I am working with Landsat 5 TM images. Unlike previous USGS landsat products that used to contain metadata in the TXT format, the new products are in the XML format and they are atmospherically corrected scenes. So I can't use ENVI 5 to do the radiometric corrections, as ENVI only supports TXT format of metadata.
As I try to use the NDWI = (Band2 - Band4)/(Band2 + Band4) calculation, I get a range of -6.2 to 6.1 instead of -1 to +1. This might be due to high spectral reflectance of the atmospherically corrected image.
How to resolve this issue?
After opening the landsat image in envi by "open external file" ,you can check the dataset variable where the scale factor must be mentioned by which you have to multiply your dataset to obtain the desired range.
Thanks for your reply. I did that. I multipled the raster with 0.0001 as per the scale factor for landsat 5 images. Unfortunately, after rescaling the range of ndwi was way below the expected range of (-1,+1).
Then I figured it out that very few pixels were outliers in the ndwi raster. I used conditional function of raster calculator of arcgis to convert the outliers with in the range (if x< -1 then -1 else if x>1 then 1 else x). The ndwi map is working fine now. Regards.
Polash Banerjee You can also perform image processing analysis on the Landsat TM image using the Erdas Imagine or ENVI, and then compute the spectral indices on any of these two software packages or on ArcGIS (using raster calculator tool). Sometimes , you might want to try out other software packages for their results.
The Normalized Difference Water Index (NDWI) is a remote sensing derived index estimating the leaf water content at canopy level. The value ranges from -1 to +1.
Please see attached document for further readings on the NDWI..
Visual or digital interpretation of the output image/raster created is similar to NDVI:
- -1 to 0 - Bright surface with no vegetation or water content
- +1 - represent water content.
You may use the NDWI product from the Google Earth Engine.
https://developers.google.com/earth-engine/datasets/catalog/LANDSAT_LT5_L1T_8DAY_NDWI
Polash Banerjee
Dear polash, I also facing same problem when i used Landsat 5 collection-1 Level-2 data. I applied MNDWI algorithm for raster calculation but DN value range from large negative value to large positive value instead of -1 to +1. Can you share the condition that is you use in your calculation.
Sir...try using ecognition developer. Use parameters like NDWI and shades of blue
Kindly read my article " Feature extraction for Urban and Agricultural Domains using eCognition Developer". This will surely help you ti get desired values
"...(Band2 - Band4)/(Band2 + Band4) calculation, I get a range of -6.2 to 6.1" - it looks impossible if Band2,4 are positive numbers. Check your equation and raw data.