This article was downloaded by: [University of Nebraska, Lincoln]

On: 06 April 2015, At: 20:57

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Geocarto International
Publication details, including instructions for authors and subscription information:
[Link]

Using remote sensing data to monitor land cover

changes near Afghan refugee camps in northern
Pakistan
a b c
Mahtab A. Lodhi , Fernando R. Echavarria & Chris Keithley
a
Center for Advanced Land Management Information Technologies, Conservation and
Survey Division, Institute of Natural Resources , University of Nebraska , Lincoln, NE,
68588–0517, U.S.A.
b
Department of Geography , University of Nebraska , Lincoln, NE, 68588–0135,
U.S.A.
c
Center for Advanced Land Management Information Technologies, Conservation and
Survey Division, Institute of Natural Resources , University of Nebraska , Lincoln, NE,
68588–0517, U.S.A.
Published online: 17 Sep 2008.

To cite this article: Mahtab A. Lodhi , Fernando R. Echavarria & Chris Keithley (1998) Using remote sensing data to
monitor land cover changes near Afghan refugee camps in northern Pakistan, Geocarto International, 13:1, 33-39, DOI:
10.1080/10106049809354626

To link to this article: [Link]

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”)
contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any
purpose of the Content. Any opinions and views expressed in this publication are the opinions and views
of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources of information.
Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs,
expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in
connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or
systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in
any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
[Link]/page/terms-and-conditions
 Using Remote Sensing Data to Monitor Land Cover Changes
Near Afghan Refugee Camps in Northern Pakistan

Mahtab A. Lodhi
Center for Advanced Land Management Information Technologies,
Conservation and Survey Division, Institute of Natural Resources,
University of Nebraska-Lincoln, NE 68588-0517, U.S.A.

Fernando R. Echavarria
Department of Geography, University of Nebraska-Lincoln, NE 68588-0135, U.S.A.

Chris Keithley
Center for Advanced Land Management Information Technologies,
Conservation and Survey Division, Institute of Natural Resources,
Downloaded by [University of Nebraska, Lincoln] at 20:57 06 April 2015

University of Nebraska-Lincoln, NE 68588-0517, U.S.A.

Abstract
By the early 1990s, the civil war in Afghanistan had forced more than three million Afghan refugees to resettle in
neighboring Pakistan. Although previous studies have noted that refugees have had a negative impact on natural
resources in the vicinity of resettlement camps, there has been little work done to quantify these impacts. The
research presented here examines the clearing of forests in the Siran Valley, Pakistan, where 111,000 refugees were
settled in ten camps between 1978 and 1993. Multi-temporal satellite images were georeferenced using 1:50,000
scale topographic maps of the study area. A digital elevation model (DEM) was created to stratify by altitude the
unsupervised classification of the two images used in the change detection analysis. The elevation data were included
in the classification process to reduce the spectral confusion between agricultural and forest land cover classes. The
comparison of the satellite images for the two dates show a dramatic decline of forested land. Procedures similar to
those presented here can be used to assess the environmental impact of refugee movements in many environments.

Introduction decade in densely populated mountainous watersheds

coincides with a period of considerable decline for the already
Satellite remote sensing technology provides a unique depleted forest resources of the region.
historical record to examine spatial and temporal changes in The main objective of this study is to explore the utility of
land cover. Using satellite images, various types of land multispectral satellite data to quantify the degree and extent
cover change analyses have been performed, including the of forest elimination in the Siran Valley. The specific
assessment of deforestation, crop stress detection and scores objectives include: 1) the assessment of pre- and post refugee
of other environmental studies (Skole and Tucker, 1993). forest area based on digital analysis of satellite data; and 2)
The integration of remotely sensed and ancillary data sources the identification of the most affected forest types.
in geographic information systems (GIS) has further
strengthened the capability of remote sensing technology for Study Area
monitoring and management of the land resources at a
multiple of scales. Integration of these approaches offers The Siran Valley is located in NWFP between 34° 21' to
new opportunities for rapid and cost-effective analysis and 34° 48' north latitude and 72° 55' to 73° 19' east longitude,
display of geographic data. and covers an area of 1,675 km2 This study is focused on
The civil war in Afghanistan caused over three million only 618 km2 area of the central Siran Valley. The valley
Afghan refugees to resettle in Pakistan (Allan, 1987; comprises a major part of the Siran River watershed. The
Weinbaum, 1993). The majority of these refugees were Siran River rises in the north of Bhogarmang Valley, it flows
settled in the North Western Frontier Province (NWFP). in a southerly direction and passing through the Pakhli Plain
This province contains over 75 % of Pakistan's forests. The and Tanawal Hills drains into the Tarbela Dam reservoir
presence of large number of refugees for more than one (Figure 1).

33
Geocarto International, Vol. 13, No. 1, March 1998
Published by Geocarto International Centre, G.P.O. Box 4122, Hong Kong.
Downloaded by [University of Nebraska, Lincoln] at 20:57 06 April 2015

0 20
$ Location of refugee camps A Elevation in meters
Figure 1 Landsat TM image showing the location of study area in Siran Valley

Forest Resources of Siran Valley lower limit to higher elevations). Major pine species are chir
(Pinus roxburghii) and khail (blue) (Pinus wallichiana).
Extending altitudinally from the plains to an elevation of Chir pine grows on warmer and drier aspects, whereas, khail
about 900 m, Acacia Ziziphus and Acacia Olea, Acacia pine occupies the cooler aspects. From an elevation of 900 m
Modesta, and Capparis Decidua collectively form the Low to 1000 m, chir pine grows mixed with subtropical dry
Hill Thorn or Tropical Dry Deciduous forests (Naqvi, 1976). evergreen species of olive {Olea cuspidate), phulai (Acaia
The predominant feature of these forests is their xerophytic modesta) and sanatha (Dodonaea viscosa). Chir pine grows
character. The trees and shrubs are mostly spiny with small in pure stands from 1000 to 1600 meters, and is found mixed
deciduous or thick leathery leaves. These forests are not a with blue pine up to an elevation of 2000 m. Khil pine grows
source of commercial timber extraction, but are present in a exclusively an elevations between 2000 m and 2300 m.
very degraded form due to increasing biotic pressure for fuel From 2300 m to 2500 m Khil pine is found mixed with
wood extraction and livestock grazing. deodar (Cedrus deodara) (Jan, 1982; Hussain and Ilahi,
1991).
Pine and fir are the two main types of timber forests in the
Siran Valley (Champion et al., 1965; Jan, 1982). The pine The major species of the fir forest type are fir (Abies sp.),
forest is found from roughly 900 to 2200 m (currently in spruce (Picea sp.) and deodar (Cedrus deodara). These
some areas, the encroachment of agricultural land has pushed species grow from about 2500 m to 3000 m. Deodar dominates

34
 between 2500 m and 2800 m, above which the proportion of in Pakistan reached to 2.8 million, and exceeded 3.5 million
fir increases. Fir and spruce dominate between 2800 and by the end of the decade. In some parts of NWFP, such as the
3000 m. Broadleaved species such as oak (Quercus sp), tribal districts of Orakzai, Kurram and North Waziristan
walnut (Juglans regia), horse chestnut (Aesculus indica) and Agencies 1 , they outnumbered the local population
maple {Acer pictum) are confined to stream beds and cool (Weinbaum, 1993). During this time, there were about 350
sheltered places between 1800 m to 3000 m (Jan, 1982). refugee camps, with population ranging from about 10,000
persons to approximately 120,000.
Impact of Refugees on Forest Resources The arrival of refugees also exerted profound impact on
the ethnic composition of the local population, which
In 1989 the population of Siran Valley was 569,000, with eventually resulted in sectarian tensions in Kurram and North
91% living in rural areas. An additional 111,000 refugees Waziristan Agencies. As a result, Pakistani authorities
were resettled in ten refugee camps, bringing the total decided to move some refugee camps away from the western
population to 680,000. The influx raised the population border areas to the north and northeastern districts of NWFP.
density of the valley to 747/km2. The rapid increase of local This resettling brought refugee camps closer to the densely
population coupled with the influx of refugees increased forested regions of Chitral, Dir, Swat, and Hazara (Figure 2).
pressure on the forest resources of the region (SFDP, 1991). Hence, these areas have sustained the greatest environmental
The major influx of refugees into Pakistan began in April impact, due to the increase in human and livestock populations
Downloaded by [University of Nebraska, Lincoln] at 20:57 06 April 2015

1978 and by March 1981, approximately 1.8 million refugees (Allan, 1987).
had been arrived in Pakistan. By 1984, the number of refugees Allan (1987), discussed the impact of refugees on the

Dorah Pan

36V

Abbottabad
V
ABBOTTABAD
I
Havdian

Principal Refugee Routn

— . , _ _ Afghanistan—Pakistan Border
—•—•—• Karokorum Hiahway,
3000m Contour lint
Refugee Camp (Abased on present study)
Seriously Deforested Areas

Figure 2 North Pakistan showing the areas impacted by Afghan Refugees (modified after Allan, 1987)

1 A term used for administrative sub-divisions in federally controlled semi-autonomous areas of Pakistan. The "Political Agent"
of the federal government generally is responsible for the agency administration.

35
 forest resources of northern Pakistan. He stated two main Image Classification
causes of forest destruction: a) the felling and burning of Landsat-MSS band 5, 6, 7 (600 - 1 lOOnm) and TM band
trees for the creation of agricultural land, and b) the lopping 3, 4, 5 (630 - 1750nm) were selected for digital analysis.
of trees for fuel wood and excessive tapping for resin These bands tend to enhance the spectral properties of
extraction. He also observed that refugees of agricultural vegetation. An unsupervised classification scheme was
background from interior and northern Afghanistan caused chosen to partition the data into forest and non-forest land
more damage to NWFP forests than refugees from the cover classes. This classification scheme utilized a non-
northeastern forested regions of Afghanistan. Therefore, the hierarchical clustering procedure known as ISODATA
refugees' impact on forests in Pakistan varied according to (Interative Self Organizing Data Analysis). The procedure
the type of environment from which they came and the type minimizes the variances within clusters while maximizing
of environment in which they resettled. the spectral distance between the clusters by adjusting the
In 1991, government of NWFP conducted a study to cluster means at each iteration, and grouping pixels of a
assess the forest resources in Siran Valley. The study indicated multiband image into N number of spectral classes, where N
the connection between the establishment of refugee camps is a function of the input parameters and the variance in
and clearing of forests (SFDP, 1991). The impact of refugees spectral signature throughout the scene (ERDAS, 1994).
was more pronounced on Guzara forests (forests for use of The operator defines the maximum number of classes, the
the local population). These forests do not have the same maximum cluster radius and the minimum distance between
Downloaded by [University of Nebraska, Lincoln] at 20:57 06 April 2015

level of legal protection against illegal cutting as do the cluster centeroids. Both radius and distance are the
public forests. The lack of regulations have made Guzara expressions of the spectral distance in Euclidean P-space,
forests highly vulnerable to rapid depletion. According to where P is the number of bands within the imagery.
this study, in 1979 approximately 850 km2 area was under For each classification, the actual number of classes created
forest cover. Ten years later, only 470 km2 of forests remained, by the clustering algorithm was 20. These clusters were then
a decrease of 380 km (or 45%) of forest cover in a single collapsed into fewer classes of interest i.e.. forest, scrub land
decade. and crops. This was performed by comparing the classified
MSS and TM images on pixel by pixel basis with aerial
Methodology photographs, maps of the forested land acquired by the
NWFP Forest Department, and the expert knowledge of the
Data Preprocessing region's vegetation. The labeling of vegetation classes into
Landsat Thematic Mapper (TM) digital data for June 17, the classes of interest within MSS image was accomplished
1993, were acquired from Space and Upper Atmosphere satisfactorily. However, difficulties encountered in labeling
Research Commission (SUPARCO), ground receiving vegetation classes within TM image due to spectral confusion
station, Islamabad, Pakistan. Landsat Multi Spectral Scanner between forests and agriculture on the Pakhli Plain.
(MSS) digital data for June 9,1979 were purchased from the Although, both MSS and TM images were acquired during
U.S. Geological Survey, EROS Data Center, Sioux Falls, the growing season (cloud and snow cover limited the choice
South Dakota, USA. of imagery acquisition date), the problem of spectral
The image processing was performed at the Center for confusion due to crops on the Pakhli Plain was encountered
Advanced Land Management Information Technologies only within the TM image. The summer crops in the Siran
(CALMIT) and the Geographic and Environmental Valley, and all mountainous watersheds in northern Pakistan
Information Systems Laboratory (GENISYS), of the are entirely depended on the timely arrival of the monsoon
University of Nebraska-Lincoln. The Environmental Systems' rains. The summer of 1979 experienced drought conditions
Research Institute's ARC/INFO GIS (ESRI, 1992), and due to late arrival of the monsoon rains. Thus in MSS image,
ERDAS's IMAGINE software were utilized for image all agricultural land was laying fallow at the time of
processing and digital analysis. A subscene of Landsat TM acquisition of satellite data used in this study. Whereas the
image of June 1993, covering Siran Valley was geometrically summer of 1993 was a wet season and consequently produced
corrected and georeferenced using Survey of Pakistan's good stand of crops in the valley. Thus, the dilemma of
1:50,000 topographic maps. Thirty-five ground control points delineating forest vegetation from non-forest vegetation was
were used to perform the rectification, with a root mean faced only within TM image. Hence, it was decided to
square (RMS) error of 1.30 pixels. The Universal Transverse isolate crops on the plains from forests on the nearby hills by
Mercator (UTM) projection was applied using a nearest incorporating the elevation data into the analysis process.
neighbor interpolation algorithm. To preserve the integrity The decision was primarily based on the premise, that
of the higher spatial resolution of the TM imagery, the MSS elevation based vegetation change exist in study area, and
image of June 1979 was resampled to 30 m pixels using a the inclusion of elevation data is viable solution to delineate
nearest neighbor algorithm. It was then registered to TM vegetation classes of interest. For the sake of consistency,
image of June 1993 using an image-to-image registration elevation data were incorporated within both MSS and TM
technique. images.

36
 Development of Digital Elevation Model forest cover occurred during this period. The analysis of
To create a digital elevation Model (DEM), Survey of satellite data revealed that, in 1979 there was 20,270 ha (203
Pakistan's 1:50,000 scale maps were used. Those maps have km2) area under forest cover. By 1993, this reduced to
a 1000 m grid overlay, which can be used to construct 11,340 ha (113km2) Hence, there is a drastic decline of about
elevation profile. First, all the grid interstices were digitized 40% in forested land in just 14 years. These findings are in
and recorded in a lattice form. An elevation for each agreement with an earlier study of NWFP forests which
intersection was recorded from the topographic contour line estimated about 45 % decline in forest cover in one decade
passing closest to the grid intersection. Thus, over 1000 (SFDP, 1991). It is evident that land, previously under forests,
elevation points (z-values) were recorded. Those elevation was cleared and eventually became either scrub or barren
points were then joined to grid interstices using ARC/INFO. land. The elimination of forest land occurred on lower
The points were recorded along the profiles. To create an elevations predominantly occupied by Chir pine forests.
elevation surface from those values, a universal Kriging These low lying forests are generally easily accessible to the
interpolation procedure was applied, which produced a local population.
(DEM) of 1000 x 1000 m resolution. Finally, the DEM was
re-sampled to match the 30 m pixel resolution of the Landsat- Conclusion
TM.
As mentioned in the previous section, the main purpose The use of multi-date satellite imagery coupled with
Downloaded by [University of Nebraska, Lincoln] at 20:57 06 April 2015

of using elevation data was to overcome the problem of ancillary data provides significant capability for forest change
spectral confusion that resulted due to juxtaposition of detection. The digital analysis of satellite data, when
forested and cultivated land in the study area. Thus, a DEM combined with a DEM, provided the capability to measure
was used to construct elevation zones where vegetation the extent of forest clearing in a small mountainous watershed.
types were most likely to change. The logical criteria with The dramatic increase in human and livestock population in
which to stratify the vegetation by altitude were formulated the Siran Valley has been associated with considerable
based on the region's physiography and vegetation. Thus, changes in the region's vegetation cover. Forests in the area
agriculture on the Pakhli Plain was restricted to maximum appear to have been cleared to meet the timber, fuel wood
elevation of 900 m. Terraced cultivation in valleys was and livestock forage needs of local population. The analysis
restricted to a maximum elevation of 1000 m. Any vegetation indicates that the low lying Chir pine forests suffered the
pixel (with higher spectral reflectance in the IR) above these most damage during the period covered in this study.
elevations was considered as forest. Therefore, using this The current study is yet one more example of
methodology, three elevation-based vegetation zones were multitemporal remote sensing data for environmental
constructed. The preliminary unsupervised classification map monitoring. Despite the restricted availability of large scale
was compared (pixel by pixel) to a DEM to evaluate and to maps and aerial photo coverages for most areas in the
assign vegetation pixels to three elevation levels to obtain developing world, the analysis of remote sensing data remains
the stratified classification product. Since, the estimation of an effective tool for monitoring human induced environmental
forest cover changes was the main purpose of this study. It changes and / or degradation. To date, few studies have
was decided to merge the agriculture on the plains and in the quantified the environmental impact of the growing number
valleys into one class, and recode the stratified vegetation of internal and external refugees flow created by political,
map into: a) forests; and b) crops. To create final maps for economic or environmental crises. The current study provides
the estimation of forest cover change, an ARC/ INFO policy makers with a prototype methodology to assess the
procedure GRID was used. The procedure utilized two data environmental impact of large scale refugee movements and
files, one containing three elevation levels, and the other concentrations.
containing unsupervised classified vegetation data. In a GRID
procedure, on a cell-by-cell basis, the Boolean evaluation is Acknowledgment
performed on the binary representation of the values of the
two input files (e.g., vegetation plus elevation). For each bit The authors gratefully acknowledge the assistance in
in the binary representation, a Boolean-exclusive OR is preliminary data processing provided by Asad Ullah and
performed. If one of the input values for the bit is TRUE and Franz Mora, of Center for Advanced Land Management
the other is FALSE, the output is 1, and if both input values Information Technologies (CALMIT), University of
are TRUE or if both are FALSE, the output is 0 for that bit. Nebraska-Lincoln. Prof. N. J. R. Allan of the University of
The result of the GRID analysis are given in Figure 3 and 4. California, Davis, provided the northern Pakistan refugee
camps map. The funding for fieldwork and the purchase of
Results and Discussion satellite data was provided by John F. Shroder, Jr. Department
of Geography /Geology, University of Nebraska at Omaha,
The vegetation maps based on digital analysis of Landsat- from National Geographic Society Grant # 4509-91. The
MSS and TM images depict the location of forests before funding for computer processing and the production of
and after the settlement of refugees in the region. It is graphics was provided by the Conservation and Survey
evident from these maps that significant changes in the Division of the University of Nebraska-Lincoln.

37
Downloaded by [University of Nebraska, Lincoln] at 20:57 06 April 2015

Agriculture Forests Scrubland Barren Land/ water

Riparian Fa»°w Land
Figure 3 Landsat-MSS 1979 image showing the land cover classes in Siran Valley

References Hussain, F and Ilahi, I., "Ecology and vegetation of Lesser Himalayas,
Pakistan", Department of Botany, University of Peshawar,
Allan, N. J. R., "Impact of Afghan refugees on the vegetation resources University of Peshawar Press, Pakistan, 1991, 187p.
of Pakistan's Hindukush Himalaya", Mountain Research and Jan, A., "Land use survey of Siran and Daur Rivers watershed".
Development, Vol. 7:3, 1987, pp. 200-204. Northwest Frontier Records, Inventory Series 3, Pakistan Forest
Champion, H. G., Seth, S. K., and G. M. Khattak, "Forest types of Institute, Peshawar, 1982, 87 p.
Pakistan", Pakistan Forest Institute, Peshawar, 1965, 238 p. Naqvi, H. H., "Vegetation zonation of Murree Hazara Hills". Final
ERDAS, "Field Guide", ERDAS, Inc, Atlanta, Georgia, USA. Project Report, University Grants Commission, Islamabad, Pakistan,
1976, 55p.
ESRI, "Cell-based modeling with grid", Environmental Systems
Research Institute, Redlands, CA, 1992. Rock, B.N., D.L. Skole, and B.J. Choudhury, "Monitoring Vegetation

38
Downloaded by [University of Nebraska, Lincoln] at 20:57 06 April 2015

• "'*•"'-• ^*X?f«;( •-"''• i v ® * 5 !

mim*%%m

Agriculture Forests Scrubland Barren Land / Water

Riparian Fa|low Lar>d
Figure 4 Landsat-TM 1993 image showing the land cover classes in Siran Valley

Change Using Satellite Data", In A.M. Solomon and H.H. Shugart Global Change, New York: Chapman Hall, 1993, pp. 168-189.
(eds.) Vegetation Dynamics & Global Change, New York: Chapman
Hall, 1993, pp. 153- 167. Tucker, C. J., "Red and photographic infrared linear combinations for
monitoring vegetation, Remote Sensing of Environment, Vol. 8,
SFDP, "Siran Forest Development Project PC-1", Forest Department, 1979, pp. 127-150.
Government of NWFP, 1991, pp. 2-10.
.... "A critical review of remote sensing and other methods for
Skole, D. and C. Tucker, "Tropical Deforestation and Habitat nondestructive estimation of standing crop biomass. Grass Forage
Fragmentation in the Amazon: Satellite Data from 1978 to 1988", Sci, Vol. 35, 1980, pp. 177-182.
Science, Vol. 260, 1993, pp. 1905-1910.
Weinbaum, M. G., "The impact and the legacy of the Afghan refugees
Skole, D.L., B. Moore and W.H. Chomentowski, "Global geographic in Pakistan", In Contemporary Problems of Pakistan (ed) J. H.
information systems and databases for vegetation change studies". Korson), Westview Press, USA. Pak Book Corp, Pakistan, 1993,
In A.M. Solomon and H.H. Shugart (eds.) Vegetation Dynamics & 216P.

39