Chapter 10

Forest Health Monitoring using

Hyperspectral Remote Sensing
Techniques

Narayan Kayet

Abstract Hyperspectral Remote sensing is a handy tool for forest health moni-
toring. This study focuses on forest health monitoring using hyperspectral satellite
data and validates it with tree spectral data. In the study area, increasing mining and
anthropogenic activities within and near forest lands have caused threats to forest
health. All these necessitate assessing the forest health in the areas surrounding mines.
We have used two methods for the forest health assessment: one is VIs (vegetation
indices) based model, and another is tree spectral analysis. The supervised classi-
fication (SAM) method was used for forest health classification based on spectral
data. The results showed that a healthy forest portion was located in the hilly side of
the study area while an unhealthy segment was situated alongside the mines. Hype-
rion data-based VIs model shows better accuracy than spectral based other methods.
Also, it was found that the hyperspectral data based forest health classification gave
a higher accuracy than multispectral data. Finally, forest health results were justified
by ground tree spectral data. This work provides an effective guideline for forest
planning and management.

Keywords Forest health · Hyperspectral data · Vegetation indices · GIS · Mining

area

10.1 Introduction

Hyperspectral remote sensing is a very useful tool for forest health monitoring.
Hyperspectral data are spectrally overdetermined, which means that they provide
spectral information to identify and distinguish spectrally unique materials. Also
this data has the potential for more accurate information extraction than possible
from multispectral data (Apostolescu et al. 2016; Navinkumar and Parmar 2016).
The hyperspectral sensor provides hundreds of narrow spectral bands of the Earth’s

N. Kayet (B)
Department of Mining Engineering, Indian Institute of Technology, Kharagpur, India
e-mail: narayankayet@[Link]

© The Editor(s) (if applicable) and The Author(s), under exclusive license 239
to Springer Nature Switzerland AG 2021
P. K. Shit et al. (eds.), Spatial Modeling in Forest Resources Management,
Environmental Science and Engineering,
[Link]
240 N. Kayet

surface features. NASA EO-1 satellite, equipped with the Hyperion spectrometer,
which has a sampling distance of 10 nm within the 7.7 km swath path, provides 242
spectral bands within the range of 400–2500 nm wavelengths of the EM spectrum
(Hyperion user guide). This sensor easily detects forest biochemical and biophys-
ical properties (Asner et al. 2015; Thenkabail et al. 2001; Chambers et al. 2007).
Hyperspectral data can detect pest and blight conditions in a forest as well as be
used for assessing areas for timber harvesting. Narrow-banded hyperspectral data
has been used for forest health mapping (Tuominen et al. 2009; Kayet et al. 2019a,
b), to depict the overall healthy and unhealthy portions of a forest. Recent devel-
opments in hyperspectral data acquisition from satellite-borne spectrometers have
opened new areas of research that could bring revolutionary changes in the current
approaches to forest management (Franklin 2001). Some researchers have worked
on vegetation stress from the derivative chlorophyll index and leaf area index (Broge
and Leblanc 2001; Brantley et al. 2011; Darvishzadeh et al. 2011). They have used
the airborne hyperspectral image for this analysis. Detection of vegetation stress by
hyperspectral remote sensing technique is based on the assumption that vegetation
stress factors interfere with photosynthesis or the physical structure of the vegetation
and affect the absorption of light energy and thus alter the reflectance spectrum of the
vegetation (Zarco-Tejada et al. 2009; Jacquemoud et al. 2009; Bellvert et al. 2014).
Airborne hyperspectral data is used for the estimate and distribution of different
species in the forest areas. Hyperspectral data has used to develop the Photochem-
ical Reflectance Index (PRI) for distinguishing the species-wise variations of leaves
(Cho et al. 2008; Mashimbye et al. 2012; Darvishzadeh et al. 2008). The hyperspec-
tral narrow banded data-based NDVI (normalized difference vegetation index) and
LAI (leaf area index) has been used for plant health detection by Zarco-Tejada et al.
(2005). They had used red, red edge, and NIR bands for this analysis. The Interna-
tional Institute for Geo-Information Science and Earth Observation (ITC) has studied
vegetation health and tree species classification using hyperspectral data. They have
used full pixel classification methods for vegetation health and tree species based on
ground tree spectral data (Vauhkonen et al. 2011; Dalponte et al. 2014). The deci-
sion tree classifier tool of ENVI will be used to classify the pixels of a Hyperion
image for necessary information acquisition for forest management purposes (vege-
tation indices tutorial ENVI). Ma et al. (2017) used airborne Hyperspectral data to
develop photochemical reflectance index (PRI) for distinguishing the species wise
variations of leafs. A number of researchers have assessed the vegetation stress from
derivative chlorophyll index and leaf area index estimation, using compact airborne
spectrographic image (Wu et al. 2010; Zarco-Tejada et al. 2002; Lee et al. 2004). The
detection of vegetation stress by hyperspectral remote sensing techniques is based
on the assumption that vegetation stress factors interfere with photosynthesis or the
physical structure of the vegetation and affect the absorption of light energy and thus
alter the reflectance spectrum of vegetation (Zarco-Tejada et al. 2009; Calderon et al.
2013; Li and Guo 2016).
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 241

The study area is located in the in an region that has many mines. Mining fields
are under high-stress conditions showing signs of dry and dying plant species. In the
study area, increasing mining and anthropogenic activities within and near forests
pose threats to forest health. All these necessitate monitoring of the forest health in
surrounding mining areas.

10.2 Materials and Methods

10.2.1 Study Area

The present study has been done in the Saranda Forest and its surrounding areas,
which are located in the West Singhbhum district of the Indian state of Jharkhand
(Fig. 10.1). It is famous for having Asia’s largest Sal forests and is an important
elephant habitat. Over the last few decades, in this region, many iron ore mining
towns have emerged, e.g. Gua, Chiria, Megataburu and Kiriburu. The Saranda forest
of Jharkhand is endowed with some rich iron ore deposits. The location of the
forest is within latitude 22° 3 7.98 –22°14 0.67 N and longitude 85° 21 31.52 –
85° 25 53.18 E with elevation of 850 m above the mean sea level (MSL). Saranda
forest is fed by two major rivers, Karo and Koina. The catchment of these rivers
comprises of a drainage system with stream order of up to six (Kayet et al. 2016).

Fig. 10.1 Location map of the study area

242 N. Kayet

Fig. 10.2 Trees spectral data collected from field

10.2.2 Data Source

Hyperspectral data (Hyperion) were downloaded from the USGS website and used
for forest health monitoring. Healthy and unhealthy tree spectra data were collected
by a field spectroradiometer instrument. Also, healthy and unhealthy forest locations
(latitude and longitude) were recorded by GPS. Photographs are taken for validation
purposes during the field survey (Fig. 10.2).

10.2.3 Data Pre-processing

Bad bands removal: The delivered USGS Hyperion product level LIR has 242 bands,
of which 44 were not calibrated. The main reason for not calibrating the entire
band was the decreased sensitivity of the detectors within the non-calibrated spectral
regions. Out of the total collected 242 Hyperion bands, 44 (Table 10.1) bands do not
work (Han et al. 2002).
De-stripping: Hyperion L1R data shows a severe striping effect by imprecise
co-calibration of individual detectors on the focal plane array. The first 12 visible
near-infrared bands and many short waves infrared bands are affected by striping and
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 243

Table: 10.1 List of unused bands of the hyperion sensor

Bands Description
1–7 Not illuminated
58–78 Overlap region
120–132 Water vapour absorption Band
165–182 Water vapour absorption band
185–187 Identified by hyperion bad band list
221–224 Water vapour absorption Band
225–242 Not illuminated

bad columns. An uncorrected striping effect will lead to a faulty interpretation of the
results. The vertical stripes error values were replaced by the average DN values of the
adjacent columns. Hyperspectral data are affected by different noise sources which
can be grouped into two main classes: random noise and fixed pattern noise. The
photon and thermal noise are random noise; striping noise is a fixed pattern noise and
created from push-broom sensors. Hyperspectral images are affected by those noises.
Geometric correction: Geometric distortion often has to be removed from remotely
sensed data. There are two main approaches to remove the geometric errors. One is
the systematic approach and the other is the non-systematic approach. Some of these
errors can be corrected by using the ephemeris of the platform and previously known
internal sensor distortion characteristics. Other errors can be rectified by matching
image coordinates of physical features recorded by the image to the geographic
coordinates of the same feature collected from a map or by using a Global Positioning
System (GPS).
Radiometric correction: Cross-track illumination- ENVI cross-track illumination
tool was used to remove variation of illumination of the image. Cross-track illu-
mination errors may be due to vegetating effects, instrument effects or scanning or
other non-uniform illumination effects. The EFFORT algorithm calculates the mean
values of an extended track polynomial function, and fit mean values remove this
error. We have used this algorithm to remove the variation of illumination of the
image.
Atmospheric correction (FLAASH): Atmospheric correction reduces the effects
of the atmospheric components (water vapor, dust, gasses) on the electromagnetic
radiation reflected or emitted from the surface. We have used the FLAASH model
(Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes) in ENVI soft-
ware for atmospheric correction and convert to surface reflectance. This model was
developed by Spectral Sciences, Inc., under the sponsorship of the US Air force
Research Lab. The calibration model based on the theory of atmospheric radia-
tion is according to the physical process of radiation transfer, building by radiation
transfer equation, using a theoretical formula (Cooley et al., 2002) to proceed with
atmospheric correction (Eq 1).
244 N. Kayet
   
Ap Bpe
L= + + La
1 − peS 1 − pes
 
(A + B) pe
Le = + La
1 − peS

FLAASH starts from a standard equation for spectral radiance at a sensor pixel, L;
that applies to the solar wavelength range and flat, Lambert a material or their equiv-
alents. FLAASH atmospheric correction removes this absorption feature. FLAASH
MODTRAN has outperformed other radiative transfer codes especially in the water
region 940 and 1130 nm and CO2 at 2055 nm (Pathak et al. 2016).

10.2.4 Methodology

[Link] Vegetation Indices (Vis)-Based Forest Health Mapping

The forest health analysis tool will generate a spatial map that shows the overall
health and vigor of a forested area (Tuominen et al. 2008). It is good at detecting pest
and blight conditions in a forest. The forest health tool uses the following vegetation
index categories (ENVI Forest heath tutorial):
1. Narrow band Greenness, to show the distribution of green.
2. Leaf pigments, to show the concentration of arytenoids and anthocyanin pigments
for stress levels.
3. Canopy water content, to show the concentration of water.
4. Light use efficiency, to show forest growth rate.

Greenness indices: Greenness vegetation indices generally measure the vigor and
green vegetation (Kumar et al. 2015). They measures the various aspects such as
chlorophyll concentration, canopy area, and canopy structure. Greenness vegeta-
tion indices are based on measuring the reflectance peak in the NIR region. Red
wavelengths where the chlorophyll absorption is strongest are used as a reference
(Lloret et al. 2004). Leaf pigments indices: Leaf pigment vegetation indices measure
the amount of stress-related pigment in the vegetation (Jenkins et al. 2007). In
stressed vegetation, there is a higher concentration of carotenoids and anthocyanins.
Carotenoids are the leafs pigment that prevents vegetation light conditions. Antho-
cyanin pigment contents are high in new leaves (Gamon et al. 1999). Canopy water
content indices: Water content vegetation indices are designed to estimate the canopy
water content (Colombo et al. 2008). However, water content is an important vege-
tation property that controls vegetation growth and also correlates with vegetation
health (Adam et al. 2010). The use of water content vegetation indices needs high
spectral resolution data.
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 245

Table 10.2 Narrow banded VIs for forest health mapping

Narrow
banded Indices Algorithms References
Indices
Greenness (i) Modified mNDVI705 = Datt (1999),
red edge (750nm 705nm)/(750nm + 705nm (2 ∗ 445nm)) Zarco-Tejada
normalized V R E I 1 = 740nm et al. (2013)
difference 720nm
vegetation
index
(ii) Vogelmann
red edge index
Light use Structure SIPI = (800nm−445nm)/(800nm−680nm) Penuelas
efficiency insensitive et al. (1994)
pigment index
Leaf (i) Carotenoid A R I 1 = 550nm
1
− 700nm
1
Gitelson
pigments reflectance CRI2 = (1/510nm)(1/550nm) et al. (2002)
index
(ii)
Anthocyanin
reflectance
index
Canopy Normalized N D I I = 819nm − 1649nm/819nm − 1649nm Hardisky
water difference et al. (1983)
content infrared index

Light Use Efficiency indices: Light use efficiency vegetation indices provide the
efficiency with which vegetation can use incident light for photosynthesis (Wilson
et al. 1981). It is correlated with carbon uptake efficiency and growth rate. However,
by the use of light use efficiency vegetation indices measure the growth rate and
production of vegetation. We have used six narrow banded Vis for forest health
mapping (Table 10.2).

[Link] Spectral Analysis Based Forest Health Mapping

The following procedures of Hyperspectral analysis were employed, including the

Minimum Noise Fraction (MNF) transformation for reducing the spectral data, the
Pixel Purity Index (PPI) for identifying those extreme or spectrally pure pixels,
and the n-dimensional visualizer for determining the endmember directly from the
image. Spectral Angle Mapper (SAM) was applied to estimate the abundances of
each endmember to produce the final map.
Minimum Noise Fraction (MNF): MNF can reduce the inherent dimensionality
of the dataset and reduce noise from the dataset. MNF also reduces the computa-
tional requirement for subsequent processing. The first step in MNF transforms the
246 N. Kayet

data in which the noise has a unit variance and there is also no band to band inter-
relationship (Denghui and Le 2011). Secondly, the MNF can compute/process the
principal component analysis for noise-whitened data (Gamon et al. 2004). MNF
inversion produces much smaller spectral angles than those derived in transformed
space (Peddle et al. 2008). The first ten inverse MNF bands contain 95% of the total
information.
Pixel purity index (PPI): Image transformation techniques typically use statistical
analysis and reduce the dimensionality of the data. One such transformation is done
through the principal component analysis or principal component transformation
(Chang and Plaza 2006). The pixel purity index (PPI) is a means of finding the
most spectrally pure, or extreme, pixels in multispectral and hyperspectral images
(Chaudhry et al. 2006). The most spectrally pure pixels typically correspond to
mixing endmembers (Plaza et al. 2006).
N-dimensional (N-D) Visualizer: N-D is applied after correcting the image. The
distribution of bands in N space can be used for the estimation of some spectral
endmembers and their pure spectral signature (Kruse et al. 1999). The N-dimensional
visualization is applied after gathering the data through MNF or PPI algorithms. The
pre-clustering result attempts to find in the corner pixel of N-dimensional using a
spectral scatter algorithm.
Since the purest pixel is found in the neighbourhood of the data cloud (Wang et al.
2015). The n-dimensional visualizer allows for the interactive rotation of data in n-D
space, selection of groups into different classes (Boardman et al. 1995).
Spectral classification techniques: Classification and feature extraction methods
have been commonly used for many years for the mapping of forest health and
vegetative cover from hyperspectral datasets. However, conventional classification
algorithms cannot be applied to hyperspectral data due to the high dimensionality of
the data. Spectral Angle Mapper (SAM) mapping techniques were used in the present
study to map of forest health in the study area. Spectral Angle Mapper (SAM) is an
algorithm, which is widely used for hyperspectral image correction (Petropoulos
et al. 2013). It is a supervised image correction process. A pixel with a minimum
spectral angle comparison with reference spectra is assigned to the pixel vector. This
algorithm determines the spectral similarity between two spectra by calculating the
angle between the spectra and treating them as vectors in space with dimensionality
equal to the number of bands (Girouard et al. 2004).

[Link] Accuracy Assessment

Accuracy assessment is an important work in the classification validation system.

Remote sensing technology is a great source of thematic map presentation although
accuracy assessment assists how far the classification represents the real world. Accu-
racy assessment can produce user accuracy, producer accuracy, total accuracy, and
also kappa coefficient value.
The producer accuracy of the classified pixels compared to the ground truth. The
overall research methodology is shown in Fig. 10.3.
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 247

Fig. 10.3 Overall methodology for forest health monitoring

Table 10.3 Minimum, maximum, mean, and standard deviation values of before and after FLAASH
correction
FLAASH correction Min Max Mean Stdev
Before correction −77 3383 84,444.44 66,423.37
After correction 0 70.89 11.63 10.21

10.3 Results and Discussion

10.3.1 Result of FLAASH Atmospheric Correction

FLAASH is an effective atmospheric correction process where all bands are corrected
to follow their proper geometric and radiometric characteristics. FLAASH can also
remove the de-striping of the image, path radiation of the image, and various system-
atic and non-systemic effects. Pre and post FLAASH correction statistics value are
shown in Table 10.3. Spectral variability of after FLAASH correction is shown in
Fig. 10.4.

10.3.2 Vegetation Indices (Vis) Based Forest Health Mapping

Vegetation Indices (VIs) were calculated for 60 test sample pixels. Mean and standard
deviation values for both healthy and unhealthy classes are shown in Table 10.4.
248 N. Kayet

Fig. 10.4 Image spectra after atmospheric correction

Table 10.4 Mean, standard deviation and separability values for each index test
Vegetation indices Healthy Unhealthy Separability
Mean Std. Mean Std. S
MNDVI705 0.66 0.03 0.17 0.08 4.21
VREI1 1.44 1.13 0.06 0.05 2.60
CRI1 33.22 7.92 17.17 2.24 1.57
ARI1 9.65 4.78 7.71 2.29 0.27
NDII −0.49 0.07 −1.23 0.27 2.11
SIPI 1.09 0.02 0.51 0.33 1.60

Generally, the separability values obtained for greenness and vegetation indices
were relatively high. The highest separability values were obtained for the modified
red edge normalized difference vegetation (MNDVI705) index due to good chloro-
phyll content. The MNDVI705 correlates well with good chlorophyll content, so its
good performance could be expected as the result of forest health (Kayet et al. 2019a,
b). ENVI software provides nine forest health classes (Fig. 10.5).
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 249

Fig. 10.5 Forest health map (Class 9 very healthy and class 1 unhealthy)

The classifications are relative to the particular input scene only and cannot be
generalized to other areas or other scenes. The healthy and unhealthy value range is
0.5–0.8 and 0.1–0.3 (Kayet et al. 2019a, b). The classification map rates the scene
from one, representing the least healthy forest (weakest) to the healthiest forest
(strongest) which help to assess relative forest health conditions within the scene
(ENVI forest health tool tutorial).
250 N. Kayet

Fig. 10.6 Hyperion reflectance spectra of forest health classes (1–9)

Forest health classification was done by the use of three narrow-band vegetation
indices which are MNDVI705 (Greenness or chlorophyll vegetation Index), CRI_1
(Leaf pigment Vegetation Index), and SIPI (Light efficiency vegetation Index). The
MNDVI705 index works well with the lower chlorophyll content, so it is accepted
for appreciable forest health result (Kumar et al. 2015). Test of forest health result
for leaf pigments VIs is relatively lower than any other vegetation indices. In leaf
pigment VIs, the value of ARI1 is high than other VIs (Serrano et al., 2002). For light
use efficiency, Vis value is relatively higher than that of leaf pigment VIs (Jenkins
et al. 2007). NDNI is a useful index where there exhibits a high variability in canopy
or leaf pigment structure (Rodriguez et al. 2007). The test result for canopy water
content index was low because of vegetation canopy structure (Sims et al. 2002).
The Hyperion spectral signature of each forest health class is shown in Fig. 10.6.

10.3.3 Spectral Analysis Based Forest Health Mapping

Forest health is mainly dependent on various physiological parameters such as

climate, temperature, geology, soil, slope, aspect, hill shade direction, and much
more. The study area is mainly covered by three types of forest health (healthy,
moderated healthy, and unhealthy). The healthy and unhealthy value range is 0.5–
0.8 and 0.1–0.3 (Kayet et al. 2019a, b). The spectral signature of the forest health
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 251

Fig. 10.7 Field collected healthy and unhealthy trees spectra

classes were collected from the image and these matche the spectral signature from
field collected tree spectral library (Fig. 10.7) and field sample data in ENVI software.
Forest health mapping is done by the SAM classification technique. Classification
of forest health into three classes (healthy, moderated healthy, and unhealthy) is
shown in Fig. 10.8.
Healthy forest (0.5–0.8) cover comprises 49.87% of the study area, 14.15% area is
covered by moderate healthy forest (0.3–0.5), and unhealthy forest (0.1–0.3) covers
38.03% area (Kayet et al. 2019a, b). Healthy forests are present mostly in the north
and the north-east part of the study area and unhealthy is situated in mines surrounding
area. George et al. (2014) shown the better forest classification through Hyperspectral
remote sensing and compared the classification results obtained from Hyperion and
Landsat TM sensors for the study of Western Himalaya and obtained collective
accuracies of 81.52% and 69.62% respectively. Thenkabail et al. (2004) compared
the classification results of different sensors viz., Hyperion, IKONOS, ALI, and ETM
+ sensors for the study of African rainforests and obtained collective accuracies of
93.2%, 87.46%, 81.53%, and 76.9% respectively.
252 N. Kayet

Fig. 10.8 Forest heath mapping by hyperion data

10.3.4 Accuracy Assessment

Classification accuracy was done by spectral angle mapper (SAM) classification

technique in ENVI software. SAM was implemented to the collection of the spectral
signature of healthy, moderated healthy and unhealthy forests. Forest health classi-
fication is accuracy by the USGS spectral library and sample field points. Thus, the
overall accuracy of 76.53% and 0.71 kappa coefficient were determined (Table 10.5).
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 253

Table 10.5 Accuracy assessment results of SAM based on hyperion data

SVM based on hyperion Healthy Moderated healthy Unhealthy Total UA
Healthy 11 0 1 11 81.52
Moderated healthy 0 13 0 13 74.54
Unhealthy 2 0 10 12 70.66
Total 13 13 11 36
PA 91.83 85.53 80.11
Overall accuracy: 76.53%, kappa statistics: 0.71

10.3.5 Forest Health Validation

The healthy, moderately healthy, and unhealthy components constituting the forest of
the study area were evaluated, both at ground level and pixel- level, having the highest
reflectance data from the NIR wavebands region. The correlation determination (R2 )
and RMS error values were evaluated from ground level and pixel-level spectral data
(Fig. 10.9). A correlation (R2 = 0.84) was observed between the ground level and

Fig. 10.9 Correlation between field reflected spectra and pixel reflected spectra of healthy, moderate
healthy and unhealthy forest class
254 N. Kayet

pixel-level for class healthy, and an RMS error of 3.98 was found. A correlation (R2
= 0.86) was observed between the ground level and pixel-level for class moderately
healthy, and an RMS error of 2.06 was found. And a correlation (R2 = 0.87) was
observed between the ground level and pixel-level for class unhealthy, and an RMS
error of 1.25 was found.

10.4 Conclusion

The article has summarized forest health monitoring sounding mines areas. This case
study demonstrated by hyperspectral data. Hyperspectral data has more capability
than multispectral data. In this study, a good correlation was shown between forest
health and distance from mines. It means that as the mining area increases forest as
well as environment will also get affected. This methodology would be capable of
monitoring various categories of forest region routinely irrespective of the different
climate condition, forest structure, and soil condition. Hyperspectral remote sensing-
based forest health monitoring is today’s need so that the forest department, local self-
government, and mining companies must adopt an adequate policy for reclamation
and restoration of the forest ecosystem affected by mining activities.

References

Adam E, Mutanga O, Rugege D (2010) Multispectral and hyperspectral remote sensing for
identification and mapping of wetland vegetation: a review. Wetlands Ecol Manage 18(3):281–296
Apostolescu N, Baran D (2016) Spectral similarity in hyperspectral image. Acta Electrotehnica 57
Asner GP, Ustin SL, Townsend PA, Martin RE, Chadwick KD (2015) Forest biophysical and
biochemical properties from hyperspectral and LiDAR remote sensing. Land resources moni-
toring, modeling and mapping with remote sensing. CRC Press, Taylor & Francis Group, pp
429–448.
Bellvert J, Zarco-Tejada PJ, Girona J, Fereres EJPA (2014) Mapping crop water stress index in
a ‘Pinot-noir’vineyard: comparing ground measurements with thermal remote sensing imagery
from an unmanned aerial vehicle. Precision Agric 15(4):361–376
Boardman JW (1995) Analysis, understanding, and visualization of hyperspectral data as convex sets
in n space. In: Imaging spectrometry (vol 2480). International Society for Optics and Photonics,
pp 14–22
Brantley ST, Zinnert JC, Young DR (2011) Application of hyperspectral vegetation indices to
detect variations in high leaf area index temperate shrub thicket canopies. Remote Sens Environ
115(2):514–523
Broge NH, Leblanc E (2001) Comparing prediction power and stability of broadband and hyper-
spectral vegetation indices for estimation of green leaf area index and canopy chlorophyll density.
Remote Sens Environ 76(2):156–172
Calderón R, Navas-Cortés JA, Lucena C, Zarco-Tejada PJ (2013) High-resolution airborne hyper-
spectral and thermal imagery for early detection of Verticillium wilt of olive using fluorescence,
temperature and narrow-band spectral indices. Remote Sens Environ 139:231–245
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 255

Chambers JQ, Asner GP, Morton DC, Anderson L, O, Saatchi SS, Espírito-Santo FD, Souza, C Jr.
(2007) Regional ecosystem structure and function: ecological insights from remote sensing of
tropical forests. Trends Ecol Evolut 22(8):414–423
Chang CI, Plaza A (2006) A fast iterative algorithm for implementation of pixel purity index. IEEE
Geosci Remote Sens Lett 3(1):63–67
Chaudhry F, Wu CC, Liu W, Chang CI, Plaza A (2006) Pixel purity index-based algorithms for
endmember extraction from hyperspectral imagery. Recent Adv Hyperspect Signal Image Process
37(2):29–62
Cho MA, Sobhan I, Skidmore AK, De Leeuw J (2008) Discriminating species using hyperspectral
indices at leaf and canopy scales. The International Archives of the Spatial Information Sciences,
pp 369–376
Colombo R, Meroni M, Marchesi A, Busetto L, Rossini M, Giardino C, Panigada C (2008) Esti-
mation of leaf and canopy water content in poplar plantations by means of hyperspectral indices
and inverse modeling. Remote Sens Environ 112(4):1820–1834
Cooley T, Anderson GP, Felde GW, Hoke ML, Ratkowski AJ, Chetwynd JH, Bernstein LS (2002,
June) FLAASH, a MODTRAN4-based atmospheric correction algorithm, its application and
validation. In: IEEE International Geoscience and Remote Sensing Symposium, Vol 3. IEEE, pp
1414–1418
Dalponte M, Ørka HO, Ene LT, Gobakken T, Næsset E (2014) Tree crown delineation and tree
species classification in boreal forests using hyperspectral and ALS data. Remote Sens Environ
140:306–317
Datt B (1999) Remote sensing of water content in Eucalyptus leaves. Aust J Bot 47(6):909–923
Darvishzadeh R, Atzberger C, Skidmore A, Schlerf M (2011) Mapping grassland leaf area index
with airborne hyperspectral imagery: a comparison study of statistical approaches and inversion
of radiative transfer models. ISPRS J Photogram Remote Sens 66(6):894–906
Darvishzadeh R, Skidmore A, Schlerf M, Atzberger C, Corsi F, Cho M (2008) LAI and chlorophyll
estimation for a heterogeneous grassland using hyperspectral measurements. ISPRS J Photogram
Remote Sens 63(4):409–426
Denghui Z, Le Y (2011) Support vector machine based classification for hyperspectral remote
sensing images after minimum noise fraction rotation transformation. In: 2011 International
conference on internet computing and information services. IEEE, pp 132–135
Franklin SE (2001) Remote sensing for sustainable forest management. CRC press
Gamon JA, Surfus JS (1999) Assessing leaf pigment content and activity with a reflectometer. The
New Phytol 143(1):105–117
Gamon JA, Huemmrich KF, Peddle DR, Chen J, Fuentes D, Hall FG, Miller JR (2004) Remote
sensing in BOREAS: Lessons learned. Remote Sens Environ 89(2):139–162
George R, Padalia H, Kushwaha SPS (2014) Forest tree species discrimination in western Himalaya
using EO-1 Hyperion. Int J Appl Earth Obs Geoinf 28:140–149
Girouard G, Bannari A, El Harti A, Desrochers A (2004) Validated spectral angle mapper algorithm
for geological mapping: comparative study between QuickBird and landsat-TM. In: XXth ISPRS
congress, geo-imagery bridging continents, Istanbul, Turkey, pp 12–23
Gitelson AA, Stark R, Grits U, Rundquist D, Kaufman Y, Derry D (2002) Vegetation and soil lines
in visible spectral space: a concept and technique for remote estimation of vegetation fraction.
Int J Remote Sens 23(13):2537–2562
Han T, Goodenough DG, Dyk A, Love J (2002) Detection and correction of abnormal pixels in
Hyperion images. In: IEEE international geoscience and remote sensing symposium, vol 3. IEEE,
pp 1327–1330
Jacquemoud S, Verhoef W, Baret F, Bacour C, Zarco-Tejada PJ, Asner GP, Ustin SL (2009)
PROSPECT+ SAIL models: a review of use for vegetation characterization. Remote Sens Environ
113:S56–S66
Jenkins JP, Richardson AD, Braswell BH, Ollinger SV, Hollinger DY, Smith ML (2007) Refining
light-use efficiency calculations for a deciduous forest canopy using simultaneous tower-based
carbon flux and radiometric measurements. Agric Meteorol 143(1–2):64–79
256 N. Kayet

Kayet N, Pathak K, Chakrabarty A, Sahoo S (2016) Spatial impact of land use/land cover change on
surface temperature distribution in Saranda Forest Jharkhand. Model Earth Syst Environ 2(3):127
Kayet N, Pathak K, Chakrabarty A, Singh CP, Chowdary VM, Kumar S, Sahoo S (2019) Forest
health assessment for geo-environmental planning and management in hilltop mining areas using
Hyperion and Landsat data. Ecol Ind 106:105471
Kayet N, Pathak K, Chakrabarty A, Kumar S, Chowdary VM, Singh CP, Basumatary S (2019)
Assessment of foliar dust using hyperion and landsat satellite imagery for mine environmental
monitoring in an open cast iron ore mining areas. J Clean Product 218:993–1006
Ke LIU, Qing-bo ZHOU, Wen-bin WU, Tian XIA, Hua-jun TANG (2016) Estimating the crop leaf
area index using hyperspectral remote sensing. J integrative agriculture 15(2):475–491
Kruse FA, Boardman JW, Huntington JF (1999) Fifteen years of hyperspectral data: Northern
grapevine mountains, Nevada. In: Proceedings of the 8th JPL Airborne earth science workshop:
Jet Propulsion Laboratory Publication, JPL Publication, pp 99–117
Kumar K, Singh NM, Kerr JM (2015) Decentralisation and democratic forest reforms in India:
moving to a rights-based approach. Forest Policy Econ 51:1–8
Lee KS, Cohen WB, Kennedy RE, Maiersperger TK, Gower ST (2004) Hyperspectral versus
multispectral data for estimating leaf area index in four different biomes. Remote Sens Environ
91(3–4):508–520
Li Z, Guo X (2016) Remote sensing of terrestrial non-photosynthetic vegetation using hyperspectral,
multispectral, SAR, and LiDAR data. Prog Phys Geogr 40(2):276–304
Lloret F, Peñuelas J, Ogaya R (2004) Establishment of co-existing Mediterranean tree species under
a varying soil moisture regime. J Veg Sci 15(2):237–244
Ma J, Sun DW, Qu JH, Pu H (2017) Prediction of textural changes in grass carp fillets as affected
by vacuum freeze drying using hyperspectral imaging based on integrated group wavelengths.
LWT-Food Sci Technol 82:377–385
Mashimbye ZE, Cho MA, Nell JP, De Clercq WP, Van Niekerk A, Turner DP (2012) Model-based
integrated methods for quantitative estimation of soil salinity from hyperspectral remote sensing
data: a case study of selected South African soils. Pedosphere 22(5):640–649
Navinkumar TU, Parmar SK (2016) Unmixing of hyperpectral image using nonlinear parametric
model. In: 2016 IEEE international conference on recent trends in electronics, information &
communication technology (RTEICT), IEEE, pp 1206–1211
Pathak VN (2016) Development of an atmospheric correction method for retrieval of surface
reflectance from satellite data. Doctoral dissertation, Sardar Patel University
Peddle DR, Boulton RB, Pilger N, Bergeron M, Hollinger A (2008) Hyperspectral detection of
chemical vegetation stress: evaluation for the Canadian HERO satellite mission. Can J Remote
Sens 34(sup1):S198–S216
Petropoulos GP, Vadrevu KP, Kalaitzidis C (2013) Spectral angle mapper and object-based classifi-
cation combined with hyperspectral remote sensing imagery for obtaining land use/cover mapping
in a Mediterranean region. Geocarto Int 28(2):114–129
Plaza A, Valencia D, Plaza J, Chang CI (2006) Parallel implementation of endmember extraction
algorithms from hyperspectral data. IEEE Geosci Remote Sens Lett 3(3):334–338
Rodríguez-Pérez JR, Riaño D, Carlisle E, Ustin S, Smart DR (2007) Evaluation of hyperspectral
reflectance indexes to detect grapevine water status in vineyards. Am J Enology and Viticulture
58(3):302–317
Serrano L, Penuelas J, Ustin SL (2002) Remote sensing of nitrogen and lignin in Mediterranean
vegetation from AVIRIS data: decomposing biochemical from structural signals. Remote Sens
Environ 81(2-3):355–364
Sims DA, Gamon JA (2002) Relationships between leaf pigment content and spectral reflectance
across a wide range of species, leaf structures and developmental stages. Remote Sens Environ
81(2-3):337–354
Thenkabail PS (2001) Optimal hyperspectral narrow bands for discriminating agricultural crops.
Remote Sens Rev 20(4):257–291
10 Forest Health Monitoring using Hyperspectral Remote Sensing … 257

Thenkabail PS, Enclona EA, Ashton MS, Van Der Meer B (2004) Accuracy assessments of
hyperspectral waveband performance for vegetation analysis applications. Remote Sens Environ
91(3-4):354–376
Tuominen J, Lipping T, Kuosmanen V (2008) Assesment of ENVI forest health tool in detec-
tion of dust and seepage contaminated forest areas. In: IGARSS 2008–2008 IEEE international
geoscience and remote sensing symposium, Vol 3. IEEE, pp. III–1358
Tuominen J, Lipping T, Kuosmanen V, Haapanen R (2009) Remote sensing of forest health. In: Ho
P-GP (ed) Geoscience and remote sensing. InTech, Shanghai, China, pp 29–52
Vauhkonen J, Mehtätalo L, Packalén P (2011) Combining tree height samples produced by airborne
laser scanning and stand management records to estimate plot volume in Eucalyptus plantations.
Can J Res 41(8):1649–1658
Wang W, Lawrence KC, Ni X, Yoon SC, Heitschmidt GW, Feldner P (2015) Near-infrared
hyperspectral imaging for detecting Aflatoxin B1 of maize kernels. Food Control 51:347–355
Wilson JW (1981) Analysis of growth, photosynthesis and light interception for single plants and
stands. Ann Bot 48(4):507–512
Zarco-Tejada PJ, Berni JA, Suárez L, Sepulcre-Cantó G, Morales F, Miller JR (2009) Imaging
chlorophyll fluorescence with an airborne narrow-band multispectral camera for vegetation stress
detection. Remote Sens Environ 113(6):1262–1275
Zarco-Tejada PJ, Guillén-Climent ML, Hernández-Clemente R, Catalina A, González MR, Martín
P (2013) Estimating leaf carotenoid content in vineyards using high resolution hyperspectral
imagery acquired from an unmanned aerial vehicle (UAV). Agric for Meteorol 171:281–294
Zarco-Tejada PJ, Miller JR, Mohammed GH, Noland TL, Sampson PH (2002) Vegetation stress
detection through chlorophyll a+ b estimation and fluorescence effects on hyperspectral imagery.
J Environ Qual 31(5):1433–1441
Zarco-Tejada PJ, Ustin SL, Whiting ML (2005) Temporal and spatial relationships between within-
field yield variability in cotton and high-spatial hyperspectral remote sensing imagery. Agron J
97(3):641–653
Chapter 11
Estimating Above Ground Biomass
(AGB) and Tree Density using Sentinel-1
Data

Sambhunath Roy, Sujoy Mudi, Pulakesh Das, Sujit Ghosh,

Pravat Kumar Shit, Gouri Shankar Bhunia, and John Kim

Abstract Assessment of the forest above ground biomass (AGB) and tree density
is essential for various studies related to forest structure, productivity, carbon cycle,
atmospheric processes, climate change etc. including forest cover management activ-
ities and framing the conservation policies. The freely available C-band Sentinel-1
microwave data allows to estimate forest cover biomass at high spatial resolution;
moreover, the Sentinel-2 optical data enables to integrate the biophysical attributes.
In the current study, the AGB has been estimated in a Shorea robusta (sal) dominated
forest cover in sub-tropical region employing the Sentinel-1 microwave data. The
inventory of forest cover attributes has been collected in 40 sample plots, where the
in-situ AGB was geo-statistically linked with the satellite observations. Employing
the univariate linear regression, it has been observed that the microwave backscatter
obtained in the VH band well explained (R2 = 0.63) the variability of the in-situ
AGB in comparison to the backscatter received in VV band (R2 = 0.44) and optical
data derived EVI image (R2 = 0.45). The predicted biomass map verified with the
test data points indicated an accuracy of R2 = 0.45, with low RMSE (±17 tonnes/ha)
and slight underestimation (bias = −0.024). However, the accuracy in tree density
estimation obtained from the AGB map was observed much higher (R2 = 0.87). The
field observed AGB varied between 88.56 tonnes/ha and 170.29 tonnes/ha, where the

S. Roy (B) · P. K. Shit

PG Department of Geography, Raja N. L. Khan Women’s College (Autonomous), Gope Palace,
Midnapore 721102, West Bengal, India
e-mail: sambhuroynath95@[Link]
S. Mudi · P. Das
Department of Remote Sensing & GIS, Vidyasagar University, Midnapore, India
S. Ghosh
Centre for Oceans, Rivers, Atmosphere and Land Science (CORAL), IIT Kharagpur, Kharagpur,
India
G. S. Bhunia
Seacom Silk University, Bolpur, India
J. Kim
USDA Forest Service, Pacific Northwest Research Station, Corvallis, USA

© The Editor(s) (if applicable) and The Author(s), under exclusive license 259
to Springer Nature Switzerland AG 2021
P. K. Shit et al. (eds.), Spatial Modeling in Forest Resources Management,
Environmental Science and Engineering,
[Link]
260 S. Roy et al.

satellite data derived AGB estimated the range of 44.1 tonnes/ha and 249 tonnes/ha
for the entire study area. Majority of the biomass was estimated in the range of 100–
200 tonnes/ha, which was contributed by majority of the tree density region varying
between 69/ha and 75/ha. However, few patches are observed to have much higher
and lower AGB, which could be indicating the highly dense and less dense forest
cover regions in the study area, respectively. The uniform AGB map indicates the
selected region to be more homogenous forest cover area.

Keywords Above ground biomass (AGB) · Sentinel-1 · Sentinel-2 · Tree density ·

Regression

11.1 Introduction

Forest ecosystem is one of the most important carbon sinks of the terrestrial
ecosystem and plays a very important role in the global carbon cycle by sequestering
a substantial amount of CO2 from the atmosphere (Houghton et al. 2012). Defor-
estation or natural death of plants followed by burning, decomposition or natural
decaying leads release of the stored carbon back into the terrestrial ecosystem. The
natural cycling of the carbon is preserved and controlled by a dynamic equilib-
rium between the biological and inorganic processes since the geological history of
the earth. In the 19th century, with the advent of industrial revolution, humans are
burning a huge amount of fossil fuels, and releasing the carbon stored in it back
into the atmosphere. Since the industrial insurgency, the CO2 level in the atmo-
sphere has increased tremendously causing remarkable changes in the global carbon
cycle. The CO2 concentration in the atmosphere is continuously being measured
and recorded accurately since 1957 (Keelinget al. 1976), indicates increasing trend
from about 280 ppm during the pre-industrial period to approximately 390 ppm
in the present day (Hofmann et al. 2009). Besides, the combustion of fossil fuels,
other human activities such as land use change and deforestation have a considerable
impact on the ability of the terrestrial biosphere to remove CO2 from the atmosphere
(Houghton 2012). About 60% of the observed global climate change is attributed to
this increasing CO2 concentration in the atmosphere. The main carbon pool in forest
ecosystems consists of the living biomass of trees, under story vegetation, dead mass
of litter, woody debris and soil organic matter. The tropical forest cover encompasses
majority of the earth’s terrestrial biodiversity and account for a large proportion of
terrestrial carbon stored. The world’s forest and forest soils currently store more than
one trillion tonnes of carbon, which is twice the amount found floating free in the
atmosphere. According to Global Forest Resource Assessment Report 2010 (FRA
2010), the total forest carbon stock of the world is 652 Giga tonnes (161.8 t/ha). Out
of this, the forest biomass contains 289 Giga tonnes (71.6 t/ha), the dead organic
matter contains 72 Giga tonnes (17.8 t/ha), and forest soil organic carbon contains
293 Giga tonnes (72.3 t/ha) of carbon (Fig. 11.1).
11 Estimating Above Ground Biomass (AGB) and Tree Density … 261

Fig. 11.1 Carbon pools (CIFOR, Forest Carbon Toll box 2008)

In forest ecosystem, the stored carbon is classified into five pools (Table 11.1).
The living portion of biomass carbon is classified in two pools: the ‘above ground
biomass’ (AGB) and ‘below ground biomass’ (BGB), which are stores of significant
amount of carbon. The ‘dead organic matter’ (DOM) is also classified into two
pools: ‘dead wood’ and ‘litter’. The fifth pool is ‘soil organic matter’ (SOM), which

Table 11.1 Carbon pool classification according to IPCC (2003)

Pools Description
Living biomass Above ground biomass (AGB) All living biomass above the soil
including stem, stump, branches, bark,
seeds and foliage
Below ground biomass (BGB) All living biomass of live roots
Dead organic matter Dead wood Includes all non-living woody biomass
not contained in the litter either
standing or lying on the ground
Litter Includes all non-living biomass with a
diameter less than a minimum diameter
chosen by the country (for FSI 5 cm),
lying dead, in various states of
decompositions above the mineral or
organic soil
Soil Soil organic matter (SOM) Includes organic carbon in mineral and
organic soils (including peat) to a
specific depth chosen by the country
(for FSI 30 cm)
262 S. Roy et al.

contains substantial amount of organic carbon. Among the five components, the
AGB accounts the largest carbon stored in tropical region and it is directly affected
by deforestation and degradation (Kumar and Sharma 2015). The change in the forest
areas and corresponding changes in forest biomass due to management and regrowth
greatly influence the transfer of carbon between the terrestrial forest ecosystem and
the atmosphere. Assessment of the AGB helps to quantify the carbon stocks, which
in turn enables to analyse the current status and project the near future changes.

11.2 Methods for Estimating Above Ground Biomass

(AGB)

11.2.1 Field Measurement Methods

Primarily two field measurement methods are used—(i) destructive method or

harvesting method—direct estimation of AGB and the carbon stocks stored in the
trees. Although, this method determines the biomass accurately, limited to a small tree
sample sizes being time consuming, strenuous, destructive and expensive. Usually,
this method is used for developing biomass equation to be applied for assessing
biomass on a larger area. The other method is (ii) non-destructive measurements,
involves climbing the tree to measure the various parts or by simply measuring the
DBH, tree height, volume of the tree and wood density to calculate the biomass using
allometric equations (Sullivan et al. 2018; Lucas et al. 2015). The allometric equa-
tions are developed and applied to forest inventory data to assess the biomass and
carbon stocks of forests by establishing a relationship between the various physical
parameters of the tree such as the DBH, tree height, wood density, tree species, etc.
(Vargas-Larreta et al. 2017). Haripriya (2000) used forest inventories to compute the
species-wise forest biomass and carbon stock for various forest strata in different
states. Numbers of allometric equation have been developed for estimating AGB
for different species and in different climate zones, where the relationship between
AGB and forest structural information have been statistically linked. The allometric
equation developed by Brown et al. (1989), Chamber et al. (2001) and Chave et al.
(2005) are given as follows:
 
AG B = exp −0.370 + 0.333 ln D + 0.9333 ln D 2 − 0.122 ln D 3 (Brown et al. 1989)

AG B = 13.2579 − (4.8945D) + 0.6713D 2 (Chamber et al. 2001)

 0.940
AG B = 0.0776 ρ D 2 H (Chave et al. 2005)

Chave et al. (2014) revised the parameters for the pantropical region employing
significant numbers of ground data. Nath et al. (2019) tested four models for biomass
estimation for the forest cover of North East India. Baishya et al. (2009) estimated
AGB in the humid tropical forests in North-East India employing the ground-based
11 Estimating Above Ground Biomass (AGB) and Tree Density … 263

forest inventory technique for the natural semi-evergreen and Shorea robusta (sal)
plantation forest; and observed higher AGB in plantation regions (406.4 tonnes/ha)
compared to natural forest (323.9 tonnes/ha). While assessing the AGB in seven
districts of Madhya Pradesh state, India, Salunkhe et al. (2016) utilised the species-
wise allometric equations and density provided by FSI and FRI; and observed a lower
range of average AGB as 31.8 tonnes/ha and 20.7 tonnes/ha for the tropical dry and
mixed deciduous forest, respectively.

11.2.2 Remote Sensing Approaches

The remote sensing-based approach is primarily a non-devastating method and uses

the previously developed allometric equations. The field measured plot-level forest
biomass is integrated with the remotely sensed data derived forest biophysical and
structural parameters or proxies. The developed relation is then extrapolated to
generate the continuous forest biomass map. In this process, the input data (field
plots) is divided into training and testing data points. The training data points are used
to train the model or develop the relation employing statistical or machine learning
approaches; whereas, the test data is used to validate the model performance and
derive the accuracy metrices.
The biomass estimation using optical remote sensing data is usually realized by
revealing the correlation between biomass and spectral responses (vegetation indices
derived from multispectral images). Several studies including Lu et al. (2012) and
Karna et al. (2015) have used circular plots of 0.05 ha for field data collection
while estimating AGB and carbon stock by integrating lidar and other optical remote
sensing. Sader et al. (1989) studied the relationship between forest biomass and the
Landsat data derived Normalized Difference Vegetation Index (NDVI); and observed
better performance for low biomass region with gregarious formation, which was not
suitable for high biomass and mixed forest cover regions. Lee and Nakane (1997)
also employed the Landsat data and modelled the biomass employing a number
of vegetation indices, where they observed acceptable accuracy for different forest
types for different indices. Roy and Ravan (1996) studied the relationship between
optical data derived parameters and forest biomass and suggested that the brightness
and wetness have strong contribution in explaining the AGB variability. Madugundu
et al. (2008) used the Indian Remote Sensing (IRS) LISS-IV data derived leaf area
index (LAI) and NDVI to estimate the AGB in the deciduous forests in Western
Ghats and observed an acceptable accuracy.
As, the optical remote sensing technologies have limited capability to predict
forest biomass since the recorded spectral responses in optical images are mainly
represent top canopy properties of forest stand. Thus, the satellite data derived optical
properties can model the biomass up to a certain threshold and experiences a satu-
ration at higher biomass regions. On the contrary, the microwave (radar) remote
sensing operates at much higher wavelength (1 mm to 1 m), and have unique features
compared to the optical remote sensing operates mostly in visual to middle infrared
264 S. Roy et al.

Fig. 11.2 Characteristics of microwave bands in relating to forest stand parameters

region of electromagnetic (EM) spectrum. In forestry, the microwave data are exten-
sively used in forest cover and type mapping, discrimination of forest compartments,
and estimation of forest stand parameters and biomass, etc. The most commonly used
microwave data processing technique involve conversion of digital number (DN)
values or pixel values to backscatter coefficient and generation of three-dimensional
surface using Differential SAR Interferometry (DInSAR) technique (Otukei and
Emanuel 2015). The accuracy of microwave remote sensing data for biomass and
carbon assessment depends on the choice of the suitable wavelength bands and polar-
ization. The synthetic aperture radar (SAR) data are acquired in several wavelength
bands i.e., X, C, L and P. Each of these bands has different characteristics in relating to
the forest stand parameters (Fig. 11.2). Due to its lower wavelength, the microwave
signal in ‘X-band’ is highly scattered by leaves and canopy cover and enables to
extract information about the surface layer of the trees. A microwave signal oper-
ating in a higher wavelength as ‘C-band’ penetrates through leaves and scatters by
small branches and under layer elements. In comparison, the ‘L-band’ operates at
much higher wavelength and penetrates through the surface layers and is scattered
by the trunk and main branches. The ‘P-band’ with very high operating wavelength
(around 70 cm) has very high penetration capacity through the canopy cover and the
majority backscatter in ‘P-band’ is caused by trunk and the trunk ground reflectance.
The next important parameter of microwave data is the polarization of the signals, i.e.,
the direction of electric field of the electromagnetic waves. The microwave signals
are emitted and received in horizontal (H) or vertical (V) polarizations. Depending
on the polarization of emitted and received (backscatter) signal, it has four combina-
tions: (1) HH: horizontally emitted and horizontally received, (2) HV: horizontally
emitted and vertically received, (3) VH: vertically emitted and horizontally received,
and (4) VV: both the emitted and received signals have vertical polarization. ALOS-
PALSAR is an example of quad-pol sensor commercial data operates in L band.
11 Estimating Above Ground Biomass (AGB) and Tree Density … 265

Radarsat-1, Radarsat-2 and the latest Sentinel-1 are examples C band microwave
data, where the Sentinel-1 data is available freely.
Number of local, regional and global assessments have been carried out in
mapping the AGB. Santos et al. (2002) used the L-band JERS-1 data to estimate forest
and savanna biomass in two contact zones in the Brazilian Amazonia. Pandey et al.
(2010) used the Envisat ASAR to generate the AGB map for the Dudhwa National
Park of Lakhimpur-Kheri district in Uttar Pradesh state, India, and observed higher
accuracy for the HV band than HH polarization band, where the model fitting was
found better for the low biomass region compared to the high biomass regions. Otukei
and Emanual (2015), Goh et al. (2014), Carreiras et al. (2013) used ALOS PALSAR
microwave data while estimating the AGB, where the field data was collected with
circular plot design different radius of 15 m, 20 m, and 25 m, respectively. Hamdan
et al. (2015) carried out a similar assessment using L-band ALOS PALSAR data in a
Peninsular forest and Mangrove forest in Malaysia. With the availability of the latest
no-cost Sentinel-1 microwave data, number of studies have included this data for
estimating forest biomass in different forest types and climate zones. Laurin et al.
(2018) tested the accuracy of AGB prediction using the Sentinel-1 in a Mediterranean
forested landscape and suggested to use multi-temporal data and differential anal-
ysis for various forest types, where the biomass can be reliably estimated up to 400
tonnes/ha. The advances in data processing approaches are leading to the application
of both regression-based approach as well as the incorporation on machine learning
models to develop the relation between predictor variables and field measurements.
The regression-based approach allows the use of single or multiple predictor variables
while generating the relationship and provides the statistical measures on the accu-
racy, and thereafter the developed relationship is used for extrapolation (Hamdan
et al. 2015; Laurin et al. 2018). On the contrary, although, the machine learning
models are mostly work as black box, are observed to better perform while linking
the dependent and independent variables. Castillo et al. (2017) employed both the
regression-based and machine learning models to estimate the AGB in Mangrove
forest; and, observed good performance of the optical (Sentinel-2) derived LAI
and best performance for the backscatter values from the microwave (Sentinel-1)
data using machine learning model. Numbers of researchers have also tested the
performance of integrated approach, where the optical data derived forest biophys-
ical and microwave data derived structural and textural information are merged.
Ghosh and Behera (2018) used the synergistic approach, where they integrated the
Sentinel-2 data derived vegetation indices [Normalized Difference Vegetation Index
(NDVI), Green Normalized Difference Vegetation Index (GNDVI), Renormalized
Difference Vegetation Index (RDVI), and Soil-Adjusted Vegetation Index (SAVI)]
and Sentinel-1 backscatter and textural information employing the decision tree-
based machine learning models. Sinha et al. (2020) also adopted the integration
approach, where they used the Landsat TM optical data and multiple microwave
data as COSMO-Skymed (X-band), Radarsat-2 (C-band), and ALOS-PALSAR (L-
band) data for multiple linear regression to model the AGB. Avitabile et al. (2016)
combined two datasets and generated the AGB map for the pan-tropical region.
266 S. Roy et al.

LiDAR (Light Detection and Ranging) is an active system offering tremendous

potential for estimating forest biomass with the major advantage as the acquisition of
three-dimensional data of the forest structure including canopy cover characteristics,
leaf area index, crown cover and volume, etc. The ability of the laser altimeters
to penetrate through forest canopies to the ground level better identifies the plant
structure (Rosillo-Calle and Woods 2012). Previous research has indicated that use
of LiDAR data is a promising approach for biophysical parameters estimation (Drake
et al. 2002; Hyde et al. 2007), such as timber volume and stand height (Naesset 1997),
tropical forest biomass (Drake et al. 2002), Douglas fir western hemlock biomass
(Lefsky et al. 1999) etc. Long wavelength radar data have the advantage in AGB
estimation for complex forest stand structures and LiDAR data have the potential to
provide vertical structure information (Zimble et al. 2003).
The present study is undertaken to estimate above ground biomass (AGB) and
carbon stock for Shorea robusta (sal) forest of Jhargram block, Paschim Medinipur
district, West Bengal, India. Using the no-cost high resolution Sentinel-1 microwave
data, the relationship between AGB and VH and VV polarization was derived using
regression analysis. Moreover, using the AGB map, the tree density map of the study
area was also produced.

11.3 Study Area

Jhargram is a community development block that forms an administrative division

in Jhargram district in the Indian state of West Bengal (Fig. 11.3). As per the 2011
Census of India, the total population of Jhargram CD Block was 170,097. The terrain
is regulated by the Chotanagpur Plateau, which gradually slopes down creating an
undulating area having infertile laterite soils. About 75% of the cultivated area has
lateritic soil and 20% has alluvial soil and 5% has red gravelly soil in this region. The
topography of this area is relatively flat or gently undulating with an average elevation
of 81 m above the mean sea level. The main streams in the area is Kangsabati River
also known as Kasai and Cossye in the northern part of the study area. The diurnal
temperature reaches as high as 46 °C in the hot and dry months of May and June and
drops to 4 °C during winters (December and January). The average annual rainfall
is about 1370 mm with a concentrated rainfall during monsoon (June to September)
leading to comparatively dry pre and post-monsoon. Such a climate condition lead
to dry deciduous tropical forest, where S. robusta (Sal) is dominated and partially
mixed with various species like Mahua, Palash, Kusum, Kendu and Neem, etc.
11 Estimating Above Ground Biomass (AGB) and Tree Density … 267

Fig. 11.3 Location map of the study area in West Bengal and field plots (training and testing)
overlaid on forest cover area
268 S. Roy et al.

11.4 Materials and Method

11.4.1 Data Sources

In the current study, the Sentinel-1A C-band SAR data has been used for biomass
estimation. Sentinel-1A data contains two polarization bands as VV and VH. Based
on the field data collection period, the satellite scene of 9th February 2018 was used.
The satellite scene was acquired from the ESA (European Space agency) scientific
data hub ([Link]/dhus/#/home). The Sentinel-2A optical data was also
used in this study to map the forest cover area. It consists of 13 spectral bands in
the visible, NIR, and SWIR region of the EM spectrum. The SNAP tool and ArcGIS
software was used for data processing, AGB and corresponding carbon estimation.
The statistical analysis was performed in MS excel.

11.4.2 Field Data Collection and AGB Measurement

The plot level field data was collected in 40 locations having a rectangular plot size
of 30 m × 30 m area. Within each plot, the number of trees, diameter at breast
height (DBH) and height of trees were measured, and multiple geo-locations were
recorded inside the plots and corner location using a handheld GPS device (Fig. 11.4).

Fig. 11.4 Field observations and plot level plant inventory data collection
11 Estimating Above Ground Biomass (AGB) and Tree Density … 269

The DBH was measured at height 1.3 m above the base of a tree. A revised and
improved allometric equation (Eq. 11.1) was adapted from Chave et al. (2014) for
AGB estimation; whereas the carbon fraction approach (Eq. 11.2) was used for carbon
stock estimation (Hirata et al. 2012).
 0.976
AGB = 0.0673 ρ × D2× H (11.1)

where, ρ = Wood density in gcm3 , D = DBH in cm, and H = Tree height in m

C = B × CF (11.2)

where, C = Carbon stored in tonnes, B = AGB, and CF = Carbon fraction (0.5).

11.4.3 Methodology

The field collected data on tree count, DBH and height was used to compute the AGB
using the suitable allometric equation, which was statistically linked with satellite
data derived proxies. Based on the develop relation with the training data, a contin-
uous AGB map was prepared and validated with test data points. The carbon stock
map was prepared by multiplying the carbon fraction. The overall methodology flow
diagram has been shown in Fig. 11.5.

[Link] Pre-processing of Optical Data and Generation Forest Cover

Area Map

The atmospheric correction of Sentinel-2 scene was performed using the QGIS soft-
ware. The atmospherically corrected data was then used to derive the land use forest
cover map of the study area. The maximum likelihood classifier was used for image
classification, which was validated with the field observed data points for accuracy
assessment. The forested area in the study area was extracted and used to mask the
microwave data before estimating the AGB.
Calculation of Enhanced Vegetation Index (EVI)
The Enhanced Vegetation Index (EVI) was computed using the Sentinel-2 data. The
EVI is an optimized vegetation index designed to enhance the vegetation signal with
improved sensitivity to high biomass and de-coupled the canopy background signal
and influences of atmospheric scattering using suitable correction factors.
The atmospherically corrected Sentinel-2 data was employed to derive EVI map
using the following equation:

(N I R − R E D)
EVI = G × (11.3)
(N I R + C1 × R E D − C2 × B LU E + L)
270 S. Roy et al.

Sentinel-1A SAR data Sentinel-2A Optical data Forest inventory data

(VV & VH) (30×30 m sample plot)

Radiometric, Geometric Atmospheric correction AGB estimation

correction and Speckle (allometric Equation)
filtering

Forest cover area and Training data Test data

Backscatter
EVI
co-efficient

Linear to decibel
Conversion (db)
Regression
analysis and
Comparison

AGB and Carbon Stock

map
Validation

Tree Density map

Fig. 11.5 Methodology of the study area

where, NIR, Red and Blue bands are atmospherically corrected bands, L is the
canopy background adjustment, and C1 and C2 are the coefficients factors of aerosol
resistance term, which uses the blue band to correct the influence of aerosol content.
The coefficients values used in the MODIS-EVI were utilized: L = 1, C1 = 6, C2
= 7.5, and G (gain factor) = 2.5.

[Link] Pre-processing of Microwave Data

The sentinel-1 GRD data product (level 1) was used in this study. Using the SNAP
tool, the radiometric (e.g., conversion of DN values to backscatter coefficient σ°)
and geometric correction (e.g., conversion of slant range to ground range, multi look
and azimuth compression) were performed prior to the data analysis. Thereafter, the
Gamma filter (with 5 × 5 kernel size) was applied to remove the speckles in the image.
Then, Doppler terrain correction was applied employing the shuttle radar topographic
mission (SRTM) digital elevation model (DEM) data. The pre-processed image was
then exported in GeoTIFF format for further processing in ArcGIS software.
11 Estimating Above Ground Biomass (AGB) and Tree Density … 271

[Link] Modelling and Estimation of AGB, Carbon Stock, and Tree

Density

A total of 40 plot data was available, which was segregated into training and testing to
develop the model and validation purpose, respectively. A random selection process
was applied to divide the entire data into training (25 plots or 62.5% data) and
testing (15 plots or 37.5% data). To corroborate with the pixel size and plot area,
the backscatter coefficient from 9 neighbouring pixels centring the plot location was
considered to develop the relation between backscatter values and the field measured
in situ AGB. The univariate linear regression analysis was performed to develop the
relationship between AGB and independent variables as radar backscatter bands (VV
and VH), and EVI. The best performing predictor variable was chosen by comparing
the coefficient of determination (R2 ) value. The best performing variable was then
employed to create the continuous biomass map, and the test data was utilized to
validate it. The prediction accuracy was assessed via coefficient of determination,
RMSE and bias values. The carbon stock map was prepared using the predicted
AGB map by multiplying the carbon fraction factor of 0.5 (Watson 2009; Hirata
et al. 2012). To estimate tree density map per pixel, the predicted AGB map was
used, where the equation used for biomass estimation was reversed.

Bmp × 1000
NP = (11.4)
0.0673 × (0.73 × D × H )0.976

where, Np = No. of predicted tree, Bmp is predicted biomass and 0.0673 and 0.976
is model coefficient, and D is Diameter at breast height (DBH) in cm and H is tree
height in meter.

11.5 Results and Discussion

It is essential to map the forest cover area for reliable biomass estimation. The
Sentinel-2 multi-spectral data was used to create the land use forest cover map
employing the four 10 m VNIR bands. The maximum likelihood classifier was used,
where the forest covered areas were extracted to mask the microwave data before
biomass estimation (Fig. 11.6). Field visits were carried out to collect plot level data,
where the geo-location of S. robusta trees along with the DBH, tree density and
height data were recorded in 40 sample plots (Fig. 11.6). At 30 m × 30 m plot area,
the measured DBH varied between 34 and 67 cm, tree height varied between 12 and
15 m, and the numbers of trees varied between 44 and 157 per plot. A part of plot
data was used to create signature file during image classification and rest of the data
points were used to verify the classification accuracy indicated all confirm forest area
occurrences. The total identified forest area is 55.36 km2 .
272 S. Roy et al.

Fig. 11.6 The forest cover area map of the study area. The cyan coloured oval shaped area was
not considered and masked from subsequent spatial maps being an urban settlement area (Jhargram
town)

The allometric equation developed by Chave et al. (2005) was employed to esti-
mate the AGB in the current study. The wood density value (0.73 gm/cm3 ) for S.
robusta given by the Forest Survey of India (FSI) was used in this case. The plot level
in situ AGB varied between 88.56 tonnes/ha and 170.29 tonnes/ha, with an average
biomass value of 122.93 tonnes/ha. The simple linear regression model was tested to
assess the variability explained by each of the independent parameters as backscatter
in VV and VH polarization bands, and EVI. During modelling, the obtained result
indicated the highest coefficient determination value for VH polarized band (R2 =
0.63) and a comparatively lower R2 (0.44) for VV polarized band, which was 0.45
(R2 ) for EVI (Fig. 11.7a–c).
As, the highest coefficient of determination value was obtained with the VH band,
it was thus used for estimating AGB and carbon content map. The coefficients derived
from the regression analysis was used to generate the continuous AGB map. Using
the test data points, the coefficient of determination, RMSE and bias values were
estimated for the predicted AGB map, which showed moderate correlation (R2 =
0.45) with significantly low RMSE value (±17 tonnes/ha) and slight underestimation
(bias value = −0.024) (Fig. 11.8).
The direct relationship between microwave backscatters values and plot-wise
tree density was observed comparatively weak (R2 = 0.4 for VH and 0.23 for VV),
whereas, the relation with AGB was observed much stronger (R2 = 0.87) (Fig. 11.9).
11 Estimating Above Ground Biomass (AGB) and Tree Density … 273

160 (a)
140

AGB in situ (tonnes/ ha) 120

100
80
60
40
y = 10.303x + 288.56
20 R² = 0.6333
0
-18 -17.5 -17 -16.5 -16 -15.5 -15 -14.5
VH Backscatter coefficient ( σ0) in db

160 (b)
140
AGB in situ (tonnes/ ha)

120
100
80
60
40
y = 7.0215x + 183.85
20 R² = 0.443

0
-12 -10 -8 -6 -4 -2 0
VV Backscatter coefficient ( σ0) in db

160 (c)
140
AGB in situ (tonnes/ ha)

120
100
80
60
40
y = -85.579x + 144.66
20 R² = 0.4464
0
0 0.1 0.2 0.3 0.4 0.5
EVI value

Fig. 11.7 Relationship between AGB and a backscatter coefficient (σ°) in db in VH band,
b backscatter coefficient (σ°) in db in VV band, and c EVI
274 S. Roy et al.

200
R² = 0.4493

Predicted (tonnes/ ha)

150

100

50

0
0 50 100 150 200
Observed in situ (tonnes/ ha)

Fig. 11.8 Predicted AGB is plotted against the observed AGB for validation

Fig. 11.9 Observed versus predicted tree density

Thus, the tree density map for the entire study area was produced using the AGB
map.
The estimated AGB and carbon stock maps are shown in Fig. 11.10a, b. The result
indicates that the amount of AGB varies between 44.1 tonnes/ha and 249 tonnes/ha
with the total AGB as 33,102 tonnes estimated for the study area. Subsequently,
the estimated carbon stock map follows the similar pattern of the AGB map. The
AGB map indicates majority of the areas having the AGB ranging between 100–200
tonnes/ha. On the other hand, areas with biomass below 100 tonnes/ha and above
200 tonnes/ha are comparatively less. The areas below 100 tonnes/ha are mostly
indicating the sparse and comparatively new plantation areas; whereas, the areas
11 Estimating Above Ground Biomass (AGB) and Tree Density … 275

(a)

AGB in Tonnes /ha

44.1 - 100
100.1 - 150
150.1 - 200
200.1 - 249

(b)

Carbon stock tonnes/ha

< 30
30 – 60
60 – 90
> 90

Fig. 11.10 Estimated a AGB and b Carbon stock for the study area
276 S. Roy et al.

with very high biomass are the old grown dense forest areas. This was corroborated
by the tree density map and statistics, which shows that the contribution from the
lowest and highest tree density class is about 20% only, where the intermediate range
(69/ha to 75/ha) contributes rest of the 80% AGB in this region (Fig. 11.11 and Table
11.2). The total numbers of trees estimated for the study region is 19,872,061.
The ground sampled data exhibits the variation in DBH as 34–67 cm, infers that’s
the incorporation of comparatively newer and old plants in the current analysis.
However, the variations in average tree height was comparatively less. Moreover,
the tree density mostly varies between 44/ha and 75/ha, where only two plots were
observed to have exceptional high tree density. The supervised image classifica-
tion well captured the vegetation cover area, from where the settlement region was
masked out to avoid the differential scattering from the buildings/artificial construc-
tions. The regression analysis indicted the higher performance by VH band (R2
= 0.63) in capturing the plants structural information in comparison to VV band

Tree Density/ 0.09 ha

< 68
69 - 72
73 - 75
> 75

Fig. 11.11 Tree density map of Jhargram Block

Table 11.2 Estimated tree

Density range Number of trees Total number of trees
density statistics
<68 2,194,751 19,872,061
69–72 8,684,798
72–75 7,256,195
>75 1,736,317
11 Estimating Above Ground Biomass (AGB) and Tree Density … 277

(R2 = 0.44), where higher backscatter corresponds to higher AGB. Similarly, the
biophysical characteristic as EVI, derived from the optical data could not perform
well as explained by the structural information obtained from VH band. Moreover,
the estimated AGB was found well accurate (R2 = 0.87) while deriving the tree
density. At plot level, the observed AGB varied between 88.56 tonnes/ha and 170.29
tonnes/ha, which was estimated between 44.1 tonnes/ha and 249 tonnes/ha for the
entire study region and majority of the AGB was observed in the range of 100–200
tonnes/ha. This could be indicating a uniform forest cover regime with compara-
tively lower number patches having lower and higher AGB. Forest patches below
the average AGB range could be the new plantation sites; whereas, patches with
higher AGB could be the old and dense patches. Moreover, in case of tree density,
the majority (>80%) of the estimated tree density varies between 69/ha and 75/ha,
which supporting the uniformity in tree cover in this region. Behera et al. (2016) used
the L-band ALOS-PALSAR data to estimate AGB in Katerniaghat Wildlife Sanc-
tuary (KWS), a tropical forested region and observed better performance of VV band
compared to VH band employing non-linear regression. However, their prediction
accuracy (R2 = 0.47) was similar as derived in the current study (R2 = 0.45); and,
observed a higher AGB ranges up to 625 tonnes/ha for S. robusta dominated forest
covers. Thumaty et al. (2016) also estimated the AGB for the state of Madhya Pradesh
mostly contains the moist and dry deciduous forest, Tectona grandis (teak), and S.
robusta forests employing the L-band ALOS-PALSAR data. Considering all the
forest type together, they have observed the HV band to better explain the variability
of AGB (R2 = 0.51) with a similar RMSE value (±19.32 tonnes/ha) as estimated in
the currently study (±17 tonnes/ha). Behera et al. (2016) vividly studied the forest
biomass for three distinct plant functional types (PFTs) employing filed measure-
ments in KWS and estimated the AGB ranging 290.82– 455.99 tonnes/ha, and their
principle component analysis (PCA) of microclimatic condition with forest structural
information indicates higher contribution from humidity and air-temperature. While
estimating the AGB integrating backscatter and textural information derived from
Sentinel-1 data and vegetation index from Sentinel-2 data, Ghosh and Behera (2018)
observed higher coefficient of determination value (R2 = 0.71) with poor RMSE
value (105.027 tonnes/ha) for S. robusta in KWS. The range of AGB observed in
the current study also supported by the observation of Pandey et al. (2010) estimated
the range as 119–520 tonnes/ha for sal and sal dominated forest cover in the tropical
forest of Dudhwa National Park of Lakhimpur-Kheri district in Uttar Pradesh state,
India.

11.6 Conclusion

The present study assessed the potential of Sentinel-1 microwave data for AGB and
tree density estimation over a dry deciduous S. robusta (sal) dominated forest in the
Jhargram CD block of West Bengal state, India. Moreover, the Sentinel-2 optical
remote sensing data was used to demarcate the forest area and to derive the forest
278 S. Roy et al.

biophysical parameter as EVI and to assess its relationship with AGB. The sentinel-1
provided the backscatter values in VV and VH polarizations, which were linearly
regressed with the field measured AGB data. 40 sample plot data were collected on
the field, which indicates a better variation explained by the VH band (R2 = 0.63)
compared to the VV band (R2 = 0.44), where the relationship was observed weaker
for EVI also (R2 = 0.45). Using the regression equation, the continuous forest AGB
map was thus estimated employing the VH band. The validation results showed
good accuracy (R2 = 0.45) with low RMSE value (17 tonnes/ha), and slight under-
estimation (bias = −0.024). The regression analysis and validation result clearly
indicates the applicability of Sentinel-1 microwave data for AGB estimation with
a limited accuracy. However, the accuracy for the estimated tree density map was
observed much higher (R2 = 0.87). The variability in field measurement via DBH
indicates incorporation of older and younger trees dominated plots in the analysis,
and the estimated AGB varied between 44.1 tonnes/ha and 249 tonnes/ha, which
corroborated the AGB observations from similar climate zones in India. Majority
of the estimated AGB ranged between 100–200 tonnes/ha, which could be indi-
cating the uniform forest cover in this region, which was also indicated by the tree
height and tree density data. However, few regions were observed to have lower and
higher AGB indicating the newer and densely forest cover regions, respectively. The
spatial biomass and tree density maps are highly useful for forest cover management
activities, especially to adopt the climate adaptive conservation measures. The study
outcome confirms the use of the freely available Sentinel-1 C-band microwave data
for reliable AGB estimation. It has further proposed to perform differential analysis
for other gregarious forest types and mixed forest types in different climate zones
and segregating the young and old forests.

Acknowledgments We are grateful to the PG Department of Geography, Raja N. L. Khan Women’s

College (Autonomous), affiliated to Vidyasagar University, Midnapore, West Bengal, India for
supporting this research. The author (P. K. Shit) grateful acknowledges West Bengal DSTBT for
financial support through R&D Research Project Memo no. 104(Sanc.)/ST/P/S&T/ 10G-5/2018).

References

Avitabile V, Herold M, Heuvelink GB, Lewis SL, Phillips OL, Asner GP, Armston J, Ashton PS,
Banin L, Bayol N, Berry NJ (2016) An integrated pan-tropical biomass map using multiple
reference datasets. Global Change Biol 22(4):1406–1420
Baishya R, Barik SK, Upadhaya K (2009) Distribution pattern of aboveground biomass in natural
and plantation forests of humid tropics in northeast India. Tropical Ecol 50(2):295
Behera MD, Tripathi P, Mishra B, Kumar S, Chitale VS, Behera SK (2016) Above-ground biomass
and carbon estimates of Shorea robusta and Tectona grandis forests using QuadPOL ALOS
PALSAR data. Adv Space Res 57(2):552–561
Brown S, Gillespie AJ, Lugo AE (1989) Biomass estimation methods for tropical forests with
applications to forest inventory data. For Sci 35(4):881–902
11 Estimating Above Ground Biomass (AGB) and Tree Density … 279

Carreiras J, Melo JB, Vasconcelos MJ (2013) Estimating the above-ground biomass in miombo
savanna woodlands (Mozambique, East Africa) using L-band synthetic aperture radar data.
Remote Sens 5(4):1524–1548
Castillo JAA, Apan AA, Maraseni TN, Salmo SG III (2017) Estimation and mapping of above-
ground biomass of mangrove forests and their replacement land uses in the Philippines using
Sentinel imagery. ISPRS J Photogram Remote Sens 134:70–85
Center for International Forestry Research (2008) Forest carbon toolbox. Forest and climate change
toolbox. Retrieved from [Link]
Chamber JQ, dos Santos J, Ribeiro RJ, Higuchi N (2001) Tree damage, allometric relationships, and
above-ground net primary production in central Amazon forest. For Ecol Manage 152(1–3):73–84
Chave J, Andalo C, Brown S, Cairns MA, Chambers JQ, Eamus D... Lescure JP (2005) Tree
allometry and improved estimation of carbon stocks and balance in tropical forests. Oecologia
145(1):87–99
Chave J, Réjou-Méchain M, Búrquez A, Chidumayo E, Colgan MS, Delitti WB, Duque A, Eid
T, Fearnside PM, Goodman RC, Henry M (2014) Improved allometric models to estimate the
aboveground biomass of tropical trees. Global Change Biol 20(10):3177–3190
Drake JB, Dubayah RO, Knox RG, Clark DB, Blair JB (2002) Sensitivity of large-footprint lidar to
canopy structure and biomass in a neotropical rainforest. Remote Sens Environ 81(2–3):378–392
FAR (2010) Global forest Resources assessment main report, Rome, 2010
Ghosh SM, Behera MD (2018) Aboveground biomass estimation using multi-sensor data synergy
and machine learning algorithms in a dense tropical forest. Appl Geogr 96:29–40
Goh J, Miettinen J, Chia AS, Chew PT, Liew SC (2014) Biomass estimation in humid tropical forest
using a combination of ALOS PALSAR and SPOT 5 satellite imagery. Asian J Geoinf 13(4)
Hamdan O, Hasmadi IM, Aziz HK, Norizah K, Zulhaidi MH (2015) L-band saturation level for
aboveground biomass of dipterocarp forests in peninsular Malaysia. J Tropical For Sci, 388–399
Haripriya GS (2000) Estimates of biomass in Indian forests. Biomass Bioenerg 19(4):245–258
Hirata Y, Takao G, Sato T, Toriyama J (2012) REDD-plus cookbook: How to measure and monitor
forest carbon. REDD research and development center, forestry and forest products research
institute, Tsukuba
Hofmann DJ, Butler JH, Tans PP (2009) A new look at atmospheric carbon dioxide. Atmos Environ
43(12):2084–2086
Houghton RA (2012) Carbon emissions and the drivers of deforestation and forest degradation in
the tropics. Curr Opin Environ Sustain 4(6):597–603
Houghton RA, House JI, Pongratz J, Van Der Werf GR, DeFries RS, Hansen MC, Le Quéré C,
Ramankutty N (2012) Carbon emissions from land use and land-cover change. Biogeosciences
(12):5125–5142
Hyde P, Nelson R, Kimes D, Levine E (2007) Exploring LiDAR–RaDAR synergy—predicting
aboveground biomass in a southwestern ponderosa pine forest using LiDAR, SAR and InSAR.
Remote Sens Environ 106(1):28–38
IPCC (2003) IPCC good practice guidance for land use, land-use change and forestry (adopted by
decisions 2/CP.17, 17/CP.18) IGES, Japan
Karna YK, Hussin YA, Gilani H, Bronsveld MC, Murthy MS, Qamer FM, Karky BS, Bhattarai
T, Aigong X, Baniya CB (2015). Integration of WorldView-2 and airborne LiDAR data for tree
species level carbon stock mapping in Kayar Khola watershed, Nepal. Int J Appl Earth Observ
Geoinf 38:280–291
Keeling CD, Adams JA Jr, Ekdahl CA Jr, Guenther PR (1976) Atmospheric carbon dioxide variations
at the South Pole. Tellus 28(6):552–564
Kumar A, Sharma MP (2015) Assessment of carbon stocks in forest and its implications on global
climate changes. J Mater Environ Sci 6(12):3548–3564
Laurin GV, Balling J, Corona P, Mattioli W, Papale D, Puletti N, Rizzo M, Truckenbrodt J, Urban
M (2018). Above-ground biomass prediction by Sentinel-1 multitemporal data in central Italy
with integration of ALOS2 and Sentinel-2 data. J Appl Remote Sens 12(1):016008
280 S. Roy et al.

Lee NJ, Nakane K (1997) Forest vegetation classification and biomass estimation based on Landsat
TM data in a mountainous region of west Japan. In: The use of remote sensing in the modeling
of forest productivity, pp 159–171. Springer, Dordrecht
Lefsky MA, Harding D, Cohen WB, Parker G, Shugart HH (1999) Surface lidar remote sensing of
basal area and biomass in deciduous forests of eastern Maryland, USA. Remote Sens Environ
67(1):83–98
Lu D, Chen Q, Wang G, Moran E, Batistella M, Zhang M, Vaglio Laurin G, Saah D (2012).
Aboveground forest biomass estimation with Landsat and LiDAR data and uncertainty analysis
of the estimates. Int J For Res 2012
Lucas RM, Mitchell AL, Armston J (2015) Measurement of forest above-ground biomass using
active and passive remote sensing at large (subnational to global) scales. Curr Forestry Rep
1:162–177. [Link]
Madugundu R, Nizalapur V, Jha CS (2008) Estimation of LAI and above-ground biomass in
deciduous forests: Western Ghats of Karnataka, India. Int J Appl Earth Obs Geoinf 10(2):211–219
Naesset E (1997) Estimating timber volume of forest stands using airborne laser scanner data.
Remote Sens Environ 61(2):246–253
Nath AJ, Tiwari BK, Sileshi GW, Sahoo UK, Brahma B, Deb S, Devi NB, Das AK, Reang D,
Chaturvedi SS, Tripathi OP (2019) Allometric models for estimation of forest biomass in North
East India. Forests 10(2):103
Otukei JR, Emanuel M (2015) Estimation and mapping of above ground biomass and carbon of
Bwindi impenetrable National Park using ALOS PALSAR data. S Afr J Geomat 4(1):1–13
Pandey U, Kushwaha SPS, Kachhwaha TS, Kunwar P, Dadhwal VK (2010) Potential of Envisat
ASAR data for woody biomass assessment. Tropical Ecol 51(1):117
Rosillo-Calle F, Woods J (2012) The biomass assessment handbook. Routledge
Roy PS, Ravan SA (1996) Biomass estimation using satellite remote sensing data—an investigation
on possible approaches for natural forest. J Biosci 21(4):535–561
Sader SA, Waide RB, Lawrence WT, Joyce AT (1989) Tropical forest biomass and successional age
class relationships to a vegetation index derived from Landsat TM data. Remote Sens Environ
28:143–198
Salunkhe O, Khare PK, Sahu TR, Singh S (2016) Estimation of tree biomass reserves in tropical
deciduous forests of Central India by non-destructive approach. Tropical Ecol 57(2):153–161
Santos JR, Lacruz MP, Araujo LS, Keil M (2002) Savanna and tropical rainforest biomass estimation
and spatialization using JERS-1 data. Int J Remote Sens 23(7):1217–1229
Sinha S, Mohan S, Das AK, Sharma LK, Jeganathan C, Santra A, Santra Mitra S, Nathawat MS
(2020). Multi-sensor approach integrating optical and multi-frequency synthetic aperture radar
for carbon stock estimation over a tropical deciduous forest in India. Carbon Manag 11(1):39–55
Sullivan MJP, Lewis SL, Hubau W, Qie L, Baker TR, Banin LF, Phillips OL (2018) Data from field
methods for sampling tree height for tropical forest biomass estimation. [Link]. https://
[Link]/10.5521/[Link]/2018_1
Thumaty KC, Fararoda R, Middinti S, Gopalakrishnan R, Jha CS, Dadhwal VK (2016) Estimation
of above ground biomass for central Indian deciduous forests using ALOS PALSAR L-band data.
J Indian Soc Remote Sens 44(1):31–39
Vargas-Larreta B, López-Sánchez CA, Corral-Rivas JJ, López-Martínez JO, Aguirre-Calderón CJ,
Álvarez-González JG (2017) Allometric equations for estimating biomass and Carbon stocks in
the temperate forests of north-western Mexico. Forests 8:269. [Link]
Watson C (2009) Forest carbon accounting: overview and principles. Forest carbon accounting:
overview and principles
Zimble DA, Evans DL, Carlson GC, Parker RC, Grado SC, Gerard PD (2003) Characterizing vertical
forest structure using small-footprint airborne LiDAR. Remote Sens Environ 87(2–3):171–182
Chapter 12
Forest Fire Risk Assessment for Effective
Geoenvironmental Planning
and Management using Geospatial
Techniques

Narayan Kayet

Abstract Forest are essential natural resources having the role of supporting
economic activity, which plays a significant role in regulating the climate and the
carbon cycle. Forest ecosystems increasingly threatened by fires caused by a range
of natural and anthropogenic factors. Hence, spatial assessment of fire risk is crit-
ical to reducing the impacts of wildland fires. In the current research, the evalua-
tion of forest fire risk (FFR) assessment performed by geospatial data of Melgaht
Tiger Reserve Forest (MTRS), Maharashtra, India. We have used eleven natural and
anthropogenic parameters (slope, altitude, topographic position index (TPI), aspect,
rainfall, land surface temperature (LST), air temperature, wind speed, normalized
differential vegetation index (NDVI), distance to road and distance to settlement)
for FFR assessment based on the Analytic hierarchy process (AHP) and Frequency
ratio (FR) models in a GIS framework. The results from AHP and FR models shown
similar trends. The AHP model was significantly higher accuracy than the FR model.
AHP and FR models based FHR maps were classified into five classes (very low,
low, moderate, high, and very high). According to the generated FFR maps, the
very high-risk class was found at some forest blocks (Mangtya, Kund, Gudfata,
Katharmal, Amyar). The sensitivity analysis showed that some parameters (wind
speed, air temperature, LST, slope, altitude, distance to settlement, and distance to
the road) were more sensitive to forest fire risk. The FFR results were justified by
the forest fire sample points (Forest Survey of India) and burn images (2010–2018).
This work will provide a basic guideline for effective geo-environmental planning
and management of Melgaht Tiger Reserve Forest.

Keywords Forest fire risk assessment · Analytical hierarchy process (AHP) ·

Frequency ratio (FR) · Remote sensing and GIS

N. Kayet (B)
Department of Mining Engineering, Indian Institute of Technology Kharagpur, Kharagpur, India
e-mail: narayankayet@[Link]

© The Editor(s) (if applicable) and The Author(s), under exclusive license 281
to Springer Nature Switzerland AG 2021
P. K. Shit et al. (eds.), Spatial Modeling in Forest Resources Management,
Environmental Science and Engineering,
[Link]
282 N. Kayet

12.1 Introduction

Forest is a natural resource, and it has a significant role in maintaining the balance of
the environment. Fire is a rapid combination of fuel, oxygen, and heat. The population
of India is increasing day by day, and it has become a problem, especially for the
forest. Most of the forest fires in India are caused by human activity. A forest fire
has loosely related to the socio-economic activity of the people who live in the forest
areas near the forest area. Grazing, shifting cultivation, and deforestation of forest
resources are the main reasons for a forest fire. According to the Forest Survey of
India (FSI) state-wide study in the year 2000, about 1.45 million per year of forest
area affected by the fire. So it is essential to monitor forest fire in Melgaht Tiger
Reserve (MTR) Forest for effective geo-environmental planning and management.
Due to natural or anthropogenic reasons, forest ecosystems are threatened by
increasing the rate of a forest fire. So, forest fire risk zonation or spatial assessment
of fire risk is a very important task nowadays to reduce the impact of wildfire. The
occurrence of forest fires is one of the major environmental concerns (called wild-
fires), which has an effect on forest preservation, ecological and economic damage
as well as causes to human suffering (Cortez and Morais 2007). Many environmental
and anthropogenic variables are related to fire, and recognizing them is required to
understand fire occurrences and to determine fire risk (Krivtsov et al. 2009). But most
of the forest fires are the consequences of anthropogenic activity. It may cause defor-
estation, timber harvesting, for the use of forest in another land use purpose such as
agricultural land, pastureland, etc. Wildfire can be defined as the rapid combustion
of fuel, heat, and oxygen. It is a chemical reaction of any substance that will ignite
and burn to release a lot of energy in the form of heat and light (Rawat 2003).
Remote sensing has become a very useful and reliable technology for the moni-
toring as well as for management of forest fire. Many kinds of literature have been
conducted to develop a forest fire risk zone mapping by the use of RS and GIS
techniques. MODIS data commonly applied for the identification of forest fire all
over the world; which can provide 12-bit data (Radiometric resolution) in 36 spec-
tral bands at a range of 0.4–14.4 µm (Pourghasemi et al. 2016). Roy (2003) applied
geospatial techniques that are very useful to assess the forest fire risk and degrada-
tion assessment. In this paper, he had described the present states of forest, forest
fire risk, and degradation assessment in the context of India. Sharma et al. (2012) to
find out the forest fire risk zone by the Fuzzy AHP (Analytical Hierarchy Process)
approach in Shimla Forest Division (Himachal Pradesh) in 2012. The result has a
strong agreement with actual fire occurrences in the past years.
In present study area is located in Amravati district of Maharashtra, India. Melghat
forest famous for the Tiger Reserve forest. This area has dry deciduous forests and
excellent tiger habitat. The local forest department said the Melghat forest area
destroyed in the fire would be anything between 2000 and 2200 ha of forestland in
the years. That why we chose this study area for forest fire risk assessment.
In this study, a proposed methodology is prepared for forest fire risk (FFR) assess-
ment using multi-criteria models in a GIS framework. The objectives of the study
12 Forest Fire Risk Assessment for Effective Geoenvironmental … 283

Table 12.1 Data used in the

Data layer Source of data
present study
Satellite Images (Landsat 8 OLI) Earth Explorer/GLOVIS
Cartosat 1 DEM NRSC Bhuvan
Quick Bird Image Google Earth
Forest fire location points Forest Survey of India
Climate data SWAT

were: (1) Forest fire zoning mapping using the multi-criteria model in a GIS frame-
work. (2) To identify very high forest fire risk zones in the Melghat Tiger Reserve
forest. (3) Forest risk zoning results justified by FSI training points.

12.2 Materials and Methods

12.2.1 Data Source

For the study, the Landsat satellite image of Maharashtra state was acquired for the
year 2016. The data was obtained from GLOVIS, an Earth Science Data Interface,
while that of the Digital elevation model for Melghat forest was acquired from
NRSC Bhuvan. It is also essential for the study because it provides topographical
information regarding. Climate data obtained from SWAT global weather data. The
details of the data used are given in Table 12.1.

12.2.2 Study Area

The Melghat Tiger Reserve situated in the Satpura hill ranges of Central India
and it lies in Melghat forest of Amravati district in the Vidarbha region of Maha-
rashtra, bordering Madhya Pradesh in North and East (Fig. 12.1). The MTR area of
1571.74 km2 lies in the heart of the Melghat forest and was declared the Melghat
Tiger Reserve on 22nd February 1974. The MTR is within latitude 2100 1500 to 2100
4500 N and longitude 7600 5700 to 7700 3000 E at elevations of 312 to 1178 m MSL.
The yearly maximum temperature averages 42.7 °C, with annual rainfalls 2700 mm.
Winds are generally light to moderate; the strongest winds are less than 22 km/h.
284 N. Kayet

Fig. 12.1 Location map of Melghat Tiger Reserve Forest

12.2.3 Frequency Ratio Model (FR)

The frequency ratio model, a simple geospatial assessment tool for computing the
probabilistic relationship between dependent and independent variables, including
multi-classified maps, can be obtained by an FR model (Choi et al. 2012). Based on
this assumption, the relationships between fires occurring in an area and the forest
fire-related factors can be distinguished from the relationships between fires not
occurring in an area and the forest fire-related factors. This thing can be represented
as an FR, which represents the quantitative relationship between forest fires events
and different causative parameters. The FR can be defined as the probability of
occurrence of a certain attribute (Mohammady et al. 2012).

12.2.4 Analytical Hierarchy Process (AHP)

The analytical hierarchy process is a theory of measurement for considering tangible

and intangible criteria that have been applied to numerous areas, such as decision
theory and conflict resolution (Vargas 1990; Yalcin 2008). The AHP is an eigenvalue
technique for the pair-wise comparisons approach. It is based on three principles:
decomposition, comparative judgment, and synthesis of priorities (Chen and Xu
2010). The decomposition principle is applied to structure a complex problem into a
hierarchy of clusters, sub-clusters. The AHP provides a fundamental numerical scale,
which ranges from 1 to 9 to calibrate the quantitative and qualitative performances
of priorities (Saaty 2008). This matrix ultimately enters expert choice (EC) software
and will calculate the final weight for each conditioning factor with consistency
ratio (CR). If CR is less than 10%, then the matrix can be considered as having an
acceptable consistency (Saaty 2008).
12 Forest Fire Risk Assessment for Effective Geoenvironmental … 285

Fig. 12.2 Overall methodology of forest fire risk mapping

286 N. Kayet

12.2.5 Meteorology

Figure 12.2 shows the used methodology in this study area with a flowchart. The
figure shows the factors used in the analysis and the processes applied according
to the methods. At the first stage, required data were collected, then those data will
be processed with proper correction, and various factors are prepared to performed
Frequency Ratio (FR) and Analytic hierarchy process (AHP) models. Finally, forest
fire risk zone map prepared and validate with the previous record of forest fire in
Melghat forest (2010–2016).

12.3 Results and Discussion

12.3.1 Land Use and Land Cover (LULC)

The study area is comprised mainly of forest areas, so most of the domain is belong
to different forest categories. Land use and land cover map of Melghat forest are
prepared using the Support Vector Machine (SVM) classifier technique. The LULC
map of the study area is shown in Fig. 12.3. Area of different LULC classes are
present in Table. 12.2.

Fig. 12.3 Land use and land cover map of Melghat Tiger Reserve Forest
12 Forest Fire Risk Assessment for Effective Geoenvironmental … 287

Table 12.2 Area of LULC

Classes name Area in km2 Area in percent
classes
Dense Forest 257.91 17.47
Mixed Forest 379.81 25.74
Open Mixed Forest 395.89 26.83
Agricultural Land 9.24 0.62
Fellow Land 7.33 0.49
Settlement 4.63 0.31
Water Body 7.37 0.49
Open Forest 413.33 28.01

12.3.2 Land Surface Temperature (LST)

The land surface temperature of a particular date of an area can be measured from the
satellite image by using a remote sensing technique. In this study, the land surface
temperature of Melghat forest calculated for the date of 12th April 2016, the LST of
the Melghat area is to be measured by LST model. Figure 12.4 shown in land surface
temperature in Melghat Tiger Reserve Forest.
In this Fig. 12.4, the maximum temperature is to be found 39.34 °C, and the
minimum is 180 °C. Forest fire points are plotted in this figure to find out in which

Fig. 12.4 Land surface temperature of Melghat Tiger Reserve Forest

288 N. Kayet

Fig. 12.5 Relation between forest fire occurrence and LST

temperature there is a maximum no of fire points that have occurred. The relation
between forest fire occurrence and temperature is shown in the following figure by
a simple bar graph (Fig. 12.5).

12.3.3 Criteria for Forest Fire Risk Zoning

Forest fire is one type of combustion of vegetation which mainly occurs in the forest
area. The frequent occurrence of forest fires is a significant reason for depletion
and extinction of our valuable plants and animal species. There are many natural
and anthropogenic factors that are responsible for a forest fire, here natural factors
such as elevation, slope, aspect, topographic position index, rainfall, wind speed,
temperature, DR, DS, and NDVI are used as criteria. Again forest fire is a result of
anthropogenic activity, so distance to Settlement and distance to the road are also
used to prepare forest fire risk zone mapping as anthropogenic factors. Both natural
and anthropogenic factors that are used in this study as criteria are shown in Fig. 12.6.

12.3.4 Frequency Ratio Based FFR

The results of the spatial relationship between forest fire and forest fire-related factors
using the frequency ratio model are illustrated in Table 12.3. In the table we found that
temperature, wind speed, and settlement are the highest frequency ratio value, then
other factors and slope aspects get the lowest values. So, we can say that temperature,
wind speed, and settlement factors are most responsible for a forest fire in the study
area than other factors. The forest fire risk (FFR) zoning map shown in Fig. 12.7.
12 Forest Fire Risk Assessment for Effective Geoenvironmental … 289

Fig. 12.6 All criteria maps for forest fire risk zoning mapping

12.3.5 Analytical Hierarchy Process Based FFR

The idea of multi-criteria techniques has been implemented with different uncertainty
levels of AHP. Forest fire risk zonation map by weighted mean overlay analysis using
AHP is thus obtained in GIS mode using MCDA. AHP model is actually based on
Saaty’s pairwise comparison method (Table 12.4). Here each and individual criteria
are compared as a pair, and importance between these two criteria is measured and
then put the weight as per Saaty’s comparison scale. Figure 12.8 shown the forest
fire susceptibility map by the AHP model.
290 N. Kayet

Table 12.3 Frequency ratio value for each criteria

Factors Class No. of Percentage No. of Percentage Frequency
pixels of pixels forest fire of forest fire ratio value
point points
Elevation 273–459 76,500 30.35 466,663 26.63 0.303
(m) 459–556 120,600 47.85 713,390 43.51 0.478
556–667 41,400 16.42 328,022 20.00 0.164
667–805 12,600 5 134,911 8.22 0.050
805–1114 900 0.35 26,456 1.61 0.003
Slope (°) >15 103,500 41.07 571,217 34.84 0.410
15–30 90,900 36.07 542,794 33.10 0.360
30–45 40,500 16.07 370,035 22.57 0.160
45–60 17,100 6.78 155,396 9.47 0.067
Slope aspect Flat 20,700 8.21 184,908 11.27 0.082
North 30,600 12.14 161,102 9.82 0.121
North-east 19,800 7.85 154,762 9.43 0.078
East 26,100 10.35 166,863 10.17 0.103
South-east 31,500 12.50 193,884 11.82 0.125
South 30,600 12.14 186,325 11.36 0.121
South-west 40,500 16.07 182,038 11.10 0.160
West 29,700 11.78 194,847 11.88 0.117
North-west 22,500 8.92 214,713 13.09 0.089
Topographic Canyon 54,000 21.42 376,471 22.96 0.414
position Ridges 115,200 45.71 685,056 41.78 0.307
index (TPI)
Gentle Slope 61,200 24.28 404,390 24.66 0.242
Flat 21,600 8.57 173,525 10.58 0.085
Normalized −0.563 to 0.276 153,000 60.71 486,777 29.69 0.287
differential 0.276–0.358 90,900 36.07 996,163 60.76 0.199
vegetation
index 0.358–0.830 8100 3.21 156,502 9.54 0.113
(NDVI)
Rainfall 1354.26–1606.82 119,700 47.5 621,516 37.91 0.475
(mm) 1606.82–1787.87 93,600 37.14 650,652 39.68 0.371
1787.87–1924.21 38,700 15.35 367,274 22.40 0.153
Temperature 21.24–24.26 52,200 20.71 557,364 33.99 0.207
(°C) 24.26–27.34 92,700 36.78 477,214 29.10 0.367
(continued)
12 Forest Fire Risk Assessment for Effective Geoenvironmental … 291

Table 12.3 (continued)

Factors Class No. of Percentage No. of Percentage Frequency
pixels of pixels forest fire of forest fire ratio value
point points
27.34–29.34 107,100 42.50 604,864 36.89 0.425
Wind speed 1.85–2.21 80,100 31.78 792,860 48.36 0.317
(m/s) 2.21–2.68 116,100 46.07 660,030 40.25 0.460
2.68–2.98 55,800 22.14 186,552 11.37 0.482
Distance to 0–1 18,000 7.14 180,984 11.03 0.214
settlement 1–2 54,000 21.42 381,139 23.24 0.067
(km)
2–3 61,200 24.28 385,454 23.51 0.202
3–4 55,800 22.14 291,708 17.79 0.221
4–5 30,600 12.14 183,744 11.20 0.121
5–12 32,400 12.85 216,413 13.20 0.128
Distance to 0–1 74,700 29.64 505,533 30.83 0.296
road (km) 1–2 55,800 22.14 374,338 22.83 0.221
2–3 62,100 24.64 309,287 18.86 0.246
3–4 30,600 12.14 216,808 13.22 0.121
4–5 22,500 8.92 124,243 7.57 0.089
5–12 6300 2.5 109,233 6.66 0.025

Fig. 12.7 Forest fire risk map using FR model

 292

Table 12.4 Pairwise Comparision Matrix using AHP

Elevation Slope Slope aspect TPI Temperature Rainfall Wind NDVI Road Settlement
Elevation 1.0 4 5 6 3 4 7 2 6 5
Slope 0.25 1.0 4 5 3 2 5 4 5 6
Slope Aspect 0.2 0.25 1.0 2 3 2 5 3 5 4
TPI 0.16 0.2 0.5 1.0 2 3 4 2 5 6
Temperature 0.33 0.33 0.33 0.5 1.0 4 6 2 6 5
Rainfall 0.25 0.5 0.5 0.33 0.25 1.0 4 3 5 6
Wind 0.14 0.2 0.2 0.25 0.16 0.25 1.0 4 3 2
NDVI 0.5 0.25 0.33 0.5 0.5 0.33 0.25 1.0 6 7
Road 0.16 0.2 0.2 0.2 0.16 0.2 0.33 0.16 1.0 3
Settlement 0.2 0.16 0.25 0.16 0.2 0.16 0.5 0.14 0.33 1.0
N. Kayet
12 Forest Fire Risk Assessment for Effective Geoenvironmental … 293

Fig. 12.8 Forest fire risk map using AHP model

12.3.6 Comparative Analysis Between FR and AHP Models

for FFR

FR and AHP-based fire risk pixel values are used to comparative analysis in different
forest risk classes (Fig. 12.9).

Fig. 12.9 Area of fire risk zones using FR and AHP models
294 N. Kayet

Fires in the forest have always been studied as natural phenomena, and more of the
efforts were on the way to suppress them. Not much effort was put on the inspection
and the analysis of the causes. Policy strategies were according to preventive and
suppression measures (Prasad et al. 2008) as well as, fire risk models are a great
approach for precautionary measures for the environmental protection of the forests.
Therefore, wildfires are the result of several underlying factors; model variables
were slope degree, slope aspect, altitude topographic position index, wind effect,
distance to roads, rivers and villages, normalized difference vegetation index, annual
temperature, annual rain, and land use. In the study area, forest fire susceptibility
maps were produced using the frequency ratio and Analytical Hierarchy Process
models. The findings revealed that the most important conditioning factors were
NDVI, land use, soil, and annual temperature. Therefore, preventive measures need
to be applied to ecological conditions. Prasad et al. (2008) stated that the mean annual
temperature could strongly influence on forest fires in the Deccan Plateau, India. In
another research, Motazeh (2013) indicated that based on expert choice in hardwood
Hyrcanian forests, vegetation coverage allocated the greatest weight.

12.4 Conclusion

In the current study, we find that the north, northeast, and southwest parts are in highly
sensitive zones because that area contains high temperature, unhealthy vegetation,
and the best part is that the human activity of those areas is very significant. So
the tiger reserve of Melghat forest should be placed on deep forest areas, which
is not fire porn also, human activity of those areas should be prohibited to protect
the natural environment. Again to minimize the fire activity in the Melghat forest,
it is essential to reduce the human activity and also should perform the plantation
program because the vegetation health condition is not satisfactory. These models
can be used in other cities by exchanging the variables and the weights but cannot
use the same weights and variables in different regions because forest fire in each
part of the earth has its characteristics and indices should be improved over areas
with different environmental conditions. However, more simulation results need to
compare other methods. In addition, remote-sensing data with its spatial information,
when combined with GIS and statistical models, allow fire managers and personnel
to predict ‘where and when’ forest fires will most likely. In general, the mentioned
methods can be applied to early warning fire suppression resources planning and
allocation works in the study area.

References

Chen C, Xu Z (2010) Forest ecosystem responses to environmental changes: the key regulatory role
of biogeochemical cycling, pp 210–214
12 Forest Fire Risk Assessment for Effective Geoenvironmental … 295

Choi J, Oh HJ, Lee HJ, Lee C, Lee S (2012) Combining landslide susceptibility maps obtained from
frequency ratio, logistic regression, and artificial neural network models using ASTER images
and GIS. Eng Geol 124:12–23
Cortez P, Morais ADJR (2007) A data mining approach to predict forest fires using meteorological
data
Krivtsov V, Vigy O, Legg C, Curt T, Rigolot E, Lecomte I, Pezzatti GB (2009) Fuel modelling
in terrestrial ecosystems: an overview in the context of the development of an object-orientated
database for wild fire analysis. Ecol Model 220(21):2915–2926
Mohammady M, Pourghasemi HR, Pradhan B (2012) Landslide susceptibility mapping at Golestan
Province, Iran: a comparison between frequency ratio, Dempster-Shafer, and weights-of-evidence
models. J Asian Earth Sci 61:221–236
Motazeh AG, Ashtiani EF, Baniasadi R, Choobar FM (2013) Rating and mapping fire hazard in
the hardwood Hyrcanian forests using GIS and expert choice software. Acknowledgement to
reviewers of the manuscripts submitted to Forestry Ideas in 2013, p 141
Pourghasemi HR, Beheshtirad M, Pradhan B (2016) A comparative assessment of prediction capa-
bilities of modified analytical hierarchy process (M-AHP) and Mamdani fuzzy logic models using
Netcad-GIS for forest fire susceptibility mapping. Geomat Nat Hazards Risk 7(2):861–885
Prasad VK, Badarinath KVS, Eaturu A (2008) Biophysical and anthropogenic controls of forest
fires in the Deccan Plateau, India. J Environ Manage 86(1):1–13
Rawat GS (2003) Fire risk assessment for forest fire control management in Chilla Forest Range of
Rajaji National Park, Uttaranchal, India. ITC
Roy PS (2003) Forest fire and degradation assessment using satellite remote sensing and geographic
information system. Satellite Remote sensing and GIS applications in agricultural meteorology,
p 361
Saaty TL (2008) Decision making with the analytic hierarchy process. Int J Serv Sci 1(1):83–98
Sharma LK, Kanga S, Nathawat MS, Sinha S, Pandey PC (2012) Fuzzy AHP for forest fire risk
modeling. Disast Prevent Manag Int J
Vargas LG (1990) An overview of the analytic hierarchy process and its applications. Eur J Oper
Res 48(1):2–8
Yalcin A (2008) GIS-based landslide susceptibility mapping using analytical hierarchy process
and bivariate statistics in Ardesen (Turkey): comparisons of results and confirmations. CATENA
72(1):1–12
Chapter 13
Forest Disturbance Analysis of Selected
Blocks of Midnapore Subdivision using
Digital Remote Sensing Technique

Ratnadeep Ray, Swarnali Biswas, and Ahona Bej

Abstract Change is ubiquitous in forest ecosystems. Forests undergo both season-

ality ns well as enduring escalation cycles which may vary in long term. These long-
term changes are punctuated by habitually interim disturbances from fire, insects,
disease, and harvest which strongly alter the state and functioning of the forest.
The spatio-temporal information on process-based forest loss as well as change is
essential for a wide range of applications. The disturbances over forest cover and
resulted changes alter the water and carbon cycles of forest stands as well as bang the
habitat and biodiversity of these ecosystems. To effectively understand how forest
disturbance impacts forest state and functioning, the disturbance and related effects
on forest cover is needed to be quantified at spatial scale where human management
and natural strife occur. In this present study, the spatial pattern of forest cover of the
four blocks like Garhbeta 1, Garhbeta 2, Garhbeta 3 and Salboni under Midnapore
Sadar Sub-division was analysed on temporal scale. The forested area of this region
is region is lying under the Midnapore forest division, the total area of which is
50,267.49 ha. Forest is the one of the important natural resources of this area and
the important source of rural livelihood as well as ecological sustainability. But it is
changing temporally rather is fading and maximum stress is seen onto the dense forest
and open forest. So for the restoration practice, forest disturbance indexing as well
as identification of disturbed sites and an account on forest regeneration and degen-
eration are so needful. Therefore, as per the goal of the study a Multicriteria based
Forest Disturbance Index and forest fragmentation analysis were deployed. Finally
it is so pertinent to mention that the forest covers under the blocks like Garhbeta 2
and Salboni were in a miserable state.

R. Ray (B)
Department of Geography and Environment Management, Vidyasagar University,
Midnapore, India
e-mail: ratnadiprsgis@[Link]
S. Biswas
Directorate of Science and Technology, Kolkata, India
A. Bej
Department of Geography, University of Calcutta, Kolkata, India

© The Editor(s) (if applicable) and The Author(s), under exclusive license 297
to Springer Nature Switzerland AG 2021
P. K. Shit et al. (eds.), Spatial Modeling in Forest Resources Management,
Environmental Science and Engineering,
[Link]
298 R. Ray et al.

Keywords Forest disturbance index · Fragmentation analysis · Spatio-temporal ·

Regeneration and degeneration

13.1 Introduction

Change is ubiquitous in forest ecosystems. Forests undergo both seasonality and

long-term growth cycles, which can range in span of 50–500 years or more (Waring
and Running 2007). Such long-term changes are punctuated by often short-term
fire, pest, disease, and harvest disturbances, which strongly alter the forest state
and function (He and Mlandenoff 1999). Both climate change and increasing global
demand for wood and fiber products are likely to cause forest destruction levels to
rise (Kurz et al. 2008; Nepstad et al. 2008). Forest status of a region is closely allied
to the annual amount of precipitation. Continuous and increasing exhaustive climate
change could have a explicit impact on the status of natural vegetation, particularly
water availability (Malik et al. 2019a, b).Such disturbance changes will alter the forest
stands’ water and carbon cycle as well as affect the environment and biodiversity of
those habitats (Lindenmayer et al. 2006; Gardner et al. 2009). Forest destruction is
now recognized as a significant driver of non-fossil-fuel-related terrestrial fluxes into
the atmosphere with respect to the carbon cycle (Running 2008; Amiro et al. 2010).
Unplanned deforestation in a very precarious way hampers both the human being
and the lives of the other animals (UN Report 2019). In addition, monitoring and
assessing the vegetation status is the most important aspect of environmental health
assessment and global climate change alleviation (Alberdi et al. 2019; Bjorkman
et al. 2019; UN Environmental Report 2019).
In order to understand effectively how forest disturbance affects the condition
and work of the environment, disturbance levels need to be quantified in the spatial
grain where human activity and natural disturbances occur; usually less than 10 ha
(Miller 1978; Cohen et al. 2002; Kuemmerle et al. 2007; Frolking et al. 2009). In
addition, disturbances need to be quantified at a time and spatial scale appropriate
to the affected processes (e.g. globally, annually), and over a time-limit appropriate
to the development of baselines applicable to forest policy initiatives (historically,
at least as far back as the 1990s) (Böttcher et al. 2008; Masek et al. 2008; Kennedy
et al. 2012).
Forest is the one of the important natural resources of the area under study and
the important source of rural livelihood as well as ecological sustainability. Besides
these the majesticity of forest of this area has made it unique in the map of West
Bengal. But it is changing day by day rather is deteriorating and maximum stress
is seen onto the dense forest and open forest. West Bengal (WB) has lost 1612
km2 of classified forest during last 100 years and within this, Paschim Medinipur
stands third among all the districts of WB in terms of forest degradation with a
loss of 110 km2 of classified forest. Degradation of forest cover, on the marginal
laterite regions with low potential agricultural productivity, is often linked to the soil
characteristics and associated geomorphic processes on rarh lateritic regions of West
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 299

Bengal (Shit and Pati 2012; Bhunia et al. 2012). State Forest Report of 2011–12,
shows an achievement in growing forest cover due to plantation activity in Paschim
Medinipur District but it is well accompanied by destruction at the local level as
19 km2 of forested track was degraded in 2017. The Tropical dry deciduous forest
in WB co-exists with human settlement where anthropogenic interactions with the
forest have aggravated the conflict between man and forest (Nagendra et al. 2009;
Mahapatra and Tiwari 2005). In order to keep the forest sustainable, recognizing
forest gaps within a specific area can be an important approach in order to provide
adequate strategic management (Pal et al. 2018). Considering in this respect the gap
as disturbed area forest disturbance indexing as well as identification of possible
forest disturbance zone is so important.

13.2 About the Study Area

In the present study, the study area is consisted of 4 blocks like-Garhbeta 1, Garhbeta
2, Garhbeta 3 and Salboni of Midnapore Sadar Subdivision. It is situated within the
latitude and longitudinal extensions of 22°28 N to 22°52 N and 87°5 E to 87°31
E. The total area under study is 1624.81 km2 (Fig. 13.1).

Fig. 13.1 Location of the study area

300 R. Ray et al.

The general ground configuration is having gentle slope towards east and the
slope of this region is ranging from 2° to 25°. Geomorphologically this area can be
divided into 4 categories like Deep buried pediment, Flood plain deposits, Valley fill
deposits, Deep to moderately buried pediments with laterite capping. Furthermore,
it is the flank of Chotanagpur erosional plateau area. The relative relief of the region
is varying from 5 to 28.92 m. Silaboti is the main river of this region and is the main
river to geomorphologically modify this region through gully erosion and ample
numbers of 1st order streamlets are seen as rill Chanel. The forested area of this
region is region is lying under the Midnapore forest division, the total area of which
is 50,267.49 ha. The area under reserved forest is the highest out of the other forest
divisions of this district and that is 3814.05 ha. And the area under protected forest
is 43,715.23 ha. Typically, the area occupied by sal, open scrub and plantation under
this division are 19,677 ha, 9620 ha and 4955 ha respectively. Champion and Seth
(1968) classified the forest types of Paschim.
Medinipur as Topical dry deciduous where dry peninsular Sal (Shorearobusta)
is the dominant species. Undergrowth is mostly formed during post-monsoon
season where Kurchi (Holarrhaenaantidesenterica), Kendu (Dispyrosmelanoxylan)
and Satamul (Aparagusreasemossa) dominate.

13.3 Materials Used

Three cloud-free Landsat digital data of TM, ETM+ and OLI/TIRS sensors for the
years of 1991, 2009, and 2018 covering the study area were downloaded freely from
the U.S. Geological Survey’s (USGS) Earth explorer website ([Link]
[Link]/).
The Landsat scenes were chosen for this study due to their affordability, avail-
ability, and medium to high spatial resolution. Details about the data are given in
Table 13.1. These three Landsat digital data are level-one terrain-corrected (L1T)

Table 13.1 Details of satellite data used

Satellite and Date Path/Row Nos. of Spatial resolution Radiometric
sensor bands resolution
Landsat 4 18.02.1991 139/44,45 7 Optical bands—30 m, 8 bit
(TM) Thermal band—120 m
Landsat 7 08.11.2000, 139/44,45 9 Optical bands—30 m, 8 bit
(ETM+) 14.11.2009 2 Thermal
bands—60 m (each)
Panchromatic—15 m
Landsat 8 08.01.2018 139/44,45 11 Optical bands—30 m, 16 bit
(OLI) 2 Thermal
bands—100 m (each)
Panchromatic—15 m
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 301

in WGS84 geodetic datum, Universal Transverse Mercator map projection (UTM,

Zone 45 N), and north-up image orientation. Besides SOI toposheets and SRTM
DEM (30 m) have been used for the completion of the study.

13.4 Methodology

13.4.1 Atmospheric Correction

Remote Sensing satellite digital data acquired at temporal span provide a potentially
ideal source for detecting change and analyzing trends. In view of the fact that
multi-temporal images, acquired by different sensors under variable atmospheric
conditions, solar illumination and view angles, atmospheric correction is obligatory
to remove radiometric distortions and make the images analogous using the retrieved
true reflectance values (Mahmoud et al. 2008).
The parameters used in different environmental condition detection algorithms,
it require physical units, such as at-sensor radiance or top-of-atmosphere (TOA)
reflectance, rather than the raw quantized calibrated pixel value (DN). TOA
reflectance can be obtained from the quantized calibrated pixel value, as given by
(Chander et al. 2009):
 
ρλ = π · Lλ · d 2 /E SU N λCosθ (13.1)

where ρλis the TOA reflectance of wavelength λ (unit less), d is the earth-sun distance
(astronomical units), E SU N λ is mean exoatmospheric solar irradiance (W/(m2 μm)),
θ is the solar zenith angle (degrees), and Lλ is the spectral radiance at wavelength
λ at the sensor’s aperture [W/(m2 srμm)]. Lλ can be obtained from the quantized
calibrated pixel value also as given by (Chander et al. 2009):

Lλ = (L max − L min /QC AL max − QC AL min ) ∗ (D N − QC AL min ) + L min

(13.2)

where L max is the spectral at-sensor radiance that is scaled to QC AL max [W/(m2 sr
μm)], L min is spectral at-sensor radiance that is scaled to QC AL min [W/(m2 sr μm)],
QC AL max is the maximum quantized calibrated pixel value corresponding to L max
(DN), QC AL min is the minimum quantized calibrated pixel value corresponding to
L min (DN), and QC AL is the quantized calibrated pixel value (DN). The parameters
in Eqs. (13.1) and (13.2) can be read from the header files of the ALI, TM, and ETM+
datasets or be retrieved from the USGS website.
302 R. Ray et al.

13.4.2 Forest Cover Mapping

Satellite images have been an important basis for vegetation mapping and moni-
toring and understanding of eco-system functions, primarily through relationships
between reflectance and vegetation structure and composition. Mapping is a method
of portraying nature and the classification permits the mapper to approximate the
true condition as clearly as possible. Although maps show objects with respect to
attributes, their principal purpose is to depict object in terms of their relative loca-
tion (Thakker et al. 1999). A good and useful mapping implementation requires
large amount of information that comes from various sources like satellite images,
ground truth etc. (Behera 2000). This study discusses a vegetation mapping method-
ology using Multi-criteria analysis technique. It agrees with the spectral limitation
of remote sensing but argues that their predictive mapping power can be used more
effectively using different floristic bio-physical parameters. This approach outlined
aims to relate the reflectance information contained in multi-spectral imagery and
spectral indices to traditionally accepted ecological classifications of the forests.
Since last two decades remote sensing is being used as an effective tool for forest
vegetation assessment in the country. National Remote sensing agency (Anon. 1983)
carried out nationwide forest cover mapping using satellite images from 1972–75
and 1980–82 on 1:1 million scale. National level forest cover assessment is now
being done periodically by Forest Survey of India (FSI) using visual and digital
interpretation techniques.
The advent of digital classification is that it incorporates spatial adjacency in addi-
tion to spectral signature (Argialas and Harlow 1990; Moller-Jensen 1990; Wickman
and Norton 1994).The traditional digital classification implies a loss in the internal
variability of the cover types considering the spatial mixtures of land covers (Ray
and Mondal 2014).In this respect using the vegetation bio-physical variables as a
criteria, calculated from different spectral indices an overlay analysis can classify
the forest cover of region.

13.4.3 Shadow Index (SI)

The crown arrangement in the forest stand leads to shadow pattern affecting the
spectral responses. The young and even aged stands have low Shadow Index (SI)
compared to the mature natural forest stands. The later forest stands show flat and
low spectral axis in comparison with open [Link] has been calculate using equation
(Jennings et al. 1999).

S.I. = {(QC AL max − G R E E N ) ∗ (QC AL max − R E D)} (13.3)
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 303

13.4.4 Bare Soil Index (BI)

The bare soil areas, fallow lands, vegetation with marked background response and
enhanced using this index. Similar to the concept of AVI the bare soil index (BI)
is a normalized index of the difference sums of two separating the vegetation with
different background viz. completely bare, sparse canopy and dense canopy etc. BI
has been calculated using equation (Azizi et al. 2008).

B.I. = {(N I R + G R E E N ) − R E D} {(N I R + G R E E N ) + R E D} (13.4)

The value of this index is varying between 0 and 1, where the value towards 0
indicates the vegetation exploitation whereas value towards 1 indicates the vegetation
abundance.

13.4.5 Modified Difference Vegetation Index (MAVI)

In this present study an object oriented enhancement algorithm has been designed
using mathematical operators, which is supervised in nature and expressing the char-
acteristics of plant chlorophyll ‘a’. Two considerations have been taken in this respect
like, pixel value having chlorophyll influence will be greater in near-infrared band
than red band and pixel value having no chlorophyll influence will be greater in red
band than near-infrared band. On the basis of these considerations following super-
vised enhancement algorithm: Modified Advance Vegetation Index (MAVI) has been
applied (Ray et al. 2013). That can be calculated as

M AV I = [{(ρ N I R + L) ∗ (QC AL max − ρ R E D)} ∗ (ρ N I R − R E D)]1/3

(13.5)

(where, ρ R E D = Reflectance value of Red band of respective Sensor, ρ N I R =

Reflectance value of Near-infrared band of respective Sensor, L = Coefficient, varies
with the vegetation cover, Here L is the relational slope of NIR and RED).
It differs from NDVI in the perspective of physiognomic vegetation classes though
there is almost positive relation-ship between these two. It is more sensitive than
NDVI as it is using the power degree of Infrared response.

13.4.6 Vegetation Density (VD)

To calculate the Vegetation Density (VD), MAVI and BI are synthesized using PCA
method as there is high correlation of negative between them. The PC 1 having high
variance has been considered as Vegetation density (VD) criteria.
304 R. Ray et al.

13.4.7 Scaled Shadow Index (SSI)

Alike VD, SSI has been developed by synthesizing MAVI and SI through PCA
method as being of highly correlated of positive. The PC 1 having high variance has
been considered as Scaled Shadow Index (SSI) criteria.
After studying the above vegetation indices, multi criteria based weightage anal-
ysis using Fuzzy AHP method using geometric mean (Buckley et al. 1985) has been
applied to map the forest cover of the study area.
Fuzzy AHP (FAHP) is a artificial extension of classical AHP method, which
considers the statistical fuzziness (uncertainty or insufficient information) of the
decision makers. Fuzzy multiple attribute decision-making methods have been devel-
oped due to the elusiveness in computing the relative importance of attributes and
the performance ratings of alternatives with respect to attributes (Chang 2011).
There are varieties of reasons that may induce elusiveness: unquantifiable informa-
tion, incomplete information, exclusive information. Conventional multiple attribute
decision-making methods cannot overcome these problems successfully (Oguztimur
2011). Basically, Fuzzy-AHP method represents the embellishment of a standard
AHP method into fuzzy domain by using fuzzy numbers for calculating instead of
real numbers (Petkovic et al. 2012). Fuzzy AHP is of equal significance in the same
hierarchy for each pair of factors (Das et al. 2019b).
In this process primarily conventional analytical Hierarchy Process (AHP) has
been applied through creating pair wise comparison matrix (Table 13.2). This pair
wise comparison matrix has been created with the help of scale of relative importance
which are crisp numeric values (Table 13.2). Fuzzification and fuzzy membership
functions are the two most important approaches in fuzzy system. In fuzzification
linguistic terms are converted to fuzzy membership. Membership functions are of
different types like triangulated, trapezoidal, bell shaped etc. In this study triangulated
membership function has been conceived. The fuzzy values are generally represented
by:

μ Ã(x) = Ã = (l, m, u) (13.6)

These (l, m, u) are known together as fuzzy members and these are lower, middle
and upper ends of the triangular membership functions in the X axis.

Table 13.2 Pair wise comparison matrix with crisp values

MAVI BI SSI VD
MAVI 1 1 1 4 5 6 4 5 6 6 7 8
BI 1/6 1/5 1/4 1 1 1 4 5 6 4 5 6
SSI 1/6 1/5 1/4 1/6 1/5 1/4 1 1 1 3 4 5
VD 1/8 1/7 1/6 1/6 1/5 1/4 1/5 1/4 1/3 1 1 1
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 305

Table 13.3 Fuzzy geometric mean

Geometric mean
MAVI 3.130169 3.637136 4.119534
BI 1.277886 1.495349 1.732051
SSI 0.539951 0.632456 2.364354
VD 0.255327 0.290715 0.344998

5.203334 6.055655 8.560937
Reciprocal 0.192184 0.165135 0.11681

In AHP matrix crisp numeric values are converted to the fuzzy numbers following
Eq. (13.6). Whereas the reciprocal values are converted to fuzzy numbers (Table 13.2)
using:

1/ Ã = 1/(l, m, u) = (1/u, 1/m, 1/l) (13.7)

Hence fuzzified pair wise comparison matrix can be calculated (Table 13.2).
After this fuzzy geometric mean (r̂i) has been calculated (Buckley et al. 1985) to
explore the criteria weight (Table 13.3). Fuzzy geometric mean (r̂i) can be calculated
as:

Ã1 ⊗ Ã2 = (l1 , m 1 , u 1 ) ⊗ (l2 , m 2 , u 2 ) (13.8)

Finally the fuzzy weight can be calculated as:

 
ŵ j = r̂i ⊗ r̂1 ⊕ r̂2 ⊕ · · · ⊕ r̂n − 1 (13.9)

After getting the fuzzy weights per criteria (Table 13.4), crisp numerical values
can be calculated as weight (ŵ j) using defuzzification, which is known as Center of
Area (COA) (Table 13.5) using,

w j = (l + m + u)/3 (13.10)

Table 13.4 Fuzzy weight

Ŵ W Wij
MAVI 0.365634 0.600618 0.791711 0.585987 0.540399
BI 0.149269 0.246934 0.332873 0.243026 0.224119
SSI 0.063072 0.10444 0.454392 0.207301 0.191174
VD 0.029825 0.048007 0.066303 0.048045 0.044307

1.08436 1
306 R. Ray et al.

Table 13.5 Criteria weight

Geometric mean Ŵ
MAVI 3.130169 3.637136 4.119534 0.365634 0.600618 0.791711
BI 1.277886 1.495349 1.732051 0.149269 0.246934 0.332873
SSI 0.539951 0.632456 2.364354 0.063072 0.10444 0.454392
VD 0.255327 0.290715 0.344998 0.029825 0.048007 0.066303
Reciprocal 0.192184 0.165135 0.11681

All the indexed raster like MAVI, VD, SSI and BI have been assigned respective
theme weights (w j ) and class score. The individual theme weights are multiplied by
its respective class scores and then all the raster thematic layers are aggregated using
a linear combination to get forest cover map (Rikimaru et al. 2002) as:
 
Criteria ∗ w j (13.11)

The final cumulative map has been reclassified into five categories of forest cover
types viz; ‘very dense to dense’, ‘dense to open’, ‘open to scrubs’, ‘scrubs to bare’
and ‘bare land’. This forest cover mapping as well as categorization has been done
on temporal basis using digital satellite dataset of 1991, 2000, 2009 and 2018 so that
a temporal dynamics can be visualized.

13.4.8 Identification of Forest Cover Dynamics

To explore the forest cover dynamics as well as the changing scenario all the temporal
reclassified forest cover maps have been overlaid using Boolean OR operator on
one to one basis i.e. 1991–2000, 2000–2009, 2009–2018 and 1991–2018. The class
conversions have been registered as ‘regeneration’, ‘degeneration’ and ‘unchanged’.
From the attribute table of each pair of overlaid forest cover maps i.e. 1991–2000,
2000–2009, 2009–2018 and 1991–2018, combinations of ‘very dense to dense’ and
other class category have been named as ‘degeneration’, combinations of ‘bare
lands’ and other class category as have been named as ‘regeneration’ and lastly
the combinations of identical class categories are named as ‘unchanged’.
For the numerical depiction of the overall changing scenario, an index has been
formulated as Disturbance Index (DI). DI is the function of the area under no forest
cover degeneration and area with forest cover degeneration. DI can be calculated as:

D I = (Ar ea under Regeneration + Ar ea under Degeneration)

/Ar ea under Regeneration (13.12)

Higher the DI, lesser will be the distortions in the forest cover and vice versa.
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 307

13.4.9 Forest Fragmentation Analysis

Forest fragmentation is disintegration of contiguous forest cover to large numbers of

forest patches on temporal basis as consequences of the edge effects (Rahman et al.
2016). In this present study forest fragmentation has been studied using one of the
pattern metric tools like Area Weighted Mean Patch Fractal Dimension (AWMPFD)
Index. It is analogous to the Mean Patch Fractal Dimension (MPFD) Index. Only
difference is the use of area as an weighting parameter to each patch. Higher the
value of AWMPFD lower will be the fragmentation and vice versa. The value ranges
between 1 and 2. It can be calculated as:
  
 m  n
2 ln 0.25Pi j ai j
P=   (13.13)
i=1 i=1
ln ai j A

where, Pi j is patch number, ai j is patch area and A is total landscape area.

[Link] Multi-distance Spatial Cluster Analysis (Ripley’s K Function)

Ripley’s K function definition is a tool to portray the spatial structure of a point or

polygon patterns by graph. It determines the values associated with features, exhibit
statistically significant clustering or dispersion over a range of distances. It has a wide
applicability in vegetation studies. The tool gives the output as a table with fields
expected and observed K values, respectively. Mathematically, the Multi-Distance
Spatial Cluster Analysis tool uses a common transformation of Ripley’s k-function,
which can also be said as L(d). With this L(d) transformed Distance values the
Expected K values will always be matched. The L(d) transformation can be done as:

 
 n 
n
L(d) =  A K i, j /π n(n − 1) (13.14)
i=1 j=1,i= j

Larger the observed K value than the expected K value indicates the clustered
distribution than the random, whereas, smaller the observed K value than the expected
K value, the distribution is more dispersed than random distribution at that distance.
Furthermore, larger the observed K value is than the upper confidence envelope
(HiConfEnv) value, the spatial clustering is statistically significant for that distance
and when the observed K value is smaller than the lower confidence envelope
(LwConfEnv) value; there will be a statistically significant spatial dispersion for
that distance (Sayer et al. 2013).
308 R. Ray et al.

[Link] Optimized Hot Spot (Getis-Ord Gi*) and LISA (Local

Indicators of Spatial Association) Analysis

The spatio-temporal hotspot detection was based on the use of Getis-Ord Gi* statistic
is a typical tool to detect spatio-temporal hot-spot of any spatial phenomenon. It
statistically characterizes and captures significant spatial clusters as hotspots and cold
spots using Gi-Bin values. In this present study, it has been devised to explore the
spatial concentration of DI values in significant manner. It is considered as a useful
tool for supporting sustainable management strategies (Zhu and Newsam 2016).
Features with Gi-Bin values of ±3 were statistically significant at the 99% confidence
level; the features with Gi-Bin value of 0 was not statistically significant, and the
features with Gi-Bin values of ±2 were statistically significant at 95% confidence
level and features with Gi-Bin values of ±1 were statistically significant at 90%
confidence level (Wang 2010).
Alike the Optimized Hot Spot analysis, Local Indicators of Spatial Association or
Autocorrelation (LISA) analysis was employed to characterize the spatial arrange-
ment of Disturbance Index per spatial unit over time to generate cluster maps for
identifying the existence of hot spots. It is such a spatial statistical technique which
gives an indication of extent for significant existence of homogeneous clusters around
a particular observation and the sum of local indicators of spatial association for all
observations is proportional to a global indicator (Anselin 1995).
Local Moran I is a local indicator of spatial autocorrelation (LISA) is a decompo-
sition of global indicators, such as Moran’s I and shows the level of spatial autocorre-
lation at various individual locations within an area (Anselin 1995). LISA calculates
an index value and Z-score for a feature. A high positive Z score indicates the asso-
ciation of a feature with similar value either above or below mean. On the contrary
a high negative Z-score indicates the association of a feature with dissimilar values
(Zhang et al. 2008). The LISA can be measured as:
⎛  ⎞
 
Ii = ⎝ z i z i2 /n wi j z ⎠ (13.15)
i j

Spatial clusters consist of two categories: (i) high-high clusters indicate high
values are surrounded by high values; (ii) low-low clusters indicate clustering of low
values are surrounded by low values; (iii) high-low clusters indicate clustering of high
values are surrounded by low values and (iv) low-high clusters indicate clustering of
low values are surrounded by high values. Besides, “non-significance” is also a type
of cluster, indicating no significant local spatial autocorrelation.
The distribution of spatial clusters and outliers can be shown in a Moran scatter
plot which has four quadrants (Fig. 13.6b). The x-axis of the scatter plot represents
the standardized values (area) of the geographical object and the y-axis measures the
mean standardized values of neighboring objects. The upper-right quadrant of the
scatter plot contains the cases where the geographical objects and their neighbors
have high values. These so-called high-high (HH) situations are associated with
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 309

clusters. The lower-left quadrant shows cases where the geographical objects and
their neighbors are relatively small (i.e. a low-low or LL situation) which also signifies
clustering. The lower-right quadrant contains the high-low (HL) cases while low-high
(LH) cases are shown in the upper-left quadrant (Anselin 1995, 2003, 2005). The two
quadrants having LH and HL represent outliers. HH cases are referred to as grouped
hot spots and LL cases as grouped cold spots, whereas HL and LH are referred to as
individual hot spots and cold spots, respectively (Anselin 2005). The HH, HL, LH
and LL cases can be mapped to better understand the spatial distribution of hot and
cold spots and to see where clustering occurs.
In this present study the calculations of local spatial autocorrelation has been
performed in GeoDa platform (Anselin et al. 2006) at a significance filter of 0.01
and permutation level of 9999.

[Link] Identification of Forest Disturbance Potential Zone

In this present study, the word forest disturbance is mainly mean to say the spatial
distortions in forest cover as the function of spatial effects of different parameters like
settlement position, river, canals and road, physiographic nature in terms of relative
relief, slope etc.
Irrespective of AHP and FAHP, to calculate the weights of each parameter for
identifying the Forest Disturbance Potential Zones (FDPZ), ‘Rank Sum method’
under SMARTER (The Simple Multi Attribute Rating Technique exploiting rank)
technique as suggested by Edwards and Barron. Using the principal of SMARTER
technique the decision-makers arrange the criteria (Ci ) into an importance order like
C1 ≥ C2 ≥ C3… ≥ Cn. and assigns ‘true’ weights according to the Rank Order
Distribution.
The criterion having mutual comparison pair wise can be synthesized by AHP
or FAHP (Saaty 2008 and Kahraman et al. 2003). But in case of FDPZ the selected
criterion cannot be compared pair wise mutually so that ‘Rank Sum method’ under
SMARTER technique has been followed. The weight per criteria (Table 13.6) using
this respective method has been calculated as:
 
wj = n − rj + 1 (n − rk + 1) (13.16)

where, w j is the normalized weight for the jth criteria, n is the number of criteria
under consideration (k = 1, 2 …n) and r j is the rank position of the criterion. Each
criterion is weighted (n − r j + 1) and then normalized by the sum of all weights, i.e.

(n − rk + 1).
310 R. Ray et al.

Table 13.6 Rank sum method to calculate FDPZ

Rk Pi Rj n − Rj + 1 n − Rk + 1 Wij
1 Can 2 5 6 0.238
2 River 6 1 5 0.047
3 Sett 5 2 4 0.095
4 Road 4 3 3 0.142
5 RR 3 4 2 0.190
6 Slp 1 6 1 0.285

n=6 21 1

13.5 Result and Discussion

13.5.1 Forest Cover Dynamicity

The forest cover maps have been prepared from the Landsat digital data for the years
of 1991, 2000, 2009 and 2018 using Multicriteria Fuzzy AHP technique (Fig. 13.2)
and typical forest cover classes have been summerised in Table 13.7.
These data reveal that in 1991, about 9.33% (151.24 km2 ) area of the study area
was under bare surface area, 22.08% (357.88 km2 ) under scrubs to bare surface area,
30.27% (490.46 km2 ) under open to scrub area, 24.15% (391.39 km2 ) under dense
to open and 10.51% (170.32 km2 ) under very dense to dense forest. During 2000
the area under these land categories was found about 6.57% (106.48 km2 ) under
bare surface area, 28.50% (461.82 km2 ) under scrubs to bare surface area, 36.66%
(594.07 km2 ) under open to scrub area, 22.30% (361.41 km2 ) under dense to open
forest and 5.49% (89.01 km2 ) under very dense to dense forest. During 2009 the area
under these land categories was found about 16.85% (273.11 km2 ) under bare surface
area, 21.19% (344.38 km2 ) under scrub to bare, 22.67% (367.34 km2 ) under open to
scrub land, 19.68% (319 km2 ) under dense to open forest and 19.19% (2011.05 km2 )
under very dense to dense forest area and lastly during 2019 the area under these
land categories was found about 5.19% (84.24 km2 ) under bare surface area, 9.78%
(158.48 km2 ) under scrub to bare, 20.28% (328.71 km2 ) under open to scrub land,
48.49% (785.74 km2 ) under dense to open forest and 16.23% (262.97 km2 ) under
very dense to dense forest area (Table 13.9).
Post-classification comparison method, which is the most common approach in
change detection as well as the reconfiguration analysis (Araya 2009; Miller et al.
1998; Zhou et al. 2004), which has been applied in this study. The comparisons
between the areal accounts of typical forest cover classes from the year of 1991–
2018 reveals both the positive and negative changes in forest cover patterns. From
Table 13.9, this has been revealed that within the 27 years span, the forest classes
like open to scrub, scrubs to bare and bare land has decreased significantly by 9.98%,
12.30% and 4.13% respectively. On the other hand, the classes like very dense to
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 311

Fig. 13.2 Temporal forest cover maps of the study area

dense and dense to open forest area is seen to be increased by 5.71% and 4.34%
respectively.
312 R. Ray et al.

Table 13.7 Temporal forest cover area (km2 /%)

Classes 1991 % 2000 % 2009 % 2018 % 1991–2018
(%)
Very dense 170.33 10.51 89.02 5.49 311.05 19.20 262.97 16.23 5.72
to Dense
Dense to 391.39 24.16 361.41 22.31 319.01 19.69 785.75 48.50 24.34
open
Open to 490.46 30.27 594.07 36.67 367.34 22.67 328.71 20.29 −9.98
scrubs
Scrubs to 357.89 22.09 461.82 28.50 343.39 21.19 158.49 9.78 −12.31
bare
Bare 151.24 9.33 106.48 6.57 273.11 16.86 84.25 5.20 −4.13

In this present study the areal database registered in Table 13.8 the rate of change
in area for each forest cover classes has been calculated using the following formula:

Ci = {(Aei − Bei )/Bei } ∗ (1/t) ∗ 100% (13.17)

where, Ci is the rate of change in the forest area, Aei and Bei are the area under
respective forest cover at the ending and beginning year of the span of study respective
and t is the span of study.
Within the overall span of the study (1991–2018), the maximum areal gain is seen
for ‘dense to open forest’ (24.34%) at the rate of 3.73% followed by very dense to
dense forest (5.71%) at the rate of 2.05%. On the other hand, the maximum areal
loss is seen for the scrubs to bare (12.30%) at the rate of 2.06% followed by open to
scrubs (9.98%) and bare land (4.13%) at the rate of 1.22% and 1.64% respectively.
Forest cover dynamics can be best explained by forest cover fragmentation.
Conversion from very dense to dense to the other category will obviously show
the high fragmentation and vice versa. The database of metric analysis for every

Table 13.8 Rate of change in forest cover

Classes 1991 % 2000 % 2009 % 2018 % 1991–2018
(%)
Very dense 170.33 10.51 89.02 5.49 311.05 19.20 262.97 16.23 5.72
to Dense
Dense to 391.39 24.16 361.41 22.31 319.01 19.69 785.75 48.50 24.34
open
Open to 490.46 30.27 594.07 36.67 367.34 22.67 328.71 20.29 −9.98
scrubs
Scrubs to 357.89 22.09 461.82 28.50 343.39 21.19 158.49 9.78 −12.31
bare
Bare 151.24 9.33 106.48 6.57 273.11 16.86 84.25 5.20 −4.13
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 313

Fig. 13.3 Area weighted mean patch fractal dimension map of forest cover for the span of 27 years

square blocks (2.5 × 2.5 km2 ) polygon covering the entire area under study. The
thematic maps having the informations about AWMPFD depict the patterns of forest
cover on temporal basis. From Fig. 13.3, an increasing trend in fragmentation can
be demonstrated within the total span of the study. The happening of fragmentation
is seen to be very distinct in blocks like Garbeta II and Salboni in 2018 rather than
that of 1991. On the other hand in 2018, clustering of forest cover has been seen to
be increased very distinctively in Garbeta I rather than in 1991.

13.5.2 Status of Forest Regeneration and Degeneration

In any region forest cover dynamics is the function of area under forest cover regener-
ation and degeneration. These two phenomenon are fundamentally the consequences
of physical and anthropogenic impacts. Within the 27 years span of study i.e. from
1991 to 2019 the degenerated area was 276.46 km2 (17.06%), regenerated area was
377.35 km2 (23.29%) and the unchanged area was 82.52 km2 (11.26%) (Fig. 13.6).
So it is very pertinent that in this area under study, forest cover regeneration is
overriding within the span of 27 years (Table 13.9).
Though forest cover regeneration was so prominent in the overall temporal span,
there was a notable consistent declining state of regenerated area from 1991 to 2000
(565.48 km2 ) towards 2009–2018 (372.86 km2 ) followed by 2000–2009 (384.56
km2 ) as per areal accounts have been registered in Table 13.10. For the three sets of
spans (1991–2000, 2000–2009 and 2009–2018), in Garbeta 1 and Garbeta 3 there
314 R. Ray et al.

were a decreasing trend in regeneration coverage. Whereas in Garbeta 2 and Salboni

regenerated area is seen to be increased within the span of 2009–2018 i.e. 108.27
km2 and 92.04 km2 respectively which were 70.10 km2 and 71.60 km2 respectively
in the span of 2000–2009 (Table 13.9).
In the context of this overall forest cover dynamics within 27 years span Garbeta 1
is experiencing maximum areal extent under forest cover regeneration (197.55 km2 )
followed by Salboni (69.70 km2 ), Garbeta 3 (57.50 km2 ), and Garbeta 2 (52.60 km2 ).
On the other hand maximum area under forest cover degeneration is seen in Salboni
(91.80 km2 ) followed by Garbeta 3 (69.50 km2 ), Garbeta 1 (66.56 km2 ) and Garbeta
2 (48.60 km2 ).
However, Repley’s K function or L(d) transformation can be a wise estimate to
in interpret the nature of regeneration and degeneration in accordance to the relation
between the clustered and dispersed factors. This function truly calculates the statisti-
cally significant clustering and dispersion at smaller and larger distances respectively.
For the forest cover regeneration and degeneration area over the span of 27 years.
The resultant ‘K’ function for the regenerations showed that the observed spatial
pattern started and finished with the statistically significant dispersion at largest
distance in smaller space. On the other hand there was a statistically significant
clustering at smaller distance in larger space (Fig. 13.4). In contrast for degeneration,
statistically observed spatial patterns had statistically significant clustering at smaller
distance between the clustered patterns and dispersed pattern (Fig. 13.5).
Forest disturbance index can be an ideal way to depict the forest cover status. DI
indicates the nature of degeneration. Higher the DI value lower will be the distur-
bance and vice versa. In this respect within the overall span of the study Garbeta 3 is
experiencing the maximum DI (1.10) followed by Salboni (1.35), Garbeta 2 (1.71)
and Garbeta 1 (4.16) (Table 13.10). But as per the rendered z-score, spatial distri-
bution forest disturbance (DI) is mostly prevalent in South and South-West part of
Salboni block having the value ranges of ≥2.0 and in the Garbeta 1, there was a low
to no disturbance regime can be identified having the value of ≤−2.0 (Fig. 13.6a).
Such clustering having those ranges of values is also persistent in nature over the
other blocks under study. Gi-Bin values of Optimized Hot Spot analysis coupled with
the Z-score rendering is also proving the existence of such concentration of forest
disturbance as significant hot spot and cold spot clusters which have the spatial extent
of 573.32 km2 and 429.94 km2 respectively (Fig. 13.6b).
Furthermore, a spatial auto correlative analysis of the spatial distribution of DI hot
spot in local scale is very pertinent for the betterment in environmental management
and sustainability. In this respect LISA or cluster and outlier analysis is improvising
to describe the clustering nature of disturbance attributes. The Global Moran’s I value
of 0.578 indicates that the DI data is positively spatially auto-correlated regardless
of neighborhood definition (Fig. 13.7b). The DI database calculated for the span
of 27 years has been classified and mapped in terms of significant clusters of hot
spots (HH and HL) and cold spots (LL and LH). The cluster association like HH and
LL suggests clustering of similar values (positive spatial auto-correlation), whereas
the cluster association like HL and LH indicate spatial outliers (negative spatial
correlation). The LISA map (Fig. 13.7a) describes the situation of the prominent
Table 13.9 Areal accounts on forest cover regeneration and degeneration
Spans 1991–2000 2000–2009 2009–2018 1991–2018
Blocks Regeneration Degeneration Regeneration Degeneration Regeneration Degeneration Regeneration Degeneration
Garbeta 1 159.44 61.68 134.54 86.10 107.94 101.02 197.55 66.56
Garbeta 2 132.81 110.92 70.10 69.50 108.27 103.23 52.60 48.60
Garbeta 3 125.01 68.00 108.32 111.31 64.61 77.23 57.50 69.50
Salboni 148.23 128.27 71.60 45.90 92.04 49.69 69.70 91.80

565.48 368.87 384.56 312.81 372.86 331.17 377.35 276.46
% 34.90 22.77 23.74 19.31 23.01 20.44 23.29 17.06
13 Forest Disturbance Analysis of Selected Blocks of Midnapore …
315
316 R. Ray et al.

Table 13.10 Accounts on block wise Forest Disturbance Index

Block name 1991–2000 2000–2009 2009–2018 1991–2018
Garbeta 1 4.55 1.93 1.68 4.15
Garbeta 2 2.37 1.65 1.88 1.70
Garbeta 3 3.46 1.82 1.56 1.10
Salboni 2.46 2.73 3.70 1.34

Fig. 13.4 ‘K’ function plot of degenerated patches

Fig. 13.5 ‘K’ function plot of regenerated patches

patterns of HH and LL cluster having the concentration in Salboni and Garbeta 1

blocks respectively. These two blocks are have momentous contribution to the global
Moran’s I statistic.
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 317

Fig. 13.6 a and b Distribution of DI and consequent Hot Spots

Fig. 13.7 a and b LISA and Moran’s I plot of DI showing spatial clusters and outlier positions

13.5.3 Forest Disturbance Potential Zonation

Potential zonation of any functional environmental happening like ground water, soil
loss, landslide [Link] an important function of regional planning and management. So
that, a situational sustainability can be achieved. In this context forest disturbance
318 R. Ray et al.

potentiality zonation (FDPZ) is an important issue to sustainable management of a

forested environment as well as socio-environmental ecology.
In the present study the total area has been classified in three FDP zones like
‘High to Moderate’ (413.89 km2 ), ‘Moderate to Low’(440.62 km2 ), ‘Low to Non
Disturbed’(761.85 km2 ) on the basis of six influencing criteria like relative relief,
slope, distance from road, river, canal and settlement position (Fig. 13.8). Out of
these criteria, distance from settlement, river, road and canal have their influences
over the disturbance potentiality as 9.5%, 4.7%, 23% and 14% respectively (Table
13.6). On the other hand the two relief features like relative relief and slope have
their influences over FDP have as 19% and 28% respectively (Table 13.6). Therefore
on the basis of aforesaid accounts this can be said that canal alignment and slope
factor have effective influences over the forest disturbance potentiality zonation.
To distinguish the FDPZ in ordinal manner, all the criteria are reclassified in three
distinctive classes. For the distance based criteria like road, river, canal, settlement
position three distinct distance buffers have been considered from each feature like

Fig. 13.8 Identified Forest Disturbance Potential Zones (FDPZ)

13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 319

1 km, 3 km and 7 km. On the other hand the relief features like slope and relative
relief also classified in three categories namely 0.5°–2.03° as 1; 2.03°–2.16° as 2;
2.16°–3.26° as 3 and 18–51 m as 1; 51–71 m as 2; 71–113 m as 3 respectively.
Furthermore each criterion has the categorical importance over each FDPZ. A
data dimensionality reduction technique like Multiple Criteria Analysis (MCA) is
an ideal approach to envisage that importance.
MCA is an exploratory data analysis technique as well as multivariate graphical
technique used to analyze categorical data (Benzecri 1992; Sourial et al. 2010). A
key feature of the analysis is the joint scaling of both row and column variables to
provide information on the interrelationships between row and column variables.
Correspondence analysis can be used on both of qualitative or quantitative data.
A final step in the analysis involves rescaling of characteristic vectors into optimal
scores. Normalising these optimal scores allows for assessment of relative importance
of factors. Correspondence analysis can also be used to find the optimal ordering of
variables for a given set of characteristics (Weller et al. 1990). Homogeneity between
row element and column element specifies the correspondence between elements and
this homogeneity can be addressed by the maximum between each of row elements
and column elements (Table 13.11). In this present study the correspondence of
FDPZ with the influencing criteria has been calculated.
However the corresponding clusters can clearly be identified from the perceptual
maps (Fig. 13.9a–f). The perceptual maps are formed by plotting the row elements
against column elements and vice versa based on the inertia or the mass values. As per
the perceptual maps high FDPZ is associated to <1 km distance from canal and road,
1–3 km distance from river and settlement positions and lastly the surface having high
elevation category along with medium slope amount is prone to forest disturbance.
On the other hand low FDPZ is associated to >3 km distance from canal and road,
<1 km distance from river and 1–3 km distance from the settlement positions and
the surface having low elevation category along with high slope amount is prone to
forest disturbance.
Contextually it is mentionworthy that positions of settlement clusters have a mixed
impact over the forest disturbance potentiality. Within 1–3 km buffer distance from
the settlement position both of high and low disturbance potentiality are seen to
be associated. Besides <1 km distance from settlement position is merely effective
to forest disturbance though moderate types of disturbance potentiality is closely
associated to this range of the distance and beyond the distance of 3 km (i.e. 3–7 km)
disturbance potentiality is getting low to no disturbance.
Therefore, it can be alleged that there is no such significant impact of settle-
ment situation over the forest cover dynamicity. The symbiotic interaction between
human livelihood and forest environment has a sustainable effect over the forest
cover retention. Besides, an eco-tourism development initiative at Ganagani village
under Garbeta 1, absence of urban development in large scale except some small
Census town conversion in Garbeta 1and 3 and absence of local urban infrastructural
development as well has kept the forest degeneration inhibited. However the devel-
opment of roadways (SH or NH) and canal alignment have a significant effect over
forest cover degeneration and there are an adjacency impact over the forest cover so
320 R. Ray et al.

Table 13.11 Multiple correspondence analyses between FDPZ and driving criteria
High FDPZ Moderate FDPZ Low FDPZ Active margin
Elevation Low 102.712 160.706 308.057 571.475
Moderate 182.705 173.602 289.851 646.158
High 128.408 106.172 163.586 398.166
Active Margin 413.825 440.480 761.494 1615.799
Slope Low 93.638 193.346 242.548 529.532
Moderate 308.955 151.648 58.102 518.705
High 11.228 95.473 460.749 567.450
Active margin 413.821 440.467 761.399 1615.687
Distance from 1 km 141.835 159.881 219.813 521.529
Settlement 3 km 254.526 260.529 477.919 992.974
7 km 17.469 20.101 63.866 101.437
Active margin 413.830 440.511 761.598 1615.940
Distance from 1 km 57.779 147.325 330.448 535.552
Canal 3 km 257.226 147.875 134.532 539.633
7 km 98.832 145.321 296.611 540.764
Active margin 413.837 440.521 761.591 1615.949
Distance from 1 km 193.825 237.040 473.306 904.171
River 3 km 217.635 195.596 271.607 684.838
7 km 2.370 7.875 16.685 26.930
Active margin 413.830 440.511 761.598 1615.939
Distance from 1 km 209.030 156.303 211.151 576.484
Road 3 km 154.415 201.881 333.136 689.432
7 km 50.382 82.317 217.307 350.006
Active margin 413.827 440.501 761.594 1615.922
Bold numbers are indicating the maximum association between each of the Row and Column
elements. As an example this can be said that there is the maximum association between Low
elevation category and Low Forest Disturbance Potentiality which has been indicated by the value
308.057

that within 1 km buffer distance from road and canal high disturbance potentiality
has found. In this respect, preferences over medium to gentle slope of terrain as well
as high to medium relief height for canal alignment have defined the areas having
elevation of 71–113 m and slope of 2.03°–2.16° as high disturbance potential zone.
Area under study composed of four CD Blocks is one of the most important areas
under West Midnapore District for conservation of forest resources in West Bengal.
The significance and use of forest land must be managed for their genetic diver-
sity, productivity, regeneration capacity, vivacity and potential to fulfill the needs
relevant to ecological and socio-economic functions at the local and regional levels
(FAO 2000). FDPZ is worth of practicing to identify the suitable areas for forest
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 321

Fig. 13.9 a–f Perceptual maps showing the correspondence between FDPZ and driving factors
322 R. Ray et al.

recovery so that environmental viability can be gained as well as forest based liveli-
hood can be sustained. The conservation and restoration of forests will provide local
communities with the best decision to get the profit and restore the forests’ environ-
mental and economic functions (Sanchirico and Siikamaki 2007). The challenges
of achieving sustainable forest management and forest protection must be addressed
jointly between stakeholders and communities (Sheil et al. 2002; Gunningham 2009).
This can be said as joint forest management (JFM) and the increasing rate of regen-
eration in the forest cover is the consequence of effectiveness of it. JFM in West
Midnapore district of West Bengal has a history that initiated during 1980s while
a number of forest protection committees (FPC) and their members have been kept
on escalating with time. During 1995, wasteland management scheme was imple-
mented in several states of West Bengal. Plantation of Acacia and Eucalyptus species
has increased the vegetation cover (SFR, 2011–12). Presently most of the forest
cover is plantation forest where clearing of mature dry deciduous species was regu-
lated properly along with planting new plants. Plantation activity has helped to gain
increasing forest cover in West Medinipur district (SFR, 2011–12). All these factors
were responsible for the dynamicity of vegetation cover in this selected part of West
Midnapore district.

13.6 Conclusion

Forest disturbance analysis is very contemporary and realistic issue in forming

sustainable environment. In global environmental forestry is an important ecosystem
parameter as a source of environmental purifier fodder and shelter as well as for
economic importance. But with the progress in civilization deforestation has come
as regular practices though the act of join forest management has prevented the
reckless of deforesting activities of human being.
This research has created a theoretical framework that can be implemented when
preparing future forest management policies based on area. In the case of efforts
made through participatory forest management schemes, the overall growth of forest
cover was very convincing. Though area and intensity different vulnerable patches
were established in this study where forest recovery schemes can be implemented on
priority bases by the administration. Such a study concerning the innate dynamics of
the certain attribute is needed in the field of planning and decision making. It can be
embattled with creating sustainable forestry market opportunities for the betterment
of forest based livelihood and to promote the use of forest based commodity. The
disturbance potential areas can be set up for new forest generation to promote planta-
tions that convey reimbursement to the environment and neighboring communities.
Protecting and restoring forests could provide the best pronouncement with the local
communities to get the benefit and restore the environmental and economic functions
of the forests. The identified disturbance potential areas can be reused by afforesta-
tion to balance the forest degeneration. Therefore, the change detection analysis and
periodic determination of potential forest disturbance areas could help to undertake
new policy and planning for forest restoration and ecosystem viability.
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 323

References

Alberdi I, Nunes L, Kovac M, Bonheme I, Cañellas I, Rego FC, Gasparini P (2019) The conservation
status assessment of Natura 2000 forest habitats in Europe: capabilities, potentials and challenges
of national forest inventories data. Annals of Forest Science 76(2):34
Amiro BD, Barr AG, Barr JG, Black TA, Bracho R, Brown M, Chen J, Clark KL, Davis KJ, Desai
AR, Dore S, Engel V, Fuentes JD, Goldstein AH, Goulden ML, Kolb TE, Lavigne MB, Law BE,
Margolis HA, Martin T, McCaughey JH, Misson L, Montes-Helu M, Noormets A, Randerson
JT, Starr G, Xiao J (2010) Ecosystem carbon dioxide fluxes after disturbance in forests of North
America. J Geophys Res 115:G00K02
Anon. (1983) Mapping of forest cover in India from satellite imagery 1972–75 and 1980–82:
Summery Report. Hyderabad, Nat Remote Sensing Agency, p 9
Anselin L (1995) Local indicators of spatial association-LISA. Geograph Anal 27(2):93–115
Anselin L (2003) Spatial externalities, spatial multipliers, and spatial econometrics. Int Reg Sci Rev
26:153–166
Anselin L (2005) Exloring spatial data with GeoDaTM: a workbook, p 198e218. [Link]
edu/pdf/[Link]
Anselin L, Syabri I, Kho Y (2006) GeoDa: an introduction to spatial data analysis. Geographical
Analysis 38(1):5–22
Araya Y (2009) Urban land use change analysis and modelling: a case study of Setubal-Sesimbra,
Portugal
Benzecri, J.P.(1992). Correspondence analysis Handbook, ISBN. 058536303X, 9780585363035,
CRC Press., 485–495.
Behera MD (2000) Biodiversity characterization at landscape level in Subansiri district of Arunachal
Pradesh (Eastern Himalaya) using remote sensing and GIS. Ph.D. Thesis. Gurukul Kangri
University, Hardwar, India
Bhunia GS, Samanta S, Pal DK, Pal B (2012) Assessment of Groundwater Potential Zone in Paschim
Medinipur District, West Bengal—a Meso-scale study using GIS and remote sensing approach.
Iiste 2:41–59
Bjorkman, A. D., Criado, M. G., Myers-Smith, I. H., Ravolainen, V., Jónsdóttir, I. S., Wester-
gaard, K. B., Heiðmarsson, S. (2019). Status and trends in Arctic vegetation: Evidence from
experimental warming and long-term monitoring (pp. 1–15). Ambio, Advance online publica-
tion. doi:10.1007/s13280-019-01161-6 UN report on environment [Link]
stream/handle/20.500.11822/27539/GEO6_2019.pdf? sequence=1&isAllowed=y (Accesses on
08 June, 2019)
Böttcher H, Kurz WA, Freibauer A (2008) Accounting of forest carbon sinks and sources under
a future climate protocol—factoring out past disturbance and management effects on age-class
structure. Environ Sci Policy 11:669–686
Champion HG, Seth SK (1968) A revised survey of forest types of India. Government of India
Press, New Delhi, p 404
Chander G, Markham BL, Helder DL (2009) Summary of current radiometric calibration coeffi-
cients for landsat MSS, TM, ETM+, and EO-1 ALI sensors. Remote Sensing of Environment
113(5):893–903
Cohen WB, Spies TA, Alig RJ, Oetter DR, Maiersperger TK, Fiorella M (2002) Characterizing
23 years (1972–1995) of stand replacement disturbance in western Oregon forests with landsat
imagery. Ecosystems 5:122−137
FAO (2000) Asia-Pacific forestry commission: development of national-level criteria and indica-
tors for the sustainable management of dry forests of Asia. Workshop Report, RAP Publication
2000/07, Food and Agriculture Organization. Available online: [Link]
[Link]. Accessed on 5 Feb 2019
Frolking S, Palace MW, Clark DB, Chambers JQ, Shugart HH, Hurtt GC (2009) Forest distur-
bance and recovery: a general review in the context of spaceborne remote sensing of impacts on
aboveground biomass and canopy structure. J Geophys Res 114:G00E02
324 R. Ray et al.

Gardner TA, Barlow J, Chazdon R, Ewers RM, Harvey CA, Peres CA, Sodhi NS (2009) Prospects
for tropical forest biodiversity in a human-modified world. Ecol Lett 12:561–582
Gunningham N (2009) The new collaborative environmental governance: the localization of
regulation. J Law Soc 36:145–166
He HS, Mlandenoff DJ (1999) Spatially explicit and stochastic simulation of forest landscape fire
disturbance and succession. Ecology 80:81–99
Kahraman C, Cebeci U, Ulukan Z (2003) Multi-criteria supplier selection using fuzzy AHP.
Logistics Inf Manag 16(6):382–394
Kennedy RE, Yang Z, Cohen WB, Pfaff E, Braaten J, Nelson P (2012) Spatial and temporal patterns
of forest disturbance and regrowth within the area of the Northwest Forest Plan. Remote Sens
Environ 122:117–133
Kuemmerle T, Hostert P, Radeloff VC, Perzanowski K, Kruhlov I (2007) Post-socialist forest
disturbance in the Carpathian border region of Poland, Slovakia, and Ukraine. Ecol Appl
17:1279–1295
Kurz WA, Dymond CC, Stinson G, Rampley GJ, Neilson ET, Carroll AL, Ebata T, Safranyik L
(2008) Mountain pine beetle and forest carbon feedback to climate change. Nat Lett 452:987–990
Lindenmayer DB, Franklin JF, Fischer J (2006) General management principles and a checklist of
strategies to guide forest biodiversity conservation. Biol Conserv 131:433–445
Mahapatra AK, Tewari DD (2005) Importance of non-timber forest products in the economic
valuation of dry deciduous forests of India. For Policy Econ 7:455–467. [Link]
[Link].2004.02.002
Mahmoud EH, Agnès B, Bruno L, Olivier H, Gérard D, Matthieu R (2008) Relative radiometric
normalization and atmospheric correction of a SPOT 5 time series. Sensors 8:2774–2791
Malik S, Pal SC, Das B, Chakraborty R (2019a) Assessment of vegetation status of Sali River basin,
a tributary of Damodar River in Bankura District West Bengal, using satellite data. Environ Dev
Sustain. [Link]
Malik S, Pal SC, Das B, Chakraborty R (2019b) Intra-annual variations of vegetation status in a sub-
tropical deciduous forest-dominated area using geospatial approach: a case study of Sali watershed
Bankura, West Bengal, India. Geol Ecol Landscapes. [Link]
3219
Masek JG, Huang C, Cohen WB, Kutler J, Hall FG, Wolfe R et al (2008) North American forest
disturbance mapped from a decadal Landsat record: methodology and initial results. Remote
Sensing of Environment 112:2914−2926
Miller DH (1978) The factor of scale: ecosystem, landscape mosaic and region. In: Hammond
KA, Macinko G, Fairchild WB (eds) Sourcebook on the environment: a guide to the literature.
The University of Chicago Press, Chicago. pp 63–88; natural disturbance rates could offset
forest management impacts on ecosystem carbon stocks. Forest Ecology and Management, 308,
178–187. Nature 2000, 403, 84–86.
Miller, AB, Bryant ES, Birnie RW (1998) An analysis of land cover changes in the Northern forest
of New England using multitemporal landsat MSS data. Int J Remote Sens 19:245–265
Moller-Jenson L (1990) Knowledge-based classification of an urban area using texture and context
information in LANDSAT-TM imagery. Photogram Eng Remote Sens 56:899–904
Nagendra H, Paul S, Pareeth S, Dutt S (2009) Landscapes of protection: forest change and fragmen-
tation in Northern West Bengal, India. Environ Manage 44:853–864. [Link]
267-009-9374-9
Nepstad DC, Stickler CM, Soares-Filho B, Merry F (2008) Interactions among Amazon land use,
forests and climate: prospects for a near-term forest tipping point. Philos Trans R Soc Sect B
363:1737–1746
Oguztimur S (2011) Why fuzzy analytic hierarchy process approach for transport problems? In:
Proceedings of 51st congress of European Regional Science Association—ERSA, pp 1–10.
Barcelona, Spain
Pal SC, Chakraborty R, Malik S, Das B (2018) Application of forest canopy density model for forest
cover mapping using LISS-IV satellite data: a case study of Sali watershed, West Bengal. In:
13 Forest Disturbance Analysis of Selected Blocks of Midnapore … 325

Modeling earth systems and environment, Springer International Publishing AG, part of Springer
Nature 2018. [Link]
Petkovic J, Sevarac Z, Jaksic ML, Marinkovic S (2012) Application of fuzzy AHP method for
choosing a technology within service company. Tech Technol Educ Manag 7(1):332–341
Rahman MF, Jashimuddin M, Kamrul Islam, Kumar Nath T (2016) Land use change and forest
fragmentation analysis: a geoinformatics approach on chunati wildlife sanctuary, Bangladesh. J
Civil Eng Environ Sci 2(1):020–029. [Link]
Ray R, Mondal S (2014) Multi criteria based forest cover type mapping using Landsat ETM + digital
data—a case study on Eastern part of Jungle Mahal Region, West Bengal. Indian J Regional Sci
XLVI:77–84
Rikimaru A, Roy PS, Miyatek S (2002) Tropical forest cover density mapping. Trop Ecol 1:39–47
Running SW (2008) Ecosystem disturbance, carbon, and climate. Science 321:652–653
Saaty TL (2008) Decision making with the analytic hierarchy process. Int J Serv Sci 1(1):83–98
Sanchirico JN, Siikamaki J (2007) Natural resource economics and policy in the 21st century:
conservation of ecosystem services. Resources 2007(165):8–10
Sayer D, Wienhold M (2013) A GIS-investigation of four early Anglo-Saxon cemeteries: Ripley’s
K function analysis of spatial groupings amongst graves. Soc Sci Comput Rev 31:71–89
Sheil D, Puri R, Basuki I, van Heist M, Wan M, Liswanti N, Rukmiyati Sardjono MA, Samsoedin
I, Sidiyasa K, Chrisandini, Permana E et al (2002) Exploring biological diversity, environment
and local peoples perspectives in forest landscapes, 2nd edn. CIFOR, Bogor, Indonesia
Shit PK, Pati CK (2012) Non-timber forest products for livelihood security of tribal communities :
a case study in Paschim Medinipur District, West Bengal, 40, 149–156
Sourial N, Wolfson C, Zhu B, Quail J, Fletcher J, Karunananthan S, Bandeen-Roche K, Béland
F, Bergman H (2010) Correspondence analysis is a useful tool to uncover the relationships
among categorical variables. J Clin Epidemiol 63(6):638–46. [Link]
2009.08.008
Thakker PS, Sastry KLN, Kandya AK, Kimithi MM, Jadav rN (1999) Forest vegetation mapping
using remote sensing and GIS-a casestudy in Gir (Sasan) forest. In: Proceedings of ISRS
National Symposium on Remote Sensing Application for Natural Resources: Retrospective and
Perspective, Hyderabad, January, 18-20, 1998.
United Nation (2019) Global forest goals and targets of the UN strategic plan for forests 2030
Waring R, Running S (2007) Forest ecosystems: Analysis at multiple scales. 3rd ed. Academic
Press
Wang Y (2010) Climate change and its ecological and socioeconomic impact: evidence from China’s
historical document for Qing Dynasty. Ph.D. Thesis, Rutgers University, New Brunswick, NJ,
USA, 30 May, pp 1–174
Weller SC, Romney AK, Scaling M (1990) Correspondence analysis. Sage, Newbury Park, CA
Wickman JD, Norton DJ (1994) Mapping and analyzing landscape patterns. Landscape Ecol 1:7–23
Zhu Y, Newsam S (2016) Spatio-temporal sentiment hotspot detection using geo-tagged photos. In:
Proceedings of the 24th ACM SIGSPATIAL international conference on advances in geographic
information systems, Burlingame, CA, USA, 31 October–3 November, p 76
Zhang C et al (2008) Use of local moran’s I and GIS to identify pollution hotspots of Pb in urban
soils of Galway, Ireland. The Science of the total environment, 398 (1- 3), p. 212
Zhou Q, Li B, Zhou C (2004) Detecting and modelling dynamic landuse change using multitemporal
and multi-sensor imagery. In: Paper presented at 20th ISPRS Congress, 12–23 July 2004, Istanbul,
Turkey
Chapter 14
Comparison of AHP and Maxent Model
for Assessing Habitat Suitability of Wild
Dog (Cuon alpinus) in Pench Tiger
Reserve, Madhya Pradesh

Dhruv Jain, G. Areendran, Krishna Raj, Varun Dutta Gupta,

and Mehebub Sahana

Abstract Habitat suitability is an illustrious mission to conserve the ecosystems and

its components. Human wildlife conflicts have increased tremendously due to habitat
fragmentation; reason being human encroachment for their own use or for commer-
cial purpose like forest products and unsustainable development programs leading to
major decline in Biodiversity. Pench Tiger Reserve (PTR), Madhya Pradesh has great
capacity to nurture young Wild dogs (Cuon alpinus) because it supports lots of biodi-
versity especially herbivores which will be great availability of food and have great
tree cover on the landscape. The motive of this study is to find out the landscape which
will suitable for Wild dog (Cuon alpinus) at PTR, MP using presence SDM (Species
Distribution Model) at 900 m grain Size using Maxent, Ecological Niche Factor anal-
ysis and Bioclim and compare with AHP (Analytical hierarchical process). Maxent
gives contributions of every factor in percentage also gives responses curve of every
factor separately and also give permutation importance of every factor. Results show
that the model with accuracy more than 0.90 AUC value. Bioclimatic variable 17 and
LULC contributes the moderately high in the model. For AHP only need experts,
who fill the Saaty table on the basis of environmental variables importance to wild
dog (Cuon alpinus). Calculate or predict the habitat suitability of wild dog (Cuon
alpinus) by experts rating to every environmental variable (comparison of every vari-
ables with respect to each other) from 1 to 9 or inverse like 1/9. Maxent shows that
out of the about 750 km2 areas, more than 150 km2 found to be highly suitable at
900 m grain size SDM. This shows PTR, MP has the great potential to nurture or
sustain Wild dog (Cuon alpinus) because wild dog (Cuon alpinus) need larger area to
walk, they walk daily around 20–25 km to find their food and PTR, MP is immensely
rich in large size herbivore animals like Chital (Axis axis). Strict implementation of
laws is required this can be achieved by government agencies like forest department
and others along with that awareness can be spread among the local residents of the

D. Jain · G. Areendran · K. Raj · V. D. Gupta · M. Sahana (B)

Indira Gandhi Conservation Monitoring Centre (IGCMC), WWF-India, New Delhi, India
e-mail: mehebubsahana@[Link]

© The Editor(s) (if applicable) and The Author(s), under exclusive license 327
to Springer Nature Switzerland AG 2021
P. K. Shit et al. (eds.), Spatial Modeling in Forest Resources Management,
Environmental Science and Engineering,
[Link]
328 D. Jain et al.

PTR, MP so the wildlife and their habitat can be conserve by the conservationists
and wildlife experts.

Keywords Wild dog (Cuon alpinus) · Maxent species distribution model · AUC
(area under curve) score · Pench tiger reserve (Madhya Pradesh) · AHP (analytical
hierarchical process) tool box

14.1 Introduction

Wild dog (Cuon alpinus) commonly known as Dhole and some others local names
in India like Jangli Kutta in Hindi, Balia Kukura in Oriya and many more local
names in India also. Wild dogs (Cuon alpinus) are the only species in the genus
Cuon (Hodgson 1838). Dole (Cuon alpinus) has been long placed in the subfamily
Simocyoninae, along with bush dogs (Speothos venaticus) and African hunting dogs
(Lycaon pictus). Recently it has become evident that the subfamily Simocyoninae is
not correct entity, and the position if falling into incorrect use (Sheldon 1992). As
long ago 1954, Thenius (1954, cited in Cohen 1977) pointed out those similarities
in dentition within three species placed in the simocyoninae sub-family could be
due to convergent or biological evolution and also pointed one more point i.e., other
anatomical features are exhaustively different.
In ancient time Asiatic wild dog (Cuon alpinus) range are in big portion of East
Asia and South-East Asian Countries like in Bangladesh, Cambodia, India, Kaza-
khstan, Kyrgyzstan, Lao Malaysia, Bhutan, Mongolia, Myanmar, Nepal, People’s
Democratic Republic, Russian Federation, Indonesia, Tajikistan, China, Thailand
Viet Nam and many more South-East Asian small Countries. Wild dogs (Cuon
alpinus) have two sub-Species those are Pocock (Cuon alpinus adustus) and other
is Hardwick (Cuon alpinus sumatrensis). The sub species Pocock (Cuon alpinus
adustus) is majorly found in the South-East Myanmar and South-west Thailand. and
the other sub species i.e., Hardwick (Cuon alpinus sumatrensis) are generally found
from the south of the Isthmus of Kra down to Sumatra and Java and the Hardwick
(Cuon alpinus sumatrensis) are the smallest in size and their hairs are also smallest
in the Genus Cuon (Fig. 14.1) (Kanchanasakha et al. 1998) Now Asiatic Wild dog
(Cuon alpinus) are constrain to very limited areas like in central, south and some
portion of north India also in some East Asian countries. In India wild dogs (Cuon
alpinus) are generally found in central Indian landscape, peninsular region of India
i.e., Southern India, North-East India and even they are sighted at most harsh condi-
tions of north India i.e., in Ladakh also. Wild dog (Cuon alpinus) cubs are generally
can only be confused with jackal cub, but can be recognized by bring longer and
narrower. The shapes of the body, with small waist and deep chest, and the legs and
their mode of attachment, are built for speed on the lines of greyhound as similar
to the other big carnivores. The size and coloration of wild dogs (Cuon alpinus)
varies regionally. The average full sized Asiatic wild dog (Cuon alpinus) stand
22 inches at the shoulder and weighs 19 kg on the other hand female Asiatic wild dogs
14 Comparison of AHP and Maxent Model for Assessing Habitat … 329

Fig. 14.1 Wild dogs ancient and present presence on global level

(Cuon alpinus) are 1.5 inches shorter and 2.5 kg lighter than males (Brandar 1982).
Generally ground color of the coat is little reddish brown and the color on the belly
is quite lighter than on the back as well as on the both sides. There is thick hair on the
claws and between the toes. The dental formula of wild dogs (Cuon alpinus) differs
from the usual formula in the Canidae. Usually there are three molars on the sides
of each lower jaw but wild dogs (Cuon alpinus) have two molars on each side on the
lower jaw. Thus, the dental formula for wild dog (Cuon alpinus) is 3–3 incisors on
both sides in upper and lower jaw also similarly 4 pre-molars, 2 molars and 1 canines
on both sides in the lower and upper jaw as well. Therefore, total 40 teeth [(4 + 3
+ 2 + 1) * 4] (Novikov 1962). Another unique physical character of the wild dog
(Cuon alpinus) is their projecting part of the face, including the nose and mouth i.e.,
muzzle is quiet shorter than that of most of canids with a larger projecting part of the
face. They are always live and hunting in packs of 3 (minimum) or 8–10 (maximum)
(Brander 1982) as unity is always big strength.
Most special characters of wild dog (Cuon alpinus) are they are extremely brave
or daring even they fight with tiger (Panthera tigris) for their food. Gestation period
of female wild dogs (Cuon alpinus) are 60–63 days (Cohen 1978; Sosnovskii 1967)
and their most favorable season for denning is winter i.e., October–February. Wild
Dog (Cuon alpinus) is a species which is considered to be highly social in nature and
they have this social nature from the start of their life as in Pub life. Their rearing
is considered to be a social activity. Breeding season of wild dog (Cuon alpinus) is
confined to a narrow season as respect to other Canidae family members. In India it
happens any time from end of the Monsoon season i.e., September and it goes up to
February end. When a female wild dog (Cuon alpinus) give birth to his young ones,
at a time she gives on an average 4–5 descendants: Maximum is 9 or 10 descendants
(Sosnovskii 1967). In India young ones of wild dogs (Cuon alpinus) generally or
maximum time observed that they are born in January or February (Prater 1965).
Up to 12 young ones of wild dogs (Cuon alpinus) may be live in a single den, but
330 D. Jain et al.

these young ones are probably the descendants of two or female wild dogs (Cuon
alpinus). Females may den and rear their descendants untidily. Prater (1965) stated
that a number of female wild dogs (Cuon alpinus) may select a den site and form a
breeding colony. A group of new young ones may have several different aged that
create confusion to determine which one is more younger or offspring of which
female wild dog (Cuon alpinus) then, Johnsingh (1982) clears that confusion and
he stated that breeding is restricted to only a single female wild dog (Cuon alpinus)
within a pack, as is also the case for the highly social, pack living African wild dogs
(Lycaon pictus) and for wolves (Canis lupus). In a pack of Wild dogs (Cuon alpinus)
every adult ones help to young ones to provide food at their den and protect all
descendants from the predators like Tiger (Panthera tigris) and other big carnivores
who can eat their descendants along with the lactating females. “Guards”—Wild dog
(Cuon alpinus) that remain behind at the den site with the young ones while the rest
are out for the hunting—are also fed (Johnsingh 1982; Fox 1984). Many times, in
a group, members of a group of Wild dogs (Cuon alpinus) after hunting, it is not
mandatory to all members of group to assemble again. In the big size group of wild
dogs (Cuon alpinus) is very rarely to seem to do large hunting together otherwise the
wild dogs (Cuon alpinus) are rapidly spilt into many small groups or say large size
group breaks into many small groups for the hunting. Wild dogs (Cuon alpinus) use
odor of the prey and find their prey then take turns pursue to catch or attempt for catch
their prey. Most common pattern of hunting is group of wild dogs (Cuon alpinus)
cover the prey in a circle like structure and attack on prey from all sides specially on
those moments will extremely effective when prey is busy to defend from one side
and another group member attack from back side or say when prey is unaware from
back side attack. When sometime a pack of wild dogs (Cuon alpinus) were not find
any prey for them to fulfill their food requirement then sometimes they eat their pups.
According to Davidar (1975), he has been seen group of wild dogs (Cuon alpinus)
whose strength is about 40. During the summer’s i.e., from May to mid July and in
Rainy or Monsoon Season (i.e., not in breeding Season) larger assemblage of wild
dogs (Cuon alpinus) on a single place is quite common because all are in the resting
stage of the year so that’s why quite common to assemblage of very larger of wild dogs
(Cuon alpinus) on a place. In the recent years sighting of large group of wild dogs
(Cuon alpinus) is rare or frequency of sighting is frequently decreased because of
anthropogenic pressure, encroachment of humans on their suitable habitat landscape,
loss of suitable habitat and decline in prey population. During the end of Monsoon
season i.e., end of the august or starting of the September and Start of breeding season
or denning season sighting of small packs of wild dogs (Cuon alpinus) are frequent
(Sheldon 1992). Wild dogs (Cuon alpinus) are generally amalgamate and disruptive
way in a loosely or scattered pattern. There is no proof about the influence over the
other gradable stratified class within their groups, even though in the free moving
group of wild dogs (Cuon alpinus), one male wild dog (Cuon alpinus) was clearly
show his influence over the rest members of this group. Belligerent interactions
between the group members will seen extremely rare and group member showing
willingness to each other for the killing purpose and in a group generally competition
for the hunting at very low level. The relating places where wild dogs (Cuon alpinus)
14 Comparison of AHP and Maxent Model for Assessing Habitat … 331

can live and their range of that kind of places in India is generally in a group is
around 35–40 km2 , this home range is only when they are in no breeding season but
when we talk about during their breeding season then their suitable habitat range
will rapidly decrease from 35–40 km2 to 10–12 km2 (Johnsingh 1982).
Most preferable habitat is tropical deciduous forest, open forest and grasslands
also (Brandar 1982). Sambar (Rusa unicolor), Chital (Axis axis), and Wild boar (Sus
scrofa) are considered the principal prey of Wild dog (Cuon alpinus) within India
(Davidar 1975; Johnsingh 1992; Venkataraman et al. 1995). From a study done in
India, it was found out that the prey of the wild dogs (Cuon alpinus) are as follows 73%
area spotted deer (Axis axis), 17% are sambar deer (Rusa unicolor), 5% are rodents
[Order—Rodentia (Bowdich 1821)] and 2.5% are rabbit (Oryctolagus cuniculus).
7% of their excrement is the residue of grass (Johnsingh 1982) Main thing is they
don’t require lots of water like other organisms and they are generally walked a lot
basically they eat and walk around 20–25 km per day to find their food. Generally,
Wild dogs (Cuon alpinus) have a bimodal or say in other words crepuscular activity
pattern of life style or daily routine. During heat of the day or commonly during
summers high temperature day they are usually inactive or in lazy state or in other
we can say that in rest mode, though in India during Monsoon i.e., in late June, July,
August and September, they may be in active stage because when Monsoon season
end their breeding season will about to start and they can search for their prey at
any time of the day during monsoon season. Another reason for their hunting during
Monsoon is because in Monsoon heavy rainfall help to grow small grasses on land
or help to growing of herbs and shrubs also which the primary food for herbivores
like chital (Axis axis), Sambar (Rusa unicolor), Indian bison (Bos gaurus) and other
big or small herbivore which means probability of availability of food for wild dogs
(Cuon alpinus) will be high. Hayward et al. (2014) found the prey preference index
of the wild dogs (Cuon alpinus) is aviate slightly in different areas. During Monsoon
season wild dogs (Cuon alpinus) hunt some time during night also it depends on their
group size and their hungriness’ (Johnsingh 1982; Krishnan 1972; Fox and Johnsingh
1975; Sosnovskii 1967; Cohen et al. 1978; Fox 1984; Prater 1965). Hayward et al.
(2014) concluded that the prey of wild dogs (Cuon alpinus) from 24 different studies
in the across 16 areas including habitat and distribution are Wild boar (Sus scrofa),
Sambar (Rusa unicolor), Gaur (Bos gaurus), Spotted deer (Axis axis) and Muntjac
(Muntiacus).
Wild dogs (Cuon alpinus) have threat from continuous loss of habitat, rapid
decrease in the count of their prey, persecution and also due to increase in transfer
of disease or several infected pathogens from domestic dogs (Canis lupus famil-
iaris) and feral dogs (Canis lupus familiaris). The present range of wild dogs (Cuon
alpinus) has been much reduced due to anthropogenic activities. Wild dogs (Cuon
alpinus) have become rare and are depleted entirely from the parts of central Asia,
large parts of India, and Eastern China (Müller-Using 1975). Across the globe any
government or any organization doesn’t take any attempt to census of the wild dogs
(Cuon alpinus) population, either at country level or regional level wise, Even in
332 D. Jain et al.

places where prediction of the available, because of the chances of the same animal
being counted more than once, or not at all, animal’s wandering habits and the figures
can be misleading.
The present study is important to identify most suitable landscape for wild dogs
(Cuon alpinus) in the Pench tiger reserve, Madhya Pradesh with the help of geospatial
techniques. The major research gap is most of scientists or researchers work on
mostly on wild dog (Cuon alpinus) ecology, physical characteristics like in which
pattern their body size varies during their upbringing, their distribution and habitat,
taxonomy, diet preferences, daily activities, Reproductive behavior across all seasons,
Social organization and the behavior in with their group members but present study
break the tradition of study of the wild dog (Cuon alpinus) because now first time try
to predict suitable landscape for the habitat of wild dog (Cuon alpinus) in Pench tiger
reserve, Madhya Pradesh with the help of GIS (Geographic Information System) and
remote sensing. The main objective of the study is to identify and assess the suitable
habitat of wild dog (Cuon alpinus) and to compare the different model’s prediction
for the habitat suitability of wild dogs (Cuon alpinus) in the Pench tiger reserve,
Madhya Pradesh. This study will contribute a new pillar to the house of literature of
wild dogs (Cuon alpinus) because all the literatures are old and some new literatures
also didn’t predict the suitable habitat, they are also about their diet preferences and
their percentage of food of different preys.

14.2 Study Area

This study is conducted on the Pench Tiger Reserve (PTR) Madhya Pradesh, India. It
comprises the Pench Mowgli sanctuary, the Indira Priyadarshini Pench national park,
and a buffer. The Pench Tiger Reserve, Madhya Pradesh occurs in the Karmajhiri,
Gumtara, Kurai and Satpura range (southern slopes of the Satpura range of central
India) with altitude between 300 and 671 m. The geographical extent 21° 63 to
21° 95 N latitude and 78° 94 to 79° 53 E longitude (Fig. 14.2). Satpura hills in
Pench tiger reserve, MP have almost flat, gently, slopping top and steep sides. Pench
Tiger Reserve, Madhya Pradesh is the one of the major protected areas of Satpura-
maikal range. Pench Tiger Reserve, Madhya Pradesh is one of the highest herbivores
densities in India. The Pench national Park, Madhya Pradesh is split into two parts
by river Pench. The river Pench is the lifeline of the National Park, near the Pench
River the area is flat and forming woodland and meadows. The Pench tiger reserve is
the first interstate tiger project area of the country that is between Madhya Pradesh
and Maharashtra. Pench Tiger Reserve, Madhya Pradesh is among the sites notified
as important bird areas of India. Granites and amphibolites pegmatite and quartzite
later intrude these rocks. The Deccan trap, otherwise prevalent in the other parts of
both Seoni and Chhindwara districts are almost absent within the limits of the present
boundary of the National park and sanctuary.
14 Comparison of AHP and Maxent Model for Assessing Habitat … 333

Fig. 14.2 The location of Pench Tiger Reserve (PTR) in Madhya Pradesh, India

As per the all India tiger estimation report 2018, highest number of tigers found
in the Pench tiger reserve in Madhya Pradesh and according to the 4th round of the
management effectiveness evaluation of tiger reserves (MEETR), which assess the
India’s all 50 tiger reserves, which says that Pench tiger reserve, Madhya Pradesh
as best managed tiger reserve in the country with Periyar sanctuary of the Kerala.
Pench national park which occurs under the Pench tiger reserve, Madhya Pradesh
boundary, which has the maximum density of the herbivore animal’s population
like nilgai (Boselaphus tragocamelus), Sambar Deer (Rusa unicolor), Chital (Axis
axis), Barking Deer (Muntiacus muntjak), Indian spotted chevrotain or Indian mouse
deer (Moschiola meminna), Chousingha or Four horned antelope (Tetracerus quadri-
cornis) and many more small and big size herbivore animals. All the above infor-
mation simply depicts that Pench tiger reserve, Madhya Pradesh is highly rich in
herbivore diversity which indicates availability of food for wild dog (Cuon alpinus)
and other carnivores that’s why it is the one of the highest tiger population tiger
reserve in the India. Another major reason for Pench tiger reserve, Madhya Pradesh
for wild dogs (Cuon alpinus) is one of the best managed tiger reserve in the India
so that it shows there is less human interference for the wildlife as compare other
Protected areas which means wild dogs (Cuon alpinus) can live in their natural habitat
334 D. Jain et al.

with their appropriate food availability. As per wild canids India project website wild
dogs (Cuon alpinus) inhabit less than 50% of their potential habitats in the India.

14.3 Data Base and Methodology

We take Sentinel-2 satellite data of 14th May, 2019. My study area covers (with
10 km buffer boundary) four tiles of Sentinel-2 satellite i.e., T44QKK, T44QLK,
T44QLJ and T44QKJ. After the satellite image download first we processed it in
ERDAS IMAGINE 2014 software. First layer stack all the four tile in which every
tile contain 12 bands in which band 8 have two sets that are band 8 band 8 A and all
bands have different resolutions like 4 bands have 10 m resolution, 6 bands have 20 m
resolution and 3 bands have 60 m resolution (Pokhriyal et al. 2020). Then mosaic
all the tiles, mosaic is the process in which we joint or merge all tiles, again mosaic
process is done with the help of ERDAS IMAGINE 2014 software and save the file
in the *.img (Imagine) format. Then clip the Mosaic image of the all four tiles which
become one image. Open the mosaic output in the Arc Gis 10.1 software and clip the
Mosaic output image from the Study area (with 10 km buffer boundary) i.e., Pench
Tiger Reserve, Madhya Pradesh and then we get sentinel-2 Satellite image of Pench
Tiger Reserve, Madhya Pradesh.
We consider some Bioclimatic variables of change in every 30 s data like
Annual Mean temperature (BIO 1), Mean diurnal Range [Mean of monthly
(Maximum Temperature–Minimum Temperature)] (BIO 2), Isothermality (Mean
Diurnal Range/Temperature Annual Range) (*100), Maximum Temperature of the
Warmest Month, Minimum Temperature of driest Month, Annual Precipitation.
Sentinel-2 Satellite data for the year of 2018 were used to prepare the land use–
land cover map of Pench Tiger Reserve, Madhya Pradesh (Table 14.1). The study
area was extracted from the satellite imagery using 10 km buffer of Pench Tiger
Reserve, MP boundary. As we mentioned above wild dog (Cuon alpinus) walk a lot
and their movements across PTR is very fast so that we have taken 10 km buffer
boundary (Fig. 14.3) to ease to estimate suitable habitat for wild dog (Cuon alpinus)
in Pench Tiger Reserve, MP. As we mentioned above wild dog (Cuon alpinus) walk
a lot and their movements across PTR is very fast so that we have taken 10 km buffer

Table 14.1 The details of the data sources used for this study
SI No. Data Data source Resolution Resample resolution (m)
1 Sentinel 2 (2019) [Link] 10–60 m 900
gov/
2 ASTER DEM [Link] 30 m 900
gov/
3 Bioclimatic data [Link] 900 m 900
4 Occurrence data Steinmetz et al. (2012) CSV table 900
14 Comparison of AHP and Maxent Model for Assessing Habitat … 335

Fig. 14.3 Methodology for Maxent species distribution model

boundary to ease to estimate suitable habitat for wild dog (Cuon alpinus) in Pench
Tiger Reserve, Madhya Pradesh (Figs. 14.4 14.5 and 14.6).
The statistics-based decision rules were applied for determining the land use-land
cover identity of each pixel in the images. In the ERDAS IMAGINE 2014 software
analysis process, the result is an assemblage of pixels with common features without
the user giving sample classes i.e., Unsupervised Classification. For the preparation
land use classification, the spectral classes were grouped into 60 classes. Classes
like water bodies, Agricultural land, fallow land and barren land i.e., scrubland show
similar kind of spectral reflectance. Recoding and cleaning process were followed by

Fig. 14.4 Steps for making of LULC map

336 D. Jain et al.

Fig. 14.5 Land use/ Land cover map (2019) of Pench Tiger Reserve, Madhya Pradesh

unsupervised classification in ERDAS IMAGINE 2014. Different aspects or factors

which plays role in suitable habitat for wild dog (Cuon alpinus) like Pench river,
Totladoh Reservoir, Settlements and roads were digitised in Google earth Pro. Shape
file of Pench Tiger Reserve, MP was imported over the image under cleaning process
and classes like Agricultural land, Fallow land, Barren land, Forest, Settlements and
Water bodies or streams were updated on LULC using ground truth data and Google
earth pro accuracy assessment was also done. Random sample points were laid down
with a 100 points on the LULC to crosscheck the accuracy of LULC through which
we obtained around 91 points were perfectly matches with google earth pro imagine
that means 91% accuracy of LULC (Fig. 14.5 ).

14.3.1 Dactors derived form of Elevation layer

We used ASTER DEM data at 30 m resolution sample size. With the help of ASTER
DEM we can calculate Elevation of Pench Tiger Reserve, MP with the help of Arc
Gis 10.1 software in which we used Conversion tools in which first we convert DEM
data into raster file to calculate Elevation of Pench Tiger Reserve, MP through we
get Maximum elevation is 671 m in Pench tiger Reserve, MP which is nearly in
chhindwara district region.
14 Comparison of AHP and Maxent Model for Assessing Habitat … 337

Fig. 14.6 Spatial distribution of Dhole in Pench Tiger Reserve, Madhya Pradesh

In Pench Tiger Reserve, MP as such no major elevation changes but in west Pench
Tiger reserve, MP there is some portion of Satpura Hill range, but wild dogs (Cuon
alpinus) are not prefer much height they like plain surface more. They prefer mostly
plain area or slightly elevated areas not much elevated areas (Fig. 14.7).
With the help of Elevation we can calculate several other parameters which will
affect the habitat of wild dog (Cuon alpinus) like Hill Shade, Slope, and Aspect in
Arc GIs 10.1 software through surface analyst tools in which use surface option to
generate layers of Hill shade, Aspect and Slope.
We observe when we calculate hill shade with the Z-factor value 0.1 then we get
range of hill sahde is 0–183 similar way when try to calculate slope with z value is 1
and 50 interval class is maximum 83° angle in the study area Pench Tiger Reserve,
MP and Aspect which value range is between −1 and 359.433. Wild dog (Cuon
alpinus) preferred slight steep places, where slope is around 20–30° (Brander 1980).
Apart from this Wild dog (Cuon alpinus) strongly prefer grassland terrain with low
slope and low elevation and moderate slope like 10–25° (Fig. 14.8).
338 D. Jain et al.

Fig. 14.7 Spatial variation of elevation (m)

14.3.2 Preparation of Other Factors

[Link] Fire

First we take fire data of India from FSI for last decade i.e., from 2008 to 2018
then clip that data to my Study Area. We got in Pench Tiger Reserve, MP have
lots of fire incidence it indicates this Protected area is good for wild dog (Cuon
alpinus) because when fire goes down then on the place of fire firstly grows small
grasses which is the primary food of herbivores that means availability of food for
wild dog (Cuon alpinus) (Sahana and Ganaie 2017). Major fire incidence sighted in
chhindwara district region and more than 700 incidence are sighted in last decade
(2008–2018) (Fig. 14.9).

[Link] NDVI (Normalised Difference Vegetation Index)

To calculate NDVI, I use Mosaic product of Sentinel-2 Satellite Image, using unsu-
pervised classification Tool in ERDAS IMAGINE 2014. In Pench Tiger Reserve,
MP, there no as such dense forest area and Wild dogs (Cuon alpinus) are generally
14 Comparison of AHP and Maxent Model for Assessing Habitat … 339

Fig. 14.8 Spatial distribution of thematic layers

340 D. Jain et al.

Fig. 14.9 Spatial distribution of forest fire risk

prefer open forest and grassland. NDVI value is lie between −1 and +1 in which
zero indicates for water.

NDVI = NIR − RED/NIR + RED

For Sentinel-2, use band 8 (Near infrared) and band 4 (Red) (Fig. 14.10)

NDVI = Band 8 − Band 4/Band 8 + Band 4

[Link] Prey Occurrence

Major Prey species of Wild dog [(Cuon alpinus) are sambar (Rusa unicolor), Wild
boar (Sus scrofa) and Chital (Axis axis)]. We take only those places ranges where
maximum probability of sightings of all these major preys of wild dog (Cuon alpinus)
from annual forest report of Madhya Pradesh state government annual forest report
on their biodiversity 2018 (Fig 14.6; Fig. 14.11).
14 Comparison of AHP and Maxent Model for Assessing Habitat … 341

Fig. 14.10 Spatial distribution of NDVI

[Link] Tree Cover

Tree cover plays vital role for Wild dog (Cuon alpinus) because if tree cover will
high or dense forest it will create some difficult to hunt prey for food but some times
it is very good to hide and hunt preys. Tree cover data is taken by Hansen tree cover
from tile which has data of 20–30 N and 70–80 E coordinates (Fig. 14.12).

14.3.3 Maxent Species Distribution Model

Wild dogs (Cuon alpinus) are poorly or little studied species in India except some
protected area and biosphere reserve such as Central India specially in Madhya
Pradesh and Southern India like in Tamil Nadu and nearby states such type of area
covers less than 5% of potential suitable habitat of Wild dog (Cuon alpinus) in India
(Johnsingh 1989; Yadav et al. 2020).
Maxent software are excessively use for assessing the habitat suitability of these
elusive species.
342 D. Jain et al.

Fig. 14.11 Geographical area of prey abundance

Maxent is a sophisticated machine learning model, with very explicit statistical

and mathematical formula (Jayne’s 1957). It’s widely used for predicting the habitat
suitability and species distribution. Maxent provide the probability of species distri-
bution and their suitable area which are based on concept of maximum deterioration
(i.e., heavily diffuse out or close to uniformity). It uses the covariates from the species
occupancy point for predicting the species distribution.
Maxent uses the occurrence or presence only data for species distribution. In this
study occurrence data for wild dog (Cuon alpinus) were collected from Aniruddha
Majumdhar et al. (2012). Recent updated version of Maxent 3.4.1. were used for
predicting the habitat suitability of Wild dog (Cuon alpinus) in Pench Tiger Reserve,
MP.
A total 150 plus occurrence data were used for wild dog (Cuon alpinus), 30% of
occurrence data were used for verification of model, rest 70% were used for testing or
build the model for wild dog (Cuon alpinus) species. Maxent accept the occurrence
data only in environmental variables in ASCII format and *.CSV format with same
size cell size, projection and extent.
All environmental variables which are used in habitat suitability modeling for
(Cuon alpinus) were uploaded in Arc Map window then re-projected in to GCS
(Geographic Coordinate System) WGS (World Geodetic System) by using project
raster tool in Arc Map software. Re-projected variables were resembled into 30 m
14 Comparison of AHP and Maxent Model for Assessing Habitat … 343

Fig. 14.12 Spatial variation of tree cover

meter resolution to 900 m. All variables were extracted from first layer that is BIO-I
in this study, and then all variables were converted from raster to ASCII (American
Standard Code for Information Interchange) format by using conversion from raster
tool in Arc tool box.
Maxent model can run multiple of time, which enhance the model accuracy. In this
study a total of 10 replicates with 5000 iteration were chosen, it gave the sufficient
time to the model for convergence the result. If model does not have sufficient time for
convergence the variables then there will be possibility of over and under prediction
of model. Jackknife test are used for estimate the importance of each variable and
their contribution in prediction of habitat suitability.
Sensitivity of each environmental and a biotic variable was done in logistic format.
Accuracy of output were compared with ROC value, according to expert view ROC
value 0.5–0.6 represent the unacceptable result, poor result represents by ROC value
0.6–0, Normal result with 0.7–0.8, good out represent by 0.8–0.9 and excellent output
shows with ROC value 0.9–1.0 (Swets 1988).
Main thing is that when you try to run Maxent Species distribution model always
remember your all layers have same cell size, projection and extent otherwise Maxent
Species distribution model will not run and also not predict the suitable habitat for
your which you want.
344 D. Jain et al.

14.3.4 Methodology for Maxent Species Distribution Model

As we mentioned the methodology in Fig. 14.3 we make all the factors like Land
Use/Land Cover, Hill shade, Slope, Aspect, Normalized Difference in vegetation
index, Elevation, Fire, Prey Abundance, Tree cover, Annual Mean Temperature (Bio
1), Mean Diurnal Range (Bio 2), Isothermality (Bio 3), Maximum Temperature of
warmest month (Bio 5), Minimum Temperature of the Coldest Month (Bio 6), Annual
Precipitation (Bio 12), Precipitation of the wettest Month (Bio 13), Precipitation of
Driest Month (Bio 14) and Precipitation of Driest Quarter (Bio 17).
First convert the entire layer Projection into GCS (Geographic Coordinate System)
WGS (World geodetic system) 1984 with the help of Arc GIS 10.1 software. Then
convert all the layer cell size into same cell size, we prefer all the layers cell size
similar to bioclimatic variables cell size that is at 900 m cell size and then last to do
extent of all the layers should be same that’s why me make same extent of all layers
with respect of bioclimatic variable layer. In Maxent Species Distribution Model, it
requires all the layers are in same projection, same cell size and same extent. Then
convert all the raster files into ASCII (American Standard Code for Information
Interchange) with the help of raster tool box in the Arc GIS 10.1 with the help of
batch mode option. Batch mode option help to do same process of more than 1 layer
that’s why we choose batch mode because we have 18 layers to which we convert
into ASCII (American Standard Code for Information Interchange) format. ASCII
(American Standard Code for Information Interchange) layer is the only format
which Maxent Species Distribution Model can read. Then we georeferenced the
map of wild dog (Cuon alpinus) sighted in Pench Tiger Reserve, Madhya Pradesh
from Dr. Aniruddha Majumder Ph.D. Thesis (2011) and take coordinates and make
it a shape file. Then in file manager we open shape file folder and open *.dbf (Data
Base File) file and open that *.dbf (Data Base File) file in Microsoft Excel and save
as it in *.CSV (Comma Separated Values) format file. When we get all the layers
in the ASCII (American Standard Code for Information Interchange) with same cell
size, Projection and Extent with all Coordinates of the wild dogs (Cuon alpinus) in
the *.CSV (Comma Separated Values) format of Excel file and put in the Maxent
Species Distribution Model (Fig. 14.13).

14.3.5 Overview of Factors that Affect Habitat of Wild Dog

(Cuon alpinus)

See Fig. 14.14.

14 Comparison of AHP and Maxent Model for Assessing Habitat … 345

Fig. 14.13 Methodology frame work of habitat suitability assessment

Fig. 14.14 Various aspects of habitat quality

14.3.6 Methodology for AHP (Analytical Hierarchical

Process)

Analytical hierarchical process (AHP) was given by Thomas Saaty in 1997 in which
he gave a table in which we compare two factors and rate both factors in 1–9 range
346 D. Jain et al.

if first factor is more contribute or more dominant to second factor then give rating
in the range of 1–9 in 1–9 range 1 indicates to equal importance of both factors as
when rate it 9 it means first factor is extremely important then second factor but on
the other hand if second factor is extremely important then give range in reciprocal
i.e., 1/9 or 1/8 like this. This is ensured that all the environmental Variables and
other factors were used with a combination of techniques such as GIS (Geographic
Information System) and AHP (Analytical hierarchical process) in this study. The
relative importance of the selected criteria is revealed by using AHP (Analytical
Hierarchical Process) and AHP also assisted in assigning the weightage to each
criterion. The criterion was used in Index Overlay Model to arrive at the most suitable
landscape sites for the wild dogs (Cuon alpinus) in the Pench Tiger Reserve, Madhya
Pradesh. Local knowledge and field experience have been used to assign the weight
for each factor in vulnerability analysis. Pair-wise comparison matrix is provided by
AHP (Analytical Hierarchical Process), through which the criterions are structured
according to their hierarchical order. A vector of weights can obtain by using 1–9
scale in each pair-wise comparison.
Saaty table is not restricting to two factors; Saaty table gives freedom to select
factors up to nine factors. As per Saaty table compare every factor to each other and
not repeat any combination and give rating between 1–9 and rate in reciprocal if
second factor plays dominant role like 1/9 or 1/8 etc.
For the cross check the prediction of suitable habitat of wild dog (Cuon alpinus) in
Pench Tiger Reserve, MP Maxent species distribution model we did AHP (Analytical
hierarchical process) for the predication of suitable habitat of wild dog (Cuon alpinus)
in Pench Tiger Reserve, MP. First we were doing AHP (Analytical hierarchical
process) only for Bioclimatic Variables with the help of Saaty Table 14.2. Saaty
table is filled by five experts who are expert in GIS (Geographic Information System)
and remote sensing also has expertise in wildlife. Firstly I were gave empty Saaty
table to my all experts to fill it and gave ratings to all bioclimatic variables which
are Annual mean temperature (Bio-1), Mean annual Range (Bio-2), Isothermality
(Bio-3), Maximum temperature of Warmest Month (Bio-5), Maximum temperature
of coldest Month (Bio-6), Annual Precipitation (Bio-12), Precipitation of Wettest
Month (Bio-13), Precipitation of Driest Month (Bio-14) and the Precipitation of
Driest Quarter (Bio-17). Then we have two ways to further proceed Saaty table
which filled by some GIS (Geographic Information System) and remote sensing and
wildlife experts.
Firstly we renamed the all Factors name or say use the symbols like for Annual
mean Temperature (Bio-1) denoted by A1, Mean annual range (Bio-2) denoted by
A2, Isothermality (Bio-3) denoted by A3, Maximum temperature of Warmest Month
(Bio-5) denoted by A4, Maximum temperature of coldest Month (Bio-6) denoted by
A5, Annual Precipitation (Bio-12) denoted by A6, Precipitation of Wettest Month
(Bio-13) denoted by A7, Precipitation of Driest Month (Bio-14) denoted by A8 and
the Precipitation of Driest Quarter (Bio-17) denoted by A9 (Table 14.3).
Then we used Saaty table which filled by all experts for bioclimatic variables
(Table 14.2) then we use AHP tool box in Arc GIS 10.1. In AHP (Analytical hierar-
chical process) tool box first fill the data of Saaty table which field by GIS (Geographic
14 Comparison of AHP and Maxent Model for Assessing Habitat … 347

Table 14.2 Pair-wise comparison matrix (after Saaty)

Intensity of importance Definition Explanation
1 Equal importance Two activities contribute to the
objective
2 Weak or slight
3 Moderate importance Experience and judgment slightly
favor one activity over another
4 Moderate plus
5 Strong importance Experience and judgment strongly
favor one activity over another
6 Strong plus
7 Very strong or demonstrated An activity is favored very strongly
importance over another; its dominance
demonstrated in practice
8 Very, very strong
9 Extreme importance The evidence favoring one activity
over another is of the highest
possible order of affirmation
Reciprocal of above If activity i has one of the above A reasonable assumption
non-zero numbers assigned to it
when compared with activity j,
then j has the reciprocal value
when compared with i
1.1–1.9 If the activities are very close May be difficult to assign the best
value but when compared with
other contrasting activities the size
of the small numbers would not be
too noticeable, yet they can still
indicate the relative importance of
the activities

Information System) and remote sensing and Wildlife experts for bioclimatic vari-
ables then run the data and give output location to my result but there is a little bit
problem because when we get result from AHP table in the form of a predicted map
layer in *.GTF (Geographic Transformation File) file format but when you try again
to open *.GTF (Geographic Transformation File) file layer of AHP (Analytical hier-
archical process) bioclimatic map prediction which will not open again so to avoid
this problem when you get first map prediction from AHP (Analytical hierarchical
process) tool box then immediately export that layer in your desirable location in
your suitable format.
Then again use AHP (Analytical hierarchical process), now this time we use
last AHP (Analytical hierarchical process) Bioclimatic variables map projection
(Fig. 14.15) used as a factor in next AHP (Analytical hierarchical process).
348 D. Jain et al.

Table 14.3 Responds on AHP table for bioclimatic variables

Respondents Expert A Expert B Expert C Expert D Expert E Average
A1 versus A2 1 4 5 3 0.25 2.65
A1 versus A3 1 5 4 1 6 3.4
A1 versus A4 9 6 7 7 0.14 5.82
A1 versus A5 1 6 6 0.2 1 2.84
A1 versus A6 1 7 9 0.2 1 3.64
A1 versus A7 1 4 5 0.14 1 2.22
A2 versus A3 5 5 3 1 6 4
A2 versus A4 3 6 5 5 0.16 3.83
A2 versus A5 3 5 3 0.5 5 3.3
A2 versus A6 0.14 6 3 3 5 3.42
A2 versus A7 0.14 5 5 0.33 6 3.3
A3 versus A4 3 5 5 3 0.16 3.23
A3 versus A5 1 6 3 5 3 3.6
A3 versus A6 0.33 5 6 3 3 3.46
A3 versus A7 0.33 6 3 0.33 3 2.53
A4 versus A5 0.14 5 1 0.33 1 1.5
A4 versus A6 0.14 6 3 0.33 1 2.1
A4 versus A7 1 6 5 3 1 3.2
A5 versus A6 3 5 7 5 1 4.2
A5 versus A7 1 6 5 7 1 4
A6 versus A7 1 7 1 0.33 1 2.06

Now we take six factors which are Land Use/Land Cover, Prey occurrences,
AHP (Analytical hierarchical process) predication layer of Bioclimatic variables
(Fig. 14.15), Water, Tree cover or Forest canopy cover and Human Disturbances
which include roads, settlements and highways. As in previous AHP (Analytical
hierarchical process) process we again give codes to all factors which we’ve taken
this time like A1 for LULC, A2 for Bioclimatic variable layer which we previously
AHP (Analytical hierarchical process) process we got, A3 for Prey occurrences,
A4 for Tree cover or Forest canopy cover, A5 for Water and A6 for Human distur-
bance (Sahana et al. 2018). Then again AHP (Analytical hierarchical process) table
gave to experts which are this time again filled by GIS (Geographic Information
System) and remote sensing and wildlife experts fill the table with respect to habitat
suitability of wild dogs (Cuon alpinus) in Pench Tiger Reserve, Madhya Pradesh
(Table 14.4).
Again, same process, put the values of experts rating Saaty table which we put into
the AHP (Analytical hierarchical process) tool box in the Arc GIS 10.1 software and
select the output file location in the desirable folder. When we precede the process
we get our final product of Habitat suitability map of wild dogs (Cuon alpinus) in
14 Comparison of AHP and Maxent Model for Assessing Habitat … 349

Fig. 14.15 AHP prediction for bio climatic variables

Table 14.4 Responds on AHP table for all variables

Respondents Expert A Expert B Expert C Expert D Average
A1 versus A2 8 6 6 1 5.25
A1 versus A3 0.13 7 8 0.14 3.81
A1 versus A4 6 7 9 0.2 5.55
A1 versus A5 0.11 7 4 0.14 2.81
A1 versus A6 5 7 5 9 6.5
A2 versus A3 0.11 4 7 0.14 2.81
A2 versus A4 0.17 5 5 3 3.29
A2 versus A5 0.13 5 4 0.2 2.33
A2 versus A6 5 2 2 9 4.5
A3 versus A4 6 5 5 5 5.25
A3 versus A5 1 6 4 1 3
A3 versus A6 6 2 5 9 5.5
A4 versus A5 0.11 4 7 0.2 2.83
A4 versus A6 5 1 7 9 5.5
A5 versus A6 8 6 5 9 7
350 D. Jain et al.

the Pench Tiger Reserve, Madhya Pradesh in the *.GTF format file then export the
final product of the map in *.GTF file into suitable format which supports in the
system like in *.img format or *.tiff format file which we can easily open in the any
software. We get the result after export the final map layer prediction with the help
of AHP (Analytical hierarchical process) in Sect. 14.4.2.

14.4 Results

14.4.1 Maxent Species Distribution Model Result

As we can see in Fig. 14.16 major Core area (No disturbance zone) of Pench Tiger
Reserve, Madhya Pradesh light green color shows highly moderately suitable which
covers most of core area of Seoni District of Madhya Pradesh, Pench Tiger Reserve
and North-East region of Core region of Pench Tiger Reserve, Madhya Pradesh is

Fig. 14.16 Habitat suitability of Dhole in Pench Tiger Reserve, MP (Maxent prediction)
14 Comparison of AHP and Maxent Model for Assessing Habitat … 351

highly suitable zone which shows in yellow color and light blue color zone which
covers the light green color (highly moderately suitable zone). The light blue color
zone shows slightly suitable area. The last dark blue zone which indicates highly
unsuitable zone for wild dogs (Cuon alpinus). We can observe some highly moder-
ately area which shows by light green color which goes out from Pench Tiger Reserve,
Madhya Pradesh and enters into Pench Tiger Reserve, Maharashtra and other exten-
sion we see in east region of Pench Tiger Reserve, Madhya Pradesh or say towards
Kanha-Pench Corridor in Madhya Pradesh.
With the help of Jackknife of AUC for wild dog (Cuon alpinus) we can relate or
predict the how much every variable importance with only variable shown by light
blue color histogram, without variable denoted by dark blue color histogram and
with all variables denoted by red color histogram, all these environmental variables
conditions with respect to AUC value jackknifing create all histograms (Fig. 14.17).
We can see in the Upper Image of Fig. 14.18a which shows the curve between
Sensitivity versus Specificity for wild dog (Cuon alpinus) in the Pench Tiger Reserve,
Madhya Pradesh, AUC (Area Under Curve) value is 0.938 in training data, As per
Maxent Species distribution model tutorial only those models are acceptable which
AUC value for training data is more than 0.75 and AUC value for Test data also be

Fig. 14.17 Jackknife of AUC for Cuon alpinus

352 D. Jain et al.

Fig. 14.18 a Sensitivity versus specificity for Cuon alpinus. b Omission and predicted areas for
Cuon alpinus

more than 0.75 similarly we get 0.838 AUC value for test data that means Maxent
Species Model for wild dog (Cuon alpinus) in the Pench Tiger Reserve, Madhya
Pradesh is acceptable. In the Lower image of Fig. 14.18b which represents the curve
between Omission and Predicted area for wild dogs (Cuon alpinus), we can see the
red line which indicates the fraction of background prediction that is hyperbolic
14 Comparison of AHP and Maxent Model for Assessing Habitat … 353

(A hyperbola is formed by the intersection of a plane perpendicular to the bases

of a double cone) curve which indicates the Fractional value is mirror image of
Cumulative threshold value or in other words we can say that the set of all values
of Cumulative threshold and fractional value of this curve have such ratio of the
distance to a single focal point divided by the distance to the either Fractional value
or Cumulative Threshold value is greater than one.
Final outputs for training and test data with respect to the regularized and un-
regularized value. Regularized Training gain is 1.537, training AUC is 0.938. Un-
regularized training gain is 1.772 and un-regularized training gain is 1.129.
Table 14.5 interpret the relative advantages of environmental variables to the
Maxent species distribution model. Table 14.5 depicts the results which are useful
to decide the utterance of training algorithm, i.e., an addition to the usual advantage
beside the gain in the relative variables or if we deduct the gain in the relative
advantages from the result, the value of lambda will be in the negative in numbers.
Likewise, Table 14.5 depicts the values of environmental variables which are
training presence; the secondary data were randomly permuted. Further Table 14.5
also depicts that often extracting permuted data and consequently rapidly drop in
training AUC (Area under Curve) data were shown in percentage in which observe
that Bioclimatic variable 17 i.e., Precipitation of the driest quarter permuted 47.1%,
which is highest but when we go on second highest which bioclimatic variable 13 i.e.,
precipitation of the wettest month which permuted 13.3%. As we can observe there

Table 14.5 Percent contribution and permutation importance (in %) of every factor
Variable Percent contribution Permutation importance
Bio 17 27 47.1
LU/LC 25.3 9.1
Tree cover 11.3 2.5
Bio 13 10.8 13.3
Forest fire 8.3 3.3
Bio 1 5.8 3.3
NDVI 5.1 11.3
Bio 14 1.7 0.6
Aspect 1.3 1.6
Slope 1.3 2.5
Hill shade 0.9 1.6
Elevation 0.5 1.7
Bio 2 0.4 0.1
Bio 3 0.3 0.7
Bio 6 0 1
Bio 12 0 0
Bio 5 0 0
354 D. Jain et al.

is rapid drop in the Permutation importance of the all environmental variables. As

with the variable jackknife (Fig. 14.17), Variable contributions should be interpreted
when the predictor variables are correlated but with cautions when you interpret the
data.
Maxent Species Distribution Model also predicts the how much percent every
environmental variable in (Table 14.5) contribute in the habitat suitability of wild dog
(Cuon alpinus) and how much every environmental variable permutation importance
(in Table 14.5).
With the help of Table 14.5 we can conclude that Bioclimatic Variable 17 i.e.,
Precipitation of Driest quarter contribute 27%, Land Use/Land Cover contributes
25.3%, Tree cover contributes 11.3% are the major contributors to the suitable habitat
of wild dogs (Cuon alpinus) in the Pench Tiger Reserve, Madhya Pradesh.
With the help of Table 14.5 we can say that Permutation Importance again Biocli-
matic Variable 17 i.e., Precipitation of the Driest quarter is 47.1%, Bioclimatic
Variable 13 i.e., Precipitation of the Wettest month is 13.3% and the Normalized
Difference of vegetation Index (NDVI) which is 11.3% permutation of importance.

14.4.2 AHP (Analytical Hierarchical Process) Result

The motive of this study is to identify the suitable landscape of Pench Tiger Reserve
zones in the two districts (Chhindwara and Seoni) of the Madhya Pradesh in the area
protected for wild dog (Cuon alpinus) habitat. Weighted overlay model has been
used for the declination by integrating RS-GIS and AHP techniques. The motive of
using this integrated techniques to reveal the potentiality of the Pench Tiger Reserve,
MP through the overlay analysis of various influencing factors. In suitability model,
the result was generated on the basis of weightage given to each criteria based on
its importance. With the help of AHP (Analytical hierarchical process) tool box
technique we get various type of ranges of suitable habitat of wild dogs (Cuon
alpinus) in the Pench Tiger Reserve, Madhya Pradesh like Pink color zone which is
outside the protected area i.e., Pench tiger Reserve, Madhya Pradesh boundary. Main
reason of take 10 km buffer from Pench Tiger Reserve, Madhya Pradesh boundary
due to their large range of movement, now we can see in Fig. 14.19, Pink color zone
which shows maximum suitable landscape for the habitat of wild dogs (Cuon alpinus)
which covers East zone of buffer area (10 km from the east boundary of Pench tiger
Reserve, Madhya Pradesh) another second highest suitable landscape for the wild
dogs (Cuon alpinus) in the Pench tiger Reserve, Madhya Pradesh is represented by
dark blue color, which doesn’t cover as much area but AHP (Analytical hierarchical
process) predicts the very little area just behind the most suitable area that is in
pink color, we can say dark blue color zone is high suitable landscape for the Wild
dogs (Cuon alpinus). Another zone which occupies South-East and West region
of the study area with 10-km buffer boundary as per AHP (Analytical hierarchical
process) prediction which represents from light blue color moderately suitable habitat
landscape for the wild dogs (Con alpinus). Light blue color zone showing on three
14 Comparison of AHP and Maxent Model for Assessing Habitat … 355

Fig. 14.19 Habitat suitability of Cuon alpinus in Pench Tiger Reserve, MP (AHP prediction)

directions that is in South and South-West zone but outside the protected area i.e.,
Pench Tiger Reserve, Madhya Pradesh boundary. Most of South zone, South-West
and South-East part represents by light blue color and all these areas come under the
boundary of Pench Tiger Reserve, Maharashtra which clearly indicates that not Pench
Tiger Reserve, Madhya Pradesh region is suitable landscape for the habitat of wild
dogs (Cuon alpinus) even Pench Tiger Reserve, Maharashtra have also capability
to provide suitable landscape for the habitat of wild dogs (Cuon alpinus). Another
region which is suitable landscape for the habitat of wild dogs (Cuon alpinus) which
represented by green color, which covers the core zone of the Pench Tiger Reserve,
Madhya Pradesh. Main reason of core region of the Pench Tiger Reserve, Madhya
Pradesh is just suitable because that zone tree density is quite higher which create
hurdle to see prey and also hunting of the prey due to barriers which are trees another
reason is majority green color zone is near the Pench river and the huge Totladoh
reservoir specially on southern boundary of Pench Tiger Reserve Madhya Pradesh
and Pench Tiger Reserve, Maharashtra which share the Totladoh reservoir. We can
see very little portion of green color zone on the western boundary i.e., in Chhindwara
district region of the Pench Tiger Reserve, Madhya Pradesh. Another region which
is not as such suitable landscape for the habitat of the wild dogs (Cuon alpinus)
which represents from yellow color and occupies highest area in the among all zones
of the suitable landscape for the habitat of the wild dogs (Cuon alpinus). Majorly
356 D. Jain et al.

Yellow color zone occurs on those where human interference is high like villages,
agricultural lands and barren lands or in other words we can say built-up areas and
the last zone which predicted by AHP (Analytical hierarchical process) technique,
which represents from red color which indicates highly unsuitable landscape for
the habitat of the wild dog (Cuon alpinus) because in that zone lots of noise which
created huge human disturbance and Major source of noise is Railway. In the red
zone area Indian Railways spread their railway track network.

14.5 Discussion

The combine habitat suitability approach, as given in this study, is based on the inte-
grated use of models which are purely based on surrounded in the GIS (Geographic
Information System) and remote sensing environment and empirical evaluation
models. GIS and remote sensing were help to produce those data which needed in
the Maxent Species Distribution model and AHP (Analytical hierarchical process)
technique, as a platform to execute the models and in presenting the results of the
analysis. This study showed that many GIS (Geographic Information System) and
remote sensing approaches and AHP (Analytical hierarchical process) technique are
rapidly available for habitat suitability assessment for the wild dogs (Cuon alpinus).
The one of the major advantages of these approaches are connected to the possibilities
to consider suitable habitat factors on various scales, to combine suitability of habitat
assessment for wild dogs (Cuon alpinus), and to empirical model and the knowledge
of GIS (Geographic Information System) and remote sensing. It has been pointed
out in many recent studies that habitat requirements of certain species are affected
by factors measured on different scales (Areendran et al. 2020; Edenius and Elmberg
1996; Jokimäki and Huhta 1996; Saab 1999). Also, Wu and Smeins (2000) empha-
sized in their modeling approach the importance of consideration at various spatial
temporal scales. Wu and Smeins (2000) used various levels of multiple scales in their
study, but when the different scales were not to be incorporate; as an alternative, their
models were constructed disassociated on various different parameters. Most of the
recent suitable habitat models have been mainly constructed on a single parameter
and therefore they are not as suitable as tools in large landscape management or
conservation biology. In order to use the model’s habitat suitability evaluation, the
needed variables on appropriate parameters were calculated overall for the wild dogs
(Cuon alpinus) in the Pench tiger reserve, Madhya Pradesh In our approach, all the
environmental variables at different resolution scales.
We used the Maxent Species Distribution Model and AHP (Analytical hierar-
chical process) tool box for the prediction of the suitable landscape for the habitat of
the wild dogs (Cuon alpinus) in the Pench Tiger Reserve, Madhya Pradesh. Indian
landscapes have great capacity to nourish highly diverged biodiversity that why India
consider as one of the mega diverse country in the world among all 12 mega diverse
countries. Asiatic Wild dogs (Cuon alpinus) are found in huge range of South-East
Asia and East Asian countries but now due to excess of anthropogenic pressure
14 Comparison of AHP and Maxent Model for Assessing Habitat … 357

and climate change they found in very limited areas. In India they found majorly
in central Indian landscapes. Southern Indian landscapes and some time in north
India especially in Ladakh they sighted rarely. Wild dogs (Cuon alpinus) are play
vital role for the balancing of the ecosystem because wild dogs (Cuon alpinus) are
the top-level carnivore in the food web also wild dog (Cuon alpinus) are the only
species in his genus Cuon in the family Canidae. In India ancient time wild dogs
are found most of the part of the nation but now they are extremely confined to the
protected regions of the Indian landscapes to understand what are the factors affects
the habitat of wild dog (Cuon alpinus) in the Pench Tiger Reserve, Madhya Pradesh
with the help of GIS (Geographic Information System) and Remote sensing through
two different kind of models in which one is Maxent Species Distribution Model
and Another one is AHP (Analytical hierarchical process) technique. We found that
AHP (Analytical hierarchical process) technique give us various classes for the suit-
able landscape for the habitat of the wild dogs (Cuon alpinus) in the Pench Tiger
Reserve, Madhya Pradesh but on the other hand doesn’t give as much classes like
AHP (Analytical hierarchical process) but Maxent species distribution model give
how much Percentage contribution of each factor and give percentage permutation
importance of the every factor and Major reason to accept Maxent Species distribution
model is it gives Jackknifing of all environmental variables with respect to all other
variable in three case. First case of jackknifing of test data gain for wild dog (Cuon
alpinus), second case of jackknifing is jackknife of regularized training gain data for
wild dog (Cuon alpinus) and last third case is jackknife of AUC (Area under Curve)
for wild dogs (Cuon alpinus). Another major profit of Maxent species distribution
model is gives curves show that how each environment variable affects the Maxent
prediction. The curves (Fig. 14.20) shows how the predicted probability of presence
changes as each environmental variable is varied, keeping all other environmental
variables at their average sample value. If the variables are strongly correlated that
it is hard to interpret the curviforms. Maxent species distribution model depend on
correlation in evident ways in the curviforms. In the other words, curviforms shows
marginally changing only and only one variable whereas variables change together
in the Maxent species distribution model.
In contrary to the aforesaid marginal response curviforms, each of the following
curviforms represents a unique model, a Maxent species distribution model was
created using only corresponding environmental variables. The curviforms in the
Maxent species distribution model depicts the dependency on predicted adaptability
of the wild dogs (Cuon alpinus) in the Pench tiger reserve, Madhya Pradesh. The
curviforms also reflect selected as well as dependent environmental variables which
were weighted by correlations between the selected and other environmental variable,
which will easier to interpret if there is strong correlation between the two or more.
They may be easier to interpret if there are strong correlations between the variables
(Fig. 14.21).
358 D. Jain et al.

Fig. 14.20 a Predicted probability of presence changes as each environmental variable is varied.
b Predicted probability of presence changes as each environmental variable is varied. c Predicted
probability of presence changes as each environmental variable is varied
14 Comparison of AHP and Maxent Model for Assessing Habitat … 359

Fig. 14.21 a Plots reflect the dependence of predicted suitability both on the selected variable and on
dependencies induced by correlations between the selected variable and other variable. b Plots reflect
the dependence of predicted suitability both on the selected variable and on dependencies induced
by correlations between the selected variable and other variable. c Plots reflect the dependence
of predicted suitability both on the selected variable and on dependencies induced by correlations
between the selected variable and other variable
360 D. Jain et al.

14.6 Conclusion and Recommendations

This study gives a direct way to predict the landscape for the suitable habitat for wild
dog (Cuon alpinus) with the help of geospatial techniques with comparison of AHP
and Maxent model for assessing habitat suitability of wild dog (Cuon alpinus) in
Pench Tiger Reserve, Madhya Pradesh. Several environmental variables have been
used to predict suitable landscape for the suitable habitat for the wild dogs (Cuon
alpinus). Maxent Species Distribution model and AHP technique both depicts the
suitable habitat in the different classes but Maxent species distribution model gives
less number of classes as compare to AHP also Maxent gives some statistical data like
every environmental variable contribution for the suitable habitat for the Wild dog
(Cuon alpinus) in the Pench tiger reserve, Madhya Pradesh and another good thing
is Maxent also gives the permutation importance of every environmental factor like
for wild dogs (Cuon alpinus) in the Pench tiger reserve, Madhya Pradesh precip-
itation of the driest quarter i.e., 17th Bioclimatic variable play most vital role or
say most contributed and then Land use/Land cover respectively and when we talk
about highest permutation importance then again precipitation of driest quarter top
but now second highest is Precipitation of wettest month i.e., 13th Bioclimatic vari-
able. As we know that monsoon season is very crucial season for the wild dogs
(Cuon alpinus) because when rainy season starts wild dogs (Cuon alpinus) groups
or packs become shorter in the strength and start hunting in small packs and when
winter season is about to start, their denning season also start so they make small
packs with females to reproduce offspring’s. As per IUCN (International Union for
Conservation of Nature) Canid specialist group and IUCN Red list Wild dogs (Cuon
alpinus) are extinct from Uzbekistan, Singapore, Mongolia, Korea, Afghanistan,
Tajikistan, Russian federation, republic of Kyrgyzstan, Kazakhstan and possibly
extinct in the Vietnam also IUCN red list says that in some countries like Demo-
cratic People’s republic of Pakistan and Korea, in those countries presence of wild
dogs (Cuon alpinus) presence is uncertain or say not sure about their presence but
on the other hand there are some South-east Asian countries and Asians countries
those have Wild dogs in their forests or Protected areas like in Thailand, Myanmar,
Lao People’s Democratic republic, India, Cambodia, Bangladesh, Nepal, Malaysia,
Indonesia, China and Bhutan. As per IUCN red list data across the globe about 950
to 2200 mature wild dogs (Cuon alpinus) are remained. To conserve wild dog (Cuon
alpinus) countries government should take help from several wildlife organizations
and wildlife experts to conserve them and some strict rules and policies in the favor
of wild dogs (Cuon alpinus).
Another major work for government agencies is tried to stop people towards the
deforestation or stop to destruct forest for agricultural land creation.
The results of this paper will help to overcome the finding of those landscapes
which are more suitable for wild dogs (Cuon alpinus) in the Pench tiger reserve,
Madhya Pradesh and it will to the conservation agencies or organizations as well as
for the government also to especially conserve those areas which are most suitable
for wild dogs (Cuon alpinus).
14 Comparison of AHP and Maxent Model for Assessing Habitat … 361

Maxent species Distribution model and AHP prediction will help to local people
also because now they were aware about those places where wild dogs (Cuon alpinus)
so that they were not going in those regions for collection of fire wood and other
daily life using forest products and most important don’t take their cattle’s in that
zone for grazing because it will very harmful for them also.
This research methodology will help in the future very well because it helps to
understand to what are those factors which affect the habitat of the wild dogs (Cuon
alpinus) in the Pench tiger reserve, Madhya Pradesh and help to overcome those
factors so that in future wild dogs (Cuon alpinus) can conserve and another major
reason to this study will help in future also because as we know only 950–2200
mature wild dogs (Cuon alpinus) are remain across the globe so India become prime
country to conserve wild dogs( Cuon alpinus) in their landscape like India is very
much successful to conserve tiger (Panther tigris) in the central Indian landscapes
so it will very much helpful to government agencies which are working in the Pench
tiger reserve, Madhya Pradesh.
Every research have certain limitations like this research is entirely based on
present scenario but it helps many years because ecosystem and its constituents
takes time to change their conditions but future prospect of this research is help to
overcome or conserve those environmental factors which will affect currently habitat
of wild dogs (Cuon alpinus) and enhance their population to stabilize ecosystem or
food web in the Pench tiger reserve, Madhya Pradesh because as we already mention
that according to all India tiger report 2018 Pench tiger reserve, Madhya Pradesh has
the highest herbivore animal density so wild dogs (Cuon alpinus) plays vital role to
stabilize their population.

Acknowledgements I am also obliged to Arun Kumar (Programme Officer), Akhil Sivadas

(Programme Officer), Samrat Deb (Programme Officer), and Kumar Ranjan (Programme Officer)
from, IGCMC, WWF-India for helping me throughout the project duration. I would also like
to extend my thanks to Mr. Rajeev Kumar (Senior Programme Officer) and Mr. Sandeep Kumar
(Programme Officer) from ENVIS, IGCMC, WWF-India for their constant encouragement. I would
also like to thank Mr. Ravi Singh, CEO, WWF India and Dr. Sejal Wohra, Programme Director,
WWF-India for allowing me to carry out my project in the esteemed organization. I would also like
to extend my gratitude to the library staff of WWF-India for helping me with the resources for the
literature review.

References

Areendran G, Sahana M, Raj K, Kumar R, Sivadas A, Kumar A, Gupta VD (2020) A systematic

review on high conservation value assessment (HCVs): challenges and framework for future
research on conservation strategy. Sci Total Environ 709:135425
Bowdich TE (1821) An analysis of the natural classification of mammalia for the use of students
and travellers. Printed by J. Smith, Paris
Cohen JA (1977) A review of the biology of the Dhole or Asiatic wild dog (Cuon alpinus, Pallas).
Anim Regul Stud 1:141–158