Hello Maria,

I am the person who developed the 'Improved DOS' and 'COST' atmospheric corrections models (the COST model is the one shown in the link given by Sajid above).  Keep in mind that either the DOS or COST corrections can be done with a simple stretch function which all software package contain somewhere within their tool boxes of processing procedures.  You can take the model and 'hand calculate' the input DN vs the output surface spectral reflectance (for say every 20 dn levels).  Then you can use the input-output pairs as input to the stretch function and this will convert from raw or TOA dns to reflectance values.

A very critical aspect of either correction is that a 'VALID DARK OBJECT' that can be used to compute the haze values exist within the image.  At times user blindly use the minimum dn value of the image as the dark object which can lead to over correcting for haze which means that the surface reflectance values will be to low and they will under estimate the reflectance.  For more information on the use of Improved-DOS and COST techniques also look the papers I published in 1988 and 1989 and are on the research gate pub list.

Good luck with your work.

Pat

Thank you for the answer, but unfortunately we have no ENVI. 

Hi

DOS is " Dark Object Subtraction" method for atmospheric correction, 1-4 represents various sets of input parameters like atmospheric transmittance (Tau)  used. DOS is one of the most common used atmospheric correction method. In i.landsat.toar (if you use DOS), you can set pixel value (Digital Number) of a dark object (often water bodies) from your scene.

The module manual here (http://grass.osgeo.org/grass64/manuals/i.landsat.toar.html) gives detailed information on how to use various options.

Yes, Landsat 8 is supported with this module (please select oli 8 as sensor), again check the above manual page for the details, go through the "Notes" section in manual  on Landsat-8 metadata file.

You can find more details on the DOS method in this paper:

http://info.asprs.org/publications/pers/96journal/september/1996_sep_1025-1036.pdf

Best

http://grass.osgeo.org/grass64/manuals/i.landsat.toar.html

http://info.asprs.org/publications/pers/96journal/september/1996_sep_1025-1036.pdf

Hello Maria,

I am the person who developed the 'Improved DOS' and 'COST' atmospheric corrections models (the COST model is the one shown in the link given by Sajid above).  Keep in mind that either the DOS or COST corrections can be done with a simple stretch function which all software package contain somewhere within their tool boxes of processing procedures.  You can take the model and 'hand calculate' the input DN vs the output surface spectral reflectance (for say every 20 dn levels).  Then you can use the input-output pairs as input to the stretch function and this will convert from raw or TOA dns to reflectance values.

A very critical aspect of either correction is that a 'VALID DARK OBJECT' that can be used to compute the haze values exist within the image.  At times user blindly use the minimum dn value of the image as the dark object which can lead to over correcting for haze which means that the surface reflectance values will be to low and they will under estimate the reflectance.  For more information on the use of Improved-DOS and COST techniques also look the papers I published in 1988 and 1989 and are on the research gate pub list.

Good luck with your work.

Pat

Thank you for the answers and useful links!

Now I understand that I know the essence, but i wonder why I didn't remember the method title....Maybe the reason is that I previously used ArcGis 10.2 soft where many nuances are hidden within its functions or named differently...Now I see .

GRASS provide  deeper insight . I don't just click the appropriate buttons, I want to understand to get better results. 

Dear Pat,

I'm familar with your 1988 and 1989 papers and other basic resources of course but found other interesting publications at your web page. The region we work at is very complicated for some reasons. So I have to pay a great attention to various details.

Dear Sajid,

Guess, the puzzle is assembling in my mind.

Your remarks and comments are precious!

Thank you!

Maria

Hi Maria

Glad that the tips helped. I forgot to share this blog post by Markus Neteler on processing Landsat 8 in GRASS GIS 7.0 -  http://courses.neteler.org/processing-landsat8-data-in-grass-gis-7/

It will give you a good start, if you are beginning with Grass 7.0

Best

Sajid

http://courses.neteler.org/processing-landsat8-data-in-grass-gis-7/

I would just add, that AFAIK the DOS based methods would be in most cases adequate for getting correct values of vegetation indices like NDVI, because these methods mostly remove additive efects of atmosphere and normalised and ratio indices are not very sensitive to remaining multiplicative efects. Of course, as these methods basicaly subtract a constant path radiance from the whole image, they would not work well if the said path radiance significantly varies spatially within the image or region of interest.

Thanks, Thomas, for the useful addition!

I'm studying a valley region. Aerosol concentration may be various in different parts of the region. Unfortunately, ground station data on atmospheric conditions are unavailible in many cases. The region is very heterogenous due to various conditions of humidity. So I still search for the best way of atmospheric correction...and its not as easy mission for me... Nevertheless I will try.

I'm very thankful for all the comments that help me to succeed in my work.

Kind regards,

Maria

Just a suggestion - if you would like to have the data corrected for atmosphere and then get the reflectance it is possible to have the reflectance values directly from USGS along with indices, TOA etc.

Dear Hristo,

I know about these possibilities and I'm sure it can be very useful for large-scale researches. But not in the case of my current  work.

Moreover, I couldn't find any similar information on the region of my interest. It' too far from civilization. And even if I could, I would repeat all the stages of image processing and quantitations concerning with a big pool of recent field samples received by myself and some backward field data, obtained by my colleagues. 

What Hristo was suggesting is, that now (from Christmas, in fact) you can download Landsat 8 data already atmospherically corrected. True, the Provisional surface reflectance product still has some minor issues, but it is created with quite detailed atmospheric properties map derived from other satellite data, so it might work for you. The product is not terrain-ilumination corrected, which could be issue in complex terrain you probably have, but it should be possible to correct with the help of DTM and for example GRASS module i.topo.corr.

http://earthexplorer.usgs.gov/redirect/collectioninfo/?collection_id=8704

http://wgbis.ces.iisc.ernet.in/grass/grass70/manuals/i.topo.corr.html

Yes, Thomas, I received some time ago a notification about Landsat 8 products from USGS. Although this data is just a minor part of images I have to analyze. Anyway, I'm to examine this more carefully to compare with images atmospherically corrected by myself and referred to the field observations if I found corrected Images for studied area anytime.

By the way, I've  received yet quite good results using DOS3 algorithm on Landsat 8 data for 2013 and 2014 and now trying other ways. But still there's a lot of work to do.

And thank you for the reference to Grass module. This soft is new for me and I'm very grateful for any information and references concerning this.

is scientifically correct to use the Dark Object  Subtraction(DOS1)  model to atmospherically Sentinel  2A satellite images ? I have found out the there is  a semi-automatic classification plugin in QGIS 2.18 that applies DOS1 model to atmospherically Sentinel 2 images. However on the European  Space Agency  page (https://sentinels.copernicus.eu/web/sentinel/user-guides/sentinel-2-msi/processing-levels) there are some guidelines which indicate  that the atmospheric correction of  Sentinel2A products can be done using Sen2Cor algorithm from  TOA reflectance values to BOA (Bottom of Atmosphere) reflectance values.

Pedzisai Kowe: There is no single "scientifically correct" method. DOS is simple method, but works for any optical data, provided you can find good dark target. If it is precise enough depends on application, also on the image - for example basic DOS methods are not good in case of atmospheric conditions strongly spatially variable within the image. More complex Sen2Cor algorithm should be more precise generally, but not necessarily in all specific application cases. It will be also more demanding on computer resources though - for me it takes between 40 minutes to several hours to correct one Sentinel-2 100x100 km granule on a Core i5 4GB RAM computer (so not the whole image).

I am not aware of publication comparing just these two algorithms, but you can find few comparing a complex 6S based method with several simple image based ones including DOS variants.

Thank you Prof.Dr. Tomáš Brunclík for your views and greatly appreciated.Im in  the stage of doing the image pre-processing of Sentinel 2 satellite data.The ultimate objective is to publish paper in a peer reviewed journal article. Beside the dark object subtraction method, i was just wondering if i pursue in using any method including 6S algorithm ,ATCOR2,  FLAASH and ToA will i not be criticized heavily for doing injustice to the Sen2Cor algorith which seems to be specific to the Sentinel satellite data. Again does the SenCor algorithm uses uses specific parameters of atmospheric of the geometry and of the sensor for improved conversion of the reflectance data. Is it based on any one of the atmosperic correction/radiometric correction method (6S algorithm ,ATCOR2, FLAASH, Dark Object Substraction method and ToA)? Thanks a lot

Pedzisai Kowe: I did not remember what algorithm is Sen2cor based on, but it was enough to put "sen2cor" into Google Scholar search, and in the article in the link, you would find:

"The Atmospheric correction module is a porting and

adaptation of the ATCOR software into python."

And ATCOR is in turn implementation of 6s, if I am not mistaken, but I might be.

I suppose (and hope) ultimate objective of any research is not an article. ;)

What atmospheric correction method would be suitable or sufficient depends on many factors as already stated. Even if the research you are doing is not focused on concrete application, trying to compare atmospheric correction methods generally, for example based on comparison of corrected image reflectance with ground truth data, the results may vary dependent  on the sites you use - ocean/inland, lowlands/highlands, humid/arid etc. In the general sense, Sen2cor should be good, because that was objective of creating it and recommending it as standard method of processing the sensor data in the first place. It also uses information only this specific sensor can provide - its cirrus band for example. But even in general, some method may be more precise yet. For example Sen2cor does not (AFAIK) need any external data except terrain (and that is optional and may make difference in complex terrain), some other methods may need in situ measurements of incoming radiance spectral profile at overpass, profiles of various gases and aerosol in atmosphere etc., making them less practicable, but possibly more accurate. Again just try to find some articles comparing and describing various methods, when I said I am not aware of any comparing Sen2cor with DOS, it does not mean there are none on the subject or other comparison with different methods.

Conference Paper Sentinel-2 Level-2A Prototype Processor: Architecture, Algor...

Thank you once again  Prof.Dr. Tomáš Brunclík. I now have a good perspective. I really appreciate your incisive thoughts.

Thank for the usefull comments