I want to identify coral reefs using multi-temporal Landsat data and Lyzenga's Transformation method. In relation to pre-processing the image, should atmospheric correction be done or not? If not, what is the impact for the next analysis?

  • The pre-processing of satellite images needs to be done to correct differences between scenes, if you don't then the scenes will not match as closely and the colours will not be as precise. I don't know much about how to apply it, I use TNTmips and punch in some numbers from the metadata for TOA and reflectance without truly understanding what's going on - only that the result is more uniform than the input. – Michael Stimson Jul 16 '15 at 4:37
  • To some degree that depends on your classification method: Do you identify coral reefs in a single acquisition and then compare the classification results or does your classification rely on having multi-temporal acquisitions as input? If the latter is true you will most likely need to apply athmospheric correction so you can compare the "true" surface reflectances instead of top-of-athmosphere reflectances or digital numbers, – Kersten Jul 16 '15 at 6:58
  • Thank you Michael Miles and Kersten, Actually I use the data not only multi-temporal acquisitions as input but also point of ground truth to asses the accuracy of classes. I use also the water column correction (Lyzenga's Algorithm) which use the DN value of scenes. Therefore if I do the atmospheric correction which will change the DN values. so that the Lyzenga's algortihm is not used. In the other hand, the scenes of atmospheric correction is not open in the ER Mapper software because the software has not to support the data in 32 bit. Let me know if you have any suggestion for this? – La Ode Muh Yasir Haya Jul 17 '15 at 2:01

Your Answer

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

Browse other questions tagged or ask your own question.