ASTER Satellite image process for mineral alteration detection

The Advance Spaceborne Thermal Emission and Reflection Radiometer

The Sensor

Is a Japanese sensor that is acquiring images for over 17 years. Although problems occurred in 2008 affecting some spectral bands, a large catalog of images taken during 2000-2008 are available for free and are priceless for geological applications. The sensor is still operational.
Stacks Image 1390
Stacks Image 1380
Spectral bands

ASTER offers 14 spectral bands grouped in 3 classes:

• Visible and Near Infrared (VNIR)

• Shortwave Infrared (SWIR)

• Thermal Infrared (TIR)

The SWIR and the TIR bands are those offering the bulk of information that permits to detect certain minerals and rocks.

ASTER is still today the most “geologically” oriented sensor available.

How does it works?

Each mineral has its own spectral signature. A Spectral Signature can be showed as a curve on a graphic presenting how much light is returned to the sensor (Reflectance) for each spectral band after being illuminated by the sun.

That curve is characterized by peaks, valleys and plateaus. Their location, shape and intensity are unique to a mineral or anything measured by the sensor; like a fingerprint.
Stacks Image 1396

You can find a catalog of spectral signatures of almost everything at Jet Propulsion Lab (JPL) or at the USGS.

Stacks Image 1402
Processing Background

Each ASTER scene covers 60 km X 60 km. At a 30 m resolution (pixel size), this represents 4.000.000 individual pixels having each one a unique spectral signature calculated from 14 spectral values.

The basic detection concept lies in calculating the probability that a pixel indicates the presence of a certain mineral by evaluating the similarities between the pixel spectral signature and the Mineral Lab Spectral

Once this probability calculated, the ultimate step is to apply a threshold to keep only pixels having a large enough probability.
This value is crucial and
has been tested on the field hundreds of time.


The first challenge concerns the pixel size and pixel purity.
We now know that a pixel from the SWIR bands has a resolution of 30 m covering an area of 900 m2. You don’t find patches of a single mineral that size.
This means that the spectral values received at the sensor, for one specific pixel, will be “contaminated’ by other surface elements (vegetation, other minerals, water, etc.). This has the effect of diluting the specific spectral characteristics of the mineral we are trying to detect.
The second challenge comes from the limited number of spectral bands available. ASTER offers 14 bands, this means that we have only 14 points on the graphic to define the spectral signature.
The red curve is a spectral signature of a mineral obtained by a sensor capable of measuring hundreds of spectral bands. The grey curve shows the spectral signature of the same mineral but measured by ASTER. From a very smooth spectral signature in Red showing subtle absorption features, the available bands present a rough signature.
Stacks Image 1408
Stacks Image 1412
Alteration Mineral Detection & Precision

Obviously not all the minerals can be detected. We have created 3 classes of alteration minerals.

The first class falls under the
Advanced Argillic group where Kaolinite, Dickite, Alunite and Illite are potentially detected. We grouped them together because they have almost the same absorption feature located on B5 and can not be detected separately with enough precision.

Muscovite (Sericite) falls under the
Phyllic group with an absorption feature at B6.
The third Class of alteration is called “Undifferenciated” and is created when the algorithm doesn’t confirmed with certainty whether its Phyllic or Advanced Argillic. But one thing is sure: I'ts alteration, one or another.

Our Precision is very high with over 95% for General Alteration, 90% for Advanced Argyllic and 80% for Sericite, checked on the field by numerous teams of exploration geologist from the principal mining companies of the world.

Chlorite, Epidote, Biotite are important minerals to identified but are undetectable using ASTER although its frequently offered by other consultants.
Alteration Legend

Product created by Infotierra

Our team is guided by two geologists with over twenty years of experience processing geospatial data:

Stephane Peloquin (LinkedIn Profile)
Geologist, Ph.D. in Remote Sensing
25 years of experience in image processing, development of specialised applications for databases management and GIS. He has developed very succesful solutions for drillhole logging and mining properties management. He has processed thousands of hyperspectral satellite images covering many countries in the world.

Speaks perfect French, English and Spanish.
Geologist, M. Sc. in Geographical Information Systems
20 years of experience processing spatial data and developing local and mobile applications for GIS. She has been GIS manager for several of biggest mining companies in the world, designing customised proprietary solutions for geological databases, also environmental and water resources.

Speaks perfect Spanish and English