Methodology for Assessment and Mapping of Forests

by Smt. Kalpana Palkhiwala, Assistant Director (M & C), PIB, New Delhi

Forests play a crucial role in the country’s ecological stability and economic development. This requires periodic monitoring of the forest cover of the country. By and large, till 1999 assessment, satellite data were interpreted visually. Shifting of interpretation from visual to digital started partially from 1997 assessment, wherein, two states were interpreted digitally. This was followed by digital interpretation of 14 states in 1999. However, the entire country was digitally interpreted at 1:50,000 scale from 2001 onwards.

Visual Interpretation
For visual interpretation, satellite data is procured at 1:250,000 scale in the form of hard copy called False Colour Composites (FCC). A base map is prepared on a tracing sheet (Mylar) using Survey of India (SOI) toposheet of corresponding scale. Selected details are taken from the toposheet. The base map depicts latitudes, longitudes, state and district boundaries, important places, drainage etc. The base map is set on the imagery of corresponding area so that the important features of the base map and imagery overlap each other completely. Thereafter, the interpretation of imagery is done using interpretation keys based on tone, texture, location, association etc. Delineation of forest patches is done on a light-table with the help of magnifying glasses. It results in separation of areas in the categories of dense forest, open forest, mangrove, scrub, non-forest etc. on a map. This forest cover map is then compared with the cover map of the preceding assessment. The changes with respect to preceding assessment are marked and a change map is prepared. These maps are used for ground verification to check the veracity of interpretation. Errors detected during ground verification are rectified and corrections incorporated. The area is calculated using transparent dot grid template. The forest cover is compiled by district and state.

Digital Interpretation
The satellite data for the entire country is procured from the National Remote Sensing Agency (NRSA), Hyderabad in digital form. It is a multispectral (LISS-III sensor) data of IRS P6 satellite with a resolution of 23.5 m. One scene of LISS III covers an area of about 20,000 km² (140 km x 140 km). A total of 327 scenes covering the entire country are procured.
The period of satellite data is of upmost importance. The reflectance from the forest is dependent on the crown foliage and its chlorophyll content. A deciduous forest would, therefore, not give proper reflectance in leafless period. Thus, data of the sprint summer season for such forests is not suitable for interpretation. Further, during the rainy season, it is difficult to find cloud-free data, moreover, agricultural and like lands give similar reflectance as forest cover during this period. The satellite data of the period October to December is therefore, most suitable for forest cover mapping of our country. While procuring the data, only those scenes are selected where cloud cover is less than 10 per cent.
Using Digital Image Processing (DIP) software, the satellite data in digital form is downloaded on the Workstations from the CDs procured from the NRSA. Radiometric and contrast corrections are applied for removing radiometric defects and for improving visual impact of the False Colour Composites (FCC).
Geometric rectification of the data is carried out with the help of scanned and geo-reference Survey of India (SOI) toposheets on 1:50,000 scale. The methodology of interpretation involves a hybrid approach in which unsupervised classification (ISODATA algorithm) aided on-screen visual interpretation of forest cover is done.
Normalized Difference Vegetation Index (NDVI) transformation is used for removing non-vegetated areas from the scene. Areas of less than one hectare, whether classified as forest cover within non-forest areas or blanks within forest cover, are excluded by appropriate DIP techniques. Degraded forests were tree canopy density is less than 10 per cent are classified as scrubs, which do not form part of the forest cover.
Shadow areas in the scenes are treated separately. Shadow regions on the images are highlighted using band ratio techniques. Forest cover classification of the totally obscure areas due to shadow or cloud cover are done using the ground truth information.
Mangrove forests have characteristic tone and texture on the satellite image. Their presence on the coastal areas make them even more conspicuous. They are, therefore, separately classified.
Interpretation is then followed by extensive ground verification which takes more than six months. All the necessary corrections are subsequently incorporated. Reference data collected through ground truth and field experience of the interpreter play an important role in delineating the forest cover patches and classifying them into the three canopy density classes.
Sheet wise mosaic of districts and States/UTs was made using SOI and census data to compute district wise and State/UT wise forest cover. The final output of the forest cover mapping is in the raster format.

Limitations of Remote Sensing Technology
There are certain limitations of remote sensing based mapping of forest cover, these are:
· Since resolution of data of LISS-III sensor is 23.5m, the linear strips of forest cover along roads, canals, bunds and railway lines of lesser width are generally not captured.
· Young plantations and species having less chlorophyll contents in their crown cannot be delineated as forest cover.
· Considerable details on ground may be obscured in areas having clouds and shadows. It is difficult to interpret such areas without the help of collateral data of ground truth.
· Gregarious occurrence of bush vegetation and certain agricultural crops, such as sugarcane, cotton, lantana etc. often poses problems in delineation of forest cover, as their reflectance is similar to that of tree canopy.