From: "Randolph L Kirk"
Subject: Feedback solicited on format of future USGS Mars products
Dear Colleague:
The purpose of this message is to describe a series of new Mars
cartographic datasets that will be produced by the USGS over the next
several years and to solicit your feedback on various aspects of the design
of these products that represent departures from our past and present Mars
mosaics.
To begin with, we are pleased to announce that we have completed the first
of several planned revisions of the global digital image mosaic of Mars, or
MDIM. The original version, MDIM 1.0, was released in 1991 and has been
widely used despite increasing evidence of global, regional, and local
positional errors compared to later, more accurate map products. The new
mosaic, MDIM 2.0, has been made from the same set of ~4600 Viking Orbiter
images as the original, but with improved photometric/cosmetic processing
and substantially improved geodetic control. The root-mean-square (RMS)
positional error between MDIM 2.0 and MOLA data is about 3 km. MDIM 2.0 is
an interim product that will be superseded by later mosaics as described
below. We have therefore mailed a limited number of copies to the NASA
Regional Planetary Image Facilities (RPIFs) worldwide and to Mars mission
teams. The compact disk (CD) volumes distributed are (as nearly as
possible) identical to the 1991 design. The mosaic is also available
through the Planetary Data System (PDS) Map-a-Planet website (
http://pdsmaps.wr.usgs.gov/maps.html), where users can order a custom map
of any region of Mars and choose the resolution, map projection, and data
format they require.
Whereas retaining the 1991 archive design was an appropriate way to speed
the release of the interim MDIM 2.0 product, we would like to improve upon
this design for the three major global mosaic products to be produced in
the next few years. These products are:
1) A final version of the Viking global mosaic (MDIM 2.1) with further
improved cosmetic appearance and, more importantly, absolute positional
accuracy on the order of the 231-m pixel size obtained by using MOLA data
to constrain the control network. This mosaic is in production and should
be finished late in 2001. It (and the products described below) will be
mass-replicated and distributed to individual NASA planetary geology and
geophysics investigators and to the RPIFs.
2) Updated versions of the Viking global color mosaic, which was produced
in the early 1990s and was widely used but never officially released.
Funding for these mosaics has been requested; if the project were approved,
the mosaics would be completed late in 2002. The color images have
resolutions close to 1 km/pixel; the mosaic would be produced at 1/64
degree (~926 m) per pixel, and a version merged with MDIM 2.1 at 1/256
degree (231 m) would be generated.
3) Color and monochrome mosaics based on MGS MOC images from the geodesy
campaign. Red-camera images are available at 240 m/pixel (nadir) and blue
at 480 m/pixel. From these a global quasi-true-color mosaic with a
synthetic green channel will be produced at 1/256 degree resolution. In
addition, nadir and oblique images from the red camera will be used to
produce a stereopair of mosaics at this resolution, filtered and normalized
to emphasize topographic features. We refer to these mosaics collectively
as MDIM 3.0. Funding for their production has been allocated, with a
target date for completion late in 2003.
In addition to these major digital datasets, we also have proposed to make
printed maps of Mars based on MOLA data. A global map on one sheet at
1:25,000,000 scale has been proposed for production in the coming year, and
additional maps at larger scales would certainly be possible given the
quality of the data. Although most of the issues we raise here concerning
digital archive design are not relevant to printed maps, the choice of
latitude-longitude coordinate system does affect printed as well as digital
products.
The desirability is obvious of making the design of the three new digital
products as similar as possible. Technical advances since the 1991 release
of MDIM 1.0 make some departures from the design of that product possible,
and experience with that product and others over the past decade indicates
additional changes that would be desirable. We summarize our best ideas
for the design of future Mars digital maps below, contrasting them with the
design of MDIM 1.0-2.0, and invite your comments and suggestions. Please
respond no later than 13 July 2001 if you have objections to any aspect of
the proposed design or wish to suggest modifications.
DISTRIBUTION MEDIUM
We will continue to archive and distribute digital cartographic products on
optical disk media and to design the disks to meet the requirements of the
PDS. Whereas MDIM 1.0 and 2.0 were released on CDs, we intend to use
Digital Versatile Disk (DVD) media for future products. DVDs offer
advantages in reduced cost (and bulk); until recently these advantages have
been offset by the limited availability of DVD readers, but we believe that
this is no longer a significant obstacle. The cost savings for
distributing large datasets on DVD are substantial. A single-side,
single-layer DVD holds 4.7 GB of data compared to 0.66 GB for a CD. This
means that a single-band, 1/256 degree per pixel MDIM can be archived on a
single such DVD rather than on six CDs. The cost of replicating a DVD is
greater than that for a single CD, but the net cost of creating and
shipping 200 copies of a dataset on DVDs is less than one-quarter of the
cost of using CDs.
DIGITAL SCALES
We will continue to use scales (in degrees per pixel) equal to an inverse
power of two as recommended by the NASA Planetary Cartography & Geologic
Mapping Working Group (PCGMWG). This convention helps ensure that
different datasets can readily be combined or compared by doubling or
halving pixels to achieve a common scale. The Viking-derived MDIM 2.1 and
MGS-derived MDIM 3.0 will both be produced at 1/256 degree/pixel scale.
The Viking color mosaics, if produced, will be made at 1/64 degree and
merged with MDIM 2.1 at 1/256 degree.
TILING SCHEME
The term "tiling scheme" refers to the division of the global mosaic into
individual files. We strongly support the idea, also developed by the
PCGMWG, that digital map products should be divided into files that
correspond directly to a series of printed maps of the same body. The
question remains as to which Mars map series is an appropriate basis for
tiling future mosaics. MDIM 1.0 was divided into almost 2000 files
corresponding to the quadrangles of the 1:500,000-scale map series. The
small size (1.6 MB) of these files was appropriate in 1991 but modern
computers can easily open much larger images and the need to piece together
many small images is a source of frustration. We propose to base the
tiling of future Mars mosaics on the 1:2,000,000-scale map series. This
series contains 140 quadrangles and yields file sizes (for a single-band,
8-bit dataproduct at 1/256 degree per pixel) of about 22 MB.
We believe that the 1:2,000,000-scale series yields the best compromise
between file size and number of files. No published map series has a scale
between 1:500,000 and 1:2,000,000, so if smaller files were desired it
would be necessary to subdivide the 140 quadrangles into quarters of ~5 MB
each. In the other direction, the 1:5,000,000-scale series has only 30
quadrangles and the largest of these are 160 MB in size, which is probably
unacceptably large.
Files will not be trimmed to the exact boundaries of the map sheet, but
will include a border of extra pixels to ensure that no gaps form when
adjacent files are mosaicked together. The number of pixels of extra data
will be chosen to ensure that gaps do not arise even when files are
rescaled by a factor of a power of two for comparison with other datasets.
MAP PROJECTION
The two projections appropriate for global digital databases are Sinusoidal
and Simple Cylindrical. Of these, Sinusoidal projection has been used for
past digital maps of Mars and other bodies, including MDIM 1.0 and 2.0.
The equal-area property of this projection is useful in some scientific
investigations and leads to a global archive that uses disk space
relatively efficiently (at least when there are a large number of files;
when the number of files covering the planet is small, the efficiency of
Sinusoidal projection approaches that of Simple Cylindrical). The main
disadvantages are that individual datafiles must be resampled before being
mosaicked together because their center longitudes of projection differ and
that for large regions the projection leads to objectionable distortions
away from the equator and center longitude. Because the sides of the
quadrangles are curvilinear in Sinusoidal projection, each file must be
padded with a few pixels of data from adjacent quadrangles to prevent gaps
from forming when files are reprojected and mosaicked. If the data are to
be used at reduced scale then even more padding will be required. The use
of Sinusoidal projection also complicates construction of the mosaic;
either each quadrangle must be built as a separate mosaic, with the same
images being map projected several times into neigboring quads, or mosaics
of larger regions must be built and then resampled to the center longitudes
of the multiple quadrangles they cover. To date the latter approach has
been used despite the undesirable degradation of the image that results
from the extra resampling step.
Like Sinusoidal projection, Simple Cylindrical projection has distortions:
both area and east-west scale are increasingly misrepresented toward the
poles. . The main advantages of this projection are that any number of
adjacent files can be joined without reprojection and that many software
systems (e.g., for visualization of map data as virtual globes) understand
Simple Cylindrical even if they handle no other map projections. The only
parameters needed to describe a quadrangle in this projection are the
digital scale and the latitude-longitude ranges. Another advantage is
that, because quadrangle b oundaries are straight lines in the row and
column directions, no padding of the files with data from adjacent
quadrangles is needed to insure seamless mosaics.
Users with access to software (e.g., ISIS, VICAR) that understands a
variety of projections can produce maps with equal-area, conformal,
equidistant, or other properties as desired from either Sinusoidal or
Simple Cylindrical databases. Because we will, in the future, produce
either type of database with a single resampling of the images, the
quality of the results will be at least as good as past Sinusoidal maps.
Unfortunately, not all users of the existing Mars data have access to such
software, and, as a result, we routinely receive requests for digital map
data in Simple Cylindrical projection. We therefore propose to make and
archive all future global digital maps in Simple Cylindrical projection.
The modest increase in required storage space is not a substantial
deterrent, given modern storage media.
Given that the proposed switch to Cylindrical projection is partly intended
to meet the needs of users outside the NASA-funded community, we will be
especially interested to hear whether there is a strong preference by
members of that community (to whom ISIS and VICAR are available) for
retaining Sinusoidal projection. If so, one possible compromise would be
to produce PDS-formatted images in Sinusoidal and another format such as
GeoTIFF (see next section) in Simple Cylindrical projection.
DATA FILE FORMAT
MDIMs 1.0 and 2.0, as well as numerous other digital map archives, have
contained data files in formats approved by the PDS and unique to the NASA
planetary cartography community. Such files include labels, readable both
by humans and by machines, that provide a complete and well-documented
description of the nature and map projection of the data in the file. The
detailed information in the PDS labels maximizes the utility of data in
this format for users with access to sophisticated planetary cartography
software such as ISIS and VICAR. Multispectral data can be stored as
multiband image "cubes" or as individual bands in separate files, and in
either case the software mentioned provides freedom in comparing and
displaying different spectral channels. The PDS standards also recommend
that the image data be uncompressed; if desired, an entire image file can
be archived in a standard losslessly-compressed (i.e., zip) format. This
recommendation addresses the need to preserve the quantitative accuracy of
the data for scientific analysis. We strongly agree with the rationale
behind the PDS data format as just described and intend to always provide
digital map archives in such a format. These PDS files would be provided
in Sinusoidal projection for the reasons outlined above.
There are uses, however, for which other formats may be more appropriate.
Many users simply wish to view an image of part of Mars or to use it to
prepare an illustration. Even if they have access to sophisticated
cartographic software, which they may well not, translating PDS files into
such a system, reprojecting, mosaicking, and combining spectral bands is a
time-consuming process that must often be followed by translating the data
into a more widely recognized format so it can be shared. We therefore
propose to include with future archives data in widely readable non-PDS as
well as PDS formatted files. The non-PDS files will have the same scales
and extents. As discussed above, these files will be in Simple Cylindrical
projection to facilitate combining them into larger regions, regardless of
whether Simple Cylindrical or Sinusoidal projection is adopted for the PDS
files. Where appropriate, three-band color data will be presented in a
single file.
We propose to adopt the GeoTIFF standard for the non-PDS image files. This
format incorporates map-projection information into the labels of the
well-documented TIFF format. Presently, only the map scale and a tiepoint
giving the latitude and longitude of the first pixel are defined for bodies
other than Earth, but this is sufficient to define a dataset in Simple
Cylindrical projection. Some Geographical Information System (GIS)
software (e.g., ArcView from ESRI) can read this projection information,
and a much larger number of software packages will be able to open the
images without comprehending the projection. The TIFF standard also
includes a variety of image compression algorithms. Use of a lossy
compression scheme can reduce the volume of a full-color image to
substantially less than the uncompressed size of a single band without
introducing visible artifacts. (If the unmodified data are needed they
will be available in the corresponding PDS-formatted file). The TIFF
standard also includes data types with more than one bit per pixel, useful
for digital elevation models (DEMs) and other geophysical datasets, though
not all software recognizes all TIFF pixel types. We will adopt a
different non-PDS file format (e.g., JPEG) if users strongly prefer it, but
the combination of basic map-projection information and flexible image
compression makes TIFF the best choice in our opinion.
EXTRAS
The GeoTIFF files just described are one form of "extra" data that we
propose to include in addition to the basic, PDS-formatted files. Past
digital map releases have incorporated several other types of "extras,"
some of which we propose to include in the future as well. We welcome
suggestions for extra datasets that would be useful without requiring large
amounts of processing and/or storage space.
Past PDS archives have included reduced-resolution versions of the data,
and we plan to continue this tradition. For example, MDIM files are
provided at 1/256 degree/pixel scale as described above, but files covering
the same areas at 1/64 and 1/16 degree/pixel have been included in previous
releases. These smaller files may be adequate for some scientific uses; in
addition, they provide a means to make a quick, visual verification of the
features covered by a given file. We will also continue to provide data
near the poles in Polar Stereographic projection as well as in Simple
Cylindrical (and possibly Sinusoidal) projection.
MDIM 2.0 includes a set of text files containing detached ISIS labels for
all the PDS images on a given disk. These files make it possible to open
the images in ISIS without running the pds2isis translation program. The
labels on the CD can be opened as if they were ISIS files, or an ISIS label
and corresponding PDS image can first be copied to a hard disk directory
and then be used. We plan to provide detached ISIS labels with future
datasets. We unfortunately do not have the capability to provide VICAR
labels as well.
One "extra" provided as part of MDIM 1.0 (and duplicated with MDIM 2.0) is
a set of files corresponding to the image mosaic files but containing
scanned airbrush shaded relief map data. Because the original airbrush map
has already been presented and there is no plan to adjust it to conform to
the modern geodetic control network, airbrush data will not be included
with future datasets. If you have a strong desire to see the airbrush map
updated to conform to modern control, please let us know so we can propose
to do this work. Similarly, the global digital topographic model on the
seventh CD volume of MDIM 1.0 has been superseded by MOLA data and will not
be included in future archives using the new format. (It is possible,
however, that we will include the MOLA gridded elevation data in a format
precisely matching future MDIMs as to scale, projection, and division of
the planet, in order to simplify the use of the image and topographic data
together.)
COORDINATE SYSTEM
We have left for last the most technically complex and potentially most
scientifically complex issue. Two competing coordinate systems exist for
Mars. In 1970, Commission 16 of the International Astronomical Union (IAU)
defined these systems and approved their use. One combines longitude
measured positive east with latititude measured from the equatorial plane
to a point through the center of the planet, so-called planetocentric
(areocentric for Mars, geocentric for Earth) latitude. This is simply a
right-handed spherical-polar coordinate system. The other system uses
longitude measured in such a direction that the sub-Earth longitude
increases with time; for Mars, this means positive west. The second system
also uses planetographic latitude, which is measured as an angle between
the local vertical at a point and the equatorial plane. Because the shape
of Mars is flattened relative to a sphere, the planetographic latitude of
any point is greater in magnitude than the corresponding planetocentric
latitude (except at the equator and poles, where the two types of latitude
are equal). The maximum difference between the two types of latitude on
Mars is about 0.3 degree or 20 km, at 45 degrees north and south. This is
small relative to the possible difference of as much as 360 degrees between
east and west longitudes of a point, but not negligible.
Our present dilemma arose because each of the two different IAU-approved
coordinate systems was adopted by one or more map producers. Mars maps
(paper and digital) produced by the USGS since the 1970s use the
west/planetographic system. Hundreds of maps have been made in this system
and thousands or tens of thousands of literature citations identify
features by their west/'ographic coordinates. More recently, the MGS MOLA
and TES teams adopted the east/planetocentric system for their
dataproducts. They have been followed by the Mars Odyssey Orbiter 2001
mission.
We are proposing to adopt this east/'ocentric system for the future digital
mosaics of Mars described above, and, indeed, for all future Mars maps.
Other groups we have contacted, such as the NASA Geological Mapping
Subcommittee (GEMS) and the ESA Mars Express High Resolution Stereo Camera
(HRSC) team have expressed a willingness to follow the lead of the USGS and
adopt our choice of coordinate system for their future products. We
believe the balance of advantages lie with the east/'ocentric system and
will adopt it unless we receive an overwhelming response to this email
indicating that a majority of the Mars research community favors retaining
the west/'ographic system.
There are several points to be raised that may clarify our proposal to
switch to the east/'ocentric coordinate system. The first, which is clear
enough to need no further comment, is that the only reasonable time to make
such a decision is now. Once we have started to produce the positionally
accurate map products described above there will be an overwhelming
incentive to use compatible latitudes and longitudes in all our future
products.
The second point is that the planetary research community cannot escape
dealing with two competing coordinate systems for Mars. In this respect
the situation is very different than before the MGS mission when only the
west/'ographic system was in use in the literature and opposition to the
use of east longitudes was vehement (though ultimately ineffectual). If
the USGS retains the west/'ographic system for future maps, these will not
be compatible with products generated by the MOLA, TES, and Odyssey teams.
If future USGS maps use the east/'ocentric system for compatibility with
these teams, they will be incompatible with the bulk of old map products
from the USGS and elsewhere. We strongly believe that the latter situation
is preferable because the accuracy of geodetic control networks of Mars has
increased greatly since the older products were made. This means that new
and old USGS map products would not align precisely even if they were
produced in the same coordinate system. New maps made in the
east/'ocentric system will agree closely with MOLA data, since their
geodetic control will be tied directly to MOLA. As the new products enter
service, the older, incompatible ones will naturally fade from use.
This problem of compatibility is especially visible for the largest scale
map series (the 1:500,000 geologic and topographic maps) where some but not
all quadrangles have been mapped. If the new maps are made with boundaries
at multiples of 5 degrees planetocentric rather than planetographic
latitude, they will mismatch adjacent quadrangles in the old system by as
much as 4 cm (scale equivalent of 0.3 deg or 20 km). This seems a serious
objection, but in fact the geodetic errors in the older maps will result in
mismatches of several centimeters even if the coordinate system is the
same. This is a known hazard of making a map series while simultaneously
refining the control network on which it is based.
A third point concerns the relative technical merits of the two systems.
This has aroused some strong statements in the past. We believe that the
margin of advantage here is slender but goes to the east/'ocentric system.
This system is "primary" in the sense that space remote sensing data are
produced in it (or actually in the Cartesian coordinate system equivalent
to it). Navigation of orbiters and landers, which is an important use for
highly precise coordinate data, is also performed in this system. In
practice, planetographic latitudes are derived by computation from the
planetocentric angles determined by remote sensing. (Planetographic
latitudes could be measured directly by a human surveyor with a theodolite
on Mars, but this possibility seems too distant to carry much weight.) The
most important result of planetographic latitudes being computed from
primary, planetocentric measurements is that the transformation depends on
the polar flattening of Mars. This is a measured quantity and our best
estimate of it has changed significantly from Viking to MGS and even during
the MGS mission. Because of this, different planetographic latitudes could
be computed for the same feature at different times, quite apart from
changes in the geodetic control net. These differences could be 1-2 km,
enough to interfere with future precision landings if confusion about the
correct planetographic coordinate system arose. The conversion from
planetocentric to planetographic latitude also formally depends on the
elevation of the point, but to date the elevation has been ignored in such
calculations. This effect is <100 m even at the extremes of martian
elevations, so the concern is more theoretical than practical at the
moment.
Another technical consideration is the compatibility of software with the
two types of coordinate systems. The west/'ographic system has been
standard for products generated in ISIS and VICAR, so these systems will
need to be modified. As noted, both of them calculate with east/'ocentric
coordinates and convert to the other system on output, so the required
modification is relatively trivial. The latest release of ISIS already
allows user control of the desired output latitude-longitude system.
Third-party GIS and photogrammetric software systems pose more of a
problem, in that their code is not accessible for modification. This does
not favor either of the IAU-approved coordinate systems, however, since the
most important such systems (ESRI ArcView, LH Systems SOCET SET) use a
mixed east/planetographic system not approved by the IAU. Thus, one must
convert data for use with these packages in either case.
Our fourth and final point concerns the measures that we plan to take to
minimize the confusion caused by a change to the east/'ocentric coordinate
system. The most important is to document clearly and unambiguously the
system used in each product. This means not only noting the direction of
longitude and type of latitude (as well as the polar flattening used to
compute planetographic latitude) in the labels of each digital map file and
in the collar text of each printed map, but also including the direction of
longitude in the name of the map. Thus, a quadrangle formerly known as
15N022 (centered at approximately 15 degrees north, 22 degrees west) would
become 15N338E. If it were necessary to refer to the old designation, this
would be clarified as 15N022W. Quadrangle designations in the 1:5,000,000
and 1:2,000,000-scale series, which are based on sequential numbering and
feature names (e.g., MC-15: Elysium), would of course not be affected.
On printed maps we also plan to aid users in locating features whose
coordinates are given in the old system by supplying a secondary grid. The
map boundary and gridlines would correspond to whole degrees in
east/'ocentric coordinates but tickmarks and labels would be provided,
probably in color, for west/'ographic coordinates. This approach was
elegantly demonstrated by the DLR in simulated products for the Mars 96
mission (though at that time the primary coordinate grid was chosen to be
east/'ographic). For digital products, the latest ISIS release already
provides the capability to resample map files and convert between the two
types of latitude and longitude.
We thank you for your attention in reading this message and welcome your
feedback on any aspects of the proposed dataproduct design. Please respond
directly to Randy Kirk (rkirk@usgs.gov) no later than 13 July 2001. We
hope you share our enthusiasm about the many high-quality Mars map products
that will become available in the next few years.
__________________________________________________
Dr. Randolph L. Kirk
Astrogeology Team Ph. 520-556-7020
U.S. Geological Survey Fax 520-556-7100
2255 N. Gemini Dr.
Flagstaff, AZ 86001 USA
NOTE new email address: ASTROG::RKIRK and
rkirk@flagmail.wr.usgs.gov are now rkirk@usgs.gov
__________________________________________________