Wasteland and forest density
mapping using IRS data D.K. Pal & I.V.
Ramana Scientists Regional Remote Sensing Service Centre, Dept. of Space, Kharagpur – 721 302, India Abstract Unscientific handling of land resources has resulted in the development of vast stretches of wastelands and also formed one of the major factors of decrease in per capita arable land besides causing ecological imbalances. Indiscriminate deforestation and faulty agricultural practices are the major reasons for land degradation culminating in the formation of wastelands. The district of Bankura in the state of West Bengal, India, has been taken latitudes 22 38’ N- 23 3’ n and longitudes 86 36’ E – 87 46’ E and has an area of 6,882 sq kms. A stratified classification of notified forests and nonforest areas has been envisaged here. Wastelands in nonforest areas ere classified by supervised technique using maximum likelihood classifier while three forest density classes within the notified forest areas were segregated by density slicing of the Normalized Difference Vegetation Indices (NDVI) scene. In this paper using VAX 11/780 environment and complemented by adequate software support, different steps leading to the generation of one agricultural season map (1988-89) by aggregation of two major agricultural season’s (Rainy seasons and winter season) classified output, have been discussed in detail. The extent of miss-classification of the individual season classified output has been reduced to a great deal using one software program enables incorporation of human interactions. District boundary and notified forest boundaries were digitized and individual mask files created there from, which were all overlaid and district was extracted as the final output. Statistics for different wasteland and forest classes have been presented here. Introduction Accurate and up to date data base is indispensable for effective management of regional wastelands and forest resources. Traditionally, in India such natural resources statistics have been compiled from village records and orthodox forest maps supplied by the respective forest administrations. All of which suffer from obvious intrinsic flaws as they are not updated on suitable time scale due to the high cost and time over run involved in conventional means of surveying and monitoring, apart from subjectivity. There is now a greater concern for the statistical reliability of the information, the spatial accuracy of the map, and timeliness of its availability. Forests and wastelands and two highly inter-related themes. Annihilation of the former culminates in the aggravation of the latter. Deforestation, being primarily due to anthropogenic exploitation, is intimately connected with the population , is intimately connected with pressure on forests is greatest in the developing countries, which account for over three quarter of global account for over three quarter of global population (Rao 1991). The commercial exploitation of the forests has continued along with the expansion of agriculture. Forests are needed to protect wildlife and soil, to sustain man, to stabilize the climate, to optimize water yields and to purify water. In order to accomplish these multiple functions, a forest has to be well stocked and its cover must be adequate (Roy et al. 1990). The major causes of land degradation and subsequent formation of waste lands can be primarily attributed to ‘faulty agricultural practice’ and ‘indiscriminate deforestation’. Faulty agricultural practices include the lack of soil conservation measured and the faulty irrigation practices that often lead to the formation of the salt affected soils. Extensive deforestation leads to environmental disasters like i) increase an atmospheric carbon di oxide contributing to global warming, ii) increases in albedo altering the energy balance between the terraferma and atmospheric resulting in a pernicious impact on global circulation pattern and rainfall statistics iii) Increase in land surface ruin off affecting not only the water budget but also substantially reducing the underground water recharging and iv) acceleration I the rte of degradation acceleration in the rate of degradation of soil, erosion that far exceeds the rate of pedogenesis the basic mantle of the lithosphere that thrives most entire biosphere. The study described here deals with mapping of water lands (level 2) in forest density classes of an individual district. With a view to alleviating to inherent limitations of using single season data as far as classification accuracy is concerned, it envisages the use two season satellite data (IRS-1A) pertaining to two major agricultural seasons inorder to generate the single output. Notified forests and nonforest areas were stratified and separately classified to mitigate spectral – vegetal confusion. The waste land classes in the district and the area-wise statistics has been presented in the paper. Study Area The district situated between latitude 22°38’ N - 23°38’ N and longitudes 86°36’ E 87°46’ E. It occupies the area of 6,88,200 Ha (location map in figure 1). The district of Bankura lies along the intermediate country between the Chotnagpur Plateau and the rice procedures alluvial plains in the lower Gangetur delta. So the district may be divided into three topographic regions viz., i) the hilly country to the west, ii) connection undulating tract in the middle, and iii) the level alluvial plains to the east. In the extreme west the elevation become still more pronounced as one approaches the Chotnagpur Plateau and this portion of the district comprise numerous groups of hills and isolated peaks and is therefore, broken and rocky in nature. Among the rivers of the district the Daamodar is most important. The chief tributaries of the damodar are Sali and Bodai. Next to Damodar, the Dwarakeshwar is the second important river. The climate of the district is charcterised by an oppressively hot summer high humidity nearly all the year round and a well distributed rainfall during the monsoon months. The average annual rainfall of the district is 1,271 mm. July and August are generally the rainiest months. The greater part of the district consists of rolling country composed of laterite and alluvium. To the east there is a wide plain of recent alluvium, while gneisses and schists of Achaean age are found to the extreme west which from the eastern boundary of similar rocks of Chotnagpur. A number of dolerite dykes cutting across Gondwana rocks as well as the Archaeans are found in the north-western parts of the district. The district is endowed with the distribution of wide range of different soils. Soils in the north-eastern segment can be termed under the Great Groups of the Taxonomic classification system as difluvents (popularly known as young alluvial soils). Following this spreading towards the west are soils that can be regarded as Paleustalfs and Haplaquents (both known as older alluvial soils), Ochraquults, Rhodustults and Haplustults (known as red and yellow soils), Plinthaquults, Plinthustults and Plinthudults (laterite soils). While there are Paleustalfs and Rhodustalfs (red gravelly soils) throughout the western segment of the district. The fertility status of the alluvial soils of the east is superior to the other soils. Data Used In view of the time and cost consideration IRS-1A LISS-1 data was opted for. The data with spatial resolution of 72.5., 4 spectral bands with high spectral resolution in the visible (three) and near infrared (one) regions of 0.45 – 0.86 micro meters range and with 7 bit radiometry aptly suited the present work. Computer compatible tapes (CCTs) of two seasons representative of the rainy season (Kharif) and winter (Rabi) were selected to cover entire land use information for the two-season crop year i.e., 1988-89. For the ideal period, in each of the two seasons, the data was selected when all the crops have passed seedling stage but not attained maturity and was also free from cloud cover. The following CCTs were processed to cover the entire district of Bankura.
Methodology The methodology involves processing of two season data for Kharif and Rabi Separately and finally aggregating them to generate a single output. It is based on stratified classification with provision for human interaction through referential refinement. It essentially comprises classification of individual scenes after stratification of Forests and Non-forest areas. The non-forest areas were subjected to supervised classification using maximum likelihood classifier while forests were classified using the slicing approach of the relevant Normalized Difference Vegetation Index (NDVI) scene of Rabi season, also utilizing the ground truth information. This is to mollify intermixing of forests, grasslands, and crop lands. Finally the classification accuracy is augmented to a large extent by subjecting the classified outputs of both the seasons to a software program called “Referential Refinement” which enables incorporation of human logic in eliminating misclassification in either of the seasons. For example, grasslands are segregated using referential refinement as most of the grasses are dry or grazed away during middle of Rabi season and Kharif misclassifications (that might have occurred) is set right (NRSA 1990).
The Dunn Camera output of the Kharif plus Rabi aggregated with spatial distribution of the categories is shown in sampled and compressed mode in Photo No 3, whereas a full resolution (512x512 pixel) output is displayed in Photo No.4. The detailed statistics pertaining to each category of waste lands and forests in level-2 are shown in Table 1. Brief description of the classes in each of the major categories of wasteland and forest is as follows :
Table 2 Forest land statistics of Bankura District (within the notified forest) for the year 1988-99
Discussion and Conclusion Remotely sensed data have been used as a proven tool for large scale mapping of waste lands (NRSA 1986, Nag et al. 1990) and also for the mapping of vegetation cover using digital and visual techniques (Roy 1987, Jadav et al. 1987). The National Remote Sensing Agency (NRSA), Dept. of Space, Govt. of India, has been the first agency to prepare nation wide waste and land map in one million scale with eight fold classification using Landsat MSS data. Subsequently, they followed up with large scale (1:50,000) mapping of waste lands of the nation using TM data, also by visual interpretation. However, there has not been much attempt to map waste lands by digital techniques. Normal multi spectral data contain composite signatures contributed by type and crown closure. However, the noramlised vegetation index has been found to be more sensitive to canopy closure (Curran 1980). It is observed that the NDVI is only related to canopy closure / physiognomic classes. The NDVI as such does not show any indication of forest type (Roy 1990). Therefore, the use of NDVI is restricted to the density mapping of forests. Reclamation of the water – logged and marshy as is to be attempted with great care and without inflicting on the eco-system of the surrounding areas. Wastes under scrub land category should be reclaimed and put to forestry or agricultural plantations consistent with their land use capabilities. Efforts to be launched in order to cover at least casurina or suitable other plantations. Detailed soil surveys are recommended for assessing the production potentials of such areas. Suitable soil surveys are recommended for assessing the production potentials of such areas. Suitable soil conservation measures are to be adopted for checking soil erosion in the undulating proton of the western region of the district by conservation measures like bunding afforestation, etc. Acknowledgement The authors are grateful to the i) Research Fellows, Dept. of Geology Calcutta University for ground truth support, ii) Scientist, RRSSC, Nagpur for providing software support, and iii) Dr. D.P. Rao, Group Head (Application-1), NRSA, Hyderabad, for his overall coordinating and funding the project. Sincere thanks are due to Shri Y. Rajoo, Shri Y.K. Srivastava and Shri Madan Rana, RRSSC, Kharagpur, for their secretarial assistance. References Curran, P. Mulitspectral Remote Sensing of vegetation amount. Progress in physical geography, 4, 1980, p. 315-341. Jadhav, R.N, V.K. Srivastava, A.K. Kandya, G.V. Sarat Babu, M.M. Kimothy, K.D. Dwivedi, K.V.R. Shetty and K. Murthy. Large scale forest type mapping using satellite data. Proceedings of the XXXVII congress of the international astronautical Federation, Brighton, 1987. Nag, P., D.K. Pal, M. Kudrat, B.K. Sinha and A . Kumar. Identification and mapping of Wastelands from enlarged Landsat thematic mapper imagery in Siwan district of Bihar. S.C. Sharma (Ed) utilisation of Wastelands for sustainable developments. Concept publishing company, New Delhi, 1990. NRSA., Project Report of mapping of waste lands in India from satellite imagery of 1980-82. National Remote Sensing Agency, Hyderabad, 1986. NRSA., Manual of Nation wide Land use/ Land cover mapping using digital techniques. Part-II. National Remote Sensing Agency, Hyderabad 1990. Rao, U.R. space Technology and Forest Manager with Specific Relevance to Developing Nations. Space and Forest Management. Special Current Event session International Astronautical Federation 41st IAF Congress, Dressen, Germany, 1990. Roy, P.S. Montane vegetation stratification through digital processing of Landsat data. Geocarto international Journal of Remote Sensing (i). 1987. p. 19-26. Roy,. P.S., P.G. Diwakar, T.P.S. Vohra and S.K. Bhan. Forest Resource Management Using Indian Remote Sensing Satellite Data. Asian Pacific Remote Sensing Journal, Vol. 3, No. 1, July 1990. p. 11-22 |