Visit PANCROMA Website

See additional satellite image processing articles at www.PANCROMA.com menu selection 'White Papers'


ASTER Surface Temperature Analysis

A prospective customer enquired about computing surface temperatures from ASTER L1B data. PANCROMA™ was not able to either handle ASTER L1B data in its native HDF file format, nor was it able to compute surface temperatures using the ASTER thermal infrared (TIR) bands. So I decided to add this capability for version 6.09.

Computing surface temperatures from the ASTER L1B digital number (DN) data was not a particular problem, since the method is fully described in the publication entitled Computing ASTER Reflectance and Temperature using ASTER Data and also in the ASTER User Handbook . The method if very similar to that used to compute surface temperatures from Landsat 7 data. A more significant issue was extracting the TIR bands from the bundled ASTER scene data. ASTER L1B data is archived in Hierarchical Data Format (HDF). HDF is a self-describing binary file format that is similar in many ways to GeoTiff. The format is comprised of header sections with public domain tags that describe the data and contain the file offsets that point to where the data is stored. The largest part of the file of course is the data itself, in this case all 15 band images. In order to extract the data, it is necessary to understand both the file format, and the way that a particular type of data (like ASTER L1B) is implemented.

I had some legacy code that I wrote some years ago to extract ASTER DEM elevation data from HDF files when HDF was the archival format. (Since that time, much ASTER DEM data is archived in GeoTiff format instead). The main difference is that ASTER DEM files contain only one set of image data, i.e. a single two dimensional array of elevation values. ASTER L1B scenes contain one array for each band in the scene, i.e. 15 in all. These include the VNIR, SWIR and TIR bands, each type having a different row and column count. Accessing all 15 L1B bands required extending the file reader algorithms. This was successfully accomplished, allowing any of the 15 ASTER bands to be read. I then implemented the surface temperature, as well as at sensor radiance and reflectance algorithms.

ASTER surface temperatures, like Landsat surface temperatures are computed from spectral radiance, so PANCROMA™ begins by converting DNs to radiance. The equation relating the two is given by:

Lλ = (DN - 1 ) X UCC

Where Ll is the spectral radiance, DN are the TIR band digital numbers, and UCC are the published Unit Conversion Coefficients. These are shown in the table below.

[UCC Table]

DN to radiance conversion coefficients.


Temperature (measured in degrees Kelvin) is then given by:

T = K2 / (ln(K1/ Lλ + 1)

Where K1 and K2 are constants derived from Planck's radiance function, shown in the table below.

[Temperature Conversion Constant Table]

Radiance to temperature conversion coefficients (degrees Kelvin).


The technique for computing surface temperatures is as follows. First, open you ASTER L1B HDF data file. (Note that ASTER L1A HDF data is implemented differently from L1B data, and currently PANCROMA will not handle L1A data successfully). Do this by selecting 'ASTER L1B HDF' from the File Type drop down box and clicking on your ASTER scene. Next select 'PreProcess' | 'Compute Surface Temperatures' | 'ASTER L1B TIR' from the PANCROMA™ menu.

[TIR Menu Selection]

ASTER TIR menu selection.


The ASTER TIR Data Form will become visible (see below). Set the TIR band number that you wish to process. Acceptable band numbers are 10, 11, 12, 13 or 14. You can also select various display and file output options. You can display the surface temperatures as a grayscale image or as a color contour image.

[TIR Data Form]

ASTER TIR data form.


These two, along with the thumbnail are shown in the figure below. The surface temperatures are indicated by the color scale bar at the upper left hand corner of the image. For the grayscale image, the darker the pixel, the cooler the temperature. Note that the ASTER TIR bands are I16 format with a 14 bit dynamic range. PANCROMA uses the full available dynamic range in converting the DNs to degrees Kelvin. As with Landsat surface temperate computations, you can also save your color and grayscale image files, and also save the computed temperatures in GeoTiff or flat binary format.


[Processed ASTER L1B Image]

Temperature contour plot and grayscale TIR image.


PANCROMA™ can also display individual ASTER bands and also can compute the at-sensor radiance and reflectance using utilities similar to those described in previous articles for Landsat data.



Web site and all contents Copyright TERRAINMAP Earth Imaging LLC 2010, 2011, 2012 All rights reserved.