The Scale Factor used in FLAASH is a value you enter that will be *divided* into the input data file to convert it to the units that FLAASH expects, which is um/(cm^2*sr*nm).  This is discussed in the ENVI help regarding FLAASH.  So you must know what the units are of the radiance data that you input to FLAASH.  The calibration equations supplied by the USGS state that if you use thier equations to convert the Landsat DNs, or raw values, to radiance, the data are output in units of W/(m^2*sr*um).  ENVI uses these equations so when you use the calibration tool in ENVI, and output your landsat data to radiance (not reflectance which is the other option), the units of the data are W/(m^2*sr*um).  So you would use a scale factor of 10 to convert to um/(cm^2*sr*nm). I would strongly suggest you upgrade your ENVI if at all possible so that it uses the latest calibration coefficients from the USGS.  The 'Get Calibration Parameters from web' will no longer work.  The scale min/max for this tool corresponds to the LMIN/LMAX values that come with the Landsat data in the '_mtl.txt' file.  The Data Acquisition Band is the input band of the TIFF - band 1,2,3 ... etc.

Thanks a lot for your help :) I´ll try to update Envi to avoid these problems. * I´m testing with a 10 scale factor (SF), but depending on what "Output Reflectance Scale Factor" is (in "Advanced Settings"), I get more or less detailed results. With 10 (SF)-10000 looks fine, but I don´t really know what means this last parameter, could you please explain it? 10000 (as default) is a good value? How can I measure results? * Also, I don´t understand why my output image changes the band order with respect to the input. I don´t get the correct band RGB combination to visualize it in a realistic appearance. * Finally, in "calibration tool", I still don´t know what the "Data Acquisition Band"  is: you said "the input band of the TIFF". I don´t understand you, my tiff has 6 bands: 1,2,3,4,5 and 7. Thank you very much for your help, I  keep going while you reply me... (I apologize for my little knowledge in this area). Regards!      

The output from FLAASH is scaled reflectance stored as integers rather than floating point reflectance.  This reduces the amount of disk space needed to store the file.  In order to retain the full precision of the many values you can get with floating point reflectance that ranges from about 0-1.0, a scale value of 10000 is applied to the data.  It is best to leave this setting at 10000 or you may reduce the precision of the output. I have never seen FLAASH reverse the band order so I don't know why you are seeing this.  Are the wavelengths of  each band actually in a different order?  Perhaps the RGB combination doesn't look correct because some incorrect parameters were used which can cause the output to look strange.  It can also happen if the data are not properly calibrated to radiance so that the scale factor used (10) does not convert the data into the expected units of um/(cm^2*sr*nm). The data acquisition band is the band you are calibrating in the tool.  The geotiff files come as '_B10.tif, _B02.tif, _B03.tif...', etc.  The _B10 is band 1 of Landsat and _B02 is band 2 of Landsat and so on.  You need to set this in the dialog so that ENVI knows which calibration parameters to use for the band of data.

Thank you again for your response, I´ll try to be more clear with respect to the band order: When I load a .BIP image, I can choose a sequence of bands to watch it in "real" color: for example: 3 - 2 - 1 After making the FLAASH correction, I choose again the same order: 3 - 2 - 1 , but the image appears in different colors than before. Do you think is because the output image is not in DNs? How can I convert in Envi from radiances to DNs? Also, I have upgraded to 4.7, and there´s a new atmospheric correction module called QUAC. It ask you no parameters, except the sensor name. The output looks correct, but what´s your opinion of QUAC´s results with respect to FLAASH?   Thanks!

The change in the appearance is likely due to incorrect parameters or an incorrect calibration, as stated in the previous forum post.  It is not because the bands have changed order. I do think the change is because you have processed the data in FLAASH.  The point of this whole FLAASH process is to convert your DNs which have no physical meaning to radiances (using the ENVI Landsat calibration tool), which is a physical measurement of the amount of energy recorded at the sensor and then to remove atmospheric effects by converting to reflectance, which is a real physical property of the materials in the image.  The QUac module is described in the ENVI help and it is a good alternative to FLAASH if you do not know all the appropriate input parameters for FLAASH.

Hi mminari,   My L7 images have very little metadata information: I have adquisition date, one coordinate, datum, projection, pixel resolution, and the name format L71233083_08320071114.tif I have seen in Landsat website some other L7 images with LMIN and LMAX data. Do you think these values can be applied to my images in the calibration tool? Where can I obtain the Gain and Offset? Also, in the calibration tool to convert to radiances: Should I process the image in single bands and then joining them all again? Finally, when I run QUAC, it ask me for wavelengths. I edit header and input one single value in nanometers for each band. The problem is that I am not sure if these values are correct, because each band has a range (for instance, B1 = 0.45-0.52) and I choose the middle value.. Where can I obtain the precise values for these fields? Thank you very much again, you are helping me a lot! :D

Hi juaneras, During Landsat calibration, ENVI should be automatically retrieving the appropriate LMIN and LMAX values, if you have the necessary metadata associated with your image files, and if you open the file using the appropriate tool for your Landsat file format under File > Open External File > Landsat.  For example, if your Landsat files are in TIFF format, you'll want to open them in ENVI using File > Open External File > Landsat > GeoTIFF with Metadata.  Then ENVI will automatically read all of the parameters it needs for calibration, if they are available. Don't use LMIN and LMAX from other images to calibrate your data.  Those values are not the same for all Landsat 7 ETM data, so that can easily give you inaccurate results.  If you do not have the necessary metadata for your Landsat 7 ETM files, then you will need to contact the data provider to get the missing information.  As the documentation for the Landsat calibration tool in ENVI explains, you can usually obtain calibration parameters from the EROS Data Center (if that is where these data originated) as CPF files and MetaData files. Yes, you leave your image in single band files, and calibrate them that way.  Then you can join them together into one multi-band file later. It's fine to use band centers for the wavelengths in the header.  I really recommend checking the ENVI documentation for the features you are using.  I think it will clear up a lot of your questions. peg

Thank you very much Peg, you were very precise in your answers!! Bye! :D

I´m sorry for keeping this post opened, but still don´t obtain good results. My steps: 1 I separate bands one to one (6). 2 Calibration of each (with automatic Gain & Offset values). 3 Stacking them in one single file in correct order (Basic Tools->Layer Stacking). 4 Convert this single file to BIP format. (Up to this step, images look similar to the original) 5 Run Flaash. (scale factor 100, visibility 100, Landsat TM7, no aerosol retrieval, water column 1, aerosol model = rural, atmospheric model = MLS, date/hour, location, elevation, and keeping default advanced settings ) With such scale factor images look fine. Changing to 10 or 1000 gives at least one wrong band (too shiny or too black). The problem I observe is that 3 first output bands are much more dark than original (the others are pretty similar). This makes my RGB combination look strange: for example: in Flaash output the 1-2-3 band order for visualization gives a purple image, but in the BIP file (and also in the original file) it gives a near-realistic representation of the scene. I don´t know what parameter in FLAASH is giving such problem.. Do you think scale factors should be applied individually band to band? Where could I obtain these values? Perhaps reflectance representation is neccesary different than in radiances...¿? I´ll post a new thread to ask how to convert reflectances to radiances or NDs, maybe this would change the visual representation. Thank you very much!