I'm seeking advice on how to effectively mosaic four Landsat tiles for analysis. Three of these tiles are from Landsat 5, with pixel depth at 16-bit unsigned, while the remaining one is from Landsat 7, with 8-bit unsigned pixel depth. Is it appropriate to directly mosaic these tiles, considering the potential discrepancy in pixel values? Should I convert the Landsat 7 tile from 8-bit to 16-bit before mosaic? If conversion is necessary, how can I accomplish it using raster calculator tools?
Thank you.
Kind regards,
Sohel
Mosaicking Landsat tiles from different satellites (5 vs. 7) and with different pixel depth (16-bit unsigned vs. 8-bit unsigned) requires careful consideration to ensure the resulting composite image maintains spatial and spectral integrity. Therefore, it is very important to bring them in same pixel depths before mosaicking.
Here's I how I would approach this problem:
Landsat 5 images have a pixel depth of 16 bits (unsigned--non-negative values), providing a range of 0-65535 possible values for each pixel. In contrast, Landsat 7 images with an 8-bit depth have a value range of 0-255. This discrepancy can lead to differences in the representation of brightness and spectral information.
I can linearly scale the 8-bit values to 16-bit values. This means mapping the 0-255 range of the 8-bit image to the 0-65535 range of a 16-bit image.
In ArcGIS: Use the Raster Calculator in the Map Algebra toolset.
- Example syntax : ("Landsat7_BandX" / 255.0) * 65535
Here, replace `"Landsat7_BandX"` with the actual name/path of my Landsat 7 raster band.
I would confirm that all tiles are georeferenced to the same coordinate system to ensure spatial alignment before mosaicking. I would also examine the mosaic for any visible seams or discrepancies in the areas where tiles overlap.
Hope this helps, bro.
Thank you Bhai. I tried the example syntax and it gives weird value. Will share the screenshot later.