Intergraph Geospatial Forum

How to convert Landsat 5 TM DN to reflectance in ERDAS?

rated by 0 users
This post has 14 Replies | 7 Followers

Top 75 Contributor
Posts 14
Yuzhen Posted: 05-04-2011 1:55 PM

Hi, All,

I onlly say the conversion tool for landsat 7, not see tool for Landsat 5. Can anyone here tell me how to convert Landsat 5 TM DN value to reflectance in ERDAS Imagine 2010? Thank you very much. 

Top 500 Contributor
Posts 2

using this model (COST model of Chavez, 1996", S. M. Skirvin)

Top 75 Contributor
Posts 14

Thank you very much, tzehuey85 and John. I downloaded the model, but unfortunatelly it appears that the file got damaged. I got the following error message from Winzip "cannot open file: it does not appear to be a valie archive, If you downloaded this file, try downloading the file again". I tried several times, it didnot work. Anyone got luck and can get the model running? Thanks.

I'll check the web download. See the attachment here.

 

JP

Not Ranked
Posts 1

Dear John Pollard,

Thanks for the information you provided here. I encountered some difficulties using the Cost model.  My problem is understanding the equation in the model that was used to convert the radiance to relectance. In the denominator where we have to multiply the solar exoatmospheric value (ESun) with the cosine of the zenith angle it is not clear for me why it was needed to square the cosine of the solar zenit angle inside the red box in the following equation. On other documents even on  the manual used for Landsat TM5 radiometric conversion this variable is not squared.

   

 On the other side in the equation that was used to derive the theorotical radiance of a dark object still the cosine of the solar zenith angle was squared. Is there a need to square this variable???

 

I am glad and thankful if u clarify it for me

 

Top 10 Contributor
Male
Posts 186

Dear zema28, I once wrote in a report from OMEGA project (http://omega.utu.fi/): "At least in the beginning of the OMEGA project, there were no simple solutions for IMAGINE users to turn Landsat 7 data into ground reflectances. First we took Skirvin’s COST model (2000) that is made for Landsat 5, started to adjust it for Landsat 7 units for and easier parameter control, and concluded an error in the model’s documentation (d squared and ESUN had been swapped). We asked her to correct this, and it’s done. Later we found out about some limitations of the COST (DOS2), and started to adjust our process more towards the DOS3 approach (Song et al 2001). ..."

Therefore, perhaps the COST documentation is nowadays still not perfect. Check the code, perhaps it does not do the the square? Also some parentheses are missing: / (ESUN * cos theta). And be careful what all symbols like ESUN, d2 or E0 include in each connection they're talked about, for example mean sun-earth distance or instantenous. I don't remember right away, but I think using the square might be justifiable only, if the light under consideration goes back and forth through the atmosphere. You're right, basically like in http://landsathandbook.gsfc.nasa.gov/data_prod/prog_sect11_3.html it is not there when calculating atmospheric reflectance.

 

 

Top 500 Contributor
Posts 4

Hi Everyone,

I am looking to apply something similar but I am wondering what is the difference of applying a model by chavez or song (in model builder i presume) compared specialized atmospheric corrections programs such as ACTOR?

Bjorn

Top 10 Contributor
Male
Posts 186

Yes in Modeler, using tables for variables to be multiplied etc. with multispectral images. The difference might be that these are somewhat simpler models than ATCOR, FLAASH and others, which are based on MODTRAN 4 or 6S detailed atmospherical transfer models, by using pre-calculated multidimensional tables for many wavelengths, viewing angles, AOD etc. from such actual transfer models. Those may also include scattering models etc.

I ended up in DOS3 but without the component from 6S.

Top 500 Contributor
Posts 4

Hi Arto,

Thanks for your reply and I will most certaintly keep that in mind. I am really looking to apply an effective but not computioanlly heavy atmospheric correction and therefore I was considering perhaps something similar to DOS3 or simply the DOS1. However do you know where I would acquire a Eo factor as in the exoatmospheric solar constant - i read of a link to NASA that converts the sun-distance relation to atmospheric units but unfortunatley that link was invalid.

Sincerely,

Bjorn

ps - how do you open the .gmd COST model extension (does not seem to be an accepted extension to ERDAS)

Top 10 Contributor
Male
Posts 186

Dear Björn

Well, below at least some info from my Omega report "Cost-effective operational methodology for satellite and airborne remote sensing data processing for monitoring glacial areas. Arto Vuorela, Novosat Ltd 2004" (which I think they did not finally use for the reports at omega.utu.fi). Ecorr I had named myself to avoid the E0 confusion explained there, but I calculated it using the Gamma value, see Iqbal. By the way ESA's d was actually some time the inverted 1/d. In any case, you also need to be aware of many units (m or um etc.) when calculating things like the correction. Song's theta0 in one equation should be thetaZ. The report might have some more useful information, let's see...

Arto

---

Note that also the symbol E0 is sometimes used in different meanings. In Chavez (1996), it means the corrected solar irradiance at the

current earth-sun distance, and E0 already includes the term d*d. In some reference (Lillesand and Kiefer, 2000:479) E0 means the same as our ESUN; the solar irradiance at mean earth-sun distance. This reference also shows the way how the d*d term is embedded into the E0 of Chavez; it is equal to ESUN/(d*d). In another connection (Iqbal, 1983), the symbol E0 was used for the correction term Ecorr described above. And the ESUN is sometimes called just E.

Lillesand & Kiefer, 2000. Lillesand, T.M.; Kiefer, R.W. 2000. Remote sensing and image interpretation. New York: John Wiley and Sons. 736 p.

Muhammad Iqbal, Introduction to Solar Radiation, Academic Press, 1983

Top 10 Contributor
Male
Posts 186

The gmd you open in the (Spatial) Modeler, Model Maker of ERDAS.

And I think the term is squared because of the attempt to model the transmittance tau, see http://earth.gis.usu.edu/imagestd/ and its FAQ.

Top 500 Contributor
Posts 4

Hi Arto,

Thanks again for all your help and suggestions. I was wondering if you had any suggestions or reasoning of why you left the Rayleigh 6s function out of the equation (as I need to justify that if I apply the DOS3 model) - besides the fact that you'd need some sort of 6S software calculation. Id actually really like to get your correspondence info if I may.

The COST model available above has the following basic equations but unfortunatley it seems greek to me as I cant understand how those variables relate to the equations used by Chavez et al 1996

 (( -3.0313 + (0.0602353 * $n1_phnx(1) - 0.15)) * PI * 1.0162  ** 2) / (195.7 * COS (PI/180 * (90 - 55.7758))  ** 2)

I suppose the first portion is convertion to at satellite radiance using the digital count (i.e. DC - offset/Gain) or part of the path radiance potion (i.e. Lp = G * DCmin +B...) but those equations I just stated dont match with that above.

Sincerely,

Bjorn

Top 10 Contributor
Male
Posts 186

6S software I hadn't and I had no more time to include that, so I took the corresponding term from DOS4. But I think I evaluated that this scattering term was not the most significant...  My contact information is forename dot surname and the rest is poyry dot com.

Well, from your message's web version (( -3.0313 + (0.0602353 * $n1_phnx(1) - 0.15)) * PI * 1.0162  ** 2) / (195.7 * COS (PI/180 * (90 - 55.7758))  ** 2) looks to me just like your equation, except the already mentioned missing parentheses.  3.0313: Lhaze. 0.0602353: Gain, yes but which units... -0.15: Offset, same units. 1.0162: d. 195.7: ESUN, again mind the units. 55.7758: solar elevation angle in degrees, then turned into solar zenith rad.

I hope zema28 noticed from the mere web version my hint why that one term was squared. 

 

Top 10 Contributor
Male
Posts 186

Note Antonio and zema28, also in this discussion it's worth using 180. or 180.0 and 90.0 rather than integers. The cosine is squared in COST model only (to simulate tau), not in basic conversion equations. Please don't create a new discussion merely to reply, but continue each discussion from the web page.

Page 1 of 1 (15 items) | RSS
Copyright 2014 Intergraph Corporation
Powered by Community Server (Commercial Edition), by Telligent Systems