I was trying to create an NDVI from Landsat8 OLI image from July 2013. I ran DOS1 and TOA. The values range up to 3. The original files were downloaded from EarthExplorer(EE), ran on QGIS and Multispec. The original data was downloaded again from EE and then ran only in QGIS (band calculator).

What am I doing wrong?

I switched to a landsat 5 image and i am having some problem. After the TOA and DOS1, these are the results for visible Red:'RT_LT50010862008320COA00_B3.sta' Number of Bad Lines = 0 Total Number of Pixels = 2,969,703 Histogram Summary Table First Last Line First Last Column Line Line Interval Column Column Interval 1 7251 20 1 8181 1 Channel Data Range Mean Median Standard Description Deviation 1 0.0075 to 0.6560 0.0409 0.0432 0.0294


NDVI values outside the valid -1 : 1 range are usually caused by negative pixel values in the red band. After your pre-processing is done you should always correct negative pixel values before calculating the NDVI.

An example:

The NDVI is calculated as

NDVI formula

Now assume NIR = 100 and VIS = -50. Your calculation now becomes

NDVI = (100 - (-50)) / (100 + (-50)) = 3

Some pixels are saturated. It is an known L8 issue. You can examine original bands - those pixels' values have to be unusual.

To solve this problem you can set some rules to handle such pixels.

In example:

if value > 1: value = 1
if value < -1: value = -1
if value < -1 OR value > 1: value = 0

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.