffen (1996) station history database provides information concerning the type of radiosonde and the temperature, humidity, and pressure sensing elements used by the various countries throughout the world. Table 1 summarizes the temperature sensing elements used in ducted and unducted models by major radiosonde manufactures since about 1955. Ducted sondes house the temperature sensor within a double duct, in which the outer duct shields the element from direct solar radiation. Temperature sensing elements in all ducted sondes, except for the early U.S. Bendix, consist of a bimetallic element that mechanically moves a pointer as the temperature changes. Unducted sondes utilize a sensor that has a solar reflective coating designed to minimize heating of the element. The sensors measure temperature by the change in resistance or capacitance of the element.

Table 1.

Major world radiosondes/temperature sensors. Bolding identifies those used in this study.

Major world radiosondes/temperature sensors. Bolding identifies those used in this study.
Major world radiosondes/temperature sensors. Bolding identifies those used in this study.

c. Comprehensive Aerological Reference Data Set

The National Climatic Data Center and the All-Russian Research Institute of Hydro Meteorological Information have initiated a joint project, Comprehensive Aerological Reference Data Set (CARDS). The goal of the CARDS project is to produce an upper-air dataset based on radiosonde observations, suitable for evaluating climate models and detecting global change (Eskridge et al. 1995). The CARDS project has produced a long-term (1948–94) daily dataset of upper-air observations for various station locations throughout the world.

2. Development of temperature correction models

Within the constraints of the CARDS program it was not possible to develop temperature correction models for all of the radiosondes listed in Table 1. A subset of radiosondes widely used after 1960 was chosen for study, and these radiosondes are highlighted in bold print. The model developed for each of these radiosondes is described in Luers (1996). This report and each of the temperature correction models are available through the National Climatic Data Center in Asheville, North Carolina. It is planned that in the future, temperature correction models will also be developed for the other major radiosondes listed in Table 1.

In developing and validating a radiosonde temperature correction model, a standard approach was used. The heat transfer processes that significantly influence the temperature of the sensor were determined based upon the appropriate physical principles. A temperature correction model was then developed that simulated the significant heat transfer processes, the object being to calculate the difference in temperature between air and sensor. Some of the significant parameters that are required for the heat transfer terms were unknown and had to be estimated. These include the balloon rise rate, the convective heat transfer coefficient for irregularly shaped temperature sensors; and for ducted sondes, the ventilation velocity through each duct. When these are not known they become model fitting parameters whose values are determined to provide consistent results with other sensor data, or with accepted reference data. A sensitivity analysis was performed using the temperature correction model to establish those sonde and environmental parameters that are in a constant state of flux and that significantly influence the temperature error (e.g., cloud cover, surface temperature, solar angle, balloon rise rate, etc.). In the sensitivity analysis the individual variables that influence the error in the radiosonde temperature sensor are perturbed about the reference values with a magnitude sufficient to include the extreme events that occur. The reference dataset is chosen to represent a moderate atmospheric environment so that the extreme atmospheric conditions that occur in nature will be on either side of the reference profiles. The reference dataset consists of the midlatitude spring standard atmosphere profiles of temperature, pressure, and trace constituents, a surface temperature of 280 K, clear sky, the LOWTRAN7 (Kneizys et al. 1988) rural 5-km visibility tropospheric aerosol profile, background stratospheric aerosol profile, vegetation ground cover, and a balloon rise of 5 m s−1. In the sensitivity analysis, the midlatitude spring reference temperature profile, for example, is varied by ±20°C at all altitudes to encompass the range of upper-air temperature profiles likely to occur in nature. Since the error in the radiosonde temperature at a given altitude is (except for the lag component) dependent only on the environmental conditions at that altitude, it is not necessary to vary the shape of the vertical profiles in the sensitivity analysis. Thus, the sensitivity analysis need only use the limiting range of each variable that occurs in nature. The one variable that does relate to the shape of a profile is the temperature gradient that influences the lag term. This parameter is adjusted in the sensitivity analysis through varying the balloon rise rate, which reflects both the influence of a changing temperature gradient and a change in the convective heat transfer to the sensor. Thus the sensitivity analysis is performed using a reference dataset and varying the sensitivity analysis parameters to include extreme environmental conditions so that nearly any atmospheric/environmental profile that occurs in nature will fall within the range of the sensitivity analysis.

Two additional comments should be made about the sensitivity analysis. The first is that the LOWTRAN 7 atmospheric transmission model used to specify the radiation environment allows for many other input parameters to be varied that have not been included in our discussion of the sensitivity analysis. Minor constituents such as ozone, carbon dioxide, stratospheric and marine aerosol profiles, surface vegetation, etc., all can be varied in the LOWTRAN 7 input and their influence on the radiation environment determined. This assessment, however, has already been performed in earlier analysis of the VIZ and RS80 radiosondes with no discernible influence on any temperature sensor. Thus the sensitivity analysis parameters addressed in this research are only those parameters for which some influence on the radiosonde temperature error is expected. A second note is that the sensitivity parameters are essentially independent of one another and influence the temperature error in an approximate linear manner. This was established in the development of a linear regression model to estimate the temperature error of the VIZ thermistor as a linear combination of sensitivity analysis parameters (Luers 1995). Thus, varying of each sensitivity analysis parameter individually, while the remaining parameters retain their reference profile values, is a valid method of isolating an individual influence on the temperature error of the radiosonde sensor.

To assess the validity of each model, two sources of data were used: international radiosonde temperature comparisons, and statistical day/night differences. If the sonde has been flown in international radiosonde intercomparisons experiments (where many radiosondes were flown on the same balloon), then comparisons with reference sonde data were used to validate the model. If validation data were not available, as was often the case with older radiosondes, then the model was tuned to provide temperature corrections that were in general agreement with the manufacturer-supplied temperature correction tables over the solar elevation range in which the correction tables are thought to be most accurate. Published reports on the mean day/night differences for various radiosondes (McInturff and Finger 1968; McInturff et al. 1979; Spackman 1978) were then used to qualitatively assess the accuracy of the manufacturer’s correction tables. This was confirmed (e.g., section 6c) by comparing the measured day/night differences with those predicted by the temperature correction models. Even though the models tuned in this manner cannot be expected to provide true temperature corrections, they do provide useful information concerning the sensitivity to environmental parameters such as solar angle, cloud cover, etc. From the sensitivity analysis results, guidelines are provided in Table 2 concerning the usefulness of the sondes’ data for climate studies, and how/if the data may be improved by applying corrections. We recommend whether temperature correction models can, or should, be used to adjust the profiles of night, day, or all flights, and discuss the resulting accuracy and reliability of the adjusted (corrected) profiles. Estimates are made of the accuracy achievable in corrected individual daytime and nighttime profiles, as well as the accuracy, in the mean, of a sequence of profiles averaged over a period of one month. These accuracy estimates, however, are not statistically derived. They are best estimates based upon the sensitivity of the radiosonde temperature error to changes in environmental parameters, the bias that may exist in estimating these parameters, and the expected variability of the environment over a period of 1 month. A general description of the temperature correction models developed for the various radiosondes is provided in the following paragraphs.

Table 2.

Summary of the findings and conclusions for the various radiosondes modeled. Here, α is the solar absorptivity, ε is the emissivity, se is the solar elevation angle, and M indicates marginal sensitivity.

Summary of the findings and conclusions for the various radiosondes modeled. Here, α is the solar absorptivity, ε is the emissivity, se is the solar elevation angle, and M indicates marginal sensitivity.
Summary of the findings and conclusions for the various radiosondes modeled. Here, α is the solar absorptivity, ε is the emissivity, se is the solar elevation angle, and M indicates marginal sensitivity.

The following description provides the basis for understanding the heat transfer processes that affect the temperature of the sensors in ducted and unducted radiosondes. The reader is referred to Luers (1989) for complete details.

a. Genealogy of model development

The parent temperature correction model, from which all other models of unducted sondes were developed, was the VIZCOR model for the U.S. VIZ rod thermistor. The equations that describe heat transfer processes affecting the temperature of the VIZ rod thermistor were developed in Luers’ dissertation (1989). This is the basic technical reference document for all temperature correction models for unducted radiosondes. In adapting the VIZCOR to other unducted radiosondes, the modules containing sensor dimensions, geometry, and thermal properties were modified, while the basic method of solving the heat balance equations remains unchanged.

b. Temperature correction models for unducted sondes

Each temperature correction model requires as input the sensor dimensions, the radiosonde temperature and humidity profiles, and balloon rise rate, along with independent estimates of the amount of radiation the sensor receives from the environment. Radiation estimates are derived from the Air Force’s LOWTRAN 7 atmospheric transmission code (Kneizys et al. 1988) with appropriate input values to define surface and atmospheric conditions. LOWTRAN 7 input parameters include surface temperature, solar elevation, ground albedo, cloud cover, aerosols, and water vapor.

c. Temperature correction models for ducted sondes

The parent temperature correction model for ducted sondes was the RS18COR model, developed for the Vaisala RS 18 radiosonde. This model has the same high-level structure as VIZCOR, but heat transfer processes significant to ducted sondes differ fundamentally from unducted sondes. For ducted sondes, the heat source that most influences the sensor is the temperature of the inside surface of the exterior duct. Heat is transferred to the ventilation air that passes through the duct and over the sensor. Thus the temperature correction for ducted sondes depends on solving the heat balance equation for the duct walls and the ventilation air passing through the ducts. The models include heat balance equations for both walls of the exterior duct, conduction through the duct walls and from the inside duct wall to the ventilation air, and convective and radiative transfer to the sensor. The heat transfer equations used for each ducted sonde were derived by Luers (1996).

3. U.S. radiosondes

Only two types of radiosondes were used by the United States in the time period from before 1950 until 1988. Before 1953 the U.S. Navy, and the U.S. Weather Bureau (until the early 1960s), used a ducted sonde with an uncoated thermistor to measure atmospheric temperature. After that time all U.S. military and Weather Bureau radiosondes were of the unducted type with a white lead carbonate coating applied to the thermistor (ML-405) to reflect solar radiation (M. Friedman 1992, personal communication). The Air Force uses a thermistor (ML-419) of slightly smaller dimensions with the same coating. Although many model changes and design improvements have been made on VIZ radiosondes over the years, the thermistor on today’s VIZ radiosondes is the same as that on the original unducted sondes. The lead carbonate used for the coating has been purchased from the same Kentucky mine since before 1950. Thus the absorption and emission properties of this coating are assumed not to have changed over the years.

In 1988, the U.S. National Weather Service (NWS) introduced the Space Data (SDD) sonde at 17 stations. This sonde uses a chip thermistor with a different solar reflective coating to measure temperature. The SDD sonde was phased out of NWS observations in 1995.

a. The VIZ radiosonde

The VIZ radiosonde uses a rod thermistor whose white coating has a solar absorptivity of α = 0.15 and an emissivity of ɛ = 0.86. Thus, the thermistor is a good reflector of solar radiation (85%) but also a strong absorber and emitter (86%) of infrared radiation. The thermistor is attached to a mount extending outward and above the body of the radiosonde by the thermistor lead wires (wires which lead to the sensor), which are of small diameter and sufficiently long for only a small amount of heat to be conducted between thermistor and mount.

A sensitivity analysis was performed with VIZCOR to establish how the temperature error of the VIZ sonde varies with environmental parameters. Nighttime and daytime reference datasets of the environmental parameters (background temperature, etc.) with the best-estimate radiosonde parameters (emissivities, balloon rise rate, etc.) were used to establish baseline profiles of the temperature error versus altitude, as shown in Fig. 1. Each parameter in the reference set was then varied and the VIZCOR model rerun to determine the influence of that parameter on the temperature error. Variables tested included the background temperature, pressure, humidity, ozone and aerosol profiles, cloud cover, cirrus clouds, solar angle, surface temperature, albedo, and balloon rise rate. A detailed discussion of the sensitivity analysis for each parameter is given in Luers (1996). These results are summarized in Fig. 1 for solar angle and in Table 2 for all other variables. Only those parameters that are listed in Table 2 were found to significantly or marginally (M) influence the temperature error. The temperature error of the VIZ sonde is significantly influenced by solar angle, background temperature profile, and cloud cover. Each of these parameters can result in a temperature error up to 1°C. The marginally sensitive parameters, surface temperature, and lag error, can cause a temperature error of 0.2°–0.3°C. Other parameters such as aerosols, humidity, albedo, etc., produce errors less than 0.2°C.

Fig. 1.

VIZ radiosonde day and night temperature errors.

Fig. 1.

VIZ radiosonde day and night temperature errors.

1) Validation of VIZCOR

VIZ radiosonde measurements collected by the NWS and Department of Defense have not been corrected for temperature errors. Archived data contain radiation errors in both daytime and nighttime soundings. The mean day/night difference in U.S. synoptic radiosonde temperature observations has been calculated by McInturff and Finger (1968) and McInturff et al. (1979) for the VIZ radiosonde. At 10 mb (≅30 km) this difference is about 2°–3°C, which is consistent with the VIZCOR model predictions of a positive 1°C heating error during the day and a negative 1°–2°C cooling error at night. The VIZCOR model has also been validated by comparing its predicted temperature error to that actually measured by the NASA multithermistor reference radiosonde. The results of this analysis (Luers and Eskridge 1995) verify an accuracy of ±0.3°C for the VIZCOR model.

2) Application of VIZ temperature data for climate studies

To use VIZ radiosonde measurements for climate studies, archived temperature data should be corrected to make them compatible with other radiosonde observations. The VIZCOR model can be utilized to apply the temperature corrections provided the appropriate environmental conditions are known for each flight. The most difficult of these to establish from historical radiosonde flights are the cloud conditions. The other parameters of influence, background and surface temperature, and solar elevation are available information. The cloud cover model developed by Chernykh and Eskridge (1996) employs the sounding to generate realistic cloud cover data that can be used as input to the VIZCOR model. It is recommended that this cloud model, along with the available surface observation for each flight, be used with VIZCOR to correct all VIZ radiosonde soundings that are to be used in the CARDS database. The resulting accuracy for properly corrected profiles should be pointwise ±1.0°C and ±0.5°C for the monthly mean of a series of profiles. This accuracy, of course, refers only to the error attributable to the difference between the temperature of the thermistor and that of the air. Other error sources, such as calibration, electronic noise, signal detection, etc., are not included in the estimate. The CARDS program is utilizing statistical methods to detect and remove these other error sources. These results and recommendations for the VIZ radiosonde are summarized, along with those from other radiosondes, in Table 2.

b. The Space Data radiosonde (SDD)

The SDD has been used by the NWS at selected locations since approximately 1988. By 1993 17 NWS stations utilized this instrument. However, it has now been phased out of operation. There were two different thermistors used on SDDs, a chip and a bead. The chip thermistor was primarily used by the NWS, while the bead was used by the military. Of primary importance in the CARDS database are the NWS flights.

The chip thermistor is mounted on a beam extending several inches above and to the side of the sonde housing, is imbedded in epoxy, and is coated with a white solar reflective coating. The coating has an emissivity of 0.85 and a solar absorptivity of 0.07, making it sensitive to infrared radiation but less sensitive than the VIZ thermistor to solar radiation. The shape of the coated thermistor is approximately spherical so its exposure to solar radiation is largely independent of solar angle.

However, it has short, highly conductive lead wires that are attached by solder joints on either side of the boom. Exposure of the solder joints to solar radiation, and to regions of separated airflow, makes their temperature considerably higher than that of the air during daytime soundings. This allows significant heating of the thermistor from conduction through the lead wires. In developing the SDCOR temperature correction model, it was not possible to analytically determine an accurate convective heat transfer coefficient due to the separated flow in which one of the solder joints is always embedded. Also, the geometry of the solder joints makes calculation of the area exposed to solar radiation nonprecise. Thus the convection heat transfer coefficient, emissivity, and solar absorptivity of the solder joints are fitting parameters to the model.

1) Validation of SDCOR

Temperature data from dual radiosonde balloon flights containing both an SDD and a VIZ radiosonde were used to validate the SDCOR model (NWS 1992). The SDD temperature measurements were found to be contaminated by conduction from the solder joints. When agreement did not exist between corrected temperatures from the SDD and VIZ thermistors, modifications to the fitting parameters were made in an attempt to bring the temperatures into agreement. The best fitting parameters that could be found for the SDCOR model, to account for the conduction problem, still resulted in differences between 0.5° and 1.0°C from corrected VIZ profiles.

In all flights, a temperature bias of about 0.5°C existed below 10 km, with the SDD being colder. Figure 2 shows the temperature error for the SDD for day and night flights using the SDCOR model with optimum fitting parameters. The negative bias of the SDD relative to VIZ below 10 km is not evident in Figs. 1 and 2—that is, it is not predicted by the models.

Fig. 2.

SDD day and night temperature errors.

Fig. 2.

SDD day and night temperature errors.

2) Application of SDD temperature data to climate studies

Temperature corrections have not been applied to archived SDD soundings. Since the temperature correction model SDCOR has been unable to generate corrections that agree with VIZ measurements, applying SDCOR-generated corrections is of questionable benefit. Nevertheless, because the magnitude of the error may be several degrees, applying SDCOR corrections is recommended. However, unless some additional factor can be determined that explains the discrepancies between SDD and VIZ temperature data, it is questionable whether SDD temperature profiles should be used in climate studies. Exclusion of the relatively small amount of SDD soundings from a climate analysis should not seriously degrade the geographic and temporal density of available data.

4. The VAISALA RS80 radiosonde

Vaisala radiosondes have been used at many locations worldwide for upper-air measurements since about 1940. Early ducted Vaisala radiosondes, RS 12/13/15 (Lindquist 1963), used a bimetal strip to sense the temperature of the atmosphere. The ducted RS 18 and RS 21 sondes introduced after 1973 used a bimetal ring as a replacement for the strip as a means of reducing the necessary radiation correction (Hörhammer 1971). These radiosondes were used until 1981. The unducted RS80 was introduced with a capacitive bead temperature sensor (Vaisala 1992). The RS80 radiosonde is used in many countries. It is the most widely used of all radiosondes.

In the Vaisala RS80 the bead, lead wires, and mount all have an aluminum coating that has a very low emissivity of ɛ = 0.02 and solar absorptivity of α = 0.15. The shape of the bead sensor is quasi ellipsoidal, making its area exposed to sunlight less variable with solar angle than the cylindrical VIZ thermistor. Because of the low emissivity of the aluminized sensor its temperature is virtually unaffected by infrared radiation. Consequently, the sensor is not affected by ground temperature or clouds. Also, because of its low emissivity it loses negligible heat, making it an excellent nighttime temperature sensor. During the daytime, however, solar energy that is absorbed cannot be dissipated efficiently by emissions. The sensor must dissipate this heat via convection to the ambient air. The sensor temperature is, therefore, somewhat sensitive to balloon rise rate. A tether length of 15 m effectively removes the sensor from most of the balloon wake effect.

a. Validation of VAICOR

The temperature correction model VAICOR, developed for the Vaisala RS80 radiosonde, has been validated by comparisons with the NASA multithermistor sonde and corrected VIZ radiosonde data (Luers and Eskridge 1995). The model has an accuracy equivalent to that of VIZCOR.

b. Sensitivity

A sensitivity analysis was performed on the RS80 sonde with VAICOR, using the same reference datasets as used for the VIZ sonde. The results are shown in Table 2 and Fig. 3. Because of the low emissivity coating on the RS80 temperature sensor, it is insensitive to sources of infrared radiation such as surface temperature, cloud cover, and the background temperature profile. The only source of nighttime temperature error is a small lag error of the sensor. The major error source in daytime flights is from solar radiation. Cloud cover and balloon rise rate have a marginal effect.

Fig. 3.

Vaisala RS80 radiosonde day and night temperature errors.

Fig. 3.

Vaisala RS80 radiosonde day and night temperature errors.

c. Application of RS80 temperature data to climate studies

The technique developed by Vaisala for processing RS80 data includes an algorithm for correcting both day and night temperatures. This consists of a correction table and a scaling parameter called the ventilation factor, which is used as a multiplier for table values. The ventilation factor accounts for changes in balloon rise rate. Most stations that have flown RS80 radiosondes utilize Vaisala software that applies these temperature corrections. The situation is complicated by the fact that the temperature correction table provided by Vaisala has changed several times since its inception in the early 1980s. The original correction table RSN80 was superseded by RSN86 in 1986 and recently replaced by RSN93. Radiosonde sites using the RS80 may or may not have adopted RSN86 and/or RSN93. RSN86 decreased the nighttime correction by 0.2°–0.7°C and changed the daytime value by up to ±0.5°C, making the maximum correction at a solar angle of ∼25°. RSN93 removed nighttime corrections altogether and made a slight modification for day flights when solar angles are low. The RSN93 table provides results in agreement with VAICOR, to ±0.2°C in nearly every table entry. It is possible to determine when and if a change in the correction table was implemented at a given site by comparing the RS80 data with satellite microwave sounding units, or by comparing the RS80 data with radiosonde data from non-Vaisala geographic neighbors.

Because RS80 temperature data has been processed using three different correction tables (since 1981), these data (except for those processed with the RSN93 table) in their present form are not amenable to climatic studies. Ideally it would be desirable to uncorrect each RS80 temperature observation and then apply VAICOR to generate accurate temperature profiles. However, in order to first uncorrect the data it is necessary to know the ventilation factor k used to scale correction table values. This ventilation factor depends on the rise rate, which is not a recoverable parameter in most archived data. Therefore, an estimated value must be used to remove RSN80 or RSN86 corrections and apply VAICOR. It is essential that all Vaisala data derived with RSN80 or RSN86 corrections be recorrected to be compatible with other data for climatic analysis. Otherwise fictitious temperature changes of 1 K or more will be introduced into the climatic time series.

5. Japanese RS2-80

The Japan Meteorological Agency has used two types of radiosonde for operational soundings from 1956 to the present (Nakamura et al. 1983). The early RSII-56 radiosonde was a double-ducted sonde using a bimetallic sensor. A single rectangular funnel-shaped duct was used to channel the airflow through 90° before passing over the sensor. The sensor was shaped in the form of a half cylinder with the airflow traversing along the axis of the cylinder. Although a temperature correction model has not been developed for this radiosonde it is a strong candidate for future model development and analysis. Considerable data have been collected with this sonde that would combine with data from the newer RS2-80 sonde to provide time series of nearly 40 yr at many Pacific stations.

With the worldwide demise of ducted sondes, Japan switched, in March of 1981, to the RS2-80 radiosonde, which has an exterior mounted rod thermistor. The cylindrical rod thermistor is mounted with its axis in the horizontal plane and attached by lead wires to terminals on a mount extending outward from the radiosonde body. The long thermistor mount makes the instrument insensitive to radiation emitted and reflected from the radiosonde body. Lead wires and thermistor have a white surface coating to provide high solar reflectance (α = 0.18). The coating, however, has high emissivity (ɛ = 0.84), making it sensitive to infrared radiation. A 15-m tether is used to attach radiosonde to balloon.

The RS2-80 is similar to the U.S. VIZ radiosonde with respect to temperature sensing ability. Both sondes use a rod thermistor, with white solar reflective coating, but the Japanese thermistor is smaller than the VIZ. However, the lead wires are longer on the Japanese sonde, minimizing heat conduction from the terminals.

a. Sensitivity analysis

Results from the sensitivity analysis on the Japanese RS2-80 sonde are summarized in Table 2 and Fig. 4. These results are similar to those from the VIZ sonde. Solar angle, cloud cover, and the background temperature profile are the major factors that affect the temperature sensor error. Balloon rise rate has a marginal influence.

Fig. 4.

Japanese RS2-80 radiosonde day and night temperature errors.

Fig. 4.

Japanese RS2-80 radiosonde day and night temperature errors.

b. Validation of JAPCOR

A recent international intercomparison of the Japanese RS2-80 sonde with the Vaisala RS80 and VIZ radiosondes will provide the only dataset available to validate the JAPCOR model. Until these data are available, the validity of JAPCOR must be based on its similarity to VIZCOR. Both RS2-80 and VIZ temperature sensors are rod thermistor, with coatings that have similar absorption and emission properties. They are mounted in a similar fashion, with the rod axis horizontal. The development of the VIZCOR and JAPCOR required no fitting parameters and so the accuracy of JAPCOR and VIZCOR should be equivalent.

c. Application of RS2-80 temperature data to climate studies

The Japanese RS2-80 radiosonde is sensitive to solar and infrared radiation; consequently corrections are required for both day and night flights. Operational temperature corrections are applied to RS2-80 data as a function of solar angle and pressure. These corrections do not rectify errors due to background temperature, cloud cover, or surface conditions. Thus, even if these operational corrections provide reasonably accurate mean values, errors in individual soundings are estimated to be ±1.5°C and ±1.0°C for monthly mean values. For accurate ascent data each profile must be corrected for background and surface temperature, cirrus clouds, and solar elevation angle. These are the same parameters as required by VIZCOR.

To provide maximum accuracy in archived data, the corrections that have been applied should be removed, before using JAPCOR on a flight-by-flight basis. Removal of corrections that have been applied is possible since they are functions of solar angle and pressure, both known parameters. Properly corrected Japanese RS2-80 data can, therefore, be expected to form a high quality radiosonde temperature database. Its accuracy should be equivalent to that of corrected VIZ and Vaisala RS80 data. The above results are summarized in Table 2.

6. The Russian MARS and RKZ radiosondes

The RKZ-1a radiosonde was the first Russian model to use a rod thermistor as its temperature-sensing element. It came into operation in the early 1960s (Zaitseva 1990). The thermistor was mounted with 1-cm lead wires that all had the same white solar reflective coating. The lead wires were attached to metal fastening plates and their diameter (≅0.65 mm) was nearly one-third of the thermistor diameter. This permitted significant conductive heat transfer to the thermistor. The fastening plates appear to have been tin coated. A solder joint on the end of the fastening plate attached to a current carrying wire and was coated with white (yellowing) paint. On the later RKZ-2 through RKZ-5 sondes the temperature sensor mount was improved by increasing the lead wire length and attaching the ends to a rigid rectangular wire support structure. The RKZ-5 radiosonde shows this structure to be partially oxidized, thereby increasing the absorptivity and emissivity. RKZ-2 through RKZ-5 series of radiosondes were used in the Soviet Union from 1967 until 1990 and all incorporated the same thermistor and mount. Prior to, and along with the RKZ radiosondes, a ducted A-22 sonde with a spiral bimetal temperature sensor was used in the USSR. The USSR used a 10-m tether length with all radiosondes.

a. The RKZCOR model

The temperature-sensing element for the RKZ-2 through RKZ-5 sondes is a rod thermistor with a white solar reflective coating. It is approximately 1 cm long and 2 mm in diameter. It has emissivity ɛ = 0.84 and solar absorptivity α = 0.15. Thus the thermistor has essentially the same sensitivity to solar and infrared radiation as the VIZ and Japanese RS2-80 radiosondes. Lead wires, attached to each end of the thermistor are approximately 2.5 cm long and 0.65 mm in diameter. The thick wires easily conduct heat absorbed by the mount to the thermistor. Consequently thermistor temperature is strongly influenced by the thermal properties of the mount.

The lead wires appear to be copper with a tin coating, and the rectangular wire thermistor mount is attached horizontally to the top of the radiosonde. This mount is a darker color, indicating probable oxidation of the wire, and is a strong absorber of solar radiation. Henninger (1984) gives absorptivity measurements of copper coatings that vary from 0.26 for polished to 0.55 for tarnished and 0.92 when blackened. The emissivity of the mount was estimated as ɛ = 0.20, which is consistent with handbook values for tarnished copper. Solar absorptivity of the RKZ mount was a required parameter for RKZCOR. Its value was chosen so that the RKZCOR temperature corrections agreed with the operational RKZ table values over intermediate ranges of solar angle. The table is thought to be most accurate in this range. A table of operationally used temperature corrections as a function of solar angle is provided in Ivanov et al. (1991). A solar absorptivity value of α = 0.50 provided consistent results. If most operational sondes were flown soon after manufacture, the tarnishing problem may not have been severe. The derived value of α = 0.50 tends to support this conjecture.

b. Sensitivity analysis

The results from the sensitivity analysis of the RKZ sonde are summarized in Table 2 and Fig. 5. Even though the emissivity of the RKZ sensor is nearly the same as the VIZ and RS2-80 sensors, the nighttime temperature error is less. This occurs because the RKZ support mount has low emissivity and does not lose heat as rapidly as the sensor, hence it heats the thermistor. During daytime the mount absorbs more solar radiation than the thermistor, thereby increasing the temperature error.

Fig. 5.

Russian RKZ radiosonde day and night temperature errors.

Fig. 5.

Russian RKZ radiosonde day and night temperature errors.

c. Validation of the RKZCOR model

RKZCOR could not be validated by comparison with reference sondes flown on the same balloon because of the lack of available data. However, an assessment of consistency between RKZCOR correction tables and those used operationally was made with the aid of published statistics of day–night differences. Spackman (1978) computed statistical averages of day/night geopotential height differences as a function of solar elevation for the RKZ sonde. Spackman showed two figures, one with morning daylight data and the second for evening daylight. The day–night differences are small at altitudes below 50 mb indicating that the daytime and nighttime temperatures are compatible. As no operational corrections are applied to nighttime data one can conclude that the operational daytime temperature corrections are reasonably accurate in the mean. At higher altitudes, and especially at solar angles between 10° and 30°, day–night differences increase, indicating operationally corrected daytime observations are too warm. A comparison of RKZCOR corrections with the operational correction table tends to substantiate this conclusion. RKZCOR indicates the correction should be larger in regions of large day–night differences in geopotential height.

McInturff et al. (1979) published statistics of day–night temperature differences for the operationally corrected RKZ radiosonde data from the 1974–76 period. Mean day–night temperature differences are small, generally less than 0.5°C, at all altitudes, for solar elevations above 25°. At lower solar angles, and higher altitudes, the mean differences increase to 1°–2°C. These results are consistent with Spackman and substantiate the validity in the RKZCOR-generated solar angle correction tables.

d. Application of RKZ temperature data to climate studies

The RKZ rod thermistor is largely insensitive to infrared radiation and is a good nighttime temperature sensor. The nighttime error is less than 0.8°C at all altitudes, under nearly all environmental conditions, with no corrections applied. Consequently, in correcting historical profiles for climate studies, it is not necessary that corrections be applied to nighttime RKZ soundings. However, improvement could possibly be achieved by applying background temperature corrections, a major contributor to the remaining temperature error. The resulting accuracy of corrected data would be about ±0.5°C for any sounding, with even more accurate means. For daytime soundings, there is considerable error variability from flight to flight, due to environmental changes and oxidation of the wire mount. Statistically, the operationally corrected daytime data below 50 mb (or above 25° solar elevation) appears reasonably accurate, based on day–night temperature and geopotential height differences. The accuracy of the daytime temperature data is estimated to be ±1.0°C in the mean. These results are summarized in Table 2.

7. The Russian MARS radiosonde

The completely automated MARS 2-2 radiosonde, first used in 1971, was in wide use in the former USSR by 1984. This sonde uses the same temperature sensor and mount as the RKZ series. However, MARS sondes have a bright surface on the support mount, characteristic of a tin or silver coating. The absorption properties of the mount are estimated to be ɛ = 0.04 and α = 0.20. With the decreased solar and infrared emission properties of the mount, the thermistor becomes less sensitive to both infrared and solar radiation. A tether length of 10 m was used on all MARS radiosondes.

a. Sensitivity analysis

The results of the sensitivity analysis for the MARS sonde are summarized in Table 2 and Fig. 6. Because the support mount for the MARS sonde has a very low emissivity and low solar absorptivity, its temperature better approximates that of the air than does the thermistor. The mount serves as a sink/source for thermistor heat. This results in a smaller thermistor temperature error than found for the RKZ sonde in both daytime and nighttime conditions.

Fig. 6.

Russian MARS radiosonde day and night temperature errors.

Fig. 6.

Russian MARS radiosonde day and night temperature errors.

b. Validation of MRSCOR

MRSCOR was validated for night soundings by comparing MARS 2-2 operationally corrected temperatures with VIZCOR corrected VIZ temperatures from radiosonde flights in which these sensors were attached to the same balloon (Ivanov et al. 1991). The agreement was excellent: within 0.5°C at all altitudes and under all environmental conditions. MARS daytime soundings are corrected, but not nighttime soundings. However, the correction algorithm used by the Russians was not available. Thus day soundings could not be used for intercomparison.

c. Application of MARS temperature data to climate studies

The MARS thermistor is an excellent sensor of nighttime temperature. It is largely insensitive to infrared radiation, with the temperature error being less than 0.5°C at any altitude, under nearly all environmental conditions. No corrections are made to operational nighttime soundings, nor is it recommended that MRSCOR be used to correct archived night soundings. Uncorrected nighttime MARS soundings should provide high quality data for climate studies.

Daytime temperature errors for the MARS radiosonde are sensitive to solar angle, and marginally sensitive to the environmental influences of cirrus clouds, background temperature profile, and lapse rate.

Daytime soundings have operational corrections applied. The methodology used is unknown, therefore the usefulness of these corrected data is not clear. Since the nighttime data are of known accuracy, it is recommended that day–night differences be calculated for several stations, under a wide range of solar elevation angles, so as to establish the relationship between daytime and nighttime soundings. If mean day–night differences approach zero under all solar elevations, then the daytime corrections can be assumed valid and all daytime MARS soundings included in climate studies. A summary of the results for the MARS radiosonde is provided in Table 2.

8. Correction model for the Vaisala RS 18 and RS 21 radiosondes

The last Vaisala ducted radiosondes, the Vaisala RS 18 and RS 21, were used operationally in many countries from about 1973 through the early 1980s. These radiosondes replaced earlier ducted RS 12 and RS 15 versions used from 1960 to 1975. The RS 18 and RS 21 radiosondes are similar in design, with the RS 21 utilizing a humicap rather than a hair hygrometer. For temperature measurement, RS 18 and RS 21 radiosondes used the same bimetallic (Ni–Fe) ring-shaped sensor. It is enclosed within two cylindrical ducts that serve as a radiation shield. The outer cylinder is made of thin mylar plastic with the interior wall aluminum coated. The inside surface is then painted black but the aluminum coating remains visible through the mylar from outside the duct. This radiation shield provides high solar reflectivity from the aluminum surface and high infrared emissivity from the mylar. The outer radiation shield is highly absorbent to both solar and infrared radiation striking the inside blackened surface. The interior duct is made of aluminum to prevent infrared heating or cooling. The (Ni–Fe) cylindrical sensor is positioned along the axis of the duct, attached to the capacitor, and mounted with Invar struts that have low thermal conductivity. A 15-m tether length removes the sensor from most of the wake effect from the balloon.

In the RS 18 sonde, solar radiation heats the outside surface of the exterior cylindrical duct and emissions from the highly emissive mylar surface provide cooling. The blackened inside surface is only partially exposed to direct solar radiation but the exposed area increases with higher solar elevation. Conduction between inside and outside surfaces of the duct wall may be in either direction, depending primarily on solar angle. Ventilation air coming through the exterior cylindrical duct is heated by contact with the inside duct wall. In theory, this heated air should be within a thin boundary layer that does not pass through the interior duct, thereby shielding the temperature sensor. However, this is not always so. Since the radiosonde oscillates below the balloon, duct orientation relative to still air is constantly changing and results in an airflow that is not parallel to duct axes. In addition, airflow obstruction by temperature and humidity sensors and mounts also inhibits the flow. Thus flow through the ducts is retarded, with extensive mixing of heated air into the rest of the ventilation flow. How much the flow rate through the duct is retarded relative to the rise rate of the balloon is not known but is likely to be significant. Ventilation rate is considered a variable of fit for RS18COR. In the RS 18 sonde, the interior cylindrical duct is not heated in the same way as the exterior cylindrical duct and does not significantly affect the temperature of the sensor or ventilation air. The interior duct, however, reduces ventilation air velocity. The temperature sensor itself transfers heat efficiently and takes on the temperature of the ventilation air.

a. Sensitivity analysis

The temperature error profiles for the reference daytime and nighttime datasets are shown in Fig. 7. Temperature error increases dramatically with increased solar elevation. Because the temperature errors themselves are large, there is also an increased sensitivity to other parameters. Ducted sondes are highly sensitive to the ventilation velocity of the air passing through the duct and over the sensor. The ventilation velocity cannot be assumed to be the balloon rise rate since the axis of a swaying radiosonde is generally not aligned with the vertical velocity vector. In addition, a short tether length places the radiosonde in the recirculation zone beneath the balloon. The velocity deficit in a balloon wake at 30 km with a 15-m tether attached to a 7.5-m diameter balloon is estimated to be 24% (see appendix). The sensitivity analysis results indicate that a 50% change in ventilation velocity can change the temperature error by 1°C or more for nighttime flights and by several degrees in daytime. The ventilation velocity, including the influence due to balloon wake, becomes a fitting parameter in model validation. The temperature error, as summarized in Table 2, is also sensitive to the background temperature profile.

Fig. 7.

Vaisala RS 18/21 radiosonde day and night temperature errors.

Fig. 7.

Vaisala RS 18/21 radiosonde day and night temperature errors.

b. Validation of RS18COR

Direct validation of the temperature correction model for the RS 18/21 sondes has not been possible. No data have been found that link the RS 18 sonde with a sonde of established accuracy. Antikainen (1973) made comparisons between the RS 18 sonde and RS 16 sonde with several different solar angles. RS 16 sondes used a platinum wire sensor, which minimized radiation errors and served as a reference standard. An operational temperature correction table, RSRN-18, was developed by Vaisala for the RS 18 sonde based upon these comparisons and is valid for a balloon rise rate of 6 m s−1 and tether length of 15 m. RS18COR sensitivity analysis suggests that, if the rise rate varies from this value, a serious error in temperature would result.

A comparison was made between the Vaisala RSRN-18 correction table and results from RS18COR. Using the daytime set of reference parameters, the ventilation velocity was varied to establish an RS18COR table comparable to the RSRN-18 table. The best results were for a ventilation velocity of 80% of the ascent rate. Agreement between RS18COR and RSRN-18 is then generally good for both nighttime and daytime conditions.

Validity of the RSRN-18 table and RS18COR temperature corrections was assessed by using statistical averages of day–night differences in temperature and geopotential height. McInturff et al. (1979) calculated mean day–night temperature and height differences for the RS 18 sonde at a number of pressure levels and solar angles over a 19-month period between 1974 and 1976. The day–night differences at altitudes above 200 mb indicate day is warmer than night temperature by 1°C to nearly 3°C. Because RS18COR night correction agrees with the RSRN-18 table, it is likely that nighttime temperature is accurate in the mean. Daytime corrections that have been applied are too small. This underestimation may be due to a balloon rise rate less than 6 m s−1 or a tether length shorter than 15 m. The McInturff et al. analysis is substantiated by Spackman’s (1978) day–night geopotential height differences. The height difference is small, or negative, for solar elevations less than 100 but increases to differences of 50–100 m at heights above 50 mb, and solar elevations above 20°. This is consistent with differences calculated by McInturff et al. (1979) and indicates that daytime temperature corrections where applied may be several degrees too small above 20-km altitude.

c. Application of RS 18/21 temperature data to climate analysis

Sensitivity analysis results show that nighttime temperature error is small below 30 km, and the RS18COR corrections agree in the mean with the corrections that were operationally applied. Thus these data are high quality and valid in their present form for climate studies. Above 30 km, the error becomes more sensitive tobackground temperature and balloon rise rate (ventilation velocity); and RSRN-18 adjustments of the data may have been incorrect.

Daytime RS 18/21 data are highly sensitive to rise rate, background temperature, and solar angle. Temperature error may vary by several degrees, from flight to flight, because of these variables. RSRN-18 operational corrections that have already been applied to daytime soundings are not consistent with nighttime soundings and the daytime data are likely undercorrected by several degrees. It is not feasible to recorrect RS 18/21 daytime data with RS18COR because rise rate is not archived and even if it were, its relation to ventilation velocity is not well established. Thus, no additional corrections are recommended for RS 18/21 day soundings. For these reasons, it is not recommended that RS 18/21 day soundings be used in climate studies.

9. Correction model for the Vaisala RS 12/15 radiosondes

Vaisala RS 12 radiosondes were introduced in Finland about 1960 and, along with the RS 15 sonde, used by many European and Asian countries until the mid 1970s. These sondes were later replaced by Vaisala RS 18 sondes because of decreased solar radiation error (Hörhammer 1971).

Vaisala RS 12, RS 13, and RS 15 radiosondes are double-ducted sondes that utilize a bimetallic corrugated flat plate sensor to measure atmospheric temperature (Väisälä 1965). All use the same sensing element, which is rigidly attached to the sonde frame at its endpoints. A rod, attached to a movable arm, is bonded to the sensor at its center. As the temperature changes, the sensor bends, resulting in a motion of the arm that varies the capacitance of calibrated circuitry.

The sensor is shielded from sunlight by two rectangular ducts. It is in the inside duct. Both ducts are made of aluminum and blackened on the inside surface. The aluminum exterior surface acts as an effective reflector of both solar and infrared radiation, while the blackened interior surface is a strong absorber of both.

a. Sensitivity analysis

Results from a sensitivity analysis of the RS12 sonde are summarized in Fig. 8 and Table 2. The response of the RS12 sensor to rise rate and solar angle is similar to that of the RS18. The RS12 sonde, because of the low emissivity of the exterior surface of the outer duct, is not sensitive to the background temperature.

Fig. 8.

Vaisala RS 12/15 radiosonde day and night temperature errors.

Fig. 8.

Vaisala RS 12/15 radiosonde day and night temperature errors.

b. Validation of RS12COR

McInturff and Finger (1968) calculated average day–night temperature differences for operationally corrected Vaisala RS 12 radiosonde observations from 1964 to 1966, at 100-, 50-, 30-, and 10-mb levels. At each pressure the mean difference was near zero or slightly negative at low solar elevation angles (−5° to +10°), but increased to 0.5°–1.0°C for solar elevation greater than 45°. This small variability is consistent with actual diurnal change in the atmosphere (McInturff et al. 1979). A similar analysis of day–night differences in height was done by Spackman (1978) with RS 12 data from Finland in the 1974–76 period. His results were similar to those of McInturff and Finger (1968).

c. Application of RS 12/13/15 temperature data to climate studies

The RS12COR sensitivity analysis established that nighttime errors are small and have largely been corrected by the Vaisala tables. Since day–night differences are near zero, we conclude that daytime temperature corrections are also accurate in the mean. Individual day flights may be biased at higher altitudes because of variations in rise rate but, if averaged over time periods of a month or more, should provide high quality data for climatic studies. These results and recommendations are summarized in Table 2.

10. Temperature correction model for the Chinese GZZ sonde

The GZZ sonde is double ducted and utilizes a bimetal spiral sensor to measure temperature. The sensor is rigidly attached to the sonde frame at its endpoints. An insulating spacer is placed between the sensor and frame to minimize conduction from the radiosonde body. A rod, attached to a movable arm, is bonded to the spiral sensor at its center. As sensor temperature changes, the spiral moves and rotates the arm. The arm position is calibrated to record temperature.

The sensor is shielded from sunlight by two rectangular ducts. Similar to other ducted sondes, the sensor is located in the smaller inside duct. Both ducts are made of polished aluminum, an excellent reflector of solar and infrared radiation. Inside surfaces are, however, blackened to prevent solar radiation reflection onto the sensor. This radiosonde is similar in design to the Soviet A-22, RS 12, and RS 18 models; the major difference being duct and sensor shape (cylindrical vs rectangular). The GZZ radiosonde was introduced in the 1960s and is still the operational instrument. A tether of 10-m length is used.

a. Sensitivity analysis

Results of a sensitivity analysis for the GZZ sonde are summarized in Fig. 9 and Table 2. The GZZ sonde, which is similar to the RS18, exhibits similar error profiles and sensitivities. The GZZ temperature error for day flights is influenced by solar angle and ventilation velocity. A change in ventilation velocity of 50% can change the temperature error by 5°C or more at 30 km. The nighttime sensitivities show a maximum temperature error due to background temperature or rise rate of less than 1°C at 30 km.

Fig. 9.

Chinese GZZ radiosonde day and night temperature errors.

Fig. 9.

Chinese GZZ radiosonde day and night temperature errors.

b. Validation of GZZCOR

GZZCOR temperature error can be compared with operational temperature corrections presently applied to GZZ data. No operational corrections are made to night soundings. GZZCOR derived results, using the fitting parameter ventilation velocity equals 0.6 times rise rate, agree well with the operational table for solar elevation angles above 20°, and altitudes above 15 km. At altitudes below 15 km, the operational corrections are 0.5°C or more larger than GZZCOR suggests, and at solar elevations below 20°, operational table values may be several degrees larger than GZZCOR corrections.

Averages of day–night differences for the GZZ were calculated by McInturff et al. (1979) for 1974–76 and Spackman (1978) for 1975. Both studies have data only up to 100 mb (≅16 km), as higher levels were not transmitted. Day–night differences in height and temperature were small at all pressure levels. Since at 100 mb errors are less than 1.5°C, it is difficult to establish the validity of corrections at higher altitudes (where corrections may increase to several degrees).

c. Application of GZZ temperature data to climate studies

For nighttime soundings, GZZCOR indicates that temperature errors are small below 25 km and not very sensitive to environmental parameters. Thus, sensor measurement should represent the atmospheric temperature to within ±0.5°C. Above 25 km, variations in ventilation velocity and environmental temperatures can cause nighttime errors to vary ±1°C at 30 km and several degrees at higher altitudes. Since no corrections are made in the archived records, the increased error above 25 km makes these data of low quality for climatic studies. For daytime flights, temperature corrections are most sensitive to changes in solar angle and ventilation velocity. Operational corrections applied to archived daytime GZZ soundings appear valid for data below 100 mb or 20 km but are of unknown accuracy above this altitude. The archived daytime data below 100 mb are probably satisfactory for climatic studies, but data above this level should not be used until further studies have been completed. If rise rate data are available for GZZ soundings (i.e., time vs altitude), then it is recommended that nighttime soundings be corrected above 25 km. Day soundings may also be corrected, pending the results of further day–night analysis.

11. The Russian A-22 radiosonde

The A-22 radiosonde was a popular radiosonde used throughout the USSR from 1957 to 1990 (Zaitseva 1990). Several modifications and improved versions were introduced during this time span. Transmitter, drive motor, and pressure units were all upgraded, but the temperature sensor remained the same. The temperature sensor is a bimetal spiral element that is rigidly fixed on both ends, with a movable arm attached at the center. As temperature changes the spiral distorts, moving the calibrated arm. The sensor is enclosed in a small aluminum duct ≅2.5 cm in diameter, which is located within a larger housing similar to the other ducted sondes. The exterior duct is made from whitish cardboard, ≅0.5 mm thick, approximately a half-cylinder. The dull whitish surface makes it a diffuse reflector, and solar absorptivity of about 0.2 is assumed. The smaller aluminum duct is located totally within the cardboard, about 6.5 cm below the top. This smaller duct has a black-coated interior wall to absorb any reflected solar radiation. The geometry of the two ducts is such that direct solar radiation cannot illuminate the sensor at solar elevation angles less than 55°–60°. A tether length of 10 m is used.

The theory of sensor performance for the A-22 is very similar to that of other double-ducted sondes. The exterior duct is designed to shield the sensor from direct solar radiation, while the interior duct protects it from reflection off the inside surface of the exterior duct. The white cardboard surface of the exterior duct reflects 60%–80% of direct solar radiation impacting its surface (Henninger 1984). Some radiation reflected from the inside surface strikes the aluminum interior duct surface, which is an efficient reflector. Both ducts are heated and they heat the air passing adjacent to their walls. Because of the different surface properties, the temperature of the duct walls may differ significantly. Thus, A22COR estimates the temperatures of both duct walls and of the ventilation air passing through. Since airflow through each duct is unknown, ventilation velocity became a model-fitting parameter.

a. Sensitivity analysis

Results from the sensitivity analysis for the A22 sonde are summarized in Fig. 10 and Table 2. The A22 temperature error profiles for the day and night reference datasets are similar to the other ducted radiosondes. The influence of solar angle and ventilation velocity, or rise rate, on the temperature error is also very large and similar to the other sondes. In addition, the A22 sonde is sensitive to infrared radiation parameters, such as the blackbody environment temperature, and to a lesser degree cloud cover and the surface temperature.

Fig. 10.

Russian A-22 radiosonde day and night temperature errors.

Fig. 10.

Russian A-22 radiosonde day and night temperature errors.

b. Validation of A22COR

Average A-22 temperature and height differences were calculated by McInturff and Finger (1968), McInturff et al. (1979), and Spackman (1978). Each analysis shows that day–night differences are positive and increase with altitude. At solar angles above 20° the mean difference goes from ≅0.8°C at 100 mb to ≅1.5°C at 30 mb, and 2.0°C at 10 mb, but it is less for low solar elevations. There is, however, a large variability in individual differences that is consistent with model results. Unfortunately, because nighttime measurements above 20 km cannot be considered accurate (no corrections were made even though they should have been), the part of the real day–night difference attributable to undercorrection is indeterminate.

c. Application of A-22 temperature data to climate studies

It is recommended that A-22 radiosonde data be used with caution in climatic studies. The most accurate temperature data are likely to be from nighttime flights at altitudes below 22–25 km. At higher altitudes, measured temperatures are probably 1°–3°C colder than reality. Nighttime temperature soundings can be improved significantly by applying A22COR corrections. It is reasonable to assume a fixed ascent rate during night flights; however, for daytime flights it is not feasible to attempt A22COR corrections without knowledge of balloon rise rate and ventilation velocity. Daytime temperature errors may be several degrees at higher altitudes, because of variations in the balloon rise rate and background temperature profile. Thus it is recommended that daytime A-22 flights in archived form not be used in climate studies. These results are summarized in Table 2.

12. Summary and conclusions

Ten of the most commonly used radiosondes since 1960 have been evaluated by modeling the temperature of the sensor relative to temperature of the air in which it is immersed. The difference, designated as the temperature error, is calculated under various environmental conditions that affect its magnitude. Validation and sensitivity analyses were performed on each radiosonde model. It is believed that the results of these analyses, summarized in Table 2, provide the best guidelines available to anyone wishing to perform climate or synoptic studies with radiosonde data. It should be cautioned, however, that the accuracy estimates of Table 2 refer only to the temperature error of the sensor. Other random and bias errors may be present in the data.

The research work presented in this paper indicates that climate trends can currently be estimated reliably with a subset of the upper-air data. Trends can be calculated for monthly averaged, nighttime soundings with some confidence for the Vaisala RS80 (not using the RSN80 and RSN86 corrections), Vaisala RS 12/15, Vaisala RS 18/21, Chinese GZZ (below 25 km), Russian RKZ, Russian MARS, and Russian A-22 (below 20 km) radiosonde models. The analysis presented in this paper shows all of the above radiosondes have very small errors in individual radiosonde soundings at night (< ±1°C) and monthly averaged errors are estimated to be less than ±0.5°C, except for the A-22 (±0.8°C). In addition, temperature data from the Japanese RS-2-80, Russian A-22, Vaisala RS80 (RSN80 and RSN86 corrections applied), and VIZ can be made suitable for climate analysis if the appropriate temperature correction models are used to correct the data.

Fig. A1. Schematic showing the wake behind a spherical balloon. Here, U is the balloon velocity, L is the lateral scale length of the wake, and x and y are the coordinates.

Fig. A1. Schematic showing the wake behind a spherical balloon. Here, U is the balloon velocity, L is the lateral scale length of the wake, and x and y are the coordinates.

Acknowledgments

The authors express their sincere appreciation to the many people and organizations that provided the data and information necessary to conduct this study. The Central Aerological Observatory, Moscow, Russia, provided a sonde of each type used in the former USSR. Vaisala provided each type sensor used on the Vaisala sondes. Information about the Chinese radiosondes was supplied by Mr. Huang Bingxum, and a Chinese sonde was supplied by Mr. Zhai Panmao of the State Meteorological Administration, China. Information about the Japanese sonde was supplied by Dr. Syoin Yagi of the Japan Meteorological Agency and by Mr. Maury Friedman for the VIZ sonde. Intercomparison data were received from Dr. John Nash of the United Kingdom, Mr. Francis Schmidlin of NASA, and the Test and Evaluation Branch of NWS. The authors especially thank Dr. Nina Zaitseva for the information she supplied about USSR sondes.

We thank Dr. Trevor W. R. Wallis and an anonymous reviewer for their meticulous reviews of this paper.

James Luers was supported in part by NOAA Contract 50EANE-2-00077 as part of the CARDS project. The CARDS project is supported by the Department of Energy under Contract DE-AI05-90ER61011, the NOAA Climate and Global Change Program, and the National Climatic Data Center.

REFERENCES

REFERENCES
Antikainen, V., 1973: Solar radiation error of Vaisala radiosonde RS 18. Vaisala News, 57, 3–19.
Bevilaqua, P. M., and P. S. Lykoudis, 1978: Turbulent memory in self-preserving wakes. J. Fluid Mech., 89, 589–606.
Chernykh, I. V., and R. E. Eskridge, 1996: Determination of cloud amount and level from radiosonde soundings. J. Appl. Meteor., 35, 1362–1369.
Eskridge, R. E., and R. S. Thompson, 1982: Experimental and theoretical study of the wake of a block-shaped vehicle in a shear-free boundary flow. Atmos. Environ., 16, 2821–2836.
——, O. A. Alduchov, I. V. Chernykh, Zh. Panmao, A. C. Polansky, and S. Doty, 1995: A Comprehensive Aerological Reference 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, 123 pp. [Available from World Meterological Organization, Case Postale 2300, CH-1211 Geneva 2, Switzerland.]
.
——, 1996: A digitized metadata set of global upper-air station histories. NOAA Tech. Memo. ERL ARL-211, Air Resources Laboratory, NOAA, 38 pp.
Henninger, J. H., 1984: Solar absorptance and thermal emittance of some common spacecraft thermal-control coatings. NASA Ref. Pub. 1121, NASA/GSFC, Greenbelt, MD, 43 pp.
Hörhammer, J., 1971: New Vaisala radiosondes. Vaisala News, 52, 3–7.
Ivanov, A., S. Kats,, N. Karnosenko, J. Nash, and N. Zaitseva, 1991:WMO international radiosonde comparison—Phase III, Dzhambul, USSR. WMO/TD No. 451, 135 pp. [Available from World Meterological Organization, Case Postale 2300, CH-1211 Geneva 2, Switzerland.]
.
Kneizys, F., E. Shettle, L. Abreu, J. Chetwynd, G. Anderson, W. Gallery, J. Selby, and S. Clough, 1988: User’s guide to LOWTRAN 7. AFGL-TR-88-0177, Air Force Geophysics Laboratory, Hanscom Air Force Base, MA, 137 pp.
Lindquist, A., 1963: Vaisala radiosonde type RS 12: Introduction to theory, construction principles and operation. Vaisala News, 19, 3–11.
Luers, J. K., 1989: The influence of environmental factors on the temperature of the radiosonde thermistor. Ph.D. dissertation, University of Tennessee, 169 pp. [Available from University of Michigan, 300 N. Zeeb Road, Ann Arbor, MI 48106.]
.
——, 1995: VIZ table regression model. UDRI-Tr-94-003, University of Dayton, 21 pp. [Available from University of Dayton, Dayton, OH 45469.]
.
——, 1996: Temperature corrections for the world’s major radiosondes. NOAA 50EANE-2-00077. [Available from NCDC, 151 Patton Ave. Asheville, NC 28801-5001.]
.
——, and R. E. Eskridge, 1995: Temperature corrections for the VIZ and Vaisala radiosondes. J. Appl. Meteor., 34, 1241–1253.
McInturff, R. M., and F. Finger, 1968: The compatibility of radiosonde data at stratospheric levels over the Northern Hemisphere. ESSA TM WBTM-DATAC 2, National Weather Service, Silver Spring, MD, 61 pp.
——, ——, K. Johnson, and R. Laver, 1979: Day–night differences in radiosonde observations of the stratosphere and troposphere, NOAA TM NWS NMC 63, National Weather Service, Silver Spring, MD, 48 pp.
Nakamura, T., N. Hayashi, and K. Masuda, 1983: Estimation of errors in measurement by the temperature sensor of RS2-80 radiosonde—A method of correction for temperature errors due to solar radiation (in Japanese). Atmos. Obs. Rep., 50, 117–138.
NWS, 1992: Test report on temperature and pressure measurement performance of the Space Data division VECP radiosonde. Test and Eval. Rep. 01-92, NWS, Sterling, VA, 15 pp. [Available from National Weather Service, Sterling, VA 22170.]
.
Oakley, T., 1993: Report by Rapparteur on radiosonde compatibility. Part A: WMO catalogue of radiosondes and upper-air wind systems in use by members. WMO/TD No. 587, Instruments and Observing Methods Report 56, 89 pp.
Spackman, E. A., 1978: Compatibility and performance of radiosonde measurements of geopotential height in the lower stratosphere for 1975–76. WMO TN No. 163, 29–99.
Vaisala, 1992: Vaisala RS80 radiosonde. Vaisala Ref. 435, Vaisala, Helsinki, Finland, 2 pp.
Väisälä, V., 1965: The radiant and convective radiation error components of the shielded temperature sensor of radiosonde RS 12. Vaisala News, 25, 8–11.
Zaitseva, N. A., 1990: Aerology (in Russian). Gidrometeoizdat, 325 pp.
Zurbenko, I., P. S. Porter, S. T. Rao, J. J. Ku, R. Gui, and R. E. Eskridge, 1996: Detecting discontinuities in time series of upper-air data: Demonstration of an adaptive filter technique. J. Climate, 9, 3548–3560.

APPENDIX

Radiosonde Ventilation Deficit in Wake of a Rising Balloon

An estimate of the ventilation velocity over a sensor is derived in this appendix. Assume the wake of the balloon is self-preserving. This means that the structure of the flow changes in scale but has a similar functional form at each stage of decay (e.g., distance below the balloon). In a turbulent, boundary free, shear flow, self-preservation means that the shape of the mean velocity deficit does not change with axial distance, and it can be normalized with scales L and U (see Fig. A1). A concise discussion of two- and three-dimensional wakes is given in Eskridge and Thompson (1982).

Thus

 
formula

where U is the wind speed (rise rate of the ballon), d is the balloon diameter, u is wind speed in the wake (uU), x is the axial distance, y is the distance from the centerline of the wake, and L is a wake length scale. Here A and α are empirically determined constants.

For wakes of three-dimensional objects in free flows

 
m = 2n.
(A3)

Experiments show that m = 2/3 and n = 1/3.

Along the center line of the wake y = 0, and the equation for the velocity deficit uD along the center line can be written as

 
formula

Bevilaqua and Lykoudis (1978) state that A is probably a weak function of the drag coefficient CD and possibly the density Y. However, as a first approximation, A can be assumed constant.

To estimate A, the published experimental data of Bevilaqua and Lykoudis (1978) were used. First, from the information given in their paper, the tunnel speed U was determined to be 6 m s−1, which is close to the average balloon speed. From Fig. 5 of Bevilaqua and Lykoudis, the velocity was estimated at x/d equal to 6 and 10. Inserting these velocity measurements into (4) yields estimates of 0.371 and 0.396 for A. Averaging and rounding off yields

 
A = 0.38.
(A5)

Hence, the magnitude of the velocity deficit can be calculated along the centerline of the wake by

 
formula

The ventilation velocity can be estimated from (A6) if the balloon vertical speed and diameter are known.

The Soviet Union, Russia, and China use tether lines from the balloon that are 10 m long. When the balloon has expanded to a diameter of 5 m, the ventilation velocity will be reduced to 76% of the balloon rate of ascent. At a diameter of 10 m (x/d = 1), the ventilation velocity decreases to 62% of the balloon’s velocity. When the balloon enters the recirculating wake x < d, reduction in ventilation velocity is even greater.

Several things must be noted: first, self-preserving wake estimates are considered to be valid at distances greater than 2d downwind of the object; second, a rising balloon is not spherical. It may be flattened into a pancake shape, acting more like a disk than a sphere, with velocity deficits that are almost certainly greater than those given by (A6). This can be inferred by looking at Figs. 2 and 3 of Bevilaqua and Lykoudis.

We conclude that a major factor in the spurious thermistor warming during daytime flights, is the reduction of ventilation velocity in the balloon wake.

Footnotes

Corresponding author address: Dr. Robert E. Eskridge, National Climatic Data Center, 151 Patton Ave., Asheville, NC 28801-5001. E-mail: beskridg@ncdc.noaa.gov