Step 1. Calculating NBR

The purpose of this section is to briefly instruct users who are given a Landsat scene that is in reflectance how to calculate a NBR image.                                      

                                                                                                  >>>> Step 2. Segmenting dNBR

Once you have a Landsat image in reflectance you my now want to calculate NBR or another spectral index. To calculate NBR we make use of the following equation:

NBR = (ρ₄ – ρ₇) / (ρ₄ + ρ₇)

Where, ρ₄ and ρ₇ denote the spectral reflectance as measured in bands 4 (0.76 - 0.90 µm) and 7 (2.08 - 2.35 µm) of the Landsat sensor. ** Spectral indices should be calculated on data that has been corrected for the effects of the atmosphere. This is especially true when comparing two images from two different dates, as is the case when calculating a difference between two time periods. The simplest way to do this is a procedure called "dark body subtraction".

In dark body subtraction we assume that the reflectance at the center of very dark objects (such as the center of a lake for instance) should have zero reflectance and therefore anything we actually measure is an erroneous effect due to atmospheric noise.

Once measured we then, for each band, subtract this erroneous value from all the pixels within our image.

This assumption is not very valid in the visible wavelengths (red, green, and blue) as from our own experiences we know that water appears green-blue and sediments can also change this apparent color. However, assuming no sediments are present at the center of the lake, this assumption is very valid for the bands associated with NBR as the reflectance of water at wavelengths greater than the near-infrared (i.e. Landsat band 4) should be near zero.

Once we have corrected each image for the effects of the atmosphere we are finally ready to calculate NBR (or any other spectral index) for our imagery. You can use most equation editors in GIS or image processing packages to calculate the NBR equation. As you are calculating a ration you will need to ensure that your final image outputs is a floating point value.

Note that as our numerator is a subtraction is may be possible that sometimes we will be trying to divide a "zero" (i.e. when the reflectance of each band are equal). In these cases, the programs will often produce erroneous values like 65556 or -99. Such value should be ignored as the actual index values should be between -1 and +1.