Institute of Forestry, Tribhuvan University, Nepal
Submission: February 18, 2020; Published: March 04, 2020
*Corresponding author: Ram Kumar KC, Tribhuvan University, Institute of Forestry, Pokhara, Nepal
How to cite this article: KKC R, Mahato DB, Yadav NK, Poudel P . Mapping Deforestation and Forest Degradation Using CLASlite Approach (A Case Study from Maya Devi Collaborative Forest of Kapilvastu District, Nepal). Int J Environ Sci Nat Res. 2020; 23(5): 556122. DOI: 10.19080/IJESNR.2020.23.556122
Mapping and monitoring of forest area suffering from deforestation and forest degradation using satellite images and remote sensing has been an achievable activity for the sustainable forest management and conservation. Mapping of disturbance and degradation of forest is gaining momentum through Remote Sensing despite, major challenges still exist. The present study was conducted to quantify the forest area of deforestation and degradation within the Maya Devi collaborative forest of Terai region located at Kapilvastu district of Nepal. This study based on the optical satellite data (Landsat 7 ETM of 2000 and Landsat 8 OLI/TIRS of 2016) and spectral un-mixing of these datasets which produced fractional cover (proportion of vegetation, non-photosynthetic vegetation and bare soil). Under the environment of CLASlite tools, mapping of damaged canopy, exposed soil and dead vegetation were produced. With the gap of 15 years two Landsat Image of 2000 and 2016 were processed to estimate. Results shows that the study area has 60% intact forest, 23% less degraded forest, 12% moderately degraded forest and 5% highly degraded forest. In sum, only 8.01 hectare of forest has been cleared and degraded. Finally, CLASlite approach based upon medium resolution satellite images and ground sample plots could practicable in order to monitor the forest degradation and deforestation.
In global, Nepal forests cover is about one-third of land which represents the 44.74% of total land of Nepal only. Comprises of 118 ecosystem that contributes essential service to the human being’s survival. Forests becomes a main hub to encroached in the name of relief, resettlement, homestay as well as playground for the government offenders, urban expansion, national development infrastructure which ultimately accelerated the forest deforestation. There are still some communities those rely completely in the illegal felling and selling that to the market for the daily survival. and poorly understanding the root causes of widespread forest degradation in developing countries . Forest resources are crucial and increasing population have heavy pressure on the forest of Nepal in order to fulfill the subsistence need like fuelwood, fodder and land use changes. Forest conservation are more directly relevant in the case of Nepal as forest resources are significant for ecosystem balance and people’s livelihood .
According to FRA  deforestation is the conversion of forest to another land use or the long-term reduction of tree canopy cover below the 10% threshold. The worldwide deforestation continues at an alarmingly high rate: 16 million ha per year . In Asia, a net loss of about 0.6 million hectares per year occurred in the 1990s; however, from 2000 to 2010 a net gain of more than 2.2 million hectares per year was reported, primarily due to large-scale afforestation efforts, particularly in China, but high rates of net loss continued in many countries in South and Southeast Asia .
Nepal has in total 5.96 million ha of forests, which represents more than one-third (40.36%) of total land . It comprises of 3 physiographical zones among which Terai is one having subtropical to tropical eco-climatic zone and it covers 17% of total land. High population growth, unmanaged settlement, unemployment, encroachment, grazing and forest fire are some of the underlying causes of the reduction of forest resources. Brampton  suggeststhat Terai forests have a higher deforestation rate than the middle
hill forests of Nepal. According to the National Forest Inventory,
annual average deforestation rate in Nepal from 1978 to 1994 was
about 1.7% . Similarly, global forest loss was estimated 1.7% of
deforestation for the period of 1990 to 2005 . A case study in 20
Terai districts showed forest cover loss with annual rate of 0.6%
between 1990 to 2000 . According to a recent forest assessment
report, the forest area in the Terai decreased by 16500 ha with
annual rate of 0.44% between 2001 and 2011 . This shows that
there is a good deal of pressure on forest especially on those in the
Terai and Inner Terai.
Remote sensing is a very powerful tool, widely accepted and
growing implication for the monitoring of forest degradation
studies. It involves the acquisition of remotely sensed data about
an object, area or phenomenon then undergoing through the
analysis of data acquired by a device that is not in contact with
the object, phenomenon or area under investigation . Remote
sensing data with sufficient spatial and high-resolution satellite
datasets are worthful and reliable resources for forest cover
mapping and monitoring .
Landsat Data with (30 by 30) m resolution free data are used
for the forest cover mapping and application of CLASlite program
was used to detect the forest deforestation and disturbance in
the tropical region . Deforestation is defined as permanent
absence of forest cover (decrease of photosynthetic vegetation
and increase of soil substrate). Meanwhile, disturbance refers
to patches of non-forest area in forested area, as well as natural
non-forested vegetation such as grasslands and shrub lands
and thinning of forested area . Automated Monte Carlo
Unmixing (AMCU) is a promising technique to generate fractional
cover of forest canopies and its time series study using CLASlite
can indicate the various classes and level of degradation. This
technique provided a reliable method of monitoring of forest
The main aim of this study is to map out the deforested and
disturbed forest area of the study site on the basis of fraction cover.
The factors which could accelerate the deforestation and forest
degradation process have not been studied well in the Nepalese
context yet. Though, many studies have been done on forest
canopy and its dynamics, afforestation, restoration, deforestation
and forest degradation in the different parts of the world, only few
studies have been carried out in Nepal. Therefore, identifying the
problems of deforestation, degradation and forest disturbance
at different spatial and temporal scales could provide useful
information for sustainable managements of forests, define proper
polices and strategies, implementation plans and addressing the
real issues of deforestation and its coping mechanism.
This study (Figure 1) has been carried out in Mayadevi
Collaborative Forest of Kapilvastu district in the Southern parts of
Province 5. It lies between 27°35’ 44.92” N to 27°44’ 35.96” N and
83°10’ 17.13” E to 83°13’ 50.47”E. Its elevation range is between
94-248m above mean sea-level. The maximum temperaturerecorded here is 43°C and minimum temperature is 4.5°C. It is
divided into 3 blocks: Pipara block, Prativa East block and Patna
block with a total area of 1721.27 hectares. Climatically, the study
area falls under a Tropical monsoon climate and mainly coarse
clay/ loamy soil is found here. Primarily this forest is a Sal forest.
However, other species like Bhotdhairo (Lagerstroemia parviflora),
Asna (Terminalia alata), Kusum, Harro (Terminalia chebula),
Barro (Terminalia bellirica), Sindure (Mallotus phillippinesis),
Karma (Adina cardifolia), Jamun (Syzygium cumini), Sissoo
(Dalbergia sissoo), Khair (Acacia catechu) etc. are also found in the
research area. The forest range is surrounded by agriculture and
rural settlements. There is disturbance due to grazing by cattle,
illicit felling, road widening and dying trees in flooding of lowlying
I order to mapping of deforestation and forest degradation
demands data from different sources such as satellite images, topo
sheets and various tools were used to carry out image analysis
and conclude the findings. Data, Tools and Software used for this
study are presented in Table 1.
Till today there are variety of methods developed and applied
in Nepal for the assessment of the forest degradation. Aerial
Photography, Field Surveys Satellite Image analysis and GIS
and Ecosystem service valuation out of this all Satellite Image
analysis and GIS are more feasible in terms of uniformity, costing,
and accuracy . Following the flowchart from Figure 2 forest
degradation and deforestation maps were produced.
The boundary of study file was been digitized from
Topographic map from Survey department of Nepal. CLASlite 3.2
was used for image processing. The Raw Digital Number (DN)
images downloaded from USGLOVIS (www.glovis.com) were
calibrated to reflectance using gain and offsets value available
in metadata. The result of radiometric calibration is an image in
units of radiance (i.e. watts per square meter per units of solid
angle), also known as the energy measured by the satellite-based
sensor. The atmospheric players are aerosols, water vapor and
other gases, like oxygen and ozone. These constituents scatter
and absorbs radiated energy to various extents at different
wavelengths. This means that the sensor cannot detect everything
that gets reflected off the Earth’s surface. CLASlite applied an
automated atmosphere correction model and converts the results
to reflectance images .
The CLASlite tool has been used to generate fractional cover
of both the years, sequentially following the methodological
framework as shown in Figure 2. CLASlite adopts a Monte Carlo
method, whereby the possible combination of the end member
spectra is pre-computed and are applied during the Automated
Monte Carlo Unmixing run. An advantage of the Monte Carlo
approach is that the per-pixel iterations produce a standard
deviation of the estimate for PV, NPV and bare substrate fractions.
The process of random selection is repeated up to 50 times or
until the solution converges to a mean value for each surface cover
fraction. This technique considers that each pixel is a combination
of certain pure elements [i.e. vegetation (PV), Soil-vegetation
(S), Non-Photosynthetic vegetation (NPV), Shade Burnt] that
combine to produce a given response per pixel. Each pixel will
be characterized by a percentage of each pure element (i.e. 80%
Vegetation, 10% Soil, 5% NPV and 5% Shade) and the interpretercan choose the thresholds of each element that will help classify
the image (i.e. those pixels with ≥60% of Vegetation-pure element,
will be classified as vegetation cover).
A technique that combines spectral and spatial information to
enhance the detection and mapping of canopy damage, exposed
soil and dead vegetation has been used. Comparing with spectral
indices, NDFI are more reactive to Tropical Forest disturbance
that make ease in the identification . Following Equation
given below has been used to calculate NDFI. The output layers
of CLASlite i.e. PV, NPV, and Bare Soil is used as input parameters.
Where, PV is photosynthetic vegetation, NPV is nonphotosynthetic
For the validation, we looked at five different measures: error
matrix, overall accuracy, user’s accuracy, producer ‘s accuracy and
kappa coefficient in order to evaluate the final maps. The manual
interpretation from 2016 was used for the accuracy assessment
since it consists of the most up-to-date information about the
forest extent in the study area. The required number sample
points have been collected from study area which has been used
as reference data.
Figure 3 shows the fractional cover of 2000 in which PV, NPV
and bare soil are expressed in percentage (0-100%). In 2000,
there is, 5-100 % vegetation, 0-72% of non-vegetation 0-49% of
Similarly, Figure 4(b) and 6(b), Band 1 represents Fractional
cover of BS which is displayed in red color, while Band 2
represents Fractional cover of PV which is displayed in green, and
Band 3 represents Fractional cover of NPV which is displayed in
Blue color. The intensities of each represent the presence of each
cover type in each pixel. For instance, green pixels have a higher
percentage of PV, red and yellow pixels indicate the presence of
BS and PV, while blue pixels represent higher fractional coverage
The RMSE image (Figure 4c) shows the geographic areas of
relatively high overall uncertainty in the modeling: for example,
roads and settlements have the largest errors approaching 6%.
Similarly, Figure 5, shows the fractional cover of 2016 in
which PV, NPV and bare soil expressed in percentages (-1 to
100%). In 2016, there is -1 to 100% of Vegetation, -1 to 84% of
Non-Vegetation and -1 to 77% of Bare which shows there is more
open and deforested area.
The RMSE image (figure 6c) of 2016 ranges from -1 to 5
indicating the highly accurate classification of PV, NPV and Bare
The range of NDFI lies between -1 to +1. A positive value
shows high PV implying less NPV and bare soil indicating less
disturbed forest while a negative value shows high NPV and bare
soil indicating highly disturbed forest. Figure 7(a) shows that the
range of NDFI in 2000 lies between -0.9 to 0.941748 while in 2016
Figure 7(b) the range is between -1 to +1.
In the Figure 8, it’s represents the final deforestation and
disturbance map which is derived from image analysis of two
different years 2000 and 2016 using ClASlite package. The result
shows that out of total area 8.01 hectare of forest has been
deforested from the total area.
From Figure 9, forest degradation which is composed by
overlaying of reclassifying change map of NDFI from 2000 and
2016. The result shows that 60% intact forest, 23% less degraded
forest, 12% moderately degraded forest and 5% highly degraded
From Table 2 it can be clearly conclude that the overall
accuracy of this study is 84.61% with kappa Coefficient of
agreement value 0.793, giving an indication of highest degree
of agreement between user defined class and referenced class
(ground truth class).
The Mayadevi Collaborative Forest which is dominated by
Sal forest is subjected to deforestation and degradation due
to various factors natural and man-made. LANDSAT images
are better suited for studying about forest as it has NIR, SWIR,
thermal bands. CLASlite is an automated tool based on Automatic
Monte Carlo Unmixing (AutoMCU) for separating PV, NPV and
Bare using pre-computed spectral signature. An advantage of the
Monte Carlo approach is that the pre-pixel iterations produce a
standard deviation of the estimate for PV, NPV and bare substrate
fractions. It also generates a root mean square error (RMSE)
image of the model versus observed reflectance signature whichexpressed in percentage. The RMSE and the standard deviation
images provides a way to assess the performance of the AutoMCU
on a pixel by pixel basis, allowing to identify areas of concerns.
Results show that study area has 60% intact forest, 23% less
degraded forest, 12% moderately degraded forest and 5% highly
degraded forest. In total, only 0.46% (8.01Ha) land was completely
deforested out of total area (1721.27Ha). The overall accuracy
of this study is 84.61% with Kappa Coefficient of agreement
value 0.793, giving an indication of highest degree of agreement
between user defined class and reference class (ground truth
class). The fractional cover of study area shows that Landsat OLI/
TIRS 8 of 2016 has low RMSE (i.e. 0 to 5%) than Landsat 7 ETM
of 2000 (i.e. 0 to 6%) because it has two thermal bands in which
one of them is uses for quality assessments (QA). The roads and
agricultural land shows largest error i.e. 6% while the degraded
areas show very low error.
The change in fractional cover from one time to other
included change in natural (felling, logging, lopping) as well as
anthropogenic change (conversion of forest to agriculture and
settlements). The study suggests that NDFI considers only PV,
NPV, and bare whereas fails to analyze the structural changes in
vegetation, this disadvantage is overcome by using SAR and highresolution
satellite images. CLASlite automated tool will be much
useful for operational purpose, which need limited classification
but has greater impact to protect forest degradation.
DFRS (1999) Forest Resources of Nepal (198 -1998). Kathmandu, Department of Forest Research and Survey, Ministry of Forests and Soil Conservation and Forest Resource Information System Project, Government of Finland.
FAO (2012) FAO Forestry Paper No. 169. Rome, FAO: Agriculture Organization of the United Nations and European Commission Joint Research Centre.
DoF (2005) Forest Cover Change Analysis of the Terai Districts, 1990/91-2000/01. kathmandu, Nepal: Department of Forest, Ministry of Forest and Soil Conservation.
Lillesand TK (2004) Remote Sensing and Image Interpretation. (7th edn), John Wiley: Chichester.