BADC Help File : Met Office TOVS (SSU) Analyses

This section contains basic background information to help you use the Met Office SSU Analyses held at the BADC.


1. Introduction

These data consist of sets of 3-dimensional gridpoint analyses of the stratosphere which are produced by the Met Office using data from the TIROS Operational Vertical Sounder instruments onboard the NOAA operational polar orbiters. The analyses are usually refered to as the SSU analyses, but this is slightly misleading since they use data from all three of the TOVS sub-instruments not just the SSU.

The primary products which are archived at the BADC are synoptic maps of measured radiances and geopotential height. The radiances are stored on a 5 degree latitude by 5 degree longitude grid, whilst the geopotential heights are stored on the same grid on the 850, 500, 300, 200, 100, 50, 20, 10, 5, 2 and 1 hPa pressure surfaces. Both datasets are produced daily at 1200 UTC. The analyses have been produced from December 1978 to June 1998.

Software is available to calculate brightness temperatures from the radiances and temperatures and geostrophic winds and potential vorticity from the geopotential height fields.

2. The Analysis System

2.1 Instrument Descriptions

The TOVS system has been flown on successive spacecraft in the TIROS-N/NOAA series of Polar Orbiting Environmental Satellites (POES) since 1978. TOVS consists of 3 sub-instruments which make measurements of radiance from the Earth over a large portion of the electromagnetic spectrum from the microwave to the visible:

All three instruments scan perpendicular to the orbital path which greatly improves their vertical resolution, although they use different scanning patterns. The SSU (which was built by the Met Office) is described in detail in Pick and Brownscombe (1981), whilst Schwalb (1978) gives a description of the TOVS system as a whole.

The Met Office analyses use a total of 8 channels from all three instruments. SSU channels 25, 26 and 27, which have weighting functions peaking near 15, 5 and 1.5 hPa respectively, HIRS channels 1, 2 and 3 which have broad weighting functions spanning much of the stratosphere and MSU channels 23 and 24 which peak in the lower stratosphere. The Spatial distribution of the weighting functions is shown in Fig.1

2.2 Analysis Technique

The Met Office uses 8 of the channels from the TOVS analyses to calculate 5 deep layer thicknesses in the stratosphere: 100-20, 100-10, 100-5, 100-2 and 100-1 hPa. Thicknesses rather than temperatures are retrieved because of the low vertical resolution of the soundings.

The analysis process may be divided into two distinct stages.

I. Calculation of regression Coefficients
A first preparatory step involves the calculation of regression coefficients to derive a linear relationship giving thicknesses in terms of radiances.
II. Operational Data Processing
This relationship is then used operationally to retrieve thicknesses from the radiances measured by TOVS. The thicknesses are interpolated to give synoptic thickness maps on a regular grid. Geopotential heights are then obtained by adding thicknesses to an independently analysed field of geopotential height at 100hPa.

These stages are outlined in more detail in the following sections.

2.3 Calculation of Regression of Coefficients

In order to use the satellite radiances to calculate thicknesses and and hence geopotential heights, a linear equation is derived which gives the thicknesses in terms of radiances. This is achieved by using a reference set of rocket profiles. For each of the reference profiles a set of simulated radiances (which would have been measured by a coincident satellite sounding) is calculated. These are used in conjunction with the measured thicknesses to derive a set of linear equations for the calculation of thickness from radiance. These equations may then be used to calculate thicknesses operationally.

The rocket profiles used in the calculation are a reference set of 1200 which were obtained from the NOAA/NMC Upper Air Branch. The profiles are divided into 7 zones by latitude, and a radiance-thickness expression is derived for each zone separately. There are three zones in the winter extra-tropics, three in the summer extra-tropics (each 20 degrees wide) and one tropical zone (60 degrees wide). There is no distinction between the hemispheres, biasing the a priori in favour of the northern hemisphere where the majority of the rocket ascents took place. Note that reference set of rocket soundings includes profiles representative of winter warming conditions.

The radiance-thickness expressions are recalculated for each new satellite and also if there is a change in the number of operational channels in the TOVS subsystem, or if there is a significant change (> 10%) in the gas-cell pressure in one of the SSU channels.

2.4 Operational Data Processing Steps

The data used to calculate the analyses are supplied to the Met Office by NOAA/NESDIS. The routine processing of the data involves the following steps.

  1. Interpolation of HIRS and MSU data
    HIRS and MSU data are interpolated onto SSU fields of view by locating the nearest 4 HIRS/MSU radiances to each SSU viewing position. This step effectively means that all three instruments can be considered to provide data at common viewing angles. (5,15,25 and 35 degrees off nadir).
  2. Quality Control
    A spike check is made by taking the mean and standard deviation of all radiances in a block of data. All radiances further than three standard deviations from the mean are rejected.
  3. Grouping into "Superobs"
    The 8x8 observations from the 8 different viewing angles are grouped into 4x4 fields of view by combining 4 adjacent observations into a single observation known as a "superob". The superobs are centred at angles 10 and 30 degrees either side of the nadir.
  4. Nadir corrections
    The off-nadir SSU radiances are corrected to a nadir view by using a pre-computed matrix correction. (Note that the HIRS and MSU radiances are corrected to nadir at NESDIS).
  5. Retrieval of thicknesses
    For each nadir-corrected 8-channel radiance vector, pre-calculated (see section 1.3 above) regression coefficients are used to calculate thicknesses for the standard layers 100-20, 100-10, 100-5, 100-2 and 100-1hPa. The regression coefficients chosen depend on the latitude band of the measurement, and the values are varied sinusoidally to allow for the seasonal cycle.
  6. Interpolation onto a global grid at 12GMT.
    Radiances and retrieved thicknesses are linearly interpolated in both space and time onto a regular 5 degree latitude by 5 degree longitude grid at 12 GMT. The value at each grid point is calculated from all observations within 500km and within 12 hours of 12 GMT.
  7. Filling in missing values
    Where there is missing radiance data, grid points may have no data assigned to them. In this case a search is performed over 20 grid lengths for valid data and linear interpolation is used. If no data is found within this range then a global mean value is substituted. Note that the number of missing grid points is recorded in the file header and if more that 650 points are missing then the day's analysis should be discarded.
  8. Analyses of Geopotential Height
    Global fields of geopotential height for the stratosphere are constructed by adding the gridded thicknesses to an objective analysis of geopotential height at 100hPa. This analysis comes from an operational weather forecasting centre (either the Met Office or the European Centre for Medium-range Weather Forecasts). This analysis also provides the data fields for the tropospheric levels. The result of this process is a set of geopotential heights at 850, 500, 300, 200, 100, 50, 20, 10, 5, 2 and 1 hPa.
  9. Smoothing
    The gridded fields of radiances, thicknesses and heights are then smoothed in the horizontal by Fourier analysis. Zonal wavenumbers greater than 12 are discarded and the Fourier coefficients of remaining zonal wavenumbers are smoothed by discarding pole-to-pole wavenumbers greater than 12. This is a compromise between almost complete coverage at high latitudes and much poorer coverage at lower latitudes (See Fig. 1a.)

3. The Data

This section contains a brief summary of the information in the Hadley Centre Climate Research Technical Note (CRTN24) on the TOVS analyses (Bailey et al., 1992). You are strongly advised to consult this document before using the data.

3.1 Spatial Coverage

3.1.1 Vertical Coverage

The Geopotential height data are available on a set of levels from 850 to 1 hPa. For the levels available see the section on resolution below. Radiance data are available for the 8 channels of the TOVS instruments which are used in the Met Office Analysis system. For the weighting functions for these channels refer to Fig. 1b.

3.1.2 Horizontal Coverage

The geopotential height and radiance data are both stored on a grid which covers the whole globe. For further information, see the section below on the data resolution.

3.2 Temporal Coverage

The analyses are available from the BADC for a continuous period from January 1979 to June 1998. There is a separate file for each month of data.

3.3 Resolution

Both the radiance and the geopotential height fields held at the BADC are on a 5 degree by 5 degree latitude-longitude global grid. There are 37 latitudes from 90 degrees N to 90 degrees S and 72 longitudes from 180 degrees W to 180 degrees E. The geopotential heights are stored on the following set of pressure levels:

850, 500, 300, 200, 100, 50, 20, 10, 5, 2 and 1 hPa

The Radiance data are supplied for up to 11 channels of the TOVS instruments.

3.4 Data Quality

The accuracy (precision and systematic biases) of stratospheric analyses produced by Met Office is influenced by the following factors:

  1. The accuracy and stability of the measuring instruments.
  2. The vertical resolution of the measurements. This is limited by the width and overlap of weighting functions.
  3. The horizontal resolution of the measurements. This is limited by the number of satellite orbits and the angular range of side scanning.
  4. The retrieval method which derives thicknesses from radiances. The reference a priori data used in the retrieval cannot represent the full range of stratospheric conditions.
  5. The interpolation of observations. Errors inevitably arise when observations from orbits spread throughout the day are interpolated onto a regular latitude longitude grid at a specified synoptic time.
  6. The accuracy of the 100 HPa analysis which is used as a base level. Though operational weather forecasting centres such as the Met Office continue to improve their tropospheric analyses, errors remain as evidenced by differences between independently produced fields, particularly in the southern hemisphere where ground-based data are sparse.

The effect of each of these factors is difficult, if not impossible, to assess definitively given the limited amounts (or complete lack of) independent data for the middle and upper stratosphere. The reliability of derived fields is strongly influenced locally by the vertical and horizontal scales of the actual circulation.

Bailey et al. (1992) discuss a number of tests of the data quality and conclude that the principal limitation is the inherently low vertical resolution, which can blur comparatively small-scale baroclinic structures in the stratosphere (especially during strong warmings), and which can lead to discrepancies in diagnostic calculations, especially when second (or higher) orders of spatial differentiation are involved.

When examining data for a particular day, users are advised to check the following indicators of data quality detailed in the appendix:

  1. The coverage code, which gives information about the data used to produce the analysis.
  2. The number of missing grid points (for gridded data). If this exceeds 650, then consider rejecting that day.
  3. Missing data indicators, which show which fields, if any, are missing on a given day.

See the file format document for the locations of these parameters.

4. Using TOVS analyses from the BADC

The TOVS analyses are covered by a data protocol between the Natural Environment Research Council (NERC) and the Met Office. This is to allow the Met Office to keep track who is using the data.

Since the dataset is restricted, it is not available anonymously and cannot be accessed through the WWW. If you want to use the data, please apply for access to the Met Office TOVS data.

4.1 Transferring Data from the BADC

All data and software is located beneath the directory /badc/ukmo-tovs. Here you will find a README file and the following subdirectories

4.4 Reading the data files

4.4.1 File structure

This section contains an overview of the file structures of the TOVS radiance data and geopotential height analyses. The detailed file formats are also available online.

The radiance and geopotential height datasets use a similar file format. In both cases, the data are stored on a global grid with a 5x5 degree resolution (37 latitudes for 90 degrees N to 90 degrees S and 72 longitudes from 180 degrees W to 180 degrees E). For each grid-point, the radiance files contain the data for the 8 channels used by the analyses, whilst the height files contain the geopotential heights on the 11 pressure surfaces.

The data are stored in files which contain the data for a single calendar month. Each day's data within the files contains 38 records, consisting header record and 37 records, one for each latitiude row. Days are stored in chronological order.

Schematically the files have the following structure:


     | Day 1 : Header Record        |
     | Day 1 : 1st Latitude (90N)   |
     | Day 1 : 2nd Latitude (85N)   |
     | Day 1 : 3rd Latitude (80N)   |
     | Day 1 : 37th Latitude (90S)  |
     | Day 2 : Header Record        |
     |______________________________|        39th record = Header for 
     | Day 2 : 1st Latitude (90N)   |        2nd day's data.
     | Day 2 : 2nd Latitude (85N)   |
     | Day N : 37th Latitude (90S)  |        N=Last day in the month

Each record contains 1080 data items of type integer*2. Missing data has a value of -32768. The data is encoded to fit into the arrays. If you are using the geopotential height analyses, then you should multiply the raw values by a factor of 2 to get heights in metres. The factors for the radiance data are channel dependent. You should consult the format document to find out how to convert the data values to radiances.

4.4.2 Access Routines

A set of FORTRAN subroutines to read the data are available from the TOVS software directory at the BADC. These are the routines supplied by the instrument team at the Met Office which should compile and run on the majority of platforms.

A set of access routines written in IDL at the BADC is also available.

4.4.3 Third Party Software

In the software directories for the TOVS data, there is a subdirectory for 3rd Party software. If you develop software to use with the TOVS data, then we encourage you to let us have a copy so that we can make it available to other researchers.