About Imaging Spectroscopy
http://speclab.cr.usgs.gov
IMAGING SPECTROSCOPY/SPECTROSCOPY INFORMATION
(The information described here is specifically written for USGS
scientists to understand imaging spectroscopy and what it takes to analyze
the data for USGS programs using our currently developed methods.
However, the information should be generally useful to all those
interested in imaging spectroscopy.)
EXECUTIVE SUMMARY
- WHAT IS IMAGING SPECTROSCOPY?
-
Imaging spectroscopy is a new tool that can be used to
map specific
materials by
detecting specific chemical bonds.
As a result it is an
excellent tool for environmental assessments, mineral mapping and
exploration, vegetation communities/species and health studies, and
general land management studies.
The premier imaging spectrometer is the NASA/JPL AVIRIS system, covering
a 10.5 km swath with 17-meter pixel spacing. AVIRIS collects data at a
rate of 2 square kilometers per second!
- DATA ACQUISITION PLANNING
-
Normally, AVIRIS data acquisition must be planned well in advance, up
to approximately a year. However, as paying customers some
acquisitions can be planned a few months in advance.
- COST
-
AVIRIS data cost about $10 per square kilometer, but must be funded in
full flights ($64k per flight + $6k per flight hour). A full flight
can cover up to about 8,000 square kilometers spread over multiple
sites.
-
CALIBRATION
-
AVIRIS data require ground support information to calibrate the data to
surface reflectance for
mapping materials.
It takes about 30 to 60 person
days to calibrate an AVIRIS data set from one area (up to 8,000 sq km)
for one flight-day. Calibration uses a combinations of field spectrometers
and laboratory spectrometers, as well as radiative-transfer calculations.
(To see our tutorial on calibration, click here).
- WHAT IT TAKES TO ANALYZE IMAGING SPECTROMETER DATA
-
Computer analysis takes a trained person only a few days per couple of
hundred square kilometers' worth of data. This can include creating
maps of many minerals.
Current standard analysis produces materials
maps for more than 260 materials and is expected to grow to hundreds of
materials in the near future (note not all materials may be found in a
given area).
- WHAT IT TAKES TO VERIFY IMAGING SPECTROMETER DATA
-
Imaging spectroscopy analysis can be done with only a few days of work
per couple of hundred square kilometers of data. The analysis can
produce a several-hundred-layer GIS database. The various layers can
be assembled into materials maps
which must begeometrically corrected and
verified
in the field to
ensure the data were analyzed correctly and all relevant materials were
included in the mapping process. It can take a couple of weeks or more
to verify the results and begin to understand the complexities mapped,
and depends on the scientific questions being posed. Minerals,
environmental materials, man-made materials, and vegetation
species/communities maps are a new paradigm and understanding this new
technology and what science questions can be answered with it will take
training and effort.
- PUTTING IT ALL TOGETHER, FROM PLANNING TO COMPLETED ANALYSIS AND REPORTS
-
Complete analysis of a full flight of AVIRIS data, from flight planning,
calibration, materials mapping and
verification,
laboratory analyses of
samples, data registration, map production, and science reports takes
on the order of seven person years of effort. Smaller areas take less
time, depending on the science to be done.
- UNIQUE FACILITIES
-
Specialized facilities are required to successfully
map materials with
imaging spectroscopy. Laboratory and field spectrometers are required
to calibrate imaging spectroscopy data, measure reference samples, and
verify mapping results.
- TRAINING BY THE SPECTROSCOPY GROUP
-
The USGS Denver spectroscopy group can train about six scientists per year,
covering a maximum of three new AVIRIS sites per year.
WHAT IS IMAGING SPECTROSCOPY?
Imaging spectroscopy is the application of reflectance/emittance
spectroscopy to every pixel in a spatial image. Spectroscopy can be
used to detect individual absorption features due to specific chemical
bonds in a solid, liquid, or gas. Solids can be either crystalline (i.e.
minerals) or amorphous (like glasses). Every material is formed by
chemical bonds, and has the potential for detection with spectroscopy.
Actual detection is dependent on the spectral coverage, spectral
resolution, and signal-to-noise of the spectrometer, the abundance of
the material and the strength of absorption features for that material
in the wavelength region measured. In remote sensing situations, the
surface
materials mapped
must be exposed in the optical surface (e.g.,
to map surface mineralogy it must not be covered with vegetation), and
the diagnostic absorption features must be in regions of the spectrum
that are reasonably transparent to the atmosphere (the atmosphere can be
corrected for all but the strongest absorptions). The optical
surface is the same as what the geologist sees in the field with
his or her eyes. Spectroscopy can be used in laboratories on hand
samples, in the field with portable field spectrometers (spatial
resolution in the millimeter to several meter range), from aircraft, and
in the future from satellites. The aircraft systems now operational can
image large areas in short time (~2 sq. km per second!), producing
spectra for each pixel that can be analyzed for specific absorption
bands and thus specific materials.
The premier sensor is the NASA/JPL AVIRIS system (Airborne Visual and
Infra-Red Imaging Spectrometer). AVIRIS currently covers the wavelength
region from 0.38 to 2.50 microns with 17-meter pixel spacing (20-meter
spot) and a 10.5-km swath. AVIRIS is flown on an ER-2 (U2) aircraft at
65,000 feet. Next year AVIRIS should have capability to fly on a C130
aircraft and have 5-meter pixel spacing and a 2.5 km swath. While
AVIRIS data can be used to make any scale maps, the data from ER-2
altitudes makes excellent 1:24,000-scale maps.
The AVIRIS spectral range is excellent for detecting electronic
transitions in minerals (e.g., iron oxides, Fe2+ bearing minerals, etc.),
vegetation (vegetation species, health, green leaf water content), and
vibrational absorptions due to lighter elements (OH, SO4, CO3, CH, etc.,
so OH-bearing minerals, carbonates, sulfates and organics are mappable).
Diagnostic absorptions also exist due to other processes.
For example, Rare Earth ions involve deep-lying electrons and are not
diagnostic of mineralogy but of the presence of the ions in the mineral,
thus specific Rare Earth Elements are detectable with the
AVIRIS spectral range. Vibrational absorptions from heavier elements
such as Si-O (quartz) occur in the mid-infrared and are not covered by
AVIRIS. Currently, only broad-band sensors (e.g., TIMS) are available
in the mid-IR, although new mid-IR systems are under development.
The NAVY/Civilian HYDICE imaging spectrometer has AVIRIS like capabilities
but 1-meter to 3-meter spatial resolution, is currently flying in a testing
mode, and should become available for use in the next year.
DATA ACQUISITION PLANNING
A NASA ER-2 flight is like launching a rocket through commercial
airspace and requires significant advance planning on their part. Thus,
NASA plans the whole spring/summer/fall flight season the previous year.
The ER-2 is only deployed from several (continental United States) places: Moffet
Field near San Francisco, Whollops Island on the east coast, and
occasionally Topeka, Kansas, or Spokane, Washington. Depending on the
data requests, the aircraft may be deployed out of one of these
locations for a several week period to acquire data near that base.
As paying customers, we do not need to give exact sites a year in
advance, but NASA would like our best ideas so they may plan accordingly
(e.g., what bases they need to deploy from). Each year, AVIRIS
acquires data equivalent to an area greater than the size of the State
of Nevada to provide data for numerous research programs, ranging from
geology, ecosystems, water/ocean, to clouds/atmosphere. Sites generally
are found all over the United States. One year AVIRIS was deployed in Europe; this year it
went to Alaska and South America.
Consequently, as soon as we can after a project's budget is approved, we
must give NASA an idea of how much data we might want in the coming
year. For example, we may want 3 sites in the western United States (deployment
from Moffet Field) covering X sq km in states A, B, and C. Then
approximately next March, at the latest, we would need to finalize the
sites and give NASA coordinates of specific flight lines.
As part of the data-acquisition planning stage, it is also necessary to
plan for
ground calibration.
When NASA flies the requested sites
(usually the May-July time frame), a team must obtain ground
calibration data as near as possible to the day of the overflight.
Depending on the ER-2 schedule and weather, the ground crew could be
sitting in the field for days waiting for the flight, or do a last-minute
scramble to get to another site. (See CALIBRATION below.)
AVIRIS data are recorded on a tape that vholds 70 minutes of data. The
ER-2 aircraft travels at 734 km/hr (12.233 km/minute). Pre-calibration
plus post calibration for each flight takes up about 5 minutes of data
tape, leaving about 65 minutes of flight data (about 795 km in length).
Some tape movement occurs at the beginning and end of each line of data,
so about 1.25 minutes are lost for each flight line (equivalent to 15.3
km). Thus the achievable flight data lengths = 795 - lines*15.3 km.
Thus the following table shows in practice what can be covered per flight:
Line Segments | Total km of data |
---|
1 | 780 |
2 | 764 |
3 | 749 |
4 | 734 |
5 | 718 |
6 | 703 |
7 | 688 |
8 | 673 |
9 | 657 |
10 | 642 |
11 | 626 |
12 | 611 |
13 | 596 |
14 | 581 |
15 | 565 |
Finally, a flight can be no longer than 6.5 hours without special
permission. So you must plan how long it takes to get to and from your
site, and plan on about 15 minutes per turn.
COST
AVIRIS data cost about $10.00 per sq. km when large areas are covered.
AVIRIS covers a 10.5 km swath and about 800 km of data can be flown per
day (~8,000 sq km). No cost has been set for HYDICE data.
In practice, NASA requires whole flights funded (the cost for a flight
is mostly fixed, regardless of how much data are collected). Cost
sharing with other researchers is possible if multiple requests are
made to fit together into a single flight. One full flight of data costs
$64,000 plus $6,000 per flight hour of the ER-2. Fortunately, at a
speed of 734 km/hr, only 1.1 hours of data collection time are needed to
acquire a full data tape (10 Gbytes; ~8,000 sq km). Additional flight
time involves getting to and from the site, and in making turns for each
flight line.
Note: one full AVIRIS tape is 10 gigabytes of raw data, but when
converted to 16-bit calibrated data for delivery, this becomes about 14
GBytes!
An additional requirement to getting a full flight's worth of data is
that the sites covered must be near each other (like a few hundred
miles) so that there is reasonable time for the plane to get to each
one. For example, when launching from Moffet Field, a site in Nevada,
Utah and Colorado might be feasible because they could be on a general
line from the base. An example that would not work would be something
like the New World Mine near Yellowstone and a site in Arizona (we would
be charged for 2 flights: $64k each + flight hours).
CALIBRATION
Imaging spectroscopy data come in radiance, and thus have the solar flux
spectrum, the earth's atmospheric absorption bands (you can map
atmospheric gas abundances), and surface reflectance in the data. The
surface reflectance is what we need for land-management studies. The
data are calibrated by a combination radiative transfer atmospheric
transmission model and by measuring the reflectance of relatively large
homogeneous areas on the ground with a field spectrometer (a playa, rock
outcrop or grassy field have all been used; playas work best). The
effort to calibrate one site (one day's worth of data in one region up to
~8,000 sq km) requires a minimum of 2 people to visit the
calibration
site(s) near the time of the overflight (within a couple of weeks;
before any rainstorm). The field spectrometer data are obtained by
walking over the site collecting a few hundred spectra averaging the
site as much as possible. Samples are obtained and measured on a
laboratory spectrometer to confirm the field data. Spectra from the
imaging spectrometer data are extracted over the calibration site(s),
and compared with the field and lab data. A set of correction
multipliers and offsets are derived and applied to the data.
(To see our tutorial on calibration, click here).
The entire calibration process takes about ~30 PERSON DAYS OF WORK (1.5
person-months), for each day/site of imaging spectrometer data, whether
the site is a small area of only a hundred sq. km, or large ~8,000 sq
km). If the calibration is complex with no uniform areas, it could take
longer.
WHAT IT TAKES TO ANALYZE IMAGING SPECTROMETER DATA
Although many groups around the country are developing methods for
analyzing imaging spectrometer data, the USGS imaging spectroscopy group
is currently leading the world in capability. The USGS
Tetracorder
algorithm can analyze for hundreds (or thousands) of materials
simultaneously solving for what is present on the surface, whether it be
minerals, environmental materials, land, water, or ecosystems related.
Note that not all materials may be found in a given area, but that too can
be important information, especially environmental contamination.
We group AVIRIS data in 10.5-km wide by 17.5-km long segments (614x1024
pixels), covering 184 sq. km. Called a double segment, it takes less than 2
hours to analyze for 250 materials on a 140 MFLOP workstation (like
our HP 9000/K250). The analysis is fairly routine in that command
files exist to automatically do such analyses. One double
segment of imaging spectrometer data takes up 598 megabytes of disk
space (282 Mbytes of radiance data, 282 Mbytes of calibrated data, and
~34 Mbytes of ancillary data = 598 Mbytes).
The output products comprise 3 files per material mapped, so if 300
materials were mapped, 566 Mbytes of results (614*1024*300*3) are
obtained (we compress them, so the output is 3 to 10 times less space).
Results of mapping for 300 materials in essence makes a 300 layer GIS data
base. Note that current GIS systems only use uncompressed data (to our
knowledge) and we are not aware of any that can handle this many layers.
The biggest bottleneck in producing final products is the scientist
evaluating and understanding what was mapped and not data processing.
After all, these mineral/material maps are a new paradigm providing
unprecedented new information. For each data set, images of all the
materials must be reviewed. We have clustering analysis tools and the
system builds a set of image display commands to quickly go through such
a huge amount of images. Typically, once you have experience in an area
with the data, you can setup, analyze, and review the data for about 1
double segment per day (184 sq. km per day).
Another day is required to construct color composite map images, and yet
another is required to register the data to a map base. Thus about 3 days
of work by a trained, experienced person are required per double segment
to analyze. This does not not include detailed evaluation or
verification
of results. That depends on the area complexity and the science questions
being asked.
WHAT IT TAKES TO VERIFY IMAGING SPECTROMETER DATA
Verification
is the most time intensive aspect. Such large areas are
covered (even one double segment of 184 sq. km) that field checking
can require a significant amount of time. The materials maps steer you
to the interesting areas, and with the help of a field spectrometer, the
mapped minerals
can, in most cases, be verified real time in the field. A
final verification will sometimes require study of hand samples from the
field using laboratory X-ray diffraction analysis or other methods as
appropriate. Verification depends on the complexity of the scene,
remoteness and topography. Basic verification can take a couple of
weeks, or more, depending on how much area needs to be checked.
Verification may result in additional minerals being identified which
were not included in the original imaging spectroscopy analysis. Thus,
a second round of analysis and verification is required (this again
depends on the site complexity, what is already in the standard
Tetracorder
mapping, and what science questions are being asked).
PUTTING IT ALL TOGETHER, FROM PLANNING TO COMPLETED ANALYSIS AND REPORTS
Complete analysis of a full flight of AVIRIS data, from flight planning,
calibration, materials mapping and verification, laboratory analyses of
samples, data registration, map production, and science reports takes
on the order of seven person years of effort. Smaller areas take less
time, depending on the science to be done. Below is a list of the
generalized steps. Verification steps take the majority of the time.
- 1) Flight planning, Data acquisition support
and
field calibration. This effort is
valid for a "site" which could be one
entire flight, or a single region of a
multi- region flight.
- 2) Apply calibration to AVIRIS radiance
data to derive surface reflectance.
- 3) Materials mapping (e.g. Tetracorder),
- 4) Verification of mapping results, including field work and laboratory
analyses. Science analyses.
- 5) Data registration and map generation for
more detailed field verifications.
- 6) Laboratory spectroscopy of additional
materials. Sometimes special research may need to be done to
characterize and understand samples and observed spectra.
- 7) Revised materials mapping based on above
verification. Science analyses.
- 8) Data registration and final map production.
- 9) Reports and papers
WHAT IS POSSIBLE
Small Area (~100 to 400 sq. km)
A team of 3 trained in imaging spectroscopy could calibrate the data in
about 1.5 person-months and then map the surface mineralogy, vegetation, and
environmental materials for a small area (100-400 sq km) for about
200-400 materials in about 0.5 person-month.
Verification
and field studies would take an additional couple of months.
Data volume for 2 double segments (368 sq km) would be 564 Mbytes of
radiance data, 564 Mbytes of calibrated data, about 68 Mbytes of
ancillary data, and about 761 Mbytes of uncompressed data products (for
400 materials). All data are stored on read/write optical disks, each
with a 1.3 Gbyte capacity. Three optical disks would be needed ($200
data storage costs). Data acquisition costs would be a few thousand
dollars, depending on cost sharing other sites covered by the AVIRIS
flight. Note: a full AVIRIS flight ($64k + flight hours) must still be
funded.
An example of a study of this size is the one we completed at
Leadville, the California Gulch Superfund Site, where we mapped mine
drainage, finding the most hazardous waste piles, saving time and
dollars in the remediation process. Leadville took a couple of
person-years of effort.
Medium area: Lessons from Summitville (12 AVIRIS double segments)
A team of 3 trained in imaging spectroscopy could map the surface
mineralogy for a Summitville class area (~2,000 sq km) including
calibration and first mapping pass of 200 to 400 materials in about 2.5
months (5 person-months). Verification and field studies would take
several months.
Data volume would be about 7.2 Gbytes (12 double scenes). Data
acquisition costs would be about $20-40k assuming cost sharing with
another study in the area. Cost per sq. km: $10 to $20,
depending on the cost sharing for the flight. Note: a full AVIRIS
flight ($64k + flight hours) must still be funded.
Large area: 1x2-degree sheet (~19,600 sq km, 115 AVIRIS double segments)
A team of 6 trained people could map a 1x2-degree area with 2 to 2.5
AVIRIS flights, and about 6 weeks of processing on a dedicated computer.
A 1x2-degree data base would comprise a data set of about 68 Gbytes for
data (32 Gbytes for raw, 32 Gbytes for calibrated, and 4 Gbytes of
ancillary) PLUS 87 Gbytes for 400 [uncompressed] materials, for a total
of approximately 155 GBytes. Compression of results should bring
products down to about 30 Gbytes, for a total of 98 Gbytes.
Approximately 85 1.3 GB optical disks would be required (storage cost
about $5k). AVIRIS data acquisition costs would be about $240k. Cost
per sq. km: ~$12 (in this scenario, there is about 0.5 AVIRIS flight
of capacity that could be used for additional areas).
Such a large area, complete with registered materials maps, verified and
scientific papers written would involve about 20 person-years of effort.
Very large area (several 1x2-degree sheets up to a state or more)
Studies of this size would require special negotiation with NASA and
JPL. A large study of this size would be a significant portion of the
AVIRIS program. The AVIRIS group is interested in doing such large
projects, even if spread over a couple of years. Such a large-scale
effort could add enough money into the AVIRIS system that components
like ground computers and airplane tape recorders could be upgraded so
the cost per sq. km could be substantially less, perhaps as low as
about $5 per sq. km.
Total time for field verification and registered maps and scientific
reports would be substantial but possible within the scope of the USGS.
LIMITS TO WHAT PRESENT STAFF CAN DO
The spectroscopy group staff is currently limited and believes a maximum
of 3 new AVIRIS sites (in one AVIRIS flight) can be calibrated, mapped and verified per year
without additional staff. See training issues section below. Laboratory and
field spectrometers are another limitation; see unique facilities
section below.
UNIQUE FACILITIES
Successful imaging spectroscopy analysis requires high quality spectra
of reference materials. Development of the reference database requires
laboratory spectrometers that can match (or better) the spectral
resolution and signal to noise of the flight instruments. In addition,
field spectrometers are needed for
calibration
of the data, and for
verification of the results. The USGS has these facilities, but the
project load is very high.
TRAINING BY THE SPECTROSCOPY GROUP
The central region spectroscopy group could probably train about 6
scientists per year in imaging spectroscopy if the trainees worked on
cooperative projects where all were interested in the results. Our best
estimates indicate that minimum training would take about 6 weeks, including
fundamentals of spectroscopy, familiarization with computer software,
calibration,
and field verification. The training would be broken up with
3 to 4 periods in 1- to 1.5-week segments. Training to become an expert
in imaging spectroscopy, such that you can lead projects and research
can take 2 or more years.
The best way to make training a success is as follows. Choose 3 sites
for which the USGS programs would like imaging spectroscopy mapping
each year. Have up to six scientists plus support people who are interested in
the 3 areas agree to learn imaging spectroscopy. We believe that
scientists will learn more and be able to apply imaging spectroscopy to
their problems if they can work directly with the imaging spectroscopy
group on sites all are interested in and on problems that benefit all
those involved. Scientists in the MRS Program seem to have a lot of
common interests so cooperative research should not be a problem.
If suitable sites and scientific problems can be addressed, then the
imaging spectroscopy group and students will learn together how to solve
interesting scientific problems with these new tools.
FOR FURTHER INFORMATION CONTACT:
U.S. Geological Survey,
a bureau of the
U.S. Department of the Interior
This page URL= http://speclab.cr.usgs.gov/aboutimsp.html
This page is maintained by: Dr. Roger N. Clark
rclark@usgs.gov
Last modified Sept 25, 2002.