3.1 Quantification of spatio temporal changes & implementation of the forest fragmentation model:

A. Land use analysis and change detection: Land use analysis was carried out at different time scales to determine the current status of forest ecosystem and the causes of transition in the land cover of the area. Temporal land use analyses show a decrease in the evergreen forests from 67.73% (1973) to 32.08% (2013). Land use changes involving the expansion of agricultural and horticultural lands are the prime causal factors of forest degradation.

The change in the land use pattern of the area can also attributed to various on-going developmental projects in Uttar Kannada. Portion of the evergreen forest is lost due to the horticultural plantations like areca and coconut plantation which have showed an increase from 2.01% to 5.25% during the time period from 1973 to 2013. The image differentiation is performed to identify the loss of forest cover from 1973 to 2013. This was done by thresholding the forest class and considering other classes as null. This visualization helps in assessing the impacts of unsustainable policy interventions, market induced changes and their implications.

3.2 Measuring LULC changes: LULC changes in Uttara Kannada district is analysed using temporal remote sensing data with ancillary data and field data. The method followed for LULC analysis is represented in figure1.2.

B. Forest fragmentation: Forest fragmentation category at pixel level is computed through Pf (the ratio of pixels that are forested to the total non-water pixels in the window) and Pff (the proportion of all adjacent (cardinal directions only) pixel pairs that include at least one forest pixel, for which both pixels are forested. In other words, The Pf is simply the proportion of non-missing pixels within the moving window with a specified size that are forest, and Pff is the ratio of the number of pixel pairs in cardinal directions that are both forest divided by the number of pixel pairs in cardinal directions where either one or both are forested. Pff estimates the conditional probability that given a pixel of forest, its neighbour is also forest. A moving window with the size of 5 X 5 pixels was used for the fragmentation analysis to maintain a fair representation of the proportion (Pf) of pixels in the window and to maintain interior forest at an appropriate level, due to the fact that the outcomes of the model are scale-dependent and threshold dependent (Riitters et al., 2000). The spatial maps of forest fragmentation components thus derived by computing ‘Pf’ and ‘Pff’ using the sliding kernel (of size 5 x 5)  based on the total extent of forest and its occurrence as adjacent pixels. A pixel is classified by the type of fragmentation and the result of the kernel is stored at the location of the center pixel (in the derived map), which represents between-pixel fragmentation around the corresponding forest location. Details of levels of fragmentation with discriminant criteria of the each component are listed in Table 1 based on two indices, Pf and Pff. Depending on these indices values, the analysis will derive different fragmentation components to assess the health of forest of a region. Forest pixels that comprise a small forested area surrounded by non-forested land cover refers to Patch forests. Forest pixels that define the boundary between interior forest and large non forested land cover features are Edge forests. Perforated forest refers to pixels that are the boundary between interior forest and relatively small clearings (perforations) within the forested landscape. Interior forest are the forest pixels that are reasonably far away from the forest-non forest boundary. Interior forested areas are surrounded by more thick forested areas. Transitional types are clearly depending on non-forest and edge pattern. These are in between edge type and non-forest types. If higher pixels are non-forest then they will be tending to non-forest cover with higher degree of edge. The water bodies or river coarse are considered as non-fragmenting features, because they act as natural corridors in forested landscape. Non-forested areas including buildings, roads, agricultural field, and barren land, along with developed land, are considered fragmenting features.

Table 1: Fragmentation components and their description:

Figure 3: KERNEL (5x5) for computation of Pf, Pff values

Computation of Pf and Pff is explained considering a 5 x 5 kernel with 25 pixels shown in Figure 3. Forest pixels are shaded (15 pixels) and non-forest pixels are not shaded and Pf  is 15/25=0. 6. Considering pairs of forest pixels in cardinal directions, the total number of adjacent pixel pairs is 40, and of these, 40 pairs include at least one forested pixel. Eight of those 40 pairs are forest-forest pairs, so Pff equals 8/40 = 0.2.

Figure 2: A theoretical frame work of Fragmentation