3

I have a satellite image and I am not sure whether an atmospheric correction has been done with the image. Is there a way to find that out by looking at some values?

A atmospheric correction was supposed to be done with the semi automatic classification plugin. The output sould be a layerstack of each Band. So I am not sure wether it is atmospherically corrected. I haven't done the atmospheric correction, I just got the outcome Image.

I added a spectrum view of the satellite image of which I am not sure whether it has been atmospheric corrected or not:

enter image description here

10
  • 2
    Asking the same question for both QGIS and ArcGIS makes it effectively two questions when the Tour says that only one should be asked per question. Please focus your question on one or other. – PolyGeo Jun 13 '16 at 13:13
  • What would be the alternative: Radiance or Top-of-atmosphere reflectances? Do you have a time series? I think, the best is to look for meta data of the satellite images! – Iris Jun 13 '16 at 13:23
  • What atmospheric correction method was chosen? It seems a lot like a DOS-correction, which is actually not a correction, but a weak calibration / normalization. – Mikkel Lydholm Rasmussen Jun 13 '16 at 14:18
  • 1
    The near-infrared values are too low, and it is suspicious that all the bands are so neatly arranged near 0. My initial guess would be that the data was calibrated to reflectance and then DOS-corrected. The semi automatic classification plugin has an option for DOS-correction under the pre-processing section, which may explain why it was applied and called atmospheric correction. – Mikkel Lydholm Rasmussen Jun 13 '16 at 15:40
  • 1
    So you run the L1C data through some algorithm, and get some output that you're not sure whether it's been AtCorr'ed or not. Can't you just compare the input and the output (esp. the blue band) and see whether they are different? – Jose Jun 13 '16 at 17:16
3

To put it bluntly, the best way is simply to look at the spectral bands of your Satellite image. If it has a near infrared band, the atmosphere should make a gap here in the case of an image without correction. The blue band can also help you find out if it has been corrected by looking for the effects of Rayleigh scattering (look for places where the blue or Near UV spectrae is higher than it should be on a corrected satellite image, forming a curve that goes higher in the blues and UVs).

3
  • How do I know wether the blue band is higher then it should be? I put a image of the spectral in the post. – obrob Jun 13 '16 at 14:10
  • I usually use other programs such as ENVI in order to collect some spectral information at specific points. Having the mean values like this doesn't help much. You could try the semi automatic classification module for QGIS since it works pretty much like ENVI and has options to visualize spectral signatures from small area (ROIs). Though from what i see thanks to the picture provided, it does seem a bit unlikely that the image you have is raw but i am not 100% sure. What are the bands 4 and 5? – Moreau Colin Jun 13 '16 at 15:01
  • 1
    The Semi Automatic Classification Plugin does not perform rigorous atmospheric corrections like ENVI's FLAASH or ATCOR. It applies Dark Object Subtraction (DOS) and scales the resultant values using quantification factor usually found in Sentinel image metadata file usually 10000. ESA's SNAP is another option to produce respectable results wherein Sen2Cor and Sen2Three are provided chiefly to calibrate Sentinel Images. The scaled values in the illustration indicate that values are scaled in contrast to original DN values. – Ben Jul 7 '17 at 7:20
3

Well, one way would be to read the data documentation. The Scientific Data Hub specifically states: "The Sentinel-2 data offer for the Scientific Data Hub will consist of Level-1C user products", which is in reflectance values.

As such, you may have actually corrupted the data in applying an atmospheric correction that, as part of the processing, corrects to at-sensor reflectance. Please note that the pre-processing tools in the QGIS semi automatic classification plugin are intended for Landsat products and should not have been applied to Sentinel data in the first place. In reading the users manual it is clear that a Dark Object Subtraction (DOS1) atmospheric is applied to the data, but only for Landsat. The only other tools I see under the Pre-processing tab are "Clip Multiple Rasters" and "Split Raster Bands".

Here is the quote from the users manual on pre-processing of Landsat:

The tab Landsat allows for the conversion of Landsat 1, 2, and 3 MSS and Landsat 4, 5, 7, and 8 images from DN (i.e. Digital Numbers) to the physical measure of Top Of Atmosphere reflectance (TOA), or the application of a simple atmospheric correction using the DOS1 method (Dark Object Subtraction 1), which is an image-based technique (for more information about the Landsat conversion to TOA and DOS1 correction, see Landsat image conversion to reflectance and DOS1 atmospheric correction).

Is there some other tool that you are using in the toolbox that indicates that an atmospheric correction is applied to your Sentinel-2 data? If so, please edit your post to provide more detail on exactly what you have done.

1
  • 1
    In QGIS semi automatic classification plugin there is an extra section for specificly Sentinel-2 satellite images. I put a more detailed description in the opening post what have been done (or was supposed to be done) – obrob Jun 15 '16 at 5:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.