Applied Geography 90 (2018) 145–154

Contents lists available at ScienceDirect

Applied Geography
journal homepage: www.elsevier.com/locate/apgeog

On the possibility of using CORONA and Landsat data for evaluating and T
mapping long-term LULC: Case study of Iraqi Kurdistan
A. Saleem∗, R. Corner, J. Awange
Department of Spatial Sciences, Curtin University, Western Australia, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: Modern land development is regarded as having begun in the 1960s and since then the pattern of land cover has
Panchromatic changed considerably in many parts of the world. The majority of change detection studies have used Landsat
Texture MSS, TM, ETM+, and Landsat LDCM imagery and other satellite imagery to quantify and map land use changes
Timeline for many parts of the globe from the early 1970s. In most applications however, (e.g., Balcik, Sanli, Goksel,
Space photograph
Ulugtekin, & Dogru, 2009) land use changes dating back to the 60s are desirable. Given that most images start
Kurdistan region
Change detection
from the 70s, one possibility of overcoming this is to extend the timeline of change detection analysis beyond the
first Landsat satellite imagery by cross-referencing with CORONA imagery to include the 1960s. The main ob-
jective of this study is to employ CORONA and Landsat images to perform change detection analysis using the
Kurdistan region as a case study. Kurdistan region covers an area of 44957.1 km2 and the study period ranges
from 1969 to 2014. The results of the maximum likelihood classification process for CORONA images show that
these images may be classified successfully and accurately, showcasing the fact that the timeline of change
detection analysis may be extended by using CORONA's unique historical images to include the period prior to
the start of Landsat missions in 1972. For Kurdistan region, CORONA played a pivot role as baseline map as
many of the changes occurred after 1974, when the Kurdish revolution ended. Its land use and land cover
changes have been significant and may be divided into three stages: the first stage involves the destruction and
demolition of the landscape between 1974 and 1991. The second stage is the stage of Kurdish freedom and peace
involving minor changes in the landscape of the Kurdistan region that began in 1992 and lasted until 2002.
Finally, the last stage is between late 2003 and the present date, period which, economic growth and population
increase were the main factors influencing urban growth in the Kurdistan region. A positive non-linear re-
lationship (R2 = 0.89) was found between population increase and urban growth using regression analysis for
the study area during period 1969 to 2014. This case study therefore highlights the possibility of combining both
CORONA and Landsat to study LULC change of any region prior to 1970s.

1. Introduction
The Earth's landscape is being significantly changed in different
ways due to human activities such as urban expansion, population
pressure and industrial development (Ridd & Liu, 1998). Major changes
in the patterns of the landscape have taken place in many parts of the
world because of increased development dating back to the 1960s.
Remote Sensing (RS) plays a key role in global change monitoring, and
in land use\land cover (LULC) dynamics and as such, is one of the most
important data sources used to determine the impacts of LULC on the
environment (Kavzoglu, 2008). RS provides spatially consistent cov-
erage of large areas with both high spatial and temporal details of LULC
(Jusoff & Senthavy, 2003). Since early 1960s, RS has been used to ex-
tract valuable information through a range of satellite imagery that
started with the CORONA mission and continued to date with many
other generations of satellites, such as Landsat series. A majority of
LULC and change detection analysis studies have used Landsat MSS, TM
and ETM+ and other satellite imagery to detect and map the land use
changes for many parts of the globe, which started in the early 1970s
(Abdulaziz, Hurtado, & Al-Douri, 2009; Alphan, Doygun, & Unlukaplan,
2009; Deng, Wang, Li, & Deng, 2009; Kesgin & Nurlu, 2009; Porter-
Bolland, Ellis, & Gholz, 2007; Seto et al., 2002; Sun, Ma, & Wang,
2009b; Torres-Vera, Prol-Ledesma, & Garcia-Lopez, 2009; Townshend
et al., 2012; Wong & Sarker, 2014; Xiao et al., 2006; Yadav, Kapoor, &
Sarma, 2012). In most cases, however, LULC changes (e.g., Onur,
Maktav, Sari, & Sonmez, 2009) spanning the period before 1970s is
desirable. For such cases, the timeline for change detection analysis can
be extended for all satellite imagery by cross referencing CORONA

∗
Corresponding author. Kent Street, Bentley, 6102, Perth, Western Australia, Australia.
E-mail addresses: ashty.saleem@postgrad.curtin.edu.au (A. Saleem), R.Corner@curtin.edu.au (R. Corner), J.Awange@curtin.edu.au (J. Awange).

https://doi.org/10.1016/j.apgeog.2017.12.007
Received 7 September 2017; Received in revised form 30 November 2017; Accepted 7 December 2017
Available online 22 December 2017
0143-6228/ © 2017 Elsevier Ltd. All rights reserved.
A. Saleem et al.
imagery to cover the 1960s as well.
Many researchers have noted that CORONA images could sig-
nificantly contribute to the analysis and understanding of environ-
mental change. In particular, CORONA imagery could help establish
new base dataset, especially for the previously undeveloped areas
around the world. The CORONA photographs make it possible to study
and map land resources at the beginning of the Space Age (Tappan,
2000). Since CORONA images were offered to the public, these images
have been used in many studies to visually detect archeological sites
(see e.g., Beck, 2007) or reference historical data in combination with
the latest high to medium resolution satellite data from IRS 1C and
Landsat TM (e.g., Lorenz, 2004).
CORONA images have some uniquely useful characteristics such as
spatial resolution (approximate ground resolution of 1.8 m), digital
image format, easy access, low price (Kennedy, 1998) and stereo
viewing capabilities, which make them attractive. However, there are
limitations in the spectral signature information from surface features
for the CORONA imagery and because they are panchromatic images
(single layer images), they create difficulties in usage during the auto
classification process. Despite the spectral signature limitation for
CORONA, the images have been used for creating land cover and land
use thematic maps via on-screen digitising (Andersen, 2006; Coskun
Hepcan, 2013; Lorenz, 2004; Ruelland, 2011; Tappan, 2000; Ur, 2003),
and for visual interpretation (Beck, 2007; Bitelli & Girelli, 2009; Challis,
2002; Fowler, 2004; Gheyle, 2004; Hritz, 2013; Philip, 2002; Rigina,
2003; Wilkinson, 2006). In some cases, stereo analysis and DEM gen-
eration has been carried out (Casana, 2008; Goossens, De Wulf,
Bourgeois, Gheyle, & Willems, 2006; Lamsal, 2011).
In most applications however (e.g., Balcik, Sanli, Goksel, Ulugtekin,
& Dogru, 2009), land use changes dating back to the 60s are desirable.
Given that most images start from the 70s when Landsat series begun,
one possibility of overcoming the data gap from the 60s to 70s is to
extend the timeline of change detection analysis beyond the first
Landsat satellite imagery via cross-referencing with CORONA imagery
to include the 1960s. In addition, even in times of peace, many Middle
Eastern governments are reluctant to provide any data that may be a
threat to their security. Because of these issues, for many researcher and
especially for Middle Eastern country, CORONA images can be a great
help to map the 1960s changes. This study explores this possibility
using Kurdistan region as a case study.
Kurdistan region is chosen due to lack of information on land cover
status for the early sixties and a lack of socio-economic data due to
decades of political turmoil. In addition, remote sensing data was only
available for military use in Iraq before 2003. Many maps and surveys
were lost during the Iraq invasion by the USA led coalition in 2003. Due
to the constraints above, CORONA imagery can play a significant role in
detecting and mapping the status of land cover for the Kurdistan region
starting from 1969 onwards. In this case, the period of change detection
for the Kurdistan region may be extended from 38 to 45 years using
CORONA satellite space photographs. Such cases are not limited to
Kurdistan region alone, but to most countries where pre 1970s data
does not exist.
The main objectives of this study, therefore, are: (i) to cross-re-
ference CORONA and Landsat images to extend the timeline of the
study period into the 1960s, (ii) to quantify and map the changes in the
study area for the period 1969 to 2014, and (iii), to implement Remote
Sensing and GIS technologies to explore the factors responsible for
landscape changes in the study area. To achieve these, CORONA pan-
chromatic images were classified with high level of accuracy by em-
ploying the method suggested by Saleem (2017). The outcome of the
image classification technique takes advantage of the overlap of dates
between the CORONA and Landsat satellite series to detect and map the
changes in the LULC categories for the Kurdistan region for the period
1969–2014.
146
Applied Geography 90 (2018) 145–154
2. Study area
Throughout the 20th century, the Kurdish people have attempted to
regain their political and cultural autonomy, but this has always been
rejected by the region's nations (KRG, 2013). The Kurdish population
today is estimated to be 35–40 million with the majority living in
Turkey, Iran, and Iraq. There are also Kurds in Syria, Armenia and
Azerbaijan (KRG, 2013). Kurdistan had been divided and re-divided
many times before and after the First World War. The Kurds have fre-
quently been influenced by regional politics and have been promised
the possibility of an independent state on many occasions (McDowall,
1992). All major Kurdish resistance in terms of cultural and political
activities has been continually criminalised by Turkey, Iraq and Iran
(Olson, 1992).
The case study focuses on Kurdistan region which is located in the
northern region of Iraq, and contains a population of more than five
million. It shares borders with Syria in the west, Turkey in the north,
Iran in the east, and the middle Governorates of Federal Iraq in the
south. Three Governorates (Duhok, Erbil, and Sulamani) constitute the
Kurdistan Regional Government (KRG, 2013) in the northern region of
Iraq as approved by the Iraqi Federal Government in 2005. The geo-
graphical bounds for this region which form the study area are between
latitudes 34° 42′ N and 37° 22′ N and between longitudes 42° 25′ and
46° 15′ east (KRG, 2013), with a total area of 44957.1 km2, shown in
Fig. 1.
Geographically, the Kurdistan region is regarded as a part of Alpine
mountain girdle (Abdula, Balaky, Nourmohamadi, & Piroui, 2015).
Iraqi Kurdistan is divided generally into the low folded (foothills);
highly folded; imbricated and Thrust (Zagros) zones. Topographically,
most of Iraqi Kurdistan is mountainous, with the highest peak, Halgurd,
located near the border with Iran and reaching 3660 metres above the
mean sea level. The lowest point in the Kurdistan region is in Kifri sub-
district, which has an elevation of 72 metres above the mean sea level.
The main water resources in south Kurdistan include both surface and
ground water. The surface water resources in the Kurdistan region
consist of rivers and streams, with approximately 40% of the surface
water for Kurdistan region originating from neighbouring countries,
such Turkey and Iran (KRG, 2013; UNDP, 2011).
3. Data and methods
3.1. CORONA
CORONA operated from August 1960 to May 1972 and acquired
photographic images covering nearly two billion square kilometres over
a 12-year period (Leachtenauer, 1997). Those space photographs were
released and offered for public use in February 1995 (MacDonald,
1995). This declassification of major U.S. satellite reconnaissance pro-
grams offered a significant source of imagery to the civil community
such as environmental researchers, archaeologists, historians, and other
users, with over 800,000 images having been taken from space for
many parts of the Earth (Leachtenauer, 1997). A rich database archive
of CORONA images was completed after a large numbers of the
CORONA images were released in three editions in 1996, 2002, and
2013. Those images are available for civilian access through the In-
ternet and can be easy accessed through a user-friendly catalogue
published on the website of the United States Geological Survey
(https://earthexplorer.usgs.gov/) with low price (Kennedy, 1998). The
main advantages of CORONA images are their spatial coverage, spatial
resolution (ranging between 1.8 and 12.2 m), cost effectiveness and
their easy availability. CORONA imagery also has Stereo capabilities,
and the systems KH-4, KH-4A, and KH-4B were equipped with forward
and backward looking cameras. They produce their images in stereo
form with the best resolution being 1.8 m (6 feet) for the KH-4B system.
The Kurdistan region have an areas of 44957.1 km2 in and therefore
a total of 39 CORONA (August 1969) images were required to complete
A. Saleem et al.
Fig. 1. Map of study area – source: Sentinel-2 MSI 2016 base map (United State Geological Survey; USGS).
the areal coverage of the study area. All geo-referenced images were
downloaded from the CORONA Atlas for the Middle East (BETA)
website (http://corona.cast.uark.edu/atlas). The geo-referenced images
downloaded from this website were in NITF format (National Imagery
Transmission Format). Each NITF file contains on unprocessed geo-re-
ferenced CORONA image along with RPC (Rational Polynomial
Coefficients) as a replacement sensor model for the ortho-rectification
process. The methodology suggested by the CORONA Atlas for the
Middle East website (Adam, 2011) which is based on DEM data was
adopted, using ArcMap (ArcGIS, ESRI) software an ortho-rectified
image for all CORONA images were obtained. All the images were cloud
free and a number of image pre-processing steps were applied, in-
cluding normalisation and colour balancing techniques
(Hamandawana, 2007) to reduce the brightness variation between each
adjacent scene. The ortho-rectified images were resampled to have a
uniform pixel size of 2.5 m. All resampled images were mosaicked with
ERDAS Imagine software and a final CORONA blanket ortho-image was
obtained. A subset image was created for CORONA ortho-images using
Kurdistan region boundary shapefile.
3.2. Landsat
Landsat data that has been readily available at no cost after 2008
and has the potential to significantly improve the characterisation the
changes that occurs because of human activities. In many studies, it has
been shown that Landsat imagery is a powerful data source and tool for
providing accurate and timely spatial information regarding the status
of the Earth's landscape. Because of the long-term digital archive since
1972, its imageries have been used in many land use and land cover
change studies over different parts of the world. All Landsat product can
147
Applied Geography 90 (2018) 145–154
be downloaded from United States Geological Survey website (https://
earthexplorer.usgs.gov/) as different Levels of products for instance
level 1 product.
Five Landsat scenes were required to cover the extent of the study
area for each year, and the Landsat satellite data were chosen according
to the following criteria: (i) a long time-series of images should be
available for the study area; (ii) the images should be acquired between
July to August, to match the CORONA date and to increase the se-
paration of land use classes from others land cover types; (iii) all images
should have less than 10% cloud cover; (iv) all imagery acquired should
have close anniversary dates (within one month only) to minimise
differences in the image due to atmosphere, sun angle effect (illumi-
nation) and seasonal variation. Because of the image availability, there
is variation in the length of the epoch for this study. A total of 25
Landsat images were needed to obtain five years coverage (1976, 1989,
1998, 2006 and 2014) for this study. Landsat images were downloaded
as a Level 1 product. Downloaded images were Landsat-2 MSS for 1976,
Landsat-4TM for 1989, Landsat-5TM for 1998 and 2006, and Landsat-8
for 2014. Since remotely sensed data is influenced by a number of
factors, including atmospheric effects, such datasets may not be used
directly for analysis. The Landsat image data is usually degraded due to
atmospheric effects (Tyagi & Bhosle, 2011). Atmospheric correction
was therefore performed using the DOS method (Chavez, 1996) for
each individual Landsat image. The mosaic process was performed
using the mosaic tool in ERDAS Imagine 2014 for each date. Since the
CORONA mosaic image is an ortho-image, this image was therefore
used as a reference to register the Landsat TM 1989 image. The selec-
tion of Ground Control Points (GCPs) was based on permanent features
representing road intersections, historical sites and landmarks. A total
of 100 GCPs were used, well distributed throughout the Landsat 1989
A. Saleem et al. Applied Geography 90 (2018) 145–154

image, and the registration process yielded a Root Mean Square Error Table 2
(RMSE) of less than 0.5 pixel. A first-order polynomial fit was applied The Jeffries-Matusita measure results for all images.
with nearest neighbour sampling method. Since the downloaded
Years Minimum Maximum
Landsat images are a Level 1 product, there is no need to perform image
geo-referencing since this has already been done by the provider using 1969 1.80 2.00
ground control points. Image co-registration between the new reference 1976 1.66 2.00
1989 1.88 2.00
Landsat TM 1989 and the rest of the Landsat images was performed
1998 1.78 2.00
using the geometric correction registration workflow tool in ENVI 2006 1.87 2.00
software. All the registered Landsat images with TM 1989 have a total 2014 1.95 2.00
RMSE of less than 0.5 pixel. A subset image was created for each
Landsat images using Kurdistan region boundary shapefile.
Jeffries-Matusita measure values for all training samples for each land
cover type were within the range of 1.66–2 (Table 2), except for water
3.3. Image classification
bodies for all years as it was below 1. These values indicate that training
sample are statistically separable from each other (except for the water
This study uses the multi-temporal Landsat data with cross refer-
bodies) and may be used for further analysis. Because the water bodies
encing of data from CORONA space photographs to detect, quantify and
in the images are highly heterogeneous mainly due to differences in
map LULC changes from 1969 to 2014. According to the land use and
water clarity, they allow for potential misidentification as some of the
land cover classification system developed by Anderson (1976) for in-
other ground classes. It is therefore desirable to mask out all water
terpretation of RS data at various scales and resolutions, Landsat images
bodies. This is also made necessary by the fact that the study area is
are only suitable for level I LULC mapping because of pixel size re-
mountainous, with a high possibility of shadows being present in the
strictions (Sun, Ma, & Wang, 2009a). CORONA data can also only be
Landsat images, and confusion can occur in the image classification
classified as level 1 due to spectral band restrictions. The study area will
process between water bodies and shadow areas. The water bodies in
be classified into five classes (Table 1) for all CORONA and Landsat
the CORONA and Landsat images represent less than 2% of the total
images: water bodies, built-up, bare land, forest, and agriculture area.
study area, and therefore were masked out from all images. Excluding
For the CORONA ortho-images, the methodology suggested by
those area may led to reduce the confusion between land cover classes
Saleem (2017) was implemented to calculate texture measures and
during image classification process.
create an input multiband image from those texture measures (least
A number of image classification techniques were performed for
correlated) and original gray-levels. The principle of this method is
both datasets (CORONA and Landsat). Image classification included
based on creation new input dataset from the texture calculation (least
unsupervised (as test technique) and supervised image classification
correlated) combined with original CORONA image (gray levels).
method. The main reason for performing unsupervised classification
Training samples were extracted for each Landsat image and multiband
was to obtain an estimate of how many classes can be extracted from
CORONA images using a region growth method. The region growth tool
the study's images.
is based on the principle of a growing function for the given region
For supervised classification, although there are a large number of
(seed pixels) within an N dimensional array (of bands) by aggregating
image classification techniques available, for instance Maximum
or grouping all pixels within the array that connect the neighbouring
Likelihood Classifier (MLC), Random Forest (RF), Minimum Distance
pixels to the seed region based on some constraints. The constraints
Classifier (MDC), and Support Vector Machine Classifier (SVMC). The
could be a threshold value for minimum and maximum pixel value or as
MLC is regarded as one of the superior method and popular one for long
a multiple of the standard deviation of the pixels within a same region
time and that has been commonly used in image classification processes
(Tang, Wu, & Chen, 2002). The region growth tool is based on the seed
in the field of RS (Erbek, Özkan, & Taberner, 2004; Foody, Campbell,
pixels locations and spatially aggregates those pixels that meet the
Trodd, & Wood, 1992; Frazier & Page, 2000; Gonzalez & Woods, 1992,
statistical criteria selected by the user (ENVI, 2015).
pp. 11–15; Maselli, Conese, & Petkov, 1994; Paola & Schowengerdt,
The seed pixels for each land cover type were identified visually.
1995; Paul, 2000; Paul, Kääb, Maisch, Kellenberger, & Haeberli, 2002;
During the region growth process, the standard deviation parameter
Strahler, 1980). Also different supervised classification techniques in-
was set to multiplier of 1 and the parameters (alteration, maximum
cluding Random Forest (NF), Artificial Neural Network (ANN) and MLC
growth size, neighbour pixels) were kept as default. The main reason
were compared, the results shows all three methods can achieve ac-
for setting the parameters in this way, is to obtain pure regions that
curacy above 85% after combining principle component with original
have the lowest standard deviation from the mean of the selected seed
data (Otukei & Blaschke, 2010). This classifier (MLC) has been used
pixels. Changing other parameters will not affect the results for the
widely because it is based on multivariate normal distribution theory
region growth tool as those regions are restricted to one standard de-
(Strahler, 1980). For this study, therefore, supervised classifiers with
viation from the mean, and the surrounding pixels will not be included
MLC were employed for all study's images.
in the region for the seed pixels unless their statistics are within the
range of one standard deviation from the mean of the seed pixels.
Before using those training samples directly in image classification,
the Jeffries-Matusita test was performed for the training samples. The

Table 1
Developed classification scheme based on Anderson (1976).

Code Land use/cover Description

1 Water bodies All areas of open water, including rivers, streams, lakes, and reservoirs
2 Built-up Urban including residential, commercial, and industrial buildings as well as open transportation facilities, airports, highways, railways, and single/
multiple family houses.
3 Bare land Uncultivated areas of sparse plant cover, including saline alkali land along river/lake beaches, barren rock or sand in sloping fields, bare land, and
burned. .areas
4 Forest Dense forest, open forest, orchards, nurseries, and all type of trees.
5 Agriculture A cultivated with dense annual crops and vegetables, cultivated land without crops including dry land and irrigable land.

148
A. Saleem et al. Applied Geography 90 (2018) 145–154

Table 3
Total random points number (RPN) and areas in km2 extracted using the region growth tool.

Years Water Built up Bare land Forest Agriculture Total Area km2 (RPN)

1969 100.98 (108) 3.56 (68) 15.74 (60) 3.32 (64) 16.45 (60) 140.05 (356)
1976 4.10 (60) 1.20 (45) 12.02 (60) 7.45 (60) 3.83 (61) 28.59 (286)
1989 18.64 (66) 3.30 (64) 10.95 (78) 5.97 (64) 7.41 (72) 46.27 (344)
1998 137.71 (76) 18.01 (70) 62.94 (59) 7.95 (69) 24.49 (59) 251.09 (333)
2006 94.09 (65) 5.89 (77) 19.53 (72) 9.19 (64) 28.04 (96) 156.75 (374)
2014 166.07 (65) 37.34 (66) 107.22 (68) 7.95 (61) 21.93 (73) 340.52 (333)

Table 4
Accuracy assessment results using points and regions reference datasets and regions.

Years Excluding water Including water

Points Regions Points Regions

Overall accuracy Kappa coefficient Overall accuracy Kappa coefficient Overall accuracy Kappa coefficient Overall accuracy Kappa coefficient

1969 85% 0.80 84% 0.76 91% 0.88 96% 0.90

1976 79% 0.72 83% 0.74 83% 0.79 86% 0.80
1989 86% 0.81 91% 0.88 89% 0.86 95% 0.93
1998 89% 0.86 84% 0.73 92% 0.90 93% 0.88
2006 81% 0.74 74% 0.60 84% 0.80 90% 0.82
2014 88% 0.85 89% 0.81 91% 0.88 95% 0.92

3.4. Accuracy assessment
Because of the lack of historical reference data for the study area,
due to the fact that the field of RS was restricted to military use in Iraq
until 2003, the reference data have been collected by visual inter-
pretation of the study's images.
Two sets of reference datasets (point and regions) were created for
all images (using Landsat and CORONA as a base map layer). For the
point reference dataset, a stratified random sampling approach was
adopted for each land cover category. 500 random points for each year
were generated. A total of 356, 286, 344, 333, 374, and 333 random
points a minimum of 45 points (Congalton, 1991) for each class
(Table 3) were then attributed to correct LULC types and code for 1969,
1976, 1989, 1998, 2006, and 2014, respectively. The point reference
datasets were converted to raster files with 2.5 and 30 m pixel size for
CORONA and Landsat images respectively, for the accuracy assessment
process.
From the generated random points, a total area of 963.27 km2
(Table 3) of five land cover type as training samples regions were
generated using the region growth tool. The ROI raster dataset for each
year contains all five types of land cover attributed to the correct LULC
type and code.
For the accuracy assessment process, the two datasets (points and
regions) were used with the unfiltered classified images separately, and
measures of accuracy, including user accuracy, procedure accuracy,
overall accuracy and Kappa coefficient, were calculated during this
process (Congalton, 1991).
3.5. Change detection
This study used two different datasets in terms of spatial resolution
(CORONA 2.5 m and Landsat 30 m) as inputs for the image classifica-
tion process. Therefore, their resolutions needed to be normalised so
that each evaluated pair of classified images had the same pixel size, in
order to perform “From-To” change detection analysis.
Since the original spatial resolution of CORONA is higher than
subsequent Landsat images, this spatial resolution was retained for
determining an accurate land cover map for 1969. A test was carried
out by resampling the CORONA image to 30 m pixel size and evaluating
against the 30 m 2014 Landsat image using the “From-To” method.
However this resampling to 30 m generalised the CORONA data and
much information was lost; therefore, it was considered preferable to
resample both Landsat images (1976 and 2014) to 2.5 m using the
nearest neighbour method, to provide comparison data for Landsat MSS
1976 (epoch 1969–1976) and Landsat LDCM 2014 (epoch 1969–2014)
pixel 2.5 m. For the remaining epochs (1976–1989, 1989–1998,
1998–2006, 2006–2014), all classified Landsat images have a uniform
pixel size of 30 m. A python script was developed in the ArcGIS en-
vironment to be able to quantify and map the LULC changes by using
“From-To” change detection analysis. This method captures the overall
changes and rates between every pair of classified images. The results of
this technique are a two-way cross-matrix describing all types of change
in the study area. The python script was implemented for the epochs
1969–1976, 1976–1989, 1989–1998, 1998–2006, 2006–2014, and
1969–2014. A new thematic change map with their statistics containing
different combinations of “From-To” change classes was produced for
each evaluated epoch.
4. Results and discussion
The results of the accuracy assessment for points and regions da-
tasets are shown in Table 4. The accuracy assessment process was done
twice, once without the water class and once with the water class for
both datasets, because the water class had been masked out from image
classification process. The inclusion of the water class did not affect the
final accuracy for each individual class, but it did increase the overall
accuracy and Kappa Coefficient for all the classified images. The ac-
curacy of the water bodies was free from commission and omission
errors as they were 100% accurate all the time. The overall accuracy
(without water) for all years ranged from 79% to 89% and the Kappa
coefficient ranged from 0.72 to 0.86.
For the regions datasets, the overall accuracy for all images ranged
between 74% and 91% and the Kappa coefficient of 0.60 and 0.88. The
results of the accuracy assessment for both datasets (points and regions)
excluding water are compatible with the standard accuracy expected
for classified images when performing post classification change de-
tection analysis. The results for both accuracy assessments comply with
the standard accuracy for LULC mapping studies for Landsat products as
recommended by Anderson (1976).
The spatial pattern and area statistics for all LULC change classes are
shown in Fig. 2 and Tables 5–7. Generally, these maps show the spatial
distribution pattern of the land use/cover change categories over the

149
A. Saleem et al.
Fig. 2. LULC change maps for all epochs.
study area of Kurdistan region from one epoch to another.
The overall analysis shows that the minimum unchanged area in
Iraqi Kurdistan amounts to approximately 53% of the total study area
for all study epochs. The minimum unchanged area was 53% for both
the 1998–2006 and 2006–2014 epochs, and for the other epochs, it
ranged from 55% to 72% (Table 5).
The bare land class always accounted for the bulk of these un-
changed areas, with a minimum of 63.3% and a maximum of 91.1%
(Table 6) for the periods 2006–2014 and 1976–1989, respectively.
Forested areas were second in importance after bare land with
minimum unchanged area of 3.7% and a maximum of 29.3% for the
periods 1989–1998 and 2006–2014, respectively. For the agricultural
areas, the minimum unchanged was 2.6% for the overall epoch
(1969–2014), and the maximum unchanged area was 11.3% for the
1998–2006 period. For water areas, unchanged areas from 1969 to
2014 amounted to less than 2.0%, and the water bodies represent less
than 2.0% of the total of the study area (Table 6). Finally, the built up
areas have maintain the unchanged area below 1% for all study epochs.
Regarding the changed areas, the overall analysis shows that the
minimum changed area in Kurdistan region amounts to approximately
Table 5
Total unchanged and changed areas in km2 and percentage (%) for each epoch.
Areas in KM2 and percentage (%) in respect to total study areas
Epochs 1969–1976 1976–1989
Unchanged 26427.9 (59) 28500.0 (63)
Changed 18529.2 (41) 16457.0 (37)
Totals 44957.1 (100) 44957.1 (100)
Applied Geography 90 (2018) 145–154
28% of the total study area for all study epochs. The minimum changed
area (increase) was 28% for 1989–1998 epoch, and for the other
epochs, it ranged from 37% to 47% (Table 5). The bare land class al-
ways accounted for the bulk of these increase changed areas as first,
with a minimum of 3.6% and a maximum of 80.7% (Table 7) for the
periods 1998–2006 and 2006–2014, respectively. For the agricultural
areas, also this class accounted as second after bare land class for the
bulk of these increase changed areas, with a minimum of 4.6% and a
maximum of 75.5% (Table 7) for the periods 2006–2014 and
1969–2014, respectively. Forested areas were regarded as third in im-
portance with minimum increase of 0.3% and a maximum of 32.8% for
the periods 1976–1989 and 1989–1998, respectively. For water areas,
have a maximum increase of 1.4% for the epoch 1989–1998 of the total
of the changed area (Table 7). Finally, the built up areas have a
minimum increase of 0.2% for 1969–1976 epochs, and maximum in-
crease of 6.0% for the 1969–2014 epoch.
The trend of LULC changes are showing in Fig. 3, the most visible
changes occurring in the study area were the remarkable interchange in
the bare land and agricultural area. The changes in the landscape in
Kurdistan region can be summarised and described using three phases
1989–1998 1998–2006 2006–2014 1969–2014
32351.3 (72) 23836.8 (53) 23783.7 (53) 24761.6 (55)
12605.7 (28) 21120.2 (47) 21173.4 (47) 20195.4 (44)
44957.1 (100) 44957.1 (100) 44957.1 (100) 44957.1 (100)
150
A. Saleem et al. Applied Geography 90 (2018) 145–154

Table 6
Total unchanged area for each LULC types and epochs.

Change detection “From-To”, areas in % in respect to unchanged areas only

Epochs 1969–1976 1976–1989 1989–1998 1998–2006 2006–2014 1969–2014

Agriculture unchanged 7.2 3.6 5.5 11.3 4.9 2.6

Bare land unchanged 80.9 91.1 89.5 67.1 63.3 81.0
Built up unchanged 0.1 0.1 0.3 0.5 0.9 0.2
Forest unchanged 10.5 4.3 3.7 19.3 29.3 15.1
Water unchanged 1.3 0.9 1.0 1.7 1.5 1.1

which are related to some landmarks in the Kurdish people's history in
northern region of Federal Iraq.
The first phase is the period of destruction and demolition of the
landscape which occurred from 1969 to 1991 (covers 1969–1976 and
1976–1989 epochs). During this period, the increase in bare lands and
decrease in agricultural areas, forest and built up areas occurred in this
period when thousands of Kurdish families migrated to other countries
after the failure of the Kurdish revolution in 1974. The forest areas in
the Kurdistan region went through a systematic decline from mid-1970s
to late 1980s, during the series of attacks known as the “Anfal cam-
paign” which was carried out from 1986 to 1988 by the Iraqi army. This
land use/cover class decreased during the first and second epochs
(1969–1976 and 1976–1989) by a total decrease of 308.3 km2 and
3928.6 km2, respectively (Fig. 3). These destructions in the landscape
also included built up areas and more than four thousand villages were
destroyed (KRG, 2013) during this period.
Water bodies during this period have experienced a decrease, al-
though the pre 2003 Iraqi government built a new dam in the Mosul
region in 1984, water areas decreased by 147.7 km2 for the period
1976–1989. The purpose of this dam was to provide water resources
and electricity to the other governorates at the expense of Kurdish land,
as many agricultural and built up areas were lost under the water in
1984 when the construction was completed. Another reason the water
bodies were decreased in this period is the Kurdistan region depends on
the neighbouring countries, especially Iran, for 40% of its water supply.
During the war between Iraq and Iran from 1980 to 1988, the Iranian
government reduced the water flowing to the Dukan and Derbandikhan
reservoirs, resulting in a decrease in the water surface areas during this
period.
The second phase has experienced fewer changes in the landscape
compared to the previous epochs, although it marked a major turning
point for Kurdish citizens, who obtained their freedom and in-
dependence from Iraq after the Gulf war in 1991, with the help of
western countries including USA, UK, and France. This period can be
regarded as phase of transition, with relatively little changes in the
landscape between 1989 and 1998 (cover 1989–1998 epoch), during
which the Kurdish territory became a free and autonomous region from
Iraq, and Kurdish people returned from exile to rebuild the devastated
region. Kurdistan during this period was a region of relatively little
change in the built up areas with total increase of 5.6 km2 but of sub-
stantial population increase (Figs. 4 and 5). A further reason for this low
increase in built up area is may be related to the fact that the KRG was
unable to provide loans to civilians for new buildings in response to the
population growth during this period and construction were financed
only by the private sector.
The increase in the bare land areas class and decrease in agricultural
areas is compatible with the KRG's limited capabilities for providing the
basic needs of Kurdish farmers for increasing agricultural production. In
addition, the Kurds did not have sufficient funds to support the agri-
cultural sector by converting bare lands to productive agriculture areas
for civilian needs in the region during this epoch. These limitations
were due to the fact that after 1991, Kurds were living under
Humanitarian programmes supported by the UN, in particular the UN
Oil for Food Programme (OIP) from 1995. In addition, the rate of LULC
class transformation was lower than previous epochs because Kurdish
areas were excluded from any sort of benefits from Iraqi resource in-
come, and that people in Kurdistan were living a very basic lifestyle
(KRG, 2013). The water surface area increased by 150.0 km2 and this
increase may be related to the level of precipitation received in the
Kurdistan region during this period (KRG, 2013). The forested areas
were increased during this period by 4059.7 km2, as during this phase
the KRG started a rehabilitation programme to increase forested areas
that were lost during1974 to early 1991.
The third phase runs from 1998 until 2014 (covers 1998–2006 and
2006–2014 epochs), when the Kurds were able to guarantee their rights
in the post- 2003 Iraqi administration and obtain their own budget from
the central government. The economic growth in the oil and gas sector
(Fig. 5) was the main factor influencing the remarkable changes in the
Kurdistan landscape, having a direct impact on urban growth, and the
improved living standards for most of the middle and low-income ci-
tizens in the Kurdistan region (KRG, 2013).
This period may be regarded as the most significant period of the
study as many previous socio-political events impacted the landscape
following Kurdish freedom in 1991, in particular the displacement of
the Iraqi regime by the USA and coalition troops in 2003. This event
provided an opportunity for Kurds to seek their rights and share in the
natural resources of Iraq, especially in the oil and gas sector (KRG
2013). The Kurds were able to make the three major cities in Kurdistan
region, in particular Erbil city, some of the best places for investment,
especially after the implementation of a liberal investment law, which
was approved by the Kurdistan regional Government in July 2006, and
offered foreign investors the freedom to expatriate profits, customs
relief, tax holidays, and regulated banking and finance sector services
(KRG 2013). Official figures indicate that there are currently 15,000

Table 7
Total changed area for each LULC type and epoch.

Change detection “From-To”, areas in % in respect to changed areas only

Epochs 1969–1976 1976–1989 1989–1998 1998–2006 2006–2014 1969–2014

Conversion to agriculture 21.4 57.7 13.6 75.5 4.6 7.4

Conversion to bare land 64.6 40.6 51.7 3.6 80.7 59.8
Conversion to built up 0.2 0.8 0.5 0.4 4.9 6.0
Conversion to forest 13.0 0.3 32.8 20.3 9.5 26.0
Conversion to water 0.8 0.5 1.4 0.1 0.2 0.8

151
A. Saleem et al. Applied Geography 90 (2018) 145–154

Fig. 3. Trends of LULC change for all epochs.

Table 8
Land parcels data from the 2009 KRG report.

Governorates Residential Commercial Industrial Others Parcels counts

Erbil 39430 1138 3216 590 44374

Sulamani 7645 1301 1456 171 10573
Duhok 12397 281 245 707 13630
Total 59472 2720 4917 1468 68577

Fig. 4. Population trends for the Kurdistan Region.

Fig. 6. Exponential relationship between urban expansion and population increase.

Fig. 5. Oil production for the Kurdistan Region.
local companies and 2300 foreign companies from 78 countries regis-
tered in the Kurdistan Region (KNA, 2013).
The dramatic decrease in bare land occurred in the 1998–2006
epoch and this decrease in bare land areas was a result of farmers being
provided with interest-free loans for increasing the agricultural pro-
duction of the Kurdistan region. The Kurds were also able to obtain
their share from the central government annual budget (11–13%) and
this gave them the ability to invest massively in the infrastructure and
agriculture sectors. Also during this epoch, built up areas increased
significantly which is one of the reasons for the reduction in bare land
especially for the areas surrounding the three major cities (Duhok,
Erbil, and Sulamani) in Kurdistan region.
Also during this period (2006–2014 epoch), the bare land have been
increased during this epoch especially in the western region of
Kurdistan, as those regions became a war zone when the Islamic ter-
rorist group ISIS appeared in the region. The increases in bare land were
balanced by decreases in other land use/cover types, especially the
agricultural areas. In addition, in recent times another factor has im-
pacted bare land increases. Farmers of Kurdistan lost interest in
growing wheat as the central government still owe Kurdistan farmers
the income for the wheat from the 2012 and 2013 seasons.

152
A. Saleem et al.
The forested areas were continuously increased during this period
(both epochs), as this phase included the KRG programme to increase
forested areas that were lost during1974 to early 1991. Forest areas
increased, with a total of 3564.2 km2 and 100.2 km2 respectively,
during the epochs, 1998–2006 and 2006–2014.
During this period (both epochs), built up area increased by totaled
of 46.3 km2 and 1035.3 km2 for 1998–2006 and 2006–2014 epochs,
respectively (Fig. 3). This dramatic increase in built areas was a result
of the amount of the funds that the KGR was injecting into the infra-
structure sector by providing extensive free loans to the residential
sector and beyond that, to include the agriculture, industrial, and
commercial sectors as well.
In addition, the KRG was able to boost the infrastructure sector by
providing the low and middle income population in the Kurdistan re-
gion with free parcels of land. Based on information released in the
activities report (KRG, 2009) for KRG in 2009 (Table 8), there were
around 68577 units of land distributed after 2003 which included
parcels for residential, commercial, industrial, and others purposes.
Based on this report, a total of 59472 residential units were distributed
to citizens of Kurdistan region, and the KRG provided those citizens
interest-free loans to build on their land units and, therefore, urban
areas were substantially increased during the last epoch (Table 8).
It is expected that the increasing in built up areas would be a re-
sponse to long term population increase and economic development
especially in Oil sector for the Kurdistan region (Figs. 4 and 5). Re-
gression analysis indicates that there is an exponential relationship
between population increase and urban expansion (Fig. 6).
5. Conclusion
This study showed that the combination of CORONA, Landsat MSS,
TM, and LC8 data was well suited to detect changes in LULC for the
Kurdistan region. Several changes in land use patterns were observed
and quantified between 1969 and 2014. The study also demonstrated
that normal human activities are not only the factors for change in the
landscape, but also that organised army operations can also contribute
to the demolition and destruction of landscape. The most visible
changes occurring in the study area were the remarkable interchange in
the bare land and agricultural area and massive expansion of the urban
areas, which may be explained by the increase in economic develop-
ment and population.
The use of CORONA imagery was a considerable help in creating a
baseline LULC map for the study area where accurate information was
required for the time prior to the persecution and displacement of the
Kurds. CORONA images are also the only available images that can
provide useful digital information for the 1960s, so that their use ex-
tended the timeline of the study period to include the 1960s. Since Iraqi
Kurdistan has experienced a dramatic change in the landscape over the
last eight years period of study, with high urban growth rate, it is
suggested that a national land use development policy is established in
the Kurdistan region for planning and development purposes. In addi-
tion, it is recommended that a continuous monitoring system for the
Kurdistan region is established to document the significant rates of
urban growth. It is also highly recommended that future studies in the
Kurdistan region go one step further to capture the relationship be-
tween LULC changes and environmental issues, in particular the study
of surface urban heat islands in the rapidly growing urban areas.
Acknowledgment
The authors would like to express gratitude to Professor Jason Ur
from Harvard University for directing us to the CORONA Atlas for the
Middle East website to obtain georeferenced CORONA images for this
project. Also, many thanks to “CORONA Atlas for the Middle East”
Centre for Advanced Spatial Technologies (CAST), University of
Arkansas and U.S. Geological Survey staff for permitting the download
153
Applied Geography 90 (2018) 145–154
of further valuable CORONA data from their websites.
References
Abdula, R. A., Balaky, S. M., Nourmohamadi, M. S., & Piroui, M. (2015). Microfacies
analysis and depositional environment of the sargelu formation (middle jurassic) from
Kurdistan region, northern Iraq.
Abdulaziz, A. M., Hurtado, J. M., & Al-Douri, R. (2009). Application of multitemporal
Landsat data to monitor land cover changes in the Eastern Nile Delta region, Egypt.
International Journal of Remote Sensing, 30(11), 2977–2996. http://dx.doi.org/10.
1080/01431160802558675.
Adam, B. (2011). On the fly and permanent orthorectification of NITF images in ArcGIS.
Alphan, H., Doygun, H., & Unlukaplan, Y. I. (2009). Post-classification comparison of land
cover using multitemporal Landsat and ASTER imagery: The case of Kahramanmara
angstrom, Turkey. Environmental Monitoring and Assessment, 151(1–4), 327–336.
http://dx.doi.org/10.1007/s10661-008-0274-x.
Andersen, G. L. (2006). How to detect desert trees using corona images: Discovering
historical ecological data. Journal of Arid Environments, 65(3), 491–511.
Anderson, J. R. (1976). A Land use and land cover classification system for use with remote
sensor data. Washington: U.S. Govt. Print. Off.
Balcik, F., Sanli, F., Goksel, C., Ulugtekin, N., & Dogru, A. (2009). A change detection
analysis in the izmir bird paradise: integration of remote sensing and geographic
information system. Fresenius Environmental Bulletin, 18(1), 51–56.
Beck, A. (2007). Evaluation of Corona and Ikonos high resolution satellite imagery for
archaeological prospection in western Syria. (Method). Antiquity, 81(311), 161.
Bitelli, G., & Girelli, V. A. (2009). Metrical use of declassified satellite imagery for an area
of archaeological interest in Turkey. Journal of Cultural Heritage, 10(Supplement
1(0)), e35–e40. http://dx.doi.org/10.1016/j.culher.2009.08.004.
Casana, J. (2008). Stereo analysis, DEM extraction and orthorectification of CORONA
satellite imagery: Archaeological applications from the near east. Antiquity, 82(317),
732–749.
Challis, K. (2002). Corona remotely-sensed imagery in dryland archaeology: The Islamic
city of al-Raqqa, Syria. Journal of Field Archaeology, 29(1 2), 139.
Chavez, P. S. (1996). Image-based atmospheric corrections-revisited and improved.
Photogrammetric Engineering & Remote Sensing, 62(9), 1025–1035.
Congalton, R. G. (1991). A review of assessing the accuracy of classifications of remotely
sensed data. Remote Sensing of Environment, 37(1), 35–46.
Coskun Hepcan, C. (2013). Quantifying landscape pattern and connectivity in a medi-
terranean coastal settlement: The case of the Urla district, Turkey. Environmental
Monitoring and Assessment, 185(1), 143–155.
Deng, J. S., Wang, K., Li, J., & Deng, Y. H. (2009). Urban land use change detection using
multisensor satellite images. Pedosphere, 19(1), 96–103.
ENVI (2015). ENvironment for visualizing images. USA: Harris Geospatial Solutions.
Retrieved from http://www.harrisgeospatial.com/ContactUs.aspx, 5.2.
Erbek, F. S., Özkan, C., & Taberner, M. (2004). Comparison of maximum likelihood
classification method with supervised artificial neural network algorithms for land
use activities. International Journal of Remote Sensing, 25(9), 1733–1748.
Foody, G. M., Campbell, N., Trodd, N., & Wood, T. (1992). Derivation and applications of
probabilistic measures of class membership from the maximum-likelihood classifi-
cation. Photogrammetric Engineering & Remote Sensing, 58(9), 1335–1341.
Fowler (2004). Cover: Declassified CORONA KH-4B satellite photography of remains from
Rome's desert frontier. International Journal of Remote Sensing, 25(18), 3549–3554.
Frazier, P. S., & Page, K. J. (2000). Water body detection and delineation with Landsat TM
data. Photogrammetric Engineering & Remote Sensing, 66(12), 1461–1468.
Gheyle, W. (2004). Evaluating CORONA: A case study in the Altai Republic (south
Siberia). (Method). Antiquity, 78(300), 391.
Gonzalez, R. C., & Woods, R. (1992). Digital image processing, Vol. 5. Addison-Wesley.
Goossens, R., De Wulf, A., Bourgeois, J., Gheyle, W., & Willems, T. (2006). Satellite
imagery and archaeology: The example of CORONA in the Altai mountains. Journal of
Archaeological Science, 33(6), 745–755. http://dx.doi.org/10.1016/j.jas.2005.10.010.
Hamandawana, H. (2007). Proposed methodology for georeferencing and mosaicking
Corona photographs. International Journal of Remote Sensing, 28(1), 5–22.
Hritz, C. (2013). A malarial-ridden swamp: Using Google Earth pro and corona to access
the southern Balikh valley, Syria. Journal of Archaeological Science, 40(4), 1975–1987.
Jusoff, K., & Senthavy, S. (2003). Land use change detection using remote sensing and
geographical information system (GIS) in Gua Musang district, Kelantan, Malaysia.
Journal of Tropical Forest Science, 15(2), 303–312.
Kavzoglu, T. (2008). Determination of environmental degradation due to urbanization
and industrialization in Gebze, Turkey. Environmental Engineering Science, 25(3),
429–438. http://dx.doi.org/10.1089/ees.2006.0271.
Kennedy, D. (1998). Declassified satellite photographs and archaeology in the Middle
East: Case studies from Turkey. Antiquity, 72(277), 553–561.
Kesgin, B., & Nurlu, E. (2009). Land cover changes on the coastal zone of Candarli Bay,
Turkey using remotely sensed data. Environmental Monitoring and Assessment,
157(1–4), 89–96.
KNA (2013). About Kurdistan. Retrieved from http://www.kurdistan-parliament.org/
Default.aspx.
KRG (2009). Annual activities report for the fifth cabinet Retrieved from Iraqi Kurdistan
region.
KRG (2013). Kurdistan's geography and climate. Retrieved from http://www.krg.org/p/p.
aspx?l=12&s=020000&r=303&p=213.
Lamsal, D. (2011). Digital terrain modelling using Corona and ALOS PRISM data to in-
vestigate the distal part of Imja Glacier, Khumbu Himal, Nepal. Journal of Mountain
Science, 8(3), 390–402.
Leachtenauer, J. C. (1997). Digitizing CORONA imagery: Quality vs. cost. CORONA:
A. Saleem et al.
Between the sun and the Earth, the first NRO reconnaissance eye in space.
Lorenz, H. (2004). Integration of corona and Landsat thematic mapper data for bedrock
geological studies in the high Arctic. International Journal of Remote Sensing, 25(22),
5143–5162.
MacDonald, R. A. (1995). Opening the cold war sky to the Public: Declassifying satellite
reconnaissance imagery. Photogrammetric Engineering B Remote Sensing, 61(4),
385–390.
Maselli, F., Conese, C., & Petkov, L. (1994). Use of probability entropy for the estimation
and graphical representation of the accuracy of maximum likelihood classifications.
ISPRS Journal of Photogrammetry and Remote Sensing, 49(2), 13–20.
McDowall, D. (1992). The Kurds: A nation denied. Harry Ransom Humanities Research
Center.
Olson, R. (1992). The Kurdish question in the aftermath of the Gulf war: Geopolitical and
geostrategic changes in the Middle East. Third World Quarterly, 13(3).
Onur, I., Maktav, D., Sari, M., & Sonmez, N. K. (2009). Change detection of land cover and
land use using remote sensing and GIS: A case study in Kemer, Turkey. International
Journal of Remote Sensing, 30(7), 1749–1757. http://dx.doi.org/10.1080/
01431160802639665.
Otukei, J. R., & Blaschke, T. (2010). Land cover change assessment using decision trees,
support vector machines and maximum likelihood classification algorithms.
International Journal of Applied Earth Observation and Geoinformation, 12, S27–S31.
Paola, J. D., & Schowengerdt, R. A. (1995). A detailed comparison of backpropagation
neural network and maximum-likelihood classifiers for urban land use classification.
IEEE Transactions on Geoscience and Remote Sensing, 33(4), 981–996.
Paul, F. (2000). Evaluation of different methods for glacier mapping using Landsat TM.
Paper presented at the proceedings, EARSeL-SIG workshop.
Paul, F., Kääb, A., Maisch, M., Kellenberger, T., & Haeberli, W. (2002). The new remote-
sensing-derived Swiss glacier inventory: I. Methods. Annals of Glaciology, 34(1),
355–361.
Philip, G. (2002). CORONA satellite photography: An archaeological application from the
Middle East. Antiquity, 76(291), 109.
Porter-Bolland, L., Ellis, E. A., & Gholz, H. L. (2007). Land use dynamics and landscape
history in La Montaña, Campeche, Mexico. Landscape and Urban Planning, 82(4),
198–207.
Ridd, M. K., & Liu, J. J. (1998). A comparison of four algorithms for change detection in
an urban environment. Remote Sensing of Environment, 63(2), 95–100.
Rigina, O. (2003). Detection of boreal forest decline with high-resolution panchromatic
satellite imagery. International Journal of Remote Sensing, 24(9), 1895–1912.
Ruelland, D. (2011). Comparison of methods for LUCC monitoring over 50 years from
aerial photographs and satellite images in a Sahelian catchment. International Journal
of Remote Sensing, 32(6), 1747–1777.
Saleem, A. (2017). Using CORONA and Landsat data for evaluating and mapping long-term
154
Applied Geography 90 (2018) 145–154
LULC changes in Iraqi Kurdistan. (Doctor of philosophy). Perth, Western Australia:
Curtin University.
Seto, K. C., Woodcock, C., Song, C., Huang, X., Lu, J., & Kaufmann, R. (2002). Monitoring
land-use change in the pearl river delta using Landsat TM. International Journal of
Remote Sensing, 23(10), 1985–2004.
Strahler, A. H. (1980). The use of prior probabilities in maximum likelihood classification
of remotely sensed data. Remote Sensing of Environment, 10(2), 135–163. http://dx.
doi.org/10.1016/0034-4257(80)90011-5.
Sun, Z., Ma, R., & Wang, Y. (2009a). Using Landsat data to determine land use changes in
Datong basin, China. Environmental Geology, 57(8), 1825–1837.
Sun, Z. Y., Ma, R., & Wang, Y. X. (2009b). Using Landsat data to determine land use
changes in Datong basin, China. Environmental Geology, 57(8), 1825–1837. http://dx.
doi.org/10.1007/s00254-008-1470-2.
Tang, L., Wu, C., & Chen, Z. (2002). Image dense matching based on region growth with
adaptive window. Pattern Recognition Letters, 23(10), 1169–1178.
Tappan, G. G. (2000). Use of Argon, Corona, and Landsat imagery to assess 30 years of
land resource changes in west-central Senegal. Photogrammetric Engineering and
Remote Sensing, 66(6), 727–735.
Torres-Vera, M. A., Prol-Ledesma, R. M., & Garcia-Lopez, D. (2009). Three decades of land
use variations in Mexico City. International Journal of Remote Sensing, 30(1), 117–138.
http://dx.doi.org/10.1080/01431160802261163.
Townshend, J. R., Masek, J. G., Huang, C., Vermote, E. F., Gao, F., Channan, S., et al.
(2012). Global characterization and monitoring of forest cover using Landsat data:
Opportunities and challenges. International Journal of Digital Earth, 5(5), 373–397.
Tyagi, P., & Bhosle, U. (2011). Atmospheric correction of remotely sensed images in
spatial and transform domain. International Journal of Image Processing, 5(5),
564–579.
UNDP (2011). Drought impact assessment, recovery and mitigation framework and regional
project design in Kurdistan region (KR). United Nations Development Programmer Iraq.
Ur, J. (2003). CORONA satellite photography and ancient road networks: A northern
mesopotamian case study. (Method). Antiquity, 77(295), 102.
Wilkinson, K. (2006). Satellite imagery as a resource in the prospection for archaeological
sites in central Syria. Geoarchaeology, 21(7), 735–750.
Wong, S., & Sarker, M. (2014). Land use/land cover mapping using multi-scale texture
processing of high resolution data. Paper presented at the IOP conference Series: Earth
and environmental science.
Xiao, J., Shen, Y., Ge, J., Tateishi, R., Tang, C., Liang, Y., et al. (2006). Evaluating urban
expansion and land use change in Shijiazhuang, China, by using GIS and remote
sensing. Landscape and Urban Planning, 75(1), 69–80.
Yadav, P., Kapoor, M., & Sarma, K. (2012). Land use land cover mapping, change de-
tection and conflict analysis of Nagzira-Navegaon corridor, Central India using
geospatial technology. International Journal of Remote Sensing and GIS, 1(2), 90–98.