Keywords

5.1 Introduction

Disturbance is responsible for triggering forest fragmentation, edge-interactions, and species migration in a landscape. The landscape level disturbance can be perceived with respect to natural and anthropogenic disturbances. Understanding the spatial configuration and the forces prevailing over a landscape helps study the possible impacts (Moloney and Levin 1996). The human being has been altering the frequency and extent of various disturbances at a faster face and greater extent (Barik and Behera 2020). In absence of strict land use and land cover policy enforcement, the settlements and transportation networks such as roads and rails add to the existing sources of biotic disturbances. The disturbances can be visualized across multiple scales and geographical space and can be studied using patchiness or interspersion matrices (Murthy et al. 2016).

Disturbances could change the homogeneity or heterogeneity of elementary patches in a landscape, thereby defining renewed or new functionalities. Any disturbance may alter the homo- or heterogeneous structural organization of patches. Individual-level disturbances may lead to homogeneous conditions at higher scale, while at lower scale multiple disturbances lead to patch heterogeneity. Any alternation in the spatial configuration of a landscape, therefore could be a reflection of the influencing disturbances itself (Das et al. 2017). Further, continuous or periodic disturbances could lead to propagation of the impacts or changes across the landscapes (Behera et al. 2018), thereby posing cascading challenges for management (Das et al. 2021). Therefore, the landscape disturbances must be studied in terms of their nature and sources including periodicity of occurrence and scale, for appropriate management and restoration.

The response of biological diversity to landscape disturbance could vary from individual species to community across a changing disturbance regime (Chitale et al. 2020). The response of species assemblages to any disturbance may demonstrate varied time-lag and reorganization periods. This may lead to alternation of established and matured ecosystems or creation of new ecosystems with renewed species assemblages with tendency of accommodating more invasive and faster growing species. Therefore, the disturbances alter the species assemblage pattern, either by loss or establishment of new species owing to the suitability of the changing environment. Alternations in disturbance regimes and their impact on biotic interactions could modify or redefine the structure and function of biodiversity in a landscape (Behera et al. 2018). However, it may be hard to estimate any quantitative or qualitative change in lieu of complexity and cascading effects of the interactions. Changes in the types, intensity, and frequency of disturbances could affect the composition of biodiversity and primary production within an ecosystem (Mahanand et al. 2021). The impact of climate change with respect to increased extreme weather events also has greater impact on landscape.

Landscape is often regarded as the best spatial scale to assess the responses as it can reflect ecosystem level alternations where the patches are the individual units. A landscape accommodates different homogeneous patches that can be studied to understand their spatial arrangement, shape and size, and neighborhood effect using various landscape matrices (Turner et al. 1989). The arrangement of patches over space such as the number of a particular patch in an area, the shape, and size of the patches, the configuration of patches, can be considered to understand different communities and species level interactions in a natural landscape (Lidicker 1995). The landscape matrices adopt a geographical window approach that could be overlapping or non-overlapping in nature (Dillworth et al. 1994). Fragmentation risks patch level disturbance having implications such as edge effects and alternations in energy balance in a landscape (Nilsson and Grelsson 1995). Therefore, disturbance induces alternations in spatial arrangement of patches and their functional characteristics including biodiversity and primary productivity.

Remote sensing images can be interpreted to define the spatial arrangement of different patches, and are therefore potentially used to analyze disturbances across time and scale. The patch or class wise information can be transferred and linked to characterize biodiversity structure and function using statistical and modeling approaches. The Geographical Information System (GIS) helps in integration of the satellite derived spatial maps such as fragmentation, patchiness, porosity, interspersion, and juxtaposition for estimating landscape disturbance, thereby useful in management decision-making. The Global Positioning System (GPS) provides information on the geographic locations of different landscape elements, thereby helpful in locating the sample plots for mapping, classification accuracy assessment, spatio-statistical analysis, and model validation (Behera et al. 2000).

Indian national level forest cover assessment is done biannually using visual and digital interpretation techniques by Forest Survey of India (FSI; SFR 2019). In a maiden attempt, Roy et al. (2015) that have prepared a nation-wide vegetation type map of India using satellite data, with 100 classes such as mixed, gregarious, locale specific, degradational and scrub formations along with plantations. Vegetation type map serves as an intermediate input for calculation of disturbance index (DI). Here, the vegetation type map was generated using satellite data that was further utilized for generating a DI image along with other information using a customized package, Bio_CAP (Behera et al. 2005).

5.2 Study Site

The Arunachal Pradesh, Assam, and Meghalaya states are located in the eastern Himalaya, and categorized under the Himalaya-East-Himalaya biogeographic zone (Rodgers and Panwar 1988), were selected as study sites (Fig. 5.2a; Table 5.1). The three states accommodate diverse tropical ecosystems comprising of mixed wet and dry evergreen, and deciduous broadleaved forests. The study site experiences very high humidity thereby supporting diverse forests with several primitive species. The three states vary as per climate and topographic conditions as well as human activities.

Table 5.1 Characteristics of the study site, three states in northeastern India
Full size table

The state of Arunachal Pradesh accommodates a wide range of ecosystems such as grasslands, dense evergreen broadleaved and needle leaf forests, alpine scrubs, open forests adjoining to settlements, and road networks, and disturbed secondary formations (Roy and Behera 2005). The state of Meghalaya conglomerates several undulating hills in east-west orientation, thereby accommodating mosaic of valleys and hillocks. The flora of Meghalaya has been reported as the richest in India with a dominant number of angiospermic plant species. Recently large-scale deforestation have been reported from the forests of Assam and Arunachal Pradesh (Pasha et al. 2020). The driving forces for forest loss and conversion in the region are deforestation and shifting cultivation activities (Table 1). In past few decades, the forest loss due to slash-and-burn agricultural practices (also called jhumming or shifting cultivation) has slowed down (Das et al. 2020). The jhumming is a process of slash-and-burn agriculture, wherein the farmers clear a patch of forest for cultivation and leave it fallow for certain number of years. Deforestation has a strong relationship with the proximity to roads and habitations. The rise in population growth, the shifting cultivation activities, and industrialization in the form of plywood factories are some of the major proximate factors responsible for deforestation and forest degradation in northeastern India (Khan et al. 1997).

5.3 Methodology

Forest fragmentation can be defined as the number of forest patches to the non-forest patches in a landscape (Monmonier 1974), whereas patchiness is estimated as the density and number of patches of all types occurring in a landscape (Romme 1982). The higher fragmentation and patchiness are indications of landscape heterogeneity, thus associated with disturbance. The measure of interspersion and juxtaposition in a landscape indicate dissimilar and proximal neighborhoods of various patches respectively (Lyon 1983). Therefore, higher interspersion indicates greater disturbance while lower juxtaposition indicates higher disturbance in a landscape. The porosity can be calculated to realize the disturbance within a native or original vegetation type in a defined region (Forman and Godron 1986). For the northeastern region, the porosity was judged for all the primary forests. The species richness information from 862 field plots provided inputs for juxtaposition calculation and correlating between different DI levels and their biodiversity content (Fig. 5.1).

Fig. 5.1
figure 1

a Disturbance index (DI) is estimated as a cumulative property of social and spatial attributes in a landscape; paradigm for b Assessment of forest disturbance regimes using remote sensing, GIS, and GPS tools (Modified from Behera et al. 2005) and c Validation of DI using field sampling plots

Full size image
Fig. 5.2
figure 2

a location map of (i) Arunachal Pradesh, (ii) Meghalaya and Assam, falling two hotspots; Arunachal Pradesh and upper part of Assam belong to Himalaya hotspot [shown in blue color]; Meghalaya and southern part of Assam belong to Indo-Burma hotspot. Depiction of b Fragmentation image, c Porosity image of sub-tropical evergreen forest, d Interspersion and e Juxtaposition image for part of the study area

Full size image

The landscape level disturbance Index (DI) was assessed as cumulative function of spatial and social attributes using a customized package Bio_CAP (Fig. 5.1). IRS LISS-III satellite data were used for classification of forest vegetation types using a pre-defined classification scheme (Roy et al. 2012). The simplified forest fragmentation map was utilized to generate a DI image along with a few other landscape matrices such as patchiness, porosity, juxtaposition and interspersion, and settlement and road buffers. Stratified random sampling following probability proportionate to the size (PPS) was practiced for field sampling and linking in-situ forest composition. A total of 862 field plots were laid in Arunachal Pradesh (405), Assam (324), and Meghalaya (133) respectively, those accounted for 0.002–0.005% of the natural vegetation area. The measurements for all tree and liana species (20′20 m), shrubs (5′5 m), and herbs (1′1 m) were taken within each sample plot (Roy et al. 2012). The location and extent of the natural vegetation types and their landscape matrices were integrated with the social attributes such as road and settlement buffer, by assigning proportionate weights (Behera et al. 2005) to derive DI (Eq. 5.1; Fig. 5.3). The DI was scaled at 10-levels and an area estimate was done (Fig. 5.4).

$$\begin{aligned} {\text{Disturbance}}\;{\text{Index}}\left( {DI} \right) & = f\int {\left\{ {{\text{disturbance}}\;{\text{buffer}}\left( B \right),{\text{interspersion}}\left( I \right),} \right.} \\ & \quad {\text{juxtaposition}}\left( J \right),{\text{fragmentation}}\left( F \right),{\text{patchiness}}\left( P \right), \\ & \quad \left. {{\text{and}}\;{\text{porosity}}\left( {PO} \right)} \right\} \\ \end{aligned}$$
(5.1)
$$DI = \left\{ {{\text{Wti}}\left( F \right) + {\text{Wti}}\left( P \right) + {\text{Wti}}\left( I \right) + {\text{Wti}}\left( {PO} \right) + {\text{Wti}}\left( J \right) + {\text{Wti}}\left( B \right)} \right\}$$
Fig. 5.3
figure 3

Disturbance Index (DI) image of (i) Arunachal Pradesh, (ii) Assam and (iii) Meghalaya; Various DI-levels range from 1 to 10

Full size image
Fig. 5.4
figure 4

Percentage area (Y-axis) distribution in (i) Arunachal Pradesh (ii) Assam and (iii) Meghalaya; X-axis represents different DI levels. Note that severity of disturbance varies from level 1 to 10

Full size image

where, \({\text{Wti}}\left( {t = 0.1 - 1.0} \right)\) indicates proportionate weight.

For validation of DI, the field plots were segregated into 10-categories using their field gathered GPS coordinates (Fig. 5.1c). Each group of sample plots pertaining to different levels of DI were enumerated further to study the distributional pattern of the life forms such as tree, shrub, herb, and liana species, along with the endemic species and mean tree BA. The regression fitness curves were plotted to study the agreement between modeled output (i.e., DI) and field conditions (species diversity; Fig. 5.5).

Fig. 5.5
figure 5

Distribution of various life forms such as a tree, b shrub, c herb, d liana species, along with e mean BA and f plant endemism across various disturbance indexes (DI) levels for (I) Arunachal Pradesh, (II) Assam, and (III) Meghalaya; Y-axis represents the number of species. DI-level is plotted from 1 to 10 along X-axis. All the regression curves have shown decreasing trend along increased levels of DI, except the shrubs in Assam state

Full size image

5.3.1 Biodiversity Characterization Package (Bio_CAP)

Bio_CAP is a semi-expert package, developed for uniformal use in the ‘Biodiversity Characterization at landscape level’ project (Roy et al. 2012). The package takes a vegetation type map as the primary input and calculates other landscape parameters as grid data. Bio_CAP was customized using pre-defined codes for different classes of vegetation type map such as for fragmentation estimate, all the class codes from 1 to 150 are automatically considered as forest, and from 151 to 255 picked as non-forest classes. This package was used for estimating the DI that provided 10-levels within forest vegetation patches (Behera et al. 2005). The non-forest areas and unclassified areas such as cloud, shadow, etc., were not considered for landscape analysis and DI estimate (Fig. 5.3).

5.4 Results

5.4.1 Landscape Analysis

The forest type map generated from satellite imagery using a hybrid classification approach offered the requisite input spatial layer for analysis of landscape indices (Roy and Behera 2005). The tropical semi-evergreen forests are distributed in the regions of Arunachal Pradesh and Assam state border areas. However, the forests are exposed to large-scale exploitation and deforestation owing to their easy accessibility with respect to adjacency to the road network and simpler topography. The sub-tropical evergreen forests are found distributed in the entire study site and are the most affected by jhumming, due to their proximity to human settlements. The degraded forests are originated through various activities such as jhumming, landslide, fire, etc. The secondary forests or the abandoned jhum lands are found dominated by varieties of scrubs, herbs, grasses, bamboos, weeds, etc., which receive higher direct sunlight and do faster primary production. The temperate and sub-tropical evergreen forests accommodate thorny bamboos forming an impenetrable canopy.

The higher level of forest fragmentation was observed adjacent to settlement and road networks owing to resource use by the local people (Fig. 5.1b). The patchiness image demonstrated higher homogeneity for the state of Arunachal Pradesh and higher heterogeneity for the state of Assam, whereas Meghalaya showed an intermediate level of patchiness. The medium level of patchiness is an indication of landscape disturbance. Porosity was calculated for both the tropical and sub-tropical forests that are semi-evergreen and evergreen in nature respectively (Fig. 5.1c). Both the forests were found to be relatively porous with very less patches in an intact state. Since any sort of disturbance start affecting the edges, intermediate level of porosity was found across the edge transitions, which could be attributed to degradation and deforestation activities. 80% area was found under different degrees of interspersion in the landscape, pointing to low patch diversity (Fig. 5.1d). This indicated that the dispersal ability of the central class is low or reduced. The juxtaposition image using the adjacency criteria of central pixel with the neighboring pixels was regrouped under five categories (Behera et al. 2005). The juxtaposition image demonstrated greater adjacency or neighborhood effects between the tropical semi-evergreen and sub-tropical evergreen forests owing to their nature of occurrence (Fig. 5.1e). Therefore, juxtaposition is found as a critical indicator revealing the patch interactions in a landscape.

5.4.2 Disturbance Index (DI) Model

The DI image reflects various levels of disturbance prevailing in the states (Fig. 5.3). Overall, the DI image demonstrates that the forests adjoining to human settlements and road networks are highly disturbed in the northeaster landscape (Fig. 5.3). The non-forest classes, snow, and cloud were not considered for estimating disturbance as they are relatively inert from anthropogenic point of view. The highest level of disturbance was found adjoining the agricultural area and natural vegetation. The forests with low level of interspersion and higher juxtaposition demonstrated the lowest level of disturbance on the DI image owing to their remoteness from habitations.

Many of the reserve forests are found having very low DI value that owes to the protection effort by the state forest departments. In Arunachal Pradesh, maximum area was found under low levels DI since most of the forest fragmentation, patchiness, porosity, and interspersion values were low too. The DI image of Assam shows highly disturbed patches in North Cachar district (Fig. 5.3). The evergreen and semi-evergreen forest belts of Karbi Anglong and North Cachar hill districts are found the worst affected. 20.87% area was disturbance-free and the man-made landscapes comprising of 52.11% area was not considered for DI analysis in Assam state. The disturbance in Meghalaya state was the highest in Garo hills followed by Khasi and Jaintia hills. It was observed that the level of disturbance was high in the areas accessible to human beings. The inaccessible areas with complex topography demonstrated lower DI values for the state of Meghalaya.

The percentage area for different levels of DI was plotted in Fig. 5.4. 36.6% forest area of the state of Arunachal Pradesh demonstrated medium (level-6) DI (Fig. 5.4a). In Assam, 40.73% forest area demonstrated intermediate (level-4) level of disturbance (Fig. 5.5b). Similarly, the state of Meghalaya accommodates 38.06% forest area under medium level of DI (level-6). These figures clearly indicate that more of the forest areas in Arunachal Pradesh and Meghalaya states suffer from low degree of disturbance (Fig. 5.4). However, for Assam, more of the forest areas experience high degree of disturbance.

5.4.3 Validation of DI Model

All the 405, 324, and 133 plots were splitted into 10-levels corresponding to their locations on DI image for Arunachal Pradesh, Assam, and Meghalaya respectively (Table 5.2). The species richness demonstrated S-shaped pattern across the DI levels and R2 value decreases gradually from trees to lianas (i.e., 0.55, 0.48, 0.37, and 0.33, whereas the mean tree BA resulted average R2 value of 0.41 (Fig. 5.5). In Assam state, the trees, herbs, lianas, and mean tree BA showed an overall decrease in species richness; whereas the shrubs demonstrated a reverse trend across DI levels (R2 = 0.06; Fig. 5.5). The deviation in case of shrubs could be explained by new formations of weed species at intermediate levels of disturbances. Herbs demonstrated a very good fit with R2 = 0.73, followed by lianas with (R2 = 0.6; Fig. 5.5).

Table 5.2 Distribution of life forms along various levels of DI (increases from 1 to 10) for Arunachal Pradesh, Assam, and Meghalaya
Full size table

The state of Meghalaya has shown a significant decrease in species richness for all life forms with high degree of correlation (i.e., trees, R2 = 0.80; shrubs, R2 = 0.77; herbs, R2 = 0.84 and lianas, R2 = 0.73). Plant endemism demonstrated a decreasing pattern along DI, with the highest regression fitness for the state of Meghalaya, followed by Arunachal Pradesh and Assam states (Fig. 5.5f). The mean tree BA demonstrated an S-shaped curve with maxima at intermediate level of DI.

5.5 Discussion and Conclusions

Satellite images provided forest vegetation cover map for the states with reasonably good accuracy for generating the disturbance index (DI). The distribution of various forest types across elevation was first attempted by Kaul and Haridasan (1987) in the state of Arunachal Pradesh. Of the 16 major forest types classified by Champion and Seth (1968), 13 types and 54-ecologically stable formations were observed in northeastern India (Roy et al. 2012). Landscape level patch analysis very well revealed the ecological patterns and processes and thereby is important in determining diversity and distribution of species assemblages. The increase in the level of forest fragmentation leads to reduction of patch size and the edge effect. The forest fragmentation, patchiness, and porosity images uniformly demonstrated lower values for dense and intact forests that are mostly located on difficult topography, remotely from the settlements and road networks. 41% area reported under intact or low levels of interspersion, well implies medium level of interaction among and between the patches. Juxtaposition, which is based on the relative adjacency of patches, demonstrates low DI values for intact patches in the landscape. The DI image reflects high disturbance level for the areas having more biotic access and vice versa, indicating that the DI has efficiently incorporated the spatial phenomena. The juxtaposition image has relevance in wildlife studies as some fauna just needs one habitat type in contrast to many by other groups. Therefore, patches with low forest fragmentation and high juxtaposition may be given prior attention for management.

Disturbance regimes could vary across a given landscape as a function of biotic interactions, proximity to settlement and transport networks, and complex topography (Turner 1989). The landscape level disturbance offers fundamental inputs to the understanding of changes in various ecological structures and functions (Pickett and White 1985). It was found that the pattern of disturbance in the landscape has influenced the species diversity of the region (Pickett and Thompson 1978). These disturbance activities such as jhumming and deforestation have changed the disturbance regimes and altered the ecosystem processes through habitat loss and forest fragmentation in northeastern India.

The deforestation level was found higher in low elevation regions of Arunachal Pradesh that have a direct relationship with settlement and transport networks (Khan et al. 1997). Shukla and Rao (1993) have attributed to jhum cultivation and industrial activities as the dominant proximate factors for landscape disturbance in Arunachal Pradesh. Assam is located in the Brahmaputra valley that is dominated with permanent cultivation, while the border regions with Arunachal Pradesh has lot of tea gardens owing to sloppy terrain. The shifting cultivation practices primarily attribute to biodiversity decline in Meghalaya (Roy and Tomar 2000). In spatial context, any pixel in DI model is the resultant function of fragmentation, patchiness, porosity, interspersion, juxtaposition, and road and settlement buffer (Roy and Tomar 2000). Few studies were done in India that establish relationships between the landscape level disturbance and the biological richness (Pandey and Shukla 1999), which could be the next step forward.

In general, the regression fitness curves demonstrated clear decreasing trend among the field-derived species richness, mean tree BA, and endemism with DI (Fig. 5.5). The tree species richness pattern across DI levels decreased prominently from the state of Meghalaya with greater degree of regression fitness (R2 = 0.80), followed by Arunachal Pradesh (R2 = 0.55) and Assam (R2 = 0.31). Shrub richness pattern along DI-levels also showed greater agreement in Meghalaya (R2 = 0.77), followed by Arunachal Pradesh (R2 = 0.48) and a reverse trend was seen for the state of Assam (R2 = 0.06). Greater degree of agreement was observed in Meghalaya for herbs and lianas, followed by Assam and Arunachal Pradesh (Fig. 5.5). The endemic subsets of plant species demonstrated the highest fit (R2 = 0.84) for Meghalaya followed by Arunachal Pradesh and Assam (Fig. 5.5). In contrast, Arunachal Pradesh has demonstrated good agreement of mean tree BA with high R2 value (0.42), followed by Assam (R2 = 0.21) and Meghalaya (R2 = 0.02; Fig. 5.5). Each field sample plot represents different spatial and ecological properties that are reflections of fine scale local variations. The regression fitness curves well demonstrate the relationship between various landscape structural indices and their species diversity content. Such regression fitness gives the first-hand assurance that intact patches with lower DI values hold higher species richness in a landscape.

The states of Arunachal Pradesh and Meghalaya accommodate higher of species endemism, while the state of Assam has more agricultural lands (Table 5.1). Almost all of the forests are found in difficult terrain, and therefore Assam state has less forest land than Arunachal Pradesh and Meghalaya states.

Therefore, utilization of field data forms an absolute basis to validate DI model. Here, the statistical tests depend very much on the types of data used; and therefore are simple and preferred for first-hand correlation studies. The models of ecological and biological systems facilitate understanding the possible consequences of human action, that might not be possible otherwise. Similarly, the Ganga River basin that accommodates 40% of India’s population is experiencing expansions at the periphery of agricultural and forest lands due to a significant increase in built-up area (Patidar and Keshari 2020).

The rate of plant endemism declined from tropics to alpine region in northeastern India owing to greater variation in microclimates with complex terrain and remoteness from biotic interferences (Heywood and Watson 1995). These findings partially agree with Myers’s (1988) reporting that tropics could harbor higher number of endemic species than non-tropics. Behera and Kushwaha (2007) reported a decrease in species richness and an increase in species endemism across the elevation gradients in Arunachal Pradesh. Further, Behera et al. (2002) reporter higher endemism in tropics than other regions that may be explained with higher plant functional activities. This study clearly indicated that species endemism decreases along the disturbance regimes in the eastern Himalaya. The potential of satellite remote sensing along with the kindred technologies such as GIS and GPS, for landscape level patch characterization, is well demonstrated in the study.