Temperature Fluctuations of Different Vertical Scales in Raw and Processed U.S. High Vertical-Resolution Radiosonde Data
1. Introduction
The 1-s U.S. high vertical-resolution radiosonde data (HVRRD) have been used by many investigators due to their ready availability, first from the Stratosphere–Troposphere Processes and Their Role in Climate (SPARC) Data Center (https://www.sparc-climate.org/data-centre/data-access/us-radiosonde/) and later from the National Oceanic and Atmospheric Administration (NOAA) National Centers for Environmental Information (NCEI, ftp://ftp.ncdc.noaa.gov/pub/data/ua/). The NCEI data stream includes two versions of temperature profiles at each U.S. station, i.e., the raw and processed data. Various publications have used these two different HVRRD profiles. For instance, Geller et al. (2021a) utilized the raw data for their analysis of atmospheric unstable layers, while Ko et al. (2019) used the processed data for their analysis of Thorpe-derived atmospheric turbulence climatology.
The processed HVRRD differs from the raw HVRRD in two aspects. One is that radiation corrections have been applied to the processed data. Such radiation corrections have been well documented (e.g., von Rohden et al. 2022; Lee et al. 2022). Another difference is that smoothing has been applied to derive the processed data. According to the U.S. National Weather Service Sterling Field Support Center, details of this smoothing are, however, “vendor specific proprietary information” (Helpdesk Staff 2023, personal communication).
Figure 1 shows a comparison of raw and processed temperature data in the pressure altitude segment of 17–18 km for a sounding taken at Riverton, Wyoming (43.06°N, 108.47°W), at 0000 UTC 1 July 2012. Both the raw and processed 1-s U.S. HVRRD profiles have a nominal vertical resolution of ∼5 m. Note that the overall raw and processed temperature curves compare well after a 0.69°C temperature correction is applied. This offset represents the radiation correction for this altitude segment, this local time, and this time of year of this sounding at this station. Note also, however, that greater small-scale fluctuations are seen in the raw data relative to the processed data. This demonstrates the smoothing applied to the raw data in this case. It is the purpose of this paper to compare various aspects of small-scale temperature variabilities of the raw and processed 1-s U.S. HVRRD so that we can better understand the differences between the two and how limitations of these data can affect analyses using those data.
Raw (black curve) and processed (blue curve) temperature data (°C) from the 1-s U.S. HVRRD over Riverton, Wyoming (43.06°N, 108.47°W), at 0000 UTC 1 Jul 2012. The red curve is the processed data offset by 0.69°C.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
Characterizing the raw and processed HVRRD is important because small-scale noise in those data can affect determinations of turbulent overturning (Wilson et al. 2010), which are then used to identify regions of turbulence and calculate turbulence parameters such as the rate of dissipation of turbulent kinetic energy and turbulent diffusion (e.g., Thorpe 1977; Clayson and Kantha 2008). Furthermore, understanding of the nature of HVRRD has become increasingly important since more global HVRRD have recently become available for scientific analysis (Ingleby et al. 2016; Geller et al. 2021b). These new HVRRD have been obtained using a number of different radiosonde instruments. The present analysis is restricted to the U.S. HVRRD, which itself has used a number of different radiosonde instruments. Although the present paper only examines the U.S. data, it does present a methodology which can be used to examine HVRRD from other nations, which use different instrumentations.
This manuscript is organized as follows. This introduction is followed by section 2, which describes our methodology. Section 3 presents the results and discussion. Summary and conclusions are provided in section 4.
2. Methodology
Root-mean-square (rms) of the normalized temperature fluctuations from fitted polynomials for vertical windows of various sizes Δz is used to quantify small-scale variability in the raw and processed temperature data in those windows in this study. For instance, Fig. 2 illustrates the procedure used for the 22–24-km pressure altitude segment of the temperature sounding taken at San Juan, Puerto Rico (18.43°N, 65.99°W), at 0000 UTC 1 January 2019 for Δz = 500 m. The black and red lines in Fig. 2 are the raw and processed temperature data, respectively, for that pressure altitude segment. Pressure altitude is used instead of geometric altitude in this illustration and the subsequent analyses in this paper because pressure is the measurement that we have in both the raw and processed temperature data while geometric height or geopotential height data are not readily available in both datasets. Using pressure altitude facilitates the comparison of the two datasets directly.
Figure to illustrate the methodology to quantify rms of the normalized temperature fluctuations in raw and processed temperature soundings. The black curve is the raw profile for the chosen pressure altitude range at San Juan, Puerto Rico (18.43°N, 65.99°W), at 0000 UTC 1 Jan 2019. The red curve is the processed profile. The blue and orange curves are second-degree polynomial fits over each Δz = 500 m segment to the processed and raw profiles, respectively. The magenta dashed line with squares is the rms value for each Δz segment of the raw profile, and the magenta solid line with triangles shows the same for the processed profile. The green line with diamonds shows the ratios of rms values for the raw profile to those of the processed profile. The values of rms and the ratios are shown in magenta and green at the top and lower axes, respectively.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
The background temperature for each Δz = 500 m window is calculated using the second-degree polynomial fit within that window (the blue and orange lines for the processed and raw data, respectively), and the rms of temperature perturbations normalized by the background temperature (solid and dashed magenta lines with symbols) is used to represent small-scale variability of temperatures (or temperature fluctuations) for that Δz. Note from Fig. 2 that the rms values for Δz = 500 m are a bit greater for the raw profile (dashed magenta line with squares) than for the processed profile (solid magenta line with triangles), indicating greater small-scale variability in the raw temperature profile. Also, in general, we have found that the rms values for all Δz values that we have explored (ranging from 100 m to 5 km) are greater for the raw profiles than for the processed profiles and that the ratios of these rms values for the raw profiles to those of the processed profiles (i.e., the green line and symbols in Fig. 2) increase as Δz decreases (not shown). These characteristics of rms values for various Δz values will be apparent in the upcoming results and discussion section. Finally, we note that we use a moving window (not to be confused with Δz) with an arbitrary increment of 100 m to obtain more gradual vertical profiles of rms values for Δz greater than 100 m. Thus, the rms results are oversampled for Δz greater than 100 m.
3. Results and discussion
The U.S. HVRRD balloon soundings are generally launched twice a day, nominally at 0000 and 1200 UTC. This means that U.S. radiosonde launches are made at many different local times given the fact that the U.S. HVRRD stations cover a very wide range of geographic locations covering different time zones in the contiguous United States, Alaska, Hawaii, the Caribbean, and the tropical western Pacific. Figure 3 shows maps of multiyear averaged rms values from the raw data (top) and processed data (bottom) for the pressure altitude range 5–15 km at 0000 UTC for Δz = 100 m for January (left) and July (right) over the contiguous United States. These plots were generated using all the available 1-s U.S. HVRRD for the years from 2006 to 2019 for January and for the years from 2005 to 2019 for July when both the raw and processed data exist. We note that the exact starting years of the 1-s HVRRD vary with station, and they range from 2005 to 2012 as NOAA’s upper-air network was progressively upgraded from the 6-s resolution Microcomputer Automatic Radio-Theodolite (MicroART) system to the 1-s global positioning system (GPS)-based radio replacement system during this period of time. Several conclusions can be immediately drawn from Fig. 3. One is that the rms values in the processed data are generally smaller than in the raw data by up to a factor of approximately 2. Also, note that the pattern of the rms values is similar for the raw and processed data in that maximum rms values are over the western and extreme northern portions of the contiguous United States in January, and these maximum rms values extend further into the central United States in July. In addition, the rms values are consistently smaller in the southeastern United States.
Maps of multiyear averaged rms of normalized temperature fluctuations for Δz = 100 m for the pressure altitude range 5–15 km, computed from the (top) raw and (bottom) processed data at 0000 UTC for (left) January and (right) July. These maps were generated using all the available 1-s U.S. HVRRD for January for the years 2006–19 and for July for the years 2005–2019 when both the raw and processed data exist. Different color scales are used for the top and bottom panels. See text for details.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
Figure 4 shows the same maps as in Fig. 3, but for the pressure altitude range 15–25 km. Note that the rms patterns in Fig. 4 are broadly similar to those in Fig. 3, but there are noticeable differences. The upper bound of the color scales in Fig. 4 is larger than that in Fig. 3 by a factor of ∼2.2 for the raw data (upper) and ∼1.9 for the processed data (lower), so the rms values in the higher pressure altitude range are ∼ twice those in the lower range for both the raw and processed data. Similar high versus low pressure altitude range and raw versus processed differences shown in Figs. 3 and 4 are also found for other months (not shown) and for 1200 UTC.
As in Fig. 3, but for the pressure altitude range 15–25 km. As in Fig. 3, different color scales are used for the (top) raw and (bottom) processed results.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
We can understand some of the geographic variability of rms values and January versus July differences for Δz = 100 m shown in Figs. 3 and 4 with the aid of Figs. 5 and 6. For easier comparison, the top panels of Fig. 5 are identical to the top panels of Fig. 4 and show again multiyear-averaged rms of normalized temperature fluctuations in the raw data for Δz = 100 m for the pressure altitude range 15–25 km for January (left) and July (right) at 0000 UTC. The lower panels of Fig. 5 show global maps of day and night for these two months at 0000 UTC. Figure 6 shows the same as Fig. 5, but for 1200 UTC. It is clear from Figs. 5 and 6 that the rms values are much greater when the radiosonde soundings occur under sunlit conditions, and such day/night effects can account for much of the geographic variability of rms values, the January versus July differences, and the 0000 versus 1200 UTC differences shown in those rms maps. Note that Wilson et al. (2018) also found that daytime raw radiosonde profiles were much noisier than nighttime profiles in their measurements from a campaign in Japan (34.85°N, 136.10°E), and it was for this reason that they used only nighttime raw radiosonde soundings in their study comparing the spectra of temperature fluctuation spectra in unstable layers to those in stable layers.
(top) Multiyear-averaged rms normalized temperature fluctuations in the raw data for Δz = 100 m for the pressure altitude range 15–25 km at 0000 UTC for (left) January and (right) July. The same color scales are used for the rms maps. (bottom) Day/night plots for (left) January and (right) July,1 in which the different shades from the darkest to the lightest and to no shade are for night (no twilight), astronomical twilight, nautical twilight, civil twilight, and day, respectively.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
As in Fig. 5, but for 1200 UTC.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
We now look at the dependence of rms values on the size of Δz. Figure 7 extends what has been shown in the upper-left panel of Fig. 3 and shows maps of multiyear-averaged rms of normalized temperature perturbations in the contiguous United States computed from the raw data for the pressure altitude range 5–15 km in January at 0000 UTC for Δz = 100, 200, 500, 1000, 3000, and 5000 m, respectively. It is evident that the rms values increase with increasing Δz values. Note also that the geographic patterns of rms values change with Δz, and that for Δz ≥ 1000 m, the rms pattern resembles the pattern of total gravity wave energy in the troposphere such as shown in Fig. 5 of Wang and Geller (2003). Figure 8 extends what has been shown in the upper-left panel of Fig. 4 and shows the same results as in Fig. 7, but for the pressure altitude range 15–25 km. Again, it is evident that as Δz increases, the rms values also increase, and the rms patterns change from being impacted strongly by the day/night effects seen in the Δz = 100 and 200 m panels (as shown in Figs. 5 and 6 for Δz = 100 m) to resembling those of total gravity wave energy in the stratosphere for Δz ≥ 1000 m, very similar to what was seen in Fig. 6 of Wang and Geller (2003). We note that the same Δz dependence shown in Figs. 7 and 8 was obtained when the rms values were calculated using the processed data, except that the rms values were significantly smaller for small Δz values (i.e., Δz = 100 or 200 m), as shown in Figs. 3 and 4 for Δz = 100 m, and that the rms values computed from the raw and processed data become similar for Δz ≥ 1000 (not shown). The resemblance of rms patterns for larger Δz values to previously published results of total gravity wave energies is no coincidence since the variability of temperatures at such larger vertical scales is expected to be dominated by gravity wave motions. These results also suggest that the climatology of inertia gravity wave variances derived from the raw and processed U.S. HVRRD should be very similar to each other.
Maps of multiyear-averaged rms of normalized temperature perturbations in the contiguous United States computed from the raw data for the pressure altitude range 5–15 km in January at 0000 UTC for Δz = 100, 200, 500, 1000, 3000, and 5000 m, respectively. Note that different color scales are used for different Δz values, and the values of color scales are scaled by 10a, where a is indicated by the number between the pair of parentheses under each color bar.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
As in Fig. 7, but for the pressure altitude range of 15–25 km. As in Fig. 7, different color scales are used for different Δz values.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
As mentioned earlier, the analysis in this study is done using the pressure altitude vertical coordinate for convenience. The pressure altitude ranges of 5–15 and 15–25 km in this paper do not correspond strictly to the geometric altitude ranges of the troposphere and lower stratosphere adopted by Wang and Geller (2003). They are chosen here as a crude reference to the two different vertical regions of the atmosphere. We have experimented with variations of the ranges for both the lower and higher regions, and the major results reported here remain the same.
Figure 9 shows the same as Fig. 8, but for July. The rms values are larger in July than in January for Δz = 100 and 200 m. This is consistent with the fact that all the contiguous U.S. stations are in daytime at 0000 UTC in July (see the lower-right panel of Fig. 5). The rms values are smaller in July than in January for Δz ≥ 500 m. This is consistent with the seasonal variations of gravity wave total energy reported by Wang and Geller (2003), i.e., gravity wave energies are distinctively greater in winter than in summer in both the troposphere and lower stratosphere.
As in Fig. 8, but for July. As in Figs. 7 and 8, different color scales are used for different Δz values.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
As mentioned earlier, radiation corrections and smoothing are applied to the raw HVRRD to derive the processed data. The 0.69°C temperature offset used in Fig. 1 represents the bulk effects of the applied radiation corrections for that sounding in that altitude segment. Evidently, the radiation effects are also directly related to the day/night effects of small-scale variability in the raw data shown in Figs. 5 and 6, as radiation corrections apply only to soundings in sunlit conditions. Also, we would expect that there would be no day/night differences in the processed data since radiation effects have been accounted for in those data. The similar geographic distributions of rms values of the raw and processed data in Figs. 3 and 4 (and Fig. 4 in particular) suggest, however, that there are still day/night effects in the processed data, though to a lesser extent. The day/night effects in the processed data are shown more clearly in plots of rms values as a function of month and pressure altitude for individual stations which can provide additional details and insights than the geographic maps shown previously. For instance, Fig. 10 shows the rms values calculated from the processed data for Δz = 100 m as a function of pressure altitude and month for Salem, Oregon (44.91°N, 123.01°W), for 0000 UTC (top) and 1200 UTC (bottom). Consistent with Figs. 3, 4, 7, and 8, the rms values are generally larger at higher altitudes. Moreover, there are local minimum rms values at pressure altitudes around 12–15 km. It is also evident that the large rms values at higher altitudes are considerably larger at 0000 UTC than at 1200 UTC. Such 0000 and 1200 UTC differences suggest day/night effects, as the local times of 0000 and 1200 UTC at Salem are 1548 and 0348 local time, respectively. The day/night effects in the processed data are even more pronounced for Barrow, Alaska (71.28°N, 156.79°W), where there are polar days and nights in summers and winters, respectively (Fig. 11). At higher altitudes, there is a clear indication of enhancement of small-scale temperature fluctuations in the summer months for both 0000 and 1200 UTC and the rms values at 0000 UTC (which corresponds to 1333 local time) are generally greater than those at 1200 UTC (or 0133 local time) in spring and fall. von Rohden et al. (2022) and Lee et al. (2022) have done careful laboratory characterizations of the radiation corrections that should be applied to the temperature sensor of the Väisälä RS41 radiosonde. They indicate the environmental factors that influence the radiation corrections that should be applied to correct measured temperatures to the actual atmospheric temperatures. Those environmental factors include temperature, pressure, ventilation speed, solar irradiance, and cloud cover. These radiation corrections are only applied under sunlit conditions. Radiosonde radiation corrections are applied using the expected solar irradiance geometry on the sensor as a function of the sensor altitude, the latitude of the radiosonde measurement, and the ventilation wind speed (see Dirksen et al. 2014). This is done using generalized assumptions about the cloud cover. These radiation corrections are done for each point of the radiosonde measured temperature profile. These radiation corrections, however, do not take into account the varying solar radiation experienced by the radiosonde sensors due to instrument pendulum and rotation motions. Also, the radiation corrections do not take into account the differing cloud conditions for each ascent. It is likely that the uncorrected radiation effects resulting from radiosonde pendulum and rotation motions may explain why there are still substantial though reduced day/night effects in the rms values in the processed data for small Δz shown in Figs. 10 and 11. At present, we cannot explain the seasonality of the stratospheric temperature fluctuations in Fig. 10. The minima in temperature fluctuations at pressure altitudes around 12–15 km in Fig. 10 may be due to ventilation effects from the sizeable jet stream winds at those altitudes.
Time series of monthly mean rms of normalized temperature fluctuations calculated using the processed HVRRD for Δz = 100 m as a function of altitude at Salem, Oregon (44.91°N, 123.01°W), at (top) 0000 UTC and (bottom) 1200 UTC. The months of April, July, and October are marked by thin vertical dashed lines, while the January months are marked by thick dashed lines with the last two digits of the corresponding years marked in magenta at the bottom. The 0000 and 1200 UTC correspond to 1600 and 0400 (Pacific standard time), respectively, at Salem, Oregon. The radiosonde type was Sippican Mark IIA until 0000 UTC 31 Oct 2013 when it was switched to Väisälä RS92-D.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
As in Fig. 10, but for Barrow, Alaska (71.28°N, 156.79°W). Note that the data extend to August 2019 for this station, but only those in and before August 2018 are shown here due to extensive missing data for 1200 UTC. The 0000 and 1200 UTC correspond to 1500 and 0300 (Alaska standard time), respectively, at Barrow. The radiosonde type was Väisälä RS92-NGP.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
We expect that the processed temperature fluctuations, particularly for Δz = 100 m, will depend on the radiosonde instrument being used since there are different radiation temperature corrections for different radiosonde instruments. This is illustrated in Fig. 12 by showing the temperature fluctuations at Jacksonville, Florida (30.48°N, 81.70°W), a station where Geller et al. (2021a) noted a discontinuity in unstable layer occurrences in 2013 when there was an instrument change from the Sippican Mark IIA instrument to the Väisälä RS92-NGP instrument. Note that while there are obvious discontinuities in the rms values at 0000 and 1200 UTC, both the rms values and the discontinuities are much greater at 1200 UTC, which corresponds to a ground release of the radiosonde instrument at about 0633 local time. Remembering that a radiosonde balloon rises at ∼5 m s−1, the observations for the pressure altitude range of 15–25 km took place over ∼ an hour later, so the 1200 UTC observations for that pressure altitude range occurred in the early morning hours. This discontinuity in the rms values was clearly the cause for the discontinuity in unstable layer occurrences noted in Geller et al. (2021a).
As in Fig. 10, but for Jacksonville, Florida (30.48°N, 81.70°W). The 0000 and 1200 UTC correspond to 1900 and 0700 (eastern standard time), respectively, at Jacksonville. The radiosonde type was Sippican Mark IIA until 0000 UTC 1 May 2013 when it was switched to Väisälä RS92-NGP.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
We should note that regions of large values of temperature fluctuations can be an indication of turbulence occurring in the stratosphere, rather than being an indication of instrumental “noise” resulting from varying incidence of solar radiation on the temperature sensor as a result of the swinging and rotation of the temperature sensor. An example of this is shown in Fig. 13, which shows rms values as a function of month and altitude for Δz = 100 m for Majuro, Republic of the Marshall Islands (7.07°N, 171.29°E). Note that, as expected, the rms values are larger at 0000 UTC than at 1200 UTC since 0000 and 1200 UTC correspond to 1125 and 2325 local times, respectively. Figure 13 also shows descending regions of large temperature fluctuations, which are separated in time by a little more than a year. We believe that this is a signature of turbulence occurring in the descending shear zones of the stratospheric quasi-biennial oscillation (QBO). QBO modeling papers (e.g., Richter et al. 2014; Geller et al. 2016) clearly show such descending regions of gravity wave momentum flux in the QBO shear zones, and Geller et al. (1975) have shown that turbulence should accompany gravity wave breaking as the waves approach their critical levels.
As in Fig. 10, but for Majuro, Republic of the Marshall Islands (7.07°N, 171.29°E). The 0000 and 1200 UTC correspond to 1200 and 0000 (Pacific standard time), respectively, at Majuro. The radiosonde type was Sippican Mark IIA until 0000 UTC 30 Mar 2015 when it was switched to Väisälä RS92-D.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
Possible evidence of turbulence signals may also be found in Fig. 14 which shows the rms values calculated from the processed data for Δz = 100 m as a function of month and altitude for Hilo, Hawaii (19.72°N, 155.05°W). There are distinctive seasonal variations with greater values of rms during the rainy winter months in Hilo, suggesting the effects of gravity waves from convection.
As in Fig. 10, but for Hilo, Hawaii (19.72°N, 155.05°W). The 0000 and 1200 UTC correspond to 1400 and 0200 (Pacific standard time), respectively, at Hilo. The radiosonde type was Sippican Mark IIA until 0000 UTC 2 Dec 2014 when it was switched to Väisälä RS92-D.
Citation: Journal of Atmospheric and Oceanic Technology 42, 3; 10.1175/JTECH-D-24-0012.1
4. Summary and conclusions
In this paper, we have shown results of temperature fluctuations of various vertical scales in the 1-s U.S. high vertical-resolution radiosonde data (HVRRD). These fluctuations were calculated by fitting a second-degree polynomial over various vertical intervals, Δz values, ranging from 100 m to 5 km. This was done for temperatures from both raw and processed data, where the raw temperature data are the actual temperatures measured by the radiosonde instrument, and the processed temperature data have undergone radiation corrections, as well as smoothing. The details of this smoothing are proprietary to the instrument manufacturer. We have sought to understand the effects of this smoothing through analysis of the temperature fluctuations of different vertical scales in the raw and processed HVRRD.
We have shown that for small Δz values, the temperature fluctuations are larger under sunlit conditions and are smaller for radiosonde soundings taken at nighttime for both the raw and processed data. This was shown in two ways. One was comparing the geographical distributions of the rms of normalized temperature fluctuations computed from the raw data for Δz = 100 m at 0000 and 1200 UTC to the geographical distributions of day/night at 0000 and 1200 UTC. The other was examining monthly time series of rms values computed from the processed data for Δz = 100 m at 0000 and 1200 UTC at various radiosonde stations where the two different universal times correspond to day and night, respectively. We have shown that the temperature fluctuations at small scales are greater in the raw HVRRD than in the processed HVRRD, and they are greater in the lower stratosphere than in the troposphere.
We have argued that one source of the small-scale temperature fluctuations in the processed HVRRD is the radiation effect related to the swing and rotation of radiosonde temperature sensors. The radiation corrections, used in producing the processed data, do not take into account the varying solar radiation experienced by the radiosonde sensors due to instrument pendulum and rotation motions. Also, they do not take into account the cloud conditions for each ascent.
Small-scale temperature fluctuations can also be a consequence of turbulence, and we have shown an example of enhanced small-scale temperature fluctuations that accompany the descending shear zones of the QBO, where we expect enhanced gravity wave breaking to produce turbulence. We have also shown an example of enhanced small-scale temperature fluctuations due to enhanced gravity wave activity from convection. We have also shown that the temperature fluctuations for Δz ≥ 1000 km correspond to those from total gravity wave energies in both the troposphere and lower stratosphere.
We have shown an example of a discontinuity in small-scale temperature fluctuations at a radiosonde station when the instrumentation was changed. This suggests that the radiosonde-measured temperature fluctuations resulting from varying amounts of solar radiation falling on the temperature sensor as the radiosonde instrumentation swings and rotates should be evaluated for each radiosonde system.
This is particularly important since global high vertical-resolution Binary Universal Form for the Representation of Meteorological Data (BUFR) are now becoming increasingly available for research use (Ingleby et al. 2016; Geller et al. 2021b). Unlike the U.S. HVRRD, only processed BUFR data are available in this more global dataset. Understanding the nature of temperature fluctuations in profiles derived from different radiosonde systems will be valuable to researchers using those data. As mentioned previously, HVRRD are valuable for examining the distribution of atmospheric unstable layers (e.g., Geller et al. 2021a) and for computing turbulence parameters (e.g., Ko et al. 2019). Better understanding of the nature and causes of spurious small-scale temperature fluctuations in the different global radiosonde systems will be important as more investigators perform scientific analyses on these global HVRRD datasets.
For the compactness of our presentation, we have only included a subset of our calculated results. Maps of rms values for Δz’s = 100, 200, 500, 1000, 3000, and 5000 m for both 0000 and 1200 UTC (similar to what was shown in Figs. 3, 4, 7–9) and also time series of monthly mean rms values for Δz = 100 m (similar to what was shown in Figs. 10–14) for all stations are available and may be obtained by contacting the first author of this paper.
All day/night plots in this paper are from https://www.timeanddate.com/worldclock/sunearth.html.
Acknowledgments.
This research was supported by National Science Foundation Grants AGS-2032678 and AGS-2129221. We acknowledge the Fine-Scale Atmospheric Processes and Structures (FISAPS) activity of the Atmospheric Processes and Their Role in Climate (APARC) project of the World Climate Research Programme (WCRP) in encouraging this research.
Data availability statement.
Radiosonde data analyzed in this study were obtained from the U.S. National Oceanic and Atmospheric Administration National Centers for Environmental Information (ftp://ftp.ncdc.noaa.gov/pub/data/ua/).
REFERENCES
Clayson, C. A., and L. Kantha, 2008: On turbulence and mixing in the free atmosphere inferred from high-resolution soundings. J. Atmos. Oceanic Technol., 25, 833–852, https://doi.org/10.1175/2007JTECHA992.1.
Dirksen, R. J., M. Sommer, F. J. Immler, D. F. Hurst, R. Kivi, and H. Vömel, 2014: Reference quality upper-air measurements: GRUAN data processing for the Vaisala RS92 radiosonde. Atmos. Meas. Tech., 7, 4463–4490, https://doi.org/10.5194/amt-7-4463-2014.
Geller, M. A., H. Tanaka, and D. C. Fritts, 1975: Production of turbulence in the vicinity of critical levels for internal gravity waves. J. Atmos. Sci., 32, 2125–2135, https://doi.org/10.1175/1520-0469(1975)032<2125:POTITV>2.0.CO;2.
Geller, M. A., and Coauthors, 2016: Modeling the QBO—Improvements resulting from higher-model vertical resolution. J. Adv. Model. Earth Syst., 8, 1092–1105, https://doi.org/10.1002/2016MS000699.
Geller, M. A., P. T. Love, B. Ingleby, and X. Yin, 2021a: New availability of high vertical-resolution radiosonde data for research. SPARC Newsletter, No. 56, SPARC Office, Toronto, Ontario, Canada, 14–15.
Geller, M. A., P. T. Love, and L. Wang, 2021b: A climatology of unstable layers in the troposphere and lower stratosphere: Some early results. Mon. Wea. Rev., 149, 1233–1245, https://doi.org/10.1175/MWR-D-20-0276.1.
Ingleby, B., and Coauthors, 2016: Progress toward high-resolution, real-time radiosonde reports. Bull. Amer. Meteor. Soc., 97, 2149–2161, https://doi.org/10.1175/BAMS-D-15-00169.1.
Ko, H.-V., H.-Y. Chun, R. Wilson, and M. A. Geller, 2019: Characteristics of atmospheric turbulence retrieved from high vertical-resolution radiosonde data in the United States. J. Geophys. Res. Atmos., 124, 7553–7579, https://doi.org/10.1029/2019JD030287.
Lee, S.-W., and Coauthors, 2022: Radiation correction and uncertainty evaluation of RS41 temperature sensors by using an upper-air simulator. Atmos. Meas. Tech., 15, 1107–1121, https://doi.org/10.5194/amt-15-1107-2022.
Richter, J. H., A. Solomon, and J. T. Bacmeister, 2014: On the simulation of the quasi-biennial oscillation in the Community Atmosphere Model, version 5. J. Geophys. Res. Atmos., 119, 3045–3062, https://doi.org/10.1002/2013JD021122.
Thorpe, S. A., 1977: Turbulence and mixing in a Scottish Loch. Philos. Trans. Roy. Soc., A286, 125–181, https://doi.org/10.1098/rsta.1977.0112.
von Rohden, C., M. Sommer, T. Naebert, V. Motuz, and R. J. Dirksen, 2022: Laboratory characterisation of the radiation temperature error of radiosondes and its application to the GRUAN data processing for the Vaisala RS41. Atmos. Meas. Tech., 15, 383–405, https://doi.org/10.5194/amt-15-383-2022.
Wang, L., and M. A. Geller, 2003: Morphology of gravity-wave energy as observed from 4 years (1998–2001) of high vertical resolution U.S. radiosonde data. J. Geophys. Res., 108, 4489, https://doi.org/10.1029/2002JD002786.
Wilson, R., H. Luce, F. Dalaudier, and J. Lefrère, 2010: Turbulence patch identification in potential density or temperature profiles. J. Atmos. Oceanic Technol., 27, 977–993, https://doi.org/10.1175/2010JTECHA1357.1.
Wilson, R., H. Hashiguchi, and M. Yabuki, 2018: Vertical spectra of temperature in the free troposphere at meso-and-small scales according to the flow regime: Observations and interpretation. Atmosphere, 9, 415, https://doi.org/10.3390/atmos9110415.
