I want to differentiate water pixels from snow pixels using Landsat 8 data. I have calculated Modified Normalised Difference Water Index, and classified the image using threshold into 3 classes as follows:
- Class shallow water pixels = -0.61 < MNDWI < -0.83
- Class Dense Water pixels = MNDWI > -0.83
- Class No water pixels = MNDWI < -0.61
But these thresholds are not working in the areas covered with snow. Any suggestions on differentiating water and snow pixels? I want to perform this is Python.
Have you looked in to using Normalised Difference Principal Component Snow Index (NDPCSI)? Some researchers have had success using this index to differentiate snow and water features.
Check the values in the blue band and try to use treshold after using ndwi.
In general, the two targets are easily differentiable using visible and near infrared bands. Snow reflects most of the radiation in this region of the spectrum while water absorbs most of it. You can use bands 1, 2, 3, 4, or 8 for this task. I do not think you need a standardized index.
If you are having confusion on shallow waters, it is because you have the soil response. You can use the SWIR bands for differentiating between shallow waters and snow.
The exact thresholds that you must use are image dependent, since for each image you have different geometry of illumination and acquisition, and also different atmospheric conditions. So, you must define it.
Thank you all.
Here is my result image after thresholds on MNDWI. Can you help me to improve the thresholds which will differentiate water from snow?
- Badges
- Science topic
- Similar topics
- Computer Science
- Images