Methodology and Approach
Vegetation Type Mapping
vegetation type can be defined as an embodiment of unique physiognomy, structure, and floristics (intrinsic factors), influenced by the climate, topography, and anthropogenic factors (extrinsic factors). Champion and Seth‘s (1968) classification scheme follows a hierarchical approach wherein climatically driven forest ecosystems systems with distinct physiognomy and phenology are primarily classified as type groups. These type groups are further subdivided into sub-groups based on dominant compositional patterns and region and location specific formations controlled by edaphic and disturbance conditions. Gadgil and Meher-Homji (1990) distinguished 42 forest types in India, based on the association and dominance of species and the prevailing bio-climate.
The existing classification systems precisely used ground data in deciphering the patterns of species assemblages but did not provide the explicit spatial boundaries of these assemblages. Such spatial explicit boundaries of vegetation types are important for studying the patterns of vegetation diversity and long term monitoring. The delineation of such boundaries for larger spatial extents based on geospatial tools and field information have become time and cost effective. The satellite remote sensing data, in conjunction with spatial information on the topography, soils, climate, and ground floristic data, are also used to delineate detailed vegetation formations (Ravan et al., 1996). The on-screen visual interpretation technique has been used for vegetation type/land use mapping (Fig. 1). The biogeography and altitude zone maps were also used to define classes. Wherever necessary, field data were used to delineate the vegetation type and locale-specific classes. State level vegetation type maps were mosaiced to generate a national level map. Edge matching was performed to produce a seamless national vegetation type map.
Selection of Optimal Season Data
Two-season IRS LISS-III satellite data of 2005-2006 were utilized optimally to map the vegetation types depending on the forest phenology, i.e., peak growth and leaf fall seasons. Satellite data pertaining to the time windows of November-early January and February-early April were used to take into account the phenological variations required for delineation of different vegetation types. In the case of grassland areas in Gujarat and Rajasthan, an additional data set covering August-October was also used. The IRS P6 LISS-III sensor was used. If no specified cloud free data were available, the best available archived data were used (Table 1)
Fig 1. Approach used in classifying vegetation using multi-temporal remote sensing.
Table 1. Optimal seasons of satellite data for different parts of India
|Region||Season of data selection|
|South||Nov-Dec and Mar-Apr|
|Central||Nov-Dec and Mar-Apr|
|North||Dec-Jan and April|
|South||Sept-Oct and Feb-Mar|
|Central||Oct-Dec and Mar-Apr|
|North||Nov-Dec and Mar-Apr|
|Central Plains||Nov-Dec and Mar-Apr|
|Eastern Himalaya||Dec-Jan and Apr-May|
|Subtropical temperate region||Nov-Jan and Mar-May|
|Cold desert||Aug-Nov and Apr-May|
|Indo-Gangetic Plains||Aug-Nov and Feb-Mar|
|Western Arid system||Aug-Oct, Nov-Dec and Mar-Apr|
Radiometric corrections were carried out using dark pixel subtraction. Scene-to-scene matching was carried out using histogram equalization/matching. For missing lines or pixels, suitable interpolation techniques were used.
Reconnaissance Survey/Ground Truth Collection
It is required to have a reconnaissance study of an area before attempting to classify the vegetation pattern. Traverses in the area of interest were made from the plains to the hill tops for collecting ground truth information. A survey of the published literature was carried out, and several interactions were held with the forest departments and educational/local institutions. At the end of the reconnaissance survey, an understanding was gained on the prevailing phenological, gregarious, locale-specific vegetation types of the study area. The information available in the forest working plans, published records, the tone and texture of satellite imagery, and the ground knowledge were used. The location specific data gathered on different vegetation types were utilized to prepare (a) a template for visual interpretation of satellite data and (b) delineate training sets for digital classification of satellite data.
Image Interpretation Key
An image interpretation key was developed prior to interpretation, which was further refined during the course of interpretation (Fig 2.).
Fig. 2. Use of tone and texture for on-screen digitization of vegetation types (IRS-P6 LISS III FCC images of part of Malkangiri district of Odisha showing phenological variability)
The vegetation classification scheme was framed that accommodates the natural and semi-natural systems were classified into forests, s crub/shrub lands, and grasslands based on the extent of green cover. The cultivated and managed systems were classified into orchards, croplands, long fallow/barren lands, and water bodies. The forest class was further sub-divided into mixed forest formations, gregarious formations, locale-specific formations, degraded/successional types, and plantations. The classification scheme, class details, and class codes are given in Table 2.
Table 2. The vegetation/land use types and their respective classes under Champion and Seth‘s classification.
|Class description||Champion and Seth (1968) class with codes|
|Evergreen||Tropical Wet Evergreen Forest (1)|
|Giant evergreen||Giant Evergreen Forest (1A/C1)|
|Andaman evergreen||Andamans Tropical Evergreen Forest (1A/C2)|
|Southern hill top||Southern Hilltop Tropical Evergreen Forest (1A/C3)|
|Subtropical broadleaved hill forest||Subtropical Broadleaved Hill Forests (8)|
|Subtropical dry evergreen||Subtropical Dry Evergreen Forests (10)|
|Montane wet temperate||Montane Wet Temperate Forests (11)|
|Himalayan moist temperate||Himalayan Moist Temperate Forests(12)|
|Himalayan dry temperate||Himalayan Dry Temperate Forests(13)|
|Semi evergreen||Tropical Semi-Evergreen Forests(2)|
|Moist deciduous||Tropical Moist Deciduous Forests(3)|
|Sal mixed moist deciduous||Moist Teak-Bearing Forests (3B/C1)|
|Teak mixed moist deciduous||Very Moist Sal-Bearing Forests (3C/C1)|
|Dry deciduous||Tropical Dry Deciduous Forests (5)|
|Sal mixed dry deciduous||Dry Sal-Bearing Forests (5B/C1)|
|Teak mixed dry deciduous||Dry Teak-Bearing Forests (5A/C1)|
|Thorn forest||Tropical Thorn Forests (6)|
|Sal||Moist Sal Bearing Forests(3C/C2)|
|Teak||Dry Teak Bearing Forests(5A/C1)|
|Mesua||Mesua Forest (1B/C2B)|
|Bamboo||Wet Bamboo Brakes (2/E2), Moist Bamboo Brakes (2/E3),
Secondary Moist Bamboo Brakes (2/2S1)
|Pine||Subtropical Pine Forests (9), Siwalik Chir Pine Forest (9/C1a), Himalayan
Chir Pine Forest (9/C1b), Western High-Level Dry Blue Pine (13/1S3)
|Fir||Fir Forest (14/C1a)|
|Oak||Montane Bamboo Brakes (12/DS1)|
|Deodar||Moist Deodar Forest (Cedrus) (12/C1c)|
|Hardwickia||Hardwickia Forest (5/E4)|
|Red sanders||Dry Red Sanders Bearing Forest (5A/C2)|
|Boswellia||Boswellia Forest (5/E2);|
|Acacia nilotica (Babul)||Babul Forest (5/E3)|
|Butea||Butea Forest (5/E5)|
|Aegle||Aegle Forest (5/E6)|
|Acacia catechu (khair)||Khair-Sissu Forest (5/1S2)|
|Anogeissus pendula (kardhai)||Anogeissus pendula Forest (5/E1)|
|Acacia senegal||Acacia Senegal Forest (6/E2)|
|Cypress||Cypress Forest (12/E1)|
|Alder||Alder Forest (12/1S1)|
|Rhododendron||Dwarf Rhododendron Scrub (15/C2/E1)|
|Hollock (Terminalia myriocarpa)|
|Mangrove||Tidal Swamp Forests (4B), Mangrove Forest (4B/TS2)|
|Mangrove scrub||Mangrove Scrub (4B/TS1)|
|Phoenix (palm swamp)||Palm Swamp (4B/TS4/E1)|
|Littoral forest\beach forest||Littoral Forest (4A)|
|Freshwater swamp forest||Tropical Freshwater Swamp Forests (4C)|
|Lowland swamp forest||Tropical Seasonal Swamp Forests (4D)|
|Myristica swamp||Myristica Swamp Forest (4C/FS1)|
|Syzygium swamp||Syzygium cumini Swamp Low Forest (4D/SS3)|
|Shola||Southern Subtropical Broadleaved Hill Forests (8A)|
|Riverine||Tropical Riparian Fringing Forests (4E)|
|Dry evergreen||Tropical Dry Evergreen Forests (7)|
|Ravine||Ravine Thorn Forest (6B/C2)|
|Red oil palm|
|Shifting cultivation (abandoned jhum)|
|Shifting cultivation (current jhum)|
|Tree savannah||Low Alluvial Savannah Woodland (Salmalia-Albizzia) (3/1S1),
Dry Savannah Forest (5/DS2)
|Shrub savannah||Dry Savannah Forest (5/DS2)|
|Dry evergreen scrub|
|Dry deciduous scrub||Dry Deciduous Scrub Forest (5/DS1)|
|Ziziphus||Southern Thorn Scrub (6A/DS1)|
|Euphorbia scrub||Euphorbia Scrub (6/E1)|
|Moist alpine scrub||Moist Alpine Scrub (15)|
|Dry alpine scrub||Dry Alpine Scrub (16)|
|Salvadora||Salvadora Scrub (6/E4)|
|Hippophae||Hippophae- Myricaria Scrub (13/1S1)|
|Desert dune scrub||Desert Dune Scrub (6/1S1)|
|Wet grasslands (upland grasslands)||Southern Montane Wet Grassland (11A/C1/DS2)|
|Riverine (lowland grasslands)|
|Moist alpine pasture||Alpine Pastures (15/C3)|
|Dry alpine pasture||Alpine Pastures (15/C3)|
|Saline grassland||Saline/Alkaline Scrub Savannah (5/E8)|
|Dry grassland||Dry Grassland (5/DS4)|
|Long fallow/barren land|
The species composition was recorded by field sampling in the respective mapped vegetation types, based on the stratified random sampling design described in section Phytosociological Analysis. The classes which were not amenable for delineation directly using remote sensing were grouped in the broad class. The hierarchical classification of the forest type will help in linking with different global classification systems and converging to the global scale. This has been undertaken to facilitate the migration of the database from the current classification system to any of the globally recognized classification systems for climate sensitive approaches and other research purposes. Descriptions of the different mapped vegetation types along with the satellite signatures and field photographs have been provided in Appendix 1.
he natural vegetation in the country has a long history of disturbance by way of grazing, fire, logging, deforestation for raising forest plantations, etc., resulting in complex habitats. In order to understand the composition and species diversity pattern in these complex habitats, the landscape characterization in terms of patch size, shape, and neighborhood, coupled with phytosociological data, has been taken into account. Vegetation strata proportions were used for determining the sample points (plots). A sample intensity of 0.002 to 0.005 was aimed at, depending upon the state of the forests in the area. Stratified random sampling with probability proportionate to stratum size was used across the country. Field information on cover type, locality, aspect, slope, geo-coordinates, signs of disturbance, and altitude were recorded. GPS receivers were used to determine the geo-coordinates and the altitude. Fig. 3 depicts the sample plots.
Fig. 3. Distribution of field sample plots across India
Sample plots (modified nested approach) of extent 0.04 (20 m x 20m) to 0.1 ha (31.62 m x 31.62 m) were randomly distributed across each stratum. Tree species were sampled using 20 m x 20 m and 31.62 m x 20 m plots, depending upon the within-stratum variability on the ground. For sampling the shrub species, two plots of size 5 m x 5 m at two opposite corners of a 20 m x 20 m tree plot were taken. For herbaceous plants, five plots of size 1 m x 1 m (four at the corners and one at the center) were laid inside the tree plot (Fig. 4). GPS and ground bearings from SOI maps were used to reach the plots. A modified nested quadrant was used for laying the tree, shrub, and herb plots. Information on trees, shrubs, herbs, climbers, epiphytes, and lianas was recorded from these plots using field forms.
Fig. 4. Tree, shrub, and herb sample plots in a modified nested quadrant.
In each sample plot, the circumference at breast height (cbh) of each tree with cbh = 30 cm was recorded. Trees with cbh > 17 cm and < 30 cm were treated as saplings, and those with cbh <17 cm were treated as seedlings. In the case of shrubs, the cbh was measured about 30 cm above the ground. The total number of seedlings of various species was counted, and the average girth of each species was recorded. The total number of tillers of each shrub species was counted, and for each species the average circumference at ground height level was worked out. The tree/stand height was also recorded.
The field data were used to analyze the spatial patterns of vegetation types, species diversity, total importance value, ecosystem uniqueness, and species richness. The frequency, abundance, density, basal area, IVI (importance value index), diversity index (Shannon-Weaver), H ‘ were computed.
Economic Valuation of Biodiversity
Each plant has its own value in terms of primary benefits such as livestock grazing/fodder, medicinal use, human food, fuel wood, timber, and charcoal and secondary benefits such as use in oil extraction, fiber, mats, ropes and baskets, and tanning leather and indirect benefits such as shaping hedges, soil stabilization, a role in nitrogen fixation, and scientific importance (Belal and Springuel, 1996). The economic importance may be related to the whole of the plant or part of it. Based on its importance, an importance value index was derived, and a 0–10 point scale was assigned for each use. Thus, the Total Importance Value (TIV), based on the potential importance of the plant to the local economy, was calculated as follows:
TIV(%) = (U1 + U2 + U3+.........Un/no. of uses x maximum value) x 100
where U is the importance value for each particular use, i.e., timber, fuel wood, food, etc.
Each plant species was valued on a 1-10 scale for its importance in fodder/grazing, medicine, human food, fuel wood, timber, charcoal, dye, oil, tannin, and other direct or indirect uses. The biodiversity attributes which need to be assessed for the above are richness (the number of species), rarity/threat (degree of harm), endemism (restricted to certain geographical locations), distinctiveness (the amount that differs from its nearest relative), representiveness (closeness of an area represents a defined ecosystem), and function (the degree to which a species or ecosystem affects the ability of other species or ecosystems to persist). These, along with economic values for the goods and services provided by a species or ecosystem or landscape, can be indicators of the importance of a particular landscape in terms of conservation or bio-prospecting.
Fragmentation was computed as the number of patches of forest and non-forest types per unit area. The forest type map was reclassified into two classes, i.e., forest and non-forest, resulting in a new spatial data layer. A user grid cell of n (e.g., n = 500 m) is convolved with the spatial data layer with a criterion of deriving the number of forest patches within the grid cell. The iteration is repeated by moving the grid cell through the entire spatial layer. An output layer with patch numbers is derived and a look-up table (LUT) associated with this is generated, which keeps the normalized data of the patches per cell in the range from 0 to 10 (IIRS, 2003).
Frag = f(nF,nNF)
where Frag = fragmentation; n = number of patches; F = forest patches; NF = non-forest patches.
The anthropogenic influence on the landscape is a discrete event through time that modifies landscapes, ecosystems, community, and population structure, changing the substrata, the physical environment, and availability of resources (White and Picket, 1985). Disturbance and fragmentation are two related processes with strong relationships, and it is difficult to distinguish the role and rate of the interactions. Ecosystems are in a continuous state of change, either due to natural succession or degradation due to anthropogenic pressure. The latter phenomenon is more prevalent in the areas to be taken up in phase III of the project study. Described below is a conceptual ecosystem processes at the landscape level to be considered (Fig. 5) while appreciating the role of factors to be included in the quantification of disturbance.
Fig. 5. Role of disturbance factors at the landscape level (Source: NRSC, 2008).
Fig. 6. Flow chart showing schematic method for computing Disturbance Index (NRSC, 2008).
The disturbance surface was prepared as a combination of different landscape matrices, viz., fragmentation, porosity, juxtaposition, and interspersion. The spatial distribution of the anthropogenic/natural forces on the landscape was used to generate the spatial distribution of disturbance factors, viz., proximity to roads, villages, fire intensity, shifting cultivation, and mines using ground based sampling data as well as ancillary databases. Using these, the disturbance surface was generated (Fig. 6). Baseline details of roads and settlements were used to create a buffer (distance from the source of disturbance). A zone of 2-5 km, based on the level of human induced factors and field knowledge, is considered for buffering. Variable buffering with respect to the radial distance from the point of disturbance is performed by imposing the condition that ‘the greater the distance, the less the weightage’. The same criterion is applied to point and polygon type data. The disturbance index (DI) is computed by adopting a linear combination of the defined parameters on the basis of probablistic weightage. The mathematical equation used for computing the Disturbance Index is as follows (NRSC, 2008):
where DI = Disturbance Index; Frag = fragmentation, Por = porosity; Patc = Patchiness; Int = interspersion; Jux = juxtaposition; Wt = weights.
The final spatial data were rescaled to a range of 0-100 for the preparing the final map.
A brief description of the indices used in computing the Disturbance Index is given below.
Fragmentation: Fragmentation has been taken as the number of forest and non-forest patches in a 500 m x 500 m grid and is the normalized index of number of patches per grid.
Porosity: Porosity is a measure of the number of patches or density of patches within a particular type, regardless of patch size. Porosity was calculated for only primary forest types or ecologically unique ecosystems, e.g., tropical wet evergreen forests, mangroves, sholas, etc.
Interspersion: Interspersion is a count of dissimilar neighbors with respect to a central pixel or a measure of the spatial intermixing of the vegetation types (Foreman and Godron, 1986). Interspersion is assessed by running a convolution window of 3 x 3 pixels (pixel size 24 m) on the forest type map to compute the number of dissimilar pixels in the nearest neighborhood. The computation is performed in an interactive mode through the entire spatial layer to derive an output interspersion layer. A normalized LUT will be made in the range from 0 to 10.
Juxtaposition: Juxtaposition is defined as measure of the proximity of vegetation types. Its measurement mostly includes a relative weightage assigned by the importance of the adjacency of two cover types for the species in question. Forest types were reclassified as natural and man-made vegetation. A grid cell of 3 pixels x 3pixels was convolved with the derived layer in an iterative manner by assigning higher weighs to natural vegetation and lower weighs to man-made vegetation.
Here we modeled the biological richness of as a function of ecosystem uniqueness, species richness, biodiversity value, terrain complexity, and disturbance and depicts the potential for harboring the maximum number of ecologically unique and important species. This helps in assigning conservation priorities to threatened, rare, endemic, and taxonomically distinct species and to different types of habitats or landscape elements on the basis of the richness and significance of threatened species. As a part of this project, the biologically rich areas were spatially identified for the purpose of conservation and saving the existing gene pool from extinction. Since the disturbance index, which is a part of the ecosystem process, is also a function of the biological richness, so the level of stress on the biologically rich areas is also ascertained and adequate remedial measures can be taken while implementing conservation strategies .
The biological richness at the landscape level was computed as a function of ecosystem uniqueness, species diversity, biodiversity value, terrain complexity, and Disturbance Index (NRSC, 2008):
where BR = biological richness, DI = Disturbance Index, SR = species richness, BV = biodiversity value, EU = ecosystem uniqueness, and Wt = weights.
A brief description of the various environmental and habitat attributes, viz., terrain complexity, ecosystem uniqueness, total value index, and species richness, that are used in computation of biological richness is described below.
Terrain Complexity (TC) was computed as the rate of variability or variance in the terrain, taking into account the rate of change of slope, aspect, and altitude in a given mask. DEM serves as the input data for computing this parameter. The contour intervals adopted for computation vary from 10-20 m in the plains to 40-60 m in moderately hilly regions and 80-120 m in mountainous regions.
Ecosystem Uniqueness (EU): The uniqueness of the ecosystem was determined based on field data, species composition, extent of the area, contiguity, importance in the landscape, anthropogenic pressure (from population layer), and the critical habitat value of the patch. This takes into account the species composition, extent of the area, contiguity, importance in the landscape, and critical habitat value of the patch. For example, a region having species localized to the area and having indigenous and endemic species, with no exotic species, has the potential to have high ecosystem uniqueness. Another criterion for ecosystem uniqueness is the presence of critically endangered, endangered, or vulnerable species in the area.
Biodiversity Value (BV) has been computed for each vegetation type using the economic value, scientific value, local needs, and knowledge base of the vegetation and the ecosystem services.
Species Richness (SR) is computed as the type-wise species richness, calculated using the Shannon-Wiener Index (H’) and integrated into the vegetation type map.
Ecosystem Uniqueness (EU) is computed for the vegetation type in terms of the fragility of the ecosystem, representativeness, species composition, endemism, contiguity, and IUCN category (such as threatened, rare, endangered, vulnerable, and intermediate).
All these sub-models have been developed and integrated into a software package named Spatial Landscape Model (SPLAM) (Fig. 7). The model was coded using C++ and has an interactive GUI. The ArcGIS engine is used for displaying the results as well as for its libraries.
Fig. 7. Geospatial model for biodiversity characterization using SPLAM (Roy et al., 2005).