Environmental Cost of Refugee Crisis: Case Study of Kutupalong-
Balukhali Rohingya Camp Site A Remote Sensing Approach.
S M Labib
1, Nazia Hossain
2 and Shahadath Hossain Patwary
1School of Environment, Education and Development (SEED), University of Manchester.
2Rajshahi University of Engineering and Technology
3Sheltech (Pvt.) Ltd
January 11, 2018
This paper explored the environmental cost of existing Rohingya Refugee crisis between Myanmar and
Bangladesh, utilizing remote sensing techniques. The focus was to determine the loss of vegetation in
setting refugee camp sites at Kutupalong-Balukhali expansion area. Change Vector Analysis has been
applied to determine the forest cover loss using LandSat 7 images. Then using carbon sequestration
capacity of these forest land, total loss has been estimated. It has been found that around 572 hectares
areas deforested to set up camp and that account for approximate loss of 365,288 GBP per year.
KEYWORDS: Rohingya, Refugee, Deforestation, Environmental Cost, Remote Sensing.
Geo-politics, wars, regional conflicts and extreme climatic events have triggered pressure on different
population groups, result in population displacement and created refugee crisis around the globe at
present. Apart from humanitarian crisis and threat to lives of millions, exodus of people from their
homeland is causing long term environmental damage and creating pressure on natural resources.
Existing literature suggest need for environmental monitoring using different methods, but majority of
them are time consuming, such as environmental impact assessments (EIAs). In contrast, Remote
Sensing tools and methods are often considered as an easy proxy to rapidly monitor of environmental
conditions and to determine possible cost for such population displacement (Price, 2017).
Sudden influx of Rohingya (a Muslim ethnic minority group in Myanmar), caused by recent event (on
25th August, 2017) of Arakan Rohingya Salvation Army (ARSA) attack against Military-police at
Rakhine state, triggered an ethnic cleansing of Rohingyas by Myanmar Military operations (Gee, 2017).
A total of 6,700 Rohingyas killed, UNHCR estimated 1.2 millions Rohingyas become refugees, and
among them 947,000 are in Bangladesh (BBC News, 2018). This sudden influx of Rohingyas cased
formation of many refugee camps (more than 10) in Teknaf cox's bazar areas (Figure 1) by cleaning
forest areas and cutting hills. Among these camps, IMO mentioned Kutupalong-Balukhali expansion
(KBE) is the largest camp accommodating 547,000 Rohingyas. In this study KBE site and surrounding
areas are considered to evaluate how much vegetation loss took place to provide shelter, the related
cost due to carbon sequestration loss.
2. Methods And Materials
Change Vector Analysis (CVA) is a radiometric difference based multivariate method, this method
identifies changes among multi-temporal images, by calculating difference vectors analysis unites, and
estimate direction as well as magnitude. CVA has the capacity to detect changes in “n” number of
bands, in contrast some other methods concentrated on selected bands only (e.g. Red, NIR). Hence, it
has abilities to overcome the cumulative errors in classification. Additionally, CVA has the capability
to provide better understanding about types of change happening within the area of interest (Johnson
and Kasischke, 1998). In this study, CVA has been applied due to better representation of all changes,
and as the author has no prior knowledge about study area.
Figure 1: Existing Rohingya Refugee Camps within Cox’s Bazar Bangladesh (Source: BBC News,
2.1 Change Detection Analysis
Step 1: Data Collection; LandSat Surface Reflectance (L7 ETM+) images of 04th December 2017 (Date
2), and 23th March 2017 (Date 1) have been downloaded from USGS website.
Step 2: Pre-processing; Several bands of the image were stacked together and then resized to focus on
study area. Radiometric correction of gain and offset value of 0.0001 applied to transform the Digital
numbers to reflectance values.
Step 3: Tasseled Cap (TC) transformation; CVA can be used of any number of bands, but as this
research is focused on vegetation change monitoring, using TC would allow for better interpretation of
change in vegetation based on brightness and greenness of TC (Fig 2). TC help reduce the redundant
information from all spectral bands, act as biophysical parameter, which can have >=95% variance
expressed within first two bands of LandSat TM; brightness and greenness (Siwe and Koch, 2008).
Step 4: CVA-Magnitude; The magnitude of change is Euclidean distance of pixel spectra between date
1 TC and date 2 TC bands of brightness and greenness. Equation 1 shows the CVA-magnitude
estimation process (Johnson and Kasischke, 1998).
Cm is Change Magnitude and brightness and greenness are for time 2 and 1.
Step 5: CVA-Direction; The direction of change indicate the type of change happed to the
corresponding pixels. This allow labeling land cover change to land type change. Direction angle (θ)is
estimated based on Equation 2 (Johnson and Kasischke, 1998).
θ=actangent ( greenness2-greenness1
Angel is initially estimated in radian, later on transformed into degrees (0-360). Essentially, angle
ranges (in four quadrants) indicate increase or decrease of brightness or greenness values between two
dates for the same pixel.
Using the knowledge of TC attributes, and range of angles associated with different changes, in this
case four land cover types were labeled after several trial and error process; to identify the change;
forest to forest (F To F), Forest to No Forest (F To NF), No water to water (NW To W) or Water To
Water (W To W), and finally No Forest to Forest (NF To F).
Step 6: Threshold Selection and apply threshold; In this study, mean (M) and standard deviation (SD)
of magnitude values been utilized to determine threshold, to indicate actual change and no change.
After experimenting with Mean+1SD, and Mean+2SD, the author selected Mean+1SD values of all
magnitude values as threshold value, Karnieli et al., (2014) showed use of 1SD provide realistic change
and no-change pixels.
2.2 Carbon Sequestration Loss Estimation
Using carbon sequestration capacities of forest cover per unit area, and per unit price of Carbon-tonne,
the estimation of overall loss in terms of carbon storing has been conducted using equation 3.
Closs = FA* S* PC
Closs= Carbon loss; FA is Forest Area in Hectare, S denotes Sequestration capacity (hectare/year), PC
is Per unit cost of Carbon-tonne in USD.
According to Bangladesh Forest Department (BFD), Sequestration capacity of the forest areas in
Teknaf, is 43.08Carbon-tonne/Hectare/Year (Bforest.gov.bd., 2018). Additionally, Luckow (2014),
indicated per unit cost of carbon-tonne under mid scenario would be 20 USD up-to year 2020, this price
has been considered in this study.
3. Results And Discussions
Figure 2 (a, c) shows the standard FCC of the study area in two different periods, and Figure 2 (b,d)
shows the TC images, and both indicate, the increase in white or brightness. CVA analysis provided
the magnitude of changes (Figure 3a), and after applying the threshold value, the real change pixels
where the intensity of change are significant have been found (Figure 3b), the mask image created from
this threshold would apply on the classification image obtained based on the classification conducted
on the direction image (Figure 4a). The final changed areas have been illustrated in Figure 4(b), the red
areas shows the areas loss major vegetation, and these are the areas where the deforestation took place
due to cutting trees and hills to set the shelter.
Figure 2: (a) FCC (NIR, R, G) of Data 1, (b) Tassel Cap (TC) bands for Date 1 (c) FCC (NIR, R, G) of
Data 1 (d) TC bands for Date 2.
Figure 5a illustrates, within 0-90 range, no change in forest cover observed, but between 90-180 range
major deforestation took place. Increase in brightness values, and decrease in greenness in the 2nd
quadrant indicate more exposed land, and loss of vegetation. Quantitatively the among 5332.32 hectares
of study area, 572.85 hectares were previously forest cover areas been transformed into temporary
shelter for refugees in KBE areas. This account for total 10.74% loss in forest cover (Figure 5b).
Previously Khatun (2017) showed the deforestation in the study area 382.42 hectares, according to
BFD, which was clearly an underestimation. When visually checked (due to no access of field level
accuracy assessment data) with high-resolution images, areas indicated deforestation in this study,
matched more accurately with the camp sites.
Figure 3: (a) Magnitude image showing intensities of change between data 2 and date 1 (c) Mask image,
using M+1SD threshold indicating the real change pixels.
Figure 4: (a) Direction Image, After classification using Analyst’s expertise and knowledge of TC. (b)
Masked classified direction changes, showing changes over true colour image of Date 1.
Figure 5: (a) Change detection from Direction image (b) Percentage of change in different classes.
The loss of 572.85 hectare account for total loss of 24,678.35 tonne of carbon sequestration per year,
and if this camp exists for 3 years, and reforestation initiatives need 5 years to revive the loss vegetation,
it would account 197,427 tonne-carbon sequestration loss. As mentioned earlier with account for
20USD per tonne-carbon, the loss would be 493,567 USD/year (365,288.94 GBP, 1 USD = 0.74 GBP),
and total loss in eight years would be 1,579,416 USD (≈1,167,204.22 GBP).
4. Conclusion and Future Works
Rohingya exodus has created massive humanitarian crisis, and the value of human lives are
unmeasurable in this context. Bangladesh has done a great job in saving lives of millions, and the
environmental cost associated with this is low, compared to the total value of saving the lives. However,
it is critical to understand that, displacement of such larger population causing immediate and also long
term negative impacts on the environment, and this need careful examination to create pressure on the
authorities to resolve such situation. In this regard, use of remote sensing could be very effective and
timely, despite the fact this study provided a crude estimation. Further studies may improve the cost
estimation by integrating ground water resource depletion, using temporary waste generation and
further validate the results applying proper accuracy assessment.
S M Labib is a first year PhD student at SEED, University of Manchester. His research interest include
environmental management, use of GIS and RS in monitoring ecosystem services, and Urban Green
Nazia Hossian is a Lecturer, at Dept. of Urban & Regional Planning in RUET. Her research interest
include disaster management, conflict and crisis analysis, and Urban Planning.
Shahadath Hossain Patwary is an Assistant Urban Planner. His research interests are urban housing,
and environmental management.
BBC News. (2018). Is refugee crisis 'textbook ethnic cleansing'?. [online] Available at:
http://www.bbc.co.uk/news/world-asia-41566561 [Accessed 8 Jan. 2018].
|Forest Department-Government of the People's
Republic of Bangladesh |
. [online] Available at:
http://www.bforest.gov.bd/site/page/8de0afe8-1bf1-4003-aff2-1c4758f545fa/ [Accessed 2 Jan.
Gee, J., (2017). Bangladesh, Southeast Asia Worried by Rohingya Crisis. The Washington Report on
Middle East Affairs, 36(7), pp.39-40.
Johnson, R. D., & Kasischke, E. S. (1998). Change vector analysis: a technique for the multispectral
monitoring of land cover and condition. International Journal of Remote Sensing, 19(3), 411-426.
Karnieli, A., Qin, Z., Wu, B., Panov, N., & Yan, F. (2014). Spatio-temporal dynamics of land-use and
land-cover in the Mu Us sandy land, China, using the change vector analysis technique. Remote
Sensing, 6(10), 9316-9339.
Khatun, F. (2017). Implications of the Rohingya Crisis for Bangladesh. Presentation, Centre for Policy
Dialogue (CPD), Bangladesh.
Luckow, P., Stanton, E.A., Biewald, B., Fields, S., Jackson, S., Fisher, J. and Ackerman, F., (2014).
CO2 Price Report, Spring 2014. Synapse Energy Economics Inc., Cambridge, MA.
Price, R.A. (2017). Environmental Impact Assessments in refugee crises. K4D Helpdesk Report.
Brighton, UK: Institute of Development Studies.
Siwe, R.N. and Koch, B., (2008). Change vector analysis to categorise land cover change processes
using the tasselled cap as biophysical indicator. Environmental monitoring and assessment, 145(1),
Tewkesbury, A. P., Comber, A. J., Tate, N. J., Lamb, A., & Fisher, P. F. (2015). A critical synthesis of
remotely sensed optical image change detection techniques. Remote Sensing of Environment, 160,