Data Set Overview
  =================
    The natural form of imaging spectrometer data is the spectral
    image cube.  It is normally in band sequential format, but has a
    dual nature.  It is a series of 'images' of the target, each in a
    different wavelength, in ascending order.  It is also a set of
    spectra, each at a particular line and sample, over the target
    area.  Each spectrum describes a small portion of the area.  When
    transformed into cubes, the data may be analyzed spatially, an
    image at a time, or spectrally, a spectrum at a time, or in more
    complex spatial-spectral fashion.
 
    NIMS Spectral Image Cubes are derived from NIMS Experiment Data
    Records (EDRs), which contain raw data from the Galileo Orbiter's
    Near Infrared Mapping Spectrometer [CARLSONETAL1992].  The
    instrument covers the spectral range 0.7 to 5.2 micrometers,
    measuring both reflected sunlight and emitted thermal radiation.
 
    The cubes are of two kinds.  TUBES contain data which have been
    converted to science units (radiance or I/F) and rearranged into
    band sequential form, but are unresampled and in NIMS instrument
    space.  Tubes are made for all NIMS observations during Jupiter
    operations, and form this dataset.  Data from most observations
    are *also* binned and resampled while being projected onto the
    target, forming MOSAICS (or G-CUBES) which comprise the separate
    MOSAIC dataset.  Geometric and other information is stored in
    backplanes of both kinds of cubes.
 
    This dataset includes radiance and I/F tubes of all Galileo NIMS
    observations of Jupiter and its satellites, and associated flight
    calibration data.  Observations began in June, 1996 and will
    continue until the (unknown) end of the mission.  Spacecraft
    clock (SCLK) times begin at 3449820:00.  The tubes are archived
    in volumes beginning with GO_1104, each of which contains both
    tubes and mosaics from (usually) a single orbit of Jupiter, as
    well as browse products (masks) which provide an overview of the
    data.  (Earlier volumes in this series contain data from Galileo
    flybys of Venus, the Earth and the Moon.)
 
 
  Parameters
  ==========
    A band in a NIMS tube is generated for each of the 17 detectors
    at each grating step.  (The detectors cover the range 0.7 to 5.2
    microns.) The motion of the grating is determined by the
    commanded instrument mode:
 
    Mode                         Grating   Grating   Bands  Samples/RIM
                                  steps   increment
 
    Fixed Map/Spectrometer          1         0        17      182
    Short Map/Spectrometer          6         4       102       26
    Full Map/Spectrometer          12         2       204       14
    Long Map/Spectrometer          24         1       408        7
 
    A secondary mirror moves through twenty cross-track positions in
    the map modes, or is fixed near the center of its scan in the
    spectrometer modes.  The number of lines in each image of a tube
    is always twenty, whether or not the mirror is moving.  The
    number of samples is determined by the mode and the duration of
    the observation.  (In the mosaic dataset, the image dimensions
    are determined by the pattern created by the motions of the
    secondary mirror and the scan platform.)
 
    The approximate wavelengths of the bands are determined by the
    mode, and by offset and start grating positions.  The true
    wavelengths are functions of the temperature of the grating and
    parameters determined from the ground calibration and frequent
    optical flight calibrations.  Known absorptions on some targets
    are also useful in determining these parameters.
 
    The commanded gain state is one of four sets of gains for the 14
    non-thermal detectors.  The three thermal detectors have two
    gains, automatically switching to the lower one near the
    mid-point of their range.  Raw data values of each detector and
    grating step are functions of the gain state and the temperature
    of the focal plane assembly (FPA).  Radiances are determined from
    raw data values using sensitivities based on the original ground
    calibration corrected by frequent photometric and radiometric
    flight calibrations.  I/F values are simply radiances divided by
    the solar absorption at the target's distance from the sun for
    the wavelength in question.
 
    The NIMS grating motion capability failed prior to the I24
    encounter.  Subsequently, NIMS returned only 12 useful
    wavelengths instead of the original 408.  Obviously, this reduced
    the usefulness of the NIMS data in certain respects.  However the
    resulting data redundancy turned out to have noise reduction
    advantages in the high radiation environment of Jupiter, and very
    useful products were generated.
 
    Further details may be found in VOLINFO.TXT in the DOCUMENT
    directory of the archive volume, and in the instrument paper
    [CARLSONETAL1992].
 
 
  Processing
  ==========
    Tube files in this dataset were generated by the Multimission
    Image Processing System (MIPS) at the Jet Propulsion Laboratory
    (JPL) from raw NIMS data on EDRs, which are available in a
    separate volume series (GO_10xx).  For each planned observation,
    raw 10-bit data numbers have been re-arranged into band
    sequential form and converted to spectral radiance and I/F units
    using the best calibration available at the time of processing.
 
    Wavelength values are based on wavelength calibrations described
    above.  Radiance values are derived from 10-bit NIMS raw data
    numbers by first subtracting average dark values (determined for
    each detector and mirror position and for each gain state), then
    dividing by band-dependent sensitivities, determined as described
    above.  I/F values are derived from radiances by dividing each
    NIMS spectrum by the solar spectrum at the target's distance from
    the sun.  Backplanes of geometric information and projection
    co-ordinates are based on the position of the target and
    spacecraft and the orientation of its scan platform, derived from
    SPICE information from the Navigation Ancillary Information
    Facility (NAIF) at JPL.
 
    A secondary hardcopy mask and its digital image (for the volume)
    were also generated, based on the mosaic (if one was produced) or
    the tube (if not).  In the latter case, the mask includes a
    footprint plot of the observation on the target, averaged spectra
    from one to six selected areas keyed to the plot, a 2-D histogram
    and annotation.  The masks serve as browse products for the
    cubes.
 
 
  Data structure
  ==============
    The tube files follow PDS structure and labeling conventions.  A
    PDS/ISIS label begins each file, and describes all the 'objects'
    within using ASCII keyword=value statements.  The first object is
    an ISIS history object which describes the various steps of the
    generation process.  The second object is a 2-D histogram of the
    tube.  A third object is a 'sample spectrum qube': a 'stack' of
    six spectral plots, each an average over a selected area of the
    tube.  (These also appear on the hardcopy and digital 'masks'.)
    The fourth and principal object is the actual NIMS spectral image
    tube.
 
    Spectral image tube structure follows PDS and ISIS 'qube' object
    standards.  The 'core' of the qube is a 3-dimensional array of
    32-bit VAX floating-point pixels arranged in band sequential
    order: sample, line and band.  The pixels are in radiance or I/F
    units.  Attached labels describe the structure and units of the
    tube and include vectors containing wavelengths and sensitivities
    of the bands.
 
    The core is followed by a set of backplanes, or 'extra' bands,
    with 32-bit VAX floating-point pixels.  The backplanes contain a
    number of geometric parameters, native time, projected line and
    sample and 0 to 10 'spectral index' bands, each a user-specified
    function of the data bands.  (The latter might be ratios of
    bands, or band depths.) The geometric backplanes are of latitude,
    longitude, incidence, emission and phase angles, slant distance
    and 'intercept altitude'.  Projected line and sample backplanes
    describe the position of each pixel had the data been actually
    projected on the target, which, in a tube, it has not.  Due to
    the way NIMS acquires spectra in modes with multiple grating
    steps, there are multiple backplanes of each of latitude,
    longitude and several projection co-ordinates, one backplane for
    each grating position (up to 24) in the instrument mode.  (See
    comments in the tube label for details.)
 
 
  Ancillary Data
  ==============
    A Postscript-format Guide to the planned observations, including
    footprint plots on the target, instrument parameters, etc.  is
    included in the data set, as are tables of parameters for each
    observation.  (Most of these parameters are also present in the
    tube labels.) A preprint of the NIMS instrument paper
    [CARLSONETAL1992] is also included.
 
    Calibration files, average dark value files and SPICE files
    (spacecraft positions, planetary positions and constants,
    processed pointing geometry, spacecraft clock versus universal
    time, etc.) were used in generating the tubes from EDRs but are
    not included in this dataset.  (They will be published in a
    separate volume at a later time.) However calibration information
    is present in the tube labels as vectors of sensitivity (for each
    wavelength) and of average dark value (for each detector).  Much
    of the geometric information is present as backplanes of the
    tube.
 
 
  Software
  ========
    NIMS tubes were designed to be accessed by the ISIS system, which
    includes extensive software for generating, manipulating,
    analyzing and displaying spectral image cubes.  ISIS exists in
    VMS and Unix versions, which must be obtained independently, as
    described in the documentation of this data set.
 
    Other software has since been developed for displaying spectral
    image cubes, notably the ENVI system, written in the IDL language
    and available from RSI, Inc.  for Unix and Windows systems.
 
    Simple multi-platform software for displaying bands, backplanes
    and spectra of cubes is being developed as an enhancement to the
    NASAVIEW image display program by PDS.  NASAVIEW supports Unix,
    PC and Mac systems.
 
 
  Media/Format
  ============
    The NIMS tubes and 'mask' images are archived on CD-ROM for
    distribution by the Planetary Data System (PDS).  Formats are
    based on standards for such products established by PDS.
    Specifically, the discs are formatted according to the ISO 9660
    level 1 Interchange Standard, and file attributes are specified
    by Extended Attribute Records (XARs).

Confidence Level Overview
  =========================
    NIMS tubes in this dataset contain unresampled data from all
    observations of Jupiter and its satellites that were successfully
    returned to earth.  The raw data numbers from the EDR dataset
    have been re-arranged into spectral image cube form in NIMS
    instrument space, and converted to radiance and I/F by applying
    the best available calibration.  Geometric information has been
    added as backplanes of the tubes.
 
    Users of NIMS tubes should be aware of (1) the imperfect nature
    of the calibration used to produce them, particularly the
    wavelength calibration, (2) the effect of pointing errors on
    their geometry and (3) the presence of radiation-induced noise in
    the data.  These factors are the principal sources of uncertainty
    in the dataset.  There are also many gaps in the planned spatial
    and/or spectral coverage of many observations due to limited
    playback time and unplanned interruptions of data receipt at
    downlink stations.  However it should be mentioned that the Rice
    compression algorithm used in playback of NIMS recorded data is
    lossless, and does *not* contribute any uncertainty to the data.
 
 
  Review
  ======
    NIMS data is reviewed by the NIMS team as it is being received,
    and again before archiving.  Since NIMS tubes are generated for
    *all* observations, and since the data may be readily viewed as
    images and as spectral plots, they are preferable to EDRs for
    checking data quality and coverage.  (Unlike data in mosaics
    (g-cubes), tube data have not been resampled.) Spatial coverage
    information is added to the pointer (footprint) plots of each
    observation, and spectral coverage and timing information are
    tabulated.  The NIMS Guides are revised to include this
    information, and the revised Guides included on the archival
    volumes.
 
    Several passes through the Galileo tape recorder are usually made
    for each orbit, so that coverage information from the first pass
    is used to fill gaps due to DSN outages, or to extend spatial
    and/or spectral coverage if playback time is available.
    Compression ratios are also checked, especially in the first
    pass, and used to more accurately predict playback time during
    the second pass, so as to maximize the amount of data that can be
    returned.  Knowledge of typical compression ratios is useful for
    planning return of similar data in subsequent orbits.
 
    Format and documentation of each archival volume is reviewed by
    several NIMS team members, then by MIPS personnel, who format the
    ancillary files according to ISO 9660 standards and write CDWOs,
    and finally by PDS before mastering.
 
 
  Data Coverage and Quality
  =========================
    Due to the failure of the Galileo high gain antenna to deploy,
    the amount of data returned from Jupiter operations was severely
    limited.  Techniques were developed to maximize the scientific
    value of the data that could be returned, including selective
    editing of data in wavelength and mirror position, lossless
    compression of the selected data, and an optional thresholding
    capability.  The resulting coverage is described in several
    places on the volumes, to which readers are directed.
 
    Information about planned observations is collected in NIMS
    Guides to each encounter before that encounter takes place.
    Coverage and quality information is added to each Guide after
    data is received and before it is archived.  The NIMS Guide for
    an encounter is then included on the same volume(s) as the data
    files for that encounter.
 
    The PDS label of each tube contains information about data
    coverage in the observation: time range, ranges of latitude and
    longitude, and of other geometric parameters, wavelengths of the
    bands in the tube, compression ratio during playback, etc.  In
    addition, the index table on each volume collects much of the
    geometric information from the tube (and mosaic) labels of all
    observations with products on that volume.  Likewise, spectral
    and spatial editing information may be found in the 'observation
    table' on the volume.  Cumulative versions of these two tables
    are also present.
 
    In addition, complex searches for data under various temporal,
    geometric and other constraints may be conducted on the World
    Wide Web using the Planetary Atlas at the PDS Imaging Node.
    These make use of the index table and labels described above.
 
    Data quality is best judged by actually viewing the data products
    themselves.  Radiation noise is the principal problem, roughly
    inversely proportional to the spacecraft's distance from Jupiter
    at the time the data is taken, and more pronounced in the higher
    wavelengths than in the lower ones.  Attempts to remove
    significant radiation noise by a despiking technique have been
    made for (resampled) mosaics of the icy satellites, but not of
    Jupiter and Io data.  Tubes, on the other hand, are never
    despiked and the quality of the observation should be apparent on
    viewing.
 
    Tubes may contain special values in place of some data values,
    indicating conditions such as missing data (lost during telemetry
    or removed by editing, in wavelength or secondary mirror
    position, on the spacecraft), instrument detector saturation,
    data below threshold (when an observation was returned using the
    thresholding capability), etc.  These special values are assigned
    according to ISIS conventions and documented in the label.
    Sizeable spatial gaps (containing NULL data) in some or all bands
    may be due to unplanned (and occasionally planned) interruptions
    in the playback process.
 
 
  Limitations
  ===========
    As mentioned in the overview above, the principal uncertainties
    in the NIMS dataset are due to errors in the calibration and in
    the pointing geometry and to the presence of radiation noise.
    Here, we discuss their effects on the NIMS Jupiter tube dataset.
 
 
    Calibration
    -----------
      A set of calibration parameters called the 1998A Calibration,
      derived early in that year, was used for NIMS archival data
      products from the Galileo Primary Mission, orbits G1 to E11.
      Subsequently, several errors and idiosyncrasies in the
      calibration were discovered and are still being analyzed.  They
      are summarized here.
 
      (1) The photometric calibration of Detectors 1 and 2 is
      uncertain, due to temporal changes in their spectral
      responsivities as well as in the onboard calibration target
      used for correction.  The spectral response for these two
      detectors is therefore incorrect.
 
      (2) The wavelength position behaviour of the instrument has
      changed over the orbital mission, likely due to radiation
      damage.  A grating rotation parameter (PSHIFT), in units of
      grating steps, reflects thermal changes in the instrument and
      mechanical creep of the grating.  A value of -1.3 was used
      during orbits G1-E4, and -1.0 during E6-E11.  However this was
      discovered to be inadequate after E4, most likely because the
      grating step size has gradually increased.  This increase is
      not reflected in the wavelengths listed in the labels of Prime
      Mission products (G1-E11).  However a new calibration was
      developed for the Galileo Europa Mission (GEM) which accounts
      for the changing grating step size.  Labels of GEM products
      (E12-I25) therefor contain more accurate wavelength values.
 
      (3) The dark levels for some detectors and gain states have
      changed with time, particularly for Detector 10 in the highest
      gain state, where the dark level has dropped by about 10 data
      numbers (DNs).  We used constant values for the 1998A
      Calibration.
 
      (4) The dark level of Detector 9 is slightly increased by the
      signal level of Detector 17, due to electronic crosstalk.  This
      occurs when Channel 17 is in the low gain state.  For most
      applications it is a small effect (a few DNs), occurring
      generally when viewing Jovian 5-micron hot spots.
 
      (5) Two detectors have become inoperative.  Detector 3 failed
      during orbit E4 and Detector 8 began producing erratic signals
      in orbit E6.
 
      (6) Three more detectors (1, 2 and 7) lost almost all of their
      sensitivity (and usefulness) after the NIMS grating stuck prior
      to the I24 encounter, and remained stuck.
 
      See section 6A of VOLINFO.TXT (found in the DOCUMENT directory
      of any NIMS Cube volume) for a description of how calibration
      parameters are determined, and a more detailed discussion of
      the limitations described above.
 
      Methods to correct for the above-listed errors and
      idiosyncrasies are being developed.  Correction methods and
      parameters (revised wavelengths, radiance corrections, revised
      dark levels) will be placed on the NIMS website when available:
 
      http://jumpy.igpp.ucla.edu/~nims/
 
      or follow the link from the Galileo website:
 
      http://www.jpl.nasa.gov/galileo/
 
      to nimsinfo@issac.jpl.nasa.gov.
 
 
    Pointing geometry
    -----------------
      The geometry of NIMS observations is determined by two classes of
      datasets:  (a) scan platform pointing data (either 'predict' or
      provided by AACS downlink) and (b) data defining the relative
      positions of spacecraft and target body (ephemerides).  The latter
      are highly reliable, with errors generally less than half a
      milliradian.  However, pointing data are subject to uncertainties
      of various kinds.
 
      In 'normal' operation (inertial mode, in which both gyros and star
      sensor are functioning well) and with cone angles less than 150
      degrees, the absolute pointing error is close to the nominal one
      of 1.0 mrad in cone and clock.  However the relative pointing
      error within a given observation is better than this, about 0.2
      mrad.  The absolute error is 'reset' whenever the scan platform is
      moved to a new 'aim point', which usually happens at the beginning
      of an observation.
 
      When the cone angle is greater than 150 degrees, wobble
      compensation is disabled and the scan platform pointing shows an
      oscillation with a period equal to that of the rotor spin (18 sec)
      and an amplitude that varies but is at most about 1 mrad.  In
      principle this is not an additional error source, since it is
      still tracked by AACS, but it degrades the NIMS data since they
      are no longer Nyquist-sampled, and it is possible that the
      relative pointing error is increased in this mode.
 
      When the gyros are turned off (cruise mode), then the relative
      error is considerably increased, but the absolute error remains
      about the same.  After orbit E11 one of the gyros started to
      degrade, adding an uncertainty in one dimension immediately after
      performing a clock or cone slew in the positive direction.  The
      errors depend on the size of the slew.  A correction was developed
      by AACS which gradually removes this error using the star scanner.
      In E14, this procedure was able to remove 10% of the error every
      spacecraft revolution (about 18 seconds); by E19 this fraction was
      increased to 40%.  Cruise mode was used increasingly as the GEM
      mission progressed.
 
      Occasionally, the star scanner was disabled, either for 'bright-
      body avoidance' (E2 encounter, Jupiter orbit G1 and others) or due
      to an anomaly such as a guide star being incorrectly acquired (G7
      orbit). In this case, the pointing error is unpredictable and can
      be very large.
 
      When a significant portion of the target body's limb is in the
      NIMS field of view, then the pointing can be corrected on the
      basis of our knowledge of the relative geometry, which is always
      much better than that of the scan platform pointing.  This can
      reduce the error to about 0.5 NIMS FOV, or about 0.25 mrad,
      although it is usually somewhat greater since the limb is not
      always sharply distinguishable due to the shape of the NIMS
      response function and the scan-platform wobble.  (If the limb is
      visible in an SSI image during the same observation, the pointing
      error can be determined to essentially zero error for that one
      point in time, but this rarely occurs in Jupiter observations.)
 
      Pointing corrections of this sort are documented to some extent in
      the labels and history objects of the cubes to which they were
      applied.  A discussion of pointing anomalies in a particular orbit
      may be found in the NIMS Guide to that orbit.
 
      See section 6B of VOLINFO.TXT (in the DOCUMENT directory of any
      NIMS Cube volume) for a detailed description of how geometry data
      is obtained and corrected, and on its limitations, and for a guide
      to the documentation within individual products.
 
 
    Noise
    -----
      Attempts to correct for radiation noise by despiking were made on
      mosaics (g-cubes, resampled products) of the icy satellites, but
      the (unresampled) data in NIMS tubes were NOT corrected to remove
      noise.  For a discussion of this topic, see the Jupiter mosaic
      dataset description (JUPMDS.CAT).