[GRASSLIST:4455] Re: Temperature from LANDSAT-TM5 ch.6

[Markus Neteler]

On Sun, Oct 03, 2004 at 04:44:27PM +1300, Hamish wrote:
> > I'm looking at some LANDSAT-TM5 data and I'd like to derive surface
> > temperature from channel 6 (thermal IR) as described in Neteler &
> > Mitasova (1st ed.) S9.4.3.
> > 
> > The results look all wrong though, I'm getting a range of -54 to +69 C
> > 
> > Using the following formulae:
> > 
> > radiance = 1.238+(15.6-1.238)*tm6/255.0;
> > temp_kelvin = 1260.56 / log(607.76/radiance + 1);
> > temp_celsius = temp_kelvin - 273.15
> > 
> > we get:
> > 
> > tm6(0) = -69.7787 deg C
> > tm6(255) = 68.6640 deg C
> > 
> > Using d.histogram on the "tm24r_b6" 8bit GeoTIFF input raster, I see a
> > bimodal distribution with peaks at 75 (land) and 220 (ocean). That
> > translates to -6deg for the land and +57deg for the ocean...
> > 
> > is my satellite data wrong or is it the conversion formulae?
> 
> 
> It appears my data is missing all calibration coefficients; I was
> assuming these were standardized. Apparently not.
> 
> ... back to the vendor

Good to hear that :-)

In the new edition of the book (apparently available now,
http://mpa.itc.it/grassbook2/  - also cheaper)
I have improved the wording of the section but the formula
for TM5 and TM7 (ETM) are almost unchanged. I can re-check
for differences if needed.

Markus