- Satellite Data Products
- Aerial Services
- EO Applications
- Training & Resources
- About Us
Automated Detection of Forest Cover Loss using IRS AWiFS
With 21 % of the country's area under forests, and regular biennial satellite remote monitoring, there is an urgent need for rapid automated detection of forest loss locations. This is essential if effective response to forest loss is to be put in place.This is accomplished by utilizing the spectral and spatial pattern of forest cover in long term IRS AWiFS data sets, prepared with the best possible geometric and spectral characteristics.Geometric accuracy is achieved using ortho correction. Pixel level change is detected at native AWiFS resolution using relative surface reflectance products ensure that atmospheric and BRDF effects are corrected. Cloud, Cloud shadow, terrain shadow and water are automatically delineated The principle that forests are darkest vegetated pixels in the peak green season is used for forest identification. The forest peak detection in Red Band Histogram on a 5km local moving window basis.
A multi-spectral index – Integrated Forest Z score – is used as an Inverse measure of the likelihood of the pixel being a forest pixel
Forest loss detection is through simultaneous analysis of all images in the temporal stack using the temporal behaviour of the Forest Z score
The fully automated process is designed to work on 2 Deg (200 × 200 km) IRS AWiFS tiles
Results for 8 states (35 tiles of 2° x 2° each, covering states of Andhra Pradesh, Chhattisgarh, Goa, Himachal Pradesh, Karnataka, Madhya Pradesh, Maharashtra and Telangana) (forested area 2,82,925 km2, 40.54 % of India’s forest cover, geographical areas 39% of TGA) have been completed and national coverage is in progress
Results are published on Bhuvan for field verification by SFD officials using the NRSC developed, android based, mobile phone field data collection software for QA and accuracy assessment. This will lead to accurate delineation of forest loss areas using IRS LISS-III/ LISS-IV data.