0

My aim is to compare two images of NBR index obtained by Landsat 8 for two dates: April the 1st and October the 26th (the beginning of the wildfires period and the ending of the period). After that I want to compare the difference NBR image with MODIS vector products (as it's shown below)enter image description here The "blue" pic is a result of ISODATA process (for detecting of the areas with the biggest difference which are supposed to have been burned). The questions are:

  1. Is it a good idea to use ISO DATA for this purpose (I obtained just 3 classes).
  2. How I can compare my vector data (MODIS areas) and raster data (Landsat 8) to find the difference between burned area?
  • Do you mind if I ask why you used ISODATA in this instance? With NBR I thought the idea was to calculate it for pre- and post-fire dates > subtract post-fire values from pre-fire values to obtain deltas > and classify burn severity along the range of deltas. I'm just wondering how ISODATA helps you here. Also, looking at your image, it looks like there are clouds, and like they are being picked up in the classification. Am I right, or are your images cloud free? – CSB Oct 9 '15 at 15:30
  • I thought to use ISODATA for excluding needless pixels and determinate just the necessary dNBR areas with wildfires. I'm a newbie in such kind of researches that's why I asked if I should use ISODATA or not. So far I understood that it's better to classify dNBR image by the special range (I'm glad to read the same in your reply), but I still have no idea how to compare the raster with vector data. And about clouds - I didn't correct the data so it's not cloud free. – Bulsara Oct 12 '15 at 7:01
  • NBR is not a common abbreviation on this site so would you be able to edit your question to write it out in full the first time you use it, please? – PolyGeo Dec 10 '15 at 1:36
1

1) No. The way how dNBR is usually used to identify the burned area is to set a threshold, beyond which point you'll considered the pixel to be burned. The threshold is not universal, and you'll need to determine it based on a series of factors (eg. forest type, difference in time between two images, etc) according your actual conditions.

2) You'll need to reconcile the different spatial resolutions of two datasets. If you have access to ArcGIS, a common way to do that is to use the resample tool (with the majority filter, since you are dealing with classification images). You'll probably want to resample the landsat-based image to the modis resolution, instead of the other way around.

PS: As CSB pointed out, you've got clouds in your scene. Clouds are extremely undesirable in disturbance mapping as they'll usually be identified as disturbances. So you definitely should deal with them first, by masking them out.

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.