I'm new to using satellite images. I'm using Landsat images from USGS (Level 1 data), so terrain-corrected. I want to manually delineate different forest classes using false color composites.

Do these satellite images need to be pre-processed prior to doing this?

I won't be doing any calculations and I don't need pixel values to show reflectance or radiance.

I'm simply wondering whether the images not being atmospherically corrected for will impact their visual look?

I don't want to do all this work if it isn't actually necessary for my purposes, so I'm wondering if someone can help me with this.

  • 1
    if by manually delineate you mean you're doing this purely by visual inspection (and not using some image classification technique) you might get away with it, but the potential effects of atmospheric conditions will vary greatly depending on where you're looking at so it's hard to tell without knowing. you might want to look in to using NDVI at least to help visual inspection – Alex Oct 3 '17 at 11:04

For manual/visual interpretation you don't need to perform anymore preprocessing beyond the Level 1 product. Though atmospheric correction will likely provide some visual enhancement it is not a requirement in this case.

Your approach of utilising different false colour composites is fine and you should also perform contrast enhancements to better distinguish features.

Additionally, use of band ratios and indices such as the Normalized Difference Vegetation Index will help enhance certain features. If you are using multiple images then bear in mind that such indices will not be directly comparable across images, but will aid in interpretation nonetheless.

I reccommend reading A survival guide to Landsat Preprocessing as it discusses the types of correction and when they are/aren't neccessary.

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.