NDVI approach is used to localise deforested areas because of the following reasons,
- Healthy vegetation like forest, reflects near infrared really well and absorbs red light for photosynthesis.
- Dying vegetation absorbs NIR and reflects red light.
Steps to compare two sets of sentinel-2 data to monitor deforestation:
- Download the data for a selected region from Sentinel Hub.
- Load the red(band4), green(band3), blue(band2) and nir(band8) images, in the form of arrays.
- Slice the high resolution image arrays into smaller patches to reduce computation power and memory.
- Normalise and stack red, green and blue patches to form true color composite.
- Generate NDVI image by calculating the NDVI value using nir and red band image arrays.
- Apply a threshold on NDVI image and create NDVI mask to differentiate forest and non-forest areas.
- Repeat steps 1 to 7 on another set of data of the same region that has a different timestamp.
- Visually compare the true color composite and NDVI mask of the corresponding sliced patches from both the data sets.
- Calculate the total area deforested in the particular patch by comparing the pixels of both the data sets.
- Save all the true color and ndvi mask patches to create GIFs if needed.