The initiative is a cooperative effort between groups in a number of countries or regions. It builds on the work of Allen and Vincent who used high-resolution radiosonde observations in the Australian sector to study temporal and spatial variations of wave activity in the lower atmosphere. The aim is to produce climatologies of wave parameters in as global a scale as possible. Using wind and temperature data it is planned to produce climatologies of energy and of direction of propagation in both the horizontal and vertical directions. By studying the spatial and temporal variability in these parameters it should be possible to determine dominant sources, especially where the network is sufficiently dense, for example North America. Since some parameterization schemes are using spectral representations of the wave field the radiosonde data will also be used to better characterizing vertical wavenumber spectra in order to investigate the "universality" of spectral indices.

The following map shows the location of radiosonde stations that are being used in the project. Appendix A lists the location of the stations and the person(s) responsible for the analysis. Please check the list and advise me of any inaccuracies and/or updates.

Reference is made to the following papers:

Allen and Vincent, J. Geophys. Res., 100, 1327-1350, 1995 (hereinafter AV95)

Vincent, Allen and Eckermann in "Gravity wave processes: Their parameterization in global climate models" (ed K. Hamilton) Springer-Verlag, 7-25, 1997 (ISBN 3-540-62036-2) (VAE97).

A package of IDL routines for carrying out the whole analysis procedure in as painless manner as possible is available (see below).

1. The radiosonde data should be analyzed at as high a vertical resolution as possible. Since the raw data are usually unevenly spaced, interpolate the data onto a 50-m grid. Analyze in at least monthly intervals if there are sufficient data, or use seasonal averages if the data rates are low.
2. In the first instance, analyze in two height intervals to estimate tropospheric and stratospheric wave activity. The preferred intervals are 2-9 km for the troposphere and 17-24 km for the stratosphere (see AV95). The 17-24km range is used because the tropopause in the tropics is at about 17 km. It may not be possible to cover precisely these intervals, but get as close as you can so that we can have a consistent set of results to compare.
3. Remove the background temperature from each profile by fitting a 2nd order polynomial. Subtract and normalize the residual profile by dividing by the background values to produce T'. The background profile can be used to provide a profile of N².
4. Compute the mean energy density (E_o) - eq 6 in AV95 - for each time interval.
5. Compute the vertical wavenumber spectra for the normalized temperature. Average the individual spectra to produce a monthly or seasonal average. For the stratospheric data it will probably be necessary to modify the spectra to account for the finite response time of the temperature probes - see discussion on page 1333 in AV95. This is a vexed problem as the response time is height dependent and not well known; the best guess in the lower stratosphere is about 8 sec for the Vaisala RS-80-15 sondes commonly used by many weather services. If the spectra are not corrected, however, then they are too steep at high wavenumbers.
6. Use a non-linear fit to compute the spectral slope, t, and m_* (AV95 eq 13) for each height range.
7. Repeat above for as wide a height range as possible in the stratosphere. At mid- to high-latitudes, where the tropopause is as low as 10 km, it should be possible to cover a 14-15 km height range (12-26 km say) and get a better estimate of the contribution of longer vertical wavelength, higher intrinsic phase speed waves to the spectra (VAE97). It may be necessary to experiment with the degree of the polynomial fit for the background e.g. a 3rd degree polynomial may be better.
8. Repeat the above using the zonal and meridional wind profiles. It should be noted that the wind profiles are usually smoothed in the data taking process. Even if the data are provided at 10 sec intervals the actually height resolution is usually only about 500m. Since the method of measuring winds is station dependent, you will need to explore for yourself by looking at the spectra and seeing if there is a sharp drop in the power at the higher wavenumbers. Compute the total wave energy E_T = KE + PE
9. Compute the fraction of upgoing energy by computing the rotary spectra. See VAE97 section 5, and references therein.
10. Produce seasonal climatologies of horizontal propagation directions (f ) of the dominant motions by combining temperature and wind data. This procedure is discussed in VAE97 section 5.
11. An important part of the project is to try to identify gravity wave sources. A possible way to do this is to compare observations made at closely spaced sites. Collaboration between participants is strongly encouraged.

IDL programs are made available for analysis of high-resolution radiosonde data. The programs were tested with IDL version 4, but should work with IDL 5. Please advise of any problems encountered.

This program reads data from radiosonde profiles. It is provided as an EXAMPLE only. Data from different locations will be provided in a different format. Use of the program is illustrated by using data provided by the Australian Bureau of Meteorology for its Cocos Islands station in the Indian Ocean (12°S, 97°E). It is assumed that the data are taken every 10 seconds (approximately 50-m height resolution), but the program can be adapted easily for profiles sampled at a different time resolution. The data are re-sampled to convert the observations to a 50-m height grid. A second-order polynomial is fitted to the data to give the mean wind and temperature profiles. The deviation from this fit gives the perturbation profiles.

Note, in the program provided only data between heights of 18 and 25 km are stored for later use. This height range was chosen because of the high tropical tropopause. The program will need to be modified to produce data in the 2-9 km range. The mean and perturbation profiles are stored in a data structure (called profile) for each year of observations, which is then written to disk as an IDL "SAVE" file.

The code provided assumes the radiosonde data for each launch are contained in separate file, which have the extension aed. Sample data for Cocos Islands from September to December 1992 are contained in a gzipped tar file. If you wish to test the programs with these data before you use them on your own data transfer the file (binary) and then unzip and extract them into your directory.

This uses the IDL save files (eg Cocos96.sav) generated by the READ_SONDE.PRO and saved as a structure called profile, which has the form documented below.

For each calendar month the program:

• Calculates the vertical wavenumber power spectrum of the normalized temperate fluctuations for each profile available in the month.
• Calculates the monthly-average spectrum and fits a modified Desaubies model spectrum to the average. The data average and the fit are plotted. The fitting routine used is a version of the MINPACK-1 routine adapted to IDL by Dr. C. B. Markwardt of NASA GSFC.
• The mean direction of horizontal propagation is calculated for each profile by correlating the wind components and Hilbert-transformed temperature.
• Histograms of the directions, weighted by the total energy (kinetic + potential), are produced for each month and plotted in polar form using 30° -wide "bins". The plotter output will come out on the default IDL device. To get Postscript output, uncomment the set_plot statement near the beginning of the main program code.
• The mean direction (phi) for each month is calculated and printed out.
• Plots m-spectra and angular spectra.
• An example of the output for Cocos Islands in 1997 is

2. lat : The latitude in degrees; south is negative.
3. pltfname: A string specifying the device name for plot output. For example ps means PostScript.

This package contains procedures and functions for computing date and time etc.

Examples of output (Cocos97.ps and Cocos97.log) and the programs themselves can be obtained from the anonymous ftp site by pointing your web browser to:

and saving the files.

Alternatively, do an anonymous ftp to bragg.physics.adelaide.edu.au

cd pub/atmos/rvincent

Retrieve files.

DISCLAIMER. There is no guarantee that the programs will work on your system. If you have any problems implementing the programs or find mistakes please not hesitate to contact me.

e-mail: rvincent@physics.adelaide.edu.au

Good luck!

Bob Vincent

The station coordinates are given in latitude (south negative) and longitude (west negative). Where available, the names and WMO station numbers are given.

Antarctica Chet Gardner cgardner@uiuc.edu

-90, 0 South Pole

Australian Stations Bob Vincent rvincent@physics.adelaide.edu.au

Canadian Stations Jim Whiteway jjw@abber.ac.uk

Finnish Stations Rigel Kivi rigel.kivi@fmi.fi

67.4 26.7 02836 Sodankyla

French Stations Claude Souprayen/Marie-Lise Chanin claude.souprayen@aerov.jussieu.fr

German Stations Werner Singer singer@iap-kborn.d400.de

Japanese Stations Kaoru Sato sato@kugi.kyoto-u.ac.jp

Hong Kong Johnny C. L. Chan apjcchan@cityu.edu.hk

Italian Stations Elisa Manzini manzini@dkrz.de

Korean Stations Hye-Yeong Chun chy@atmos.yonsei.ac.kr

New Zealand Stations Bryan Lawrence b.lawrence@phys.canterbury.ac.nz

Russian Stations Alexandre Kats alexandrekats@mtu-net.ru

Switzerland/Austria Christian Haeberli christian.haeberli@univie.ac.at

United Kingdom Stations Lesley Gray Lesley.Gray@rl.ac.uk

US Stations Marvin Geller mgeller@notes.cc.sunysb.edu