Posted on May 6th, 2014

The ASTER Spectral Library – a True Monument of Remote Sensing Data

In the first part of this series, we take a look at the nuts and bolts of this widely-used spectral library; and in the second, we will look at some of the ways in which workers in the field have made use of this tremendous resource.

The field of multispectral, remote sensing image analysis has been developed from aerial collections since the 1980’s, but it wasn’t until 1999 and the deployment of the Terra satellite that the science took to space. Since then, a number of sensors with true multispectral capabilities have been successfully established in orbit, including ESA’s MEdium Resolution Imaging Spectrometer (MERIS), the EO-1 Hyperion sensor , and DigitalGlobe’s WorldView-2 and QuickBird satellites. On board Terra is the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), a sensor which collects 15 bands of data spanning 0.520 micrometer (μm) to 11.650 μm in 15 meter (m) (for visible and near  infrared bands), 30 m (shortwave infrared), and 90 m (thermal infrared) resolutions. In support of the enormous amount of data collected by ASTER, the mission included an effort to build a tool which enhances the quality of information available from these 15 multispectral bands.

The ASTER Spectral Library categorizes the spectral reflectance signatures of more than 2,300 discrete materials in a standardized collection, compiled from laboratory measurements made by the Jet Propulsion Laboratory, Johns Hopkins University and the United States Geological Survey (USGS). The data covers an impressive bandwidth standard of  0.4 – 15.4 µm, and encompasses both a range of natural materials, such as vegetation, minerals, terrestrial and lunar soils, snow, ice, metals and meteorites, as well as a compendium of human-made materials. Many of us in the remote sensing field are familiar with the tool in that it is included in ENVI, but let’s take a look at the various components that comprise the library as well as a few particularities and distinguishing traits of each.

water_soil_grassThe spectral response of three material classes (vegetation, soil and water) overlain on a single graph. (Graph Credit: Penn State University)

Let’s start with the spectral collection that stands out most prominently, at least in my mind, which is the Johns Hopkins Library. Most well-known for excellence in the field of medicine, Johns Hopkins University is also home to the JHU Applied Physics Laboratory, which notably produced the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) that currently is collecting spectral data about the Martian surface onboard the Mars Reconnaissance Orbiter; and as such, the institution has a vested interest in spectroscopic understanding. Their spectral library offers superb coverage in the range of 0.4 to 2.5 µm, with spectral ‘slices’ usually less than .008 µm (8 nanometers) wide, making the data particularly robust for the most common range of multispectral data: visible to near-infrared.  While most of the materials in this library were measured using a standard hemispherical collection environment, which is most useful for correlation to remote sensing applications, the meteorite samples as well as some of the minerals studied were analyzed using a biconical collector for a greater range of bandwidths; and this set is represented by data spanning a range of 2.00 to 25 µm – purely a hardware issue. All spectra in this library were collected under the direction of John W. Salisbury.

The spectral library from the USGS distinguishes itself almost entirely on the basis of its probing curiosity. While the final product compiled of all three of these libraries states a standard limit of  0.4 to 15.4 µm, the USGS very tenaciously measured all 1,300 of the materials in its own library in a range of 0.2 µm to an incredible 150 µm, taking it well above the standard for the ASTER spectral project. The reasons for stretching the collection range so extensively into the far-infrared are presumably varied; but of primary stated importance is correlating data that is inbound from planetary missions, such as NASA’s Airborne Visible/Infra Red Imaging Spectrometer (which is orbiting Earth) or the NASA Cassini Visual and Infrared Mapping Spectrometer (currently orbiting Saturn). Further, the USGS states that the project helps support ongoing efforts of its own, and with the very impressive body of work that the institution has produced in its tenure since 1879, this seems most warranted.

Finally, the set of materials analyzed by JPL proper is simultaneously the smallest and most rigorously tested, thus providing the standard to which the others are correlated. While the entire JPL library includes 160 specimens, 135 of the most common igneous, metamorphic and sedimentary rocks were prepped into three size fractions (i.e. 125 – 500 µm, 45 – 125 µm and <45 µm), and then analyzed accordingly to provide a detailed look at the effects of grain size on reflectance. These samples were imaged in two distinct bandwidths, 0.4 to 2.5 µm and 2.0 to 15.5 µm, with an extremely thorough sampling resolution of 0.001 µm ‘spectral slices’ from 0.4 to 0.8 µm, and 0.004 µm from 0.8 to 2.5 µm. Every sample was measured multiple times to ensure stability in the instrumentation and minimal variance between iterations, and used both liquid water and pyrophyllite (a phyllosilicate mineral) as a standard in each case.

While this is a lot of raw data to ingest and most certainly a very dry read, the take-away points here are two-fold. First is a sense of the ranges and breadth of information contained in the entire ASTER Spectral Library to give a notion of the analytical capabilities it proffers. And second is the sheer volume of work that went into compiling it. Moreover, the library continues to grow and improve as each institution creates a more robust volume with successive generations of refinement, for which anyone with an interest in the science can be immensely grateful. Now that we have taken a look at the ASTER Spectral Library and what comprises it, in the next half of this series, we examine how the data has been used to extract specific understanding from complicated landscapes through the use of spectral unmixing.

Cameron Windham
Magellenic Paladin
(303) 210-5592

This entry was posted in The Geospatial Times and tagged , , , , by Katie. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.