Content uploaded by Younus Al-Saady
Author content
All content in this area was uploaded by Younus Al-Saady on Nov 10, 2015
Content may be subject to copyright.
FOG
Freiberg Online Geoscience
FOG
is an electronic journal registered under
ISSN
1434
-
7512
2015, Volume 43
Broder Merkel & Mandy Hoyer (Eds.)
FOG special volume: Man-made changes in land
cover, water quality and quantity – three case studies
61 pages, 3 contributions
I
List of contents
Al-Saady, Y.; Merkel, B.; Al-Tawash, B.; Al-Suhail, Q.:
Land use and land cover (LULC) mapping and change detection in the Little Zab River
Basin (LZRB), Kurdistan Region, NE Iraq and NW Iran
1
Gebere, S. B.; Merkel, B.; Agumassie, T. A.:
Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern
Ethiopia using remote sensing?
33
Kareem, A.; Merkel, B.:
The Influence of volatile organic components on the stable isotopic composition of the
groundwater in Tanjero area - Kurdistan region, Iraq
50
The images on the cover page are clippings taken from the papers from Al-Saady et al. and
Gebere et al. They show the following features:
Upper images: Change in the spatial extension of Dokan Lake in the Little Zab River
Basin in Kurdistan Region between September 1976 and 1984
Lower images: Land use and land cover changes in the dry Lake Haramaya Watershed
in eastern Ethiopia between 1985 and 1995
Land use and land cover (LULC) mapping and change
detection in the Little Zab River Basin (LZRB),
Kurdistan Region, NE Iraq and NW Iran
Al-Saady,
Younus Department of Geology, College of Science, University of
Baghdad, Baghdad, Iraq. Email: younusalsaady@yahoo.com
Merkel, Broder Department of Hydrogeology, Institute of Geology, Technische
Universität Bergakademie Freiberg, Gustav-Zeuner Str. 12,
09599 Freiberg, Germany. Email: merkel@geo.tu-freiberg.de
Al-Tawash, Balsam Department of Geology, College of Sciences, University of
Baghdad, Baghdad, Iraq. Email: balsamsalim@yahoo.com
Al-Suhail, Qusay Department of Geology, College of Science, University of
Basrah, Basrah, Iraq. Email: quab65@gmail.com
Abstract: The aim of this study was to investigate the effect of land use expansion on the natural
environment of the Little Zab River basin (LZRB). The specific objectives were to: (i) prepare a land use
land cover (LULC) map for the LZRB; (ii) detect changes between five dates within the time period
1976-2014 and to identify spatial and temporal changes that occurred within this time period using image
indices; (iii) assess LULC classes results and compare the differences between them.
Different remote sensing and GIS techniques were applied, such as digital image processing using
supervised classification and image indices. Supervised classification using a maximum likelihood
algorithm was only applied on the Landsat 8 OLI data (September 2014) to classify the LULC map. The
classified image was modified with the help of image indices and visual interpretation to produce a more
accurate map. The image obtained with supervised classification has an overall accuracy of 83.43 % and
an overall Kappa coefficient of 0.811. The results were validated using 507 ground truth points
distributed all over the study area. Six Landsat images from September 1976, 1984, 1990, 2000 and two
Landsat 8 OLI 2014 images from April and from September were georeferenced, radiometrically and
atmospherically calibrated to detect the changes in the LZRB and to identify seasonal changes. Changes
have been calculated for the vegetation cover (natural vegetation vs. cropland and pasture), surface water
features and urban and built-up land. Change detection of the LZRB was performed using three indices:
the Normalized Difference Vegetation Index (NDVI), the Normalized Difference Water Index (NDWI)
and the Normalized Difference Built-Up Index (NDBI). From the results from supervised classification
six main LULC classes have been distinguished in the LZRB. They involve barren land, agricultural land,
natural vegetation, urban and built-up land, burned land and water. Barren land is the main class in the
basin; it is subdivided into six subclasses. Agricultural land is the second largest class; it is subdivided
into three subclasses. In this study, change detection analysis over the period (1976-2014) has revealed
highly dynamic interchanges between the LULC classes. The change detection results show a rapid
increase in urban and built-up land. A significant increase of the population number, the migration from
small villages to the main cities and economic growth has fueled the rise of the urbanization rate within
the last decades. Surface water features were shown to shrink mainly due to the shrinkage of Dokan Lake
1
Freiberg Online Geoscience Vol 43, 2015
under the pressure of climate change and due to the construction of dams in the upper watershed area of
the LZRB. The natural vegetation displayed a wavy or irregular trend while cropland and pasture showed
an increasing trend.
With the NDVI two classes of natural vegetation, and one cropland and one pasture class have been
identified based on reflectance values. Areas with relatively high NDVI values indicate the cropland and
pasture class, while lower values represent natural vegetation. Two spectral water index methods were
employed in this study, an NDWI (Ji et al. 2009), which was applied onLandsat8OLIandTM5images
and an NDWI (El-Asmar et al. 2013), which was only applied on Landsat MSS images. Different water
indices have been applied for the MSS sensor due to the large difference in the spectral range between
MSS bands and other Landsat sensor bands. Both NDWI indices were successfully applied to extract
water features. Finally, the NDBI changes were calculated based on visual interpretation.
Keywords: Remote sensing, supervised classification, LULC, change detection NDVI, NDWI, NDBI.
1 Introduction
Little Zab River is one of the most important tributaries of Tigris River. In the last decade, the Little Zab
River Basin (LZRB) underwent many changes concerning land use expansions as a result of rapid
urbanization, increasing population, and climate change. Consequently, the demand for water supply
increased and the necessity for water quality monitoring and basin management became urgent. The
urbanization in the Kurdistan region of Iraq and in the LZRB as a part of this region has been increasing
rapidly within the last decade. Therefore, it is important to identify different LULC categories and to
evaluate the magnitude and spatial extent of changes within the basin to ensure best future planning and
management.
A supervised classification with the maximum likelihood algorithm was used to classify the LULC of the
Landsat 8 OLI images. Different supervised classification methods have been applied and tested
extensively for land use planning and management in arid and semi-arid environments (Ulbricht et al.
1993; Del Valle et al. 1998).
LULC mapping of the river basin is a useful tool that can be used for good planning and management of
urbanization, agricultural practices and other human activities. Landsat data is ideal for LULC mapping
and for change detection studies of large basin areas such as the LZRB. Landsat satellite data was chosen
because of its open accessibility, suitability for the aim of the study, its long time of observation and its
good spectral and spatial resolution for LULC analyses. Mapping LULC is presently the standard method
and most common approach to monitor land use changes and developments (Mancino et al. 2014).
Especially when LULC classes and their spatial and temporal changes are to be determined for categories
of small geographic extent in vast areas, high-resolution satellite images are needed (Reed et al. 1996).
The terms land use and land cover have been used interchangeably in many publications despite the
difference between these two terms. In general land cover refers to natural biophysical covers such as
forest, water bodies and barren land, while land use refers to the human utilization of land for different
purposes like agriculture and settlements, which lead to altering biogeochemical, physiographical and
hydrological conditions (Di Gregorio and Jansen 2000).
Despite the importance of the LZRB, little is known about the LULC change dynamics and its influence
on environment and water quality. No effort has been made to evaluate the effects that agricultural land
use, urban- and built-up expansion or municipal wastewater effluents could have on water quality of this
2
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
river. Identifying the dynamic change in water usage and scientifically interpreting the reasons for these
changes will lead to best management of the basin environment (Nian et al. 2014).
LULC can be considered as one of the most important determinants for the quality of water resources
(Griffith et al. 2002). Agricultural activities are the pollution sources most commonly mentioned in
literature because of their contribution in increasing the sediment load and nutrient input to water
resources (Cooper 1993). Besides that they directly contribute to the input of pesticides, pathogens and
hormones. Land use practices directly influence the hydrochemistry and quality of water and then the
aquatic organisms in receiving waters (Environment Canada, 2001). New techniques including
geographical information systems (GIS), geo-statistics, and remote sensing provide new opportunities for
the research on land surface processes. Strong population growth and consequently an increased demand
for food, grazing land and fodder as well as increasing industrial activities have essentially led to the
rapid change in LULC patterns. Agriculture, especially annual crops represent a large portion of the land
use in the LZRB. Runoff from agricultural land could have a major and direct impact on the river water
quality. The other major threat to this river is the release of municipal wastewater from the cities and
towns.
1.1 Study area
The LZRB is located in the northeastern Iraq and northwestern Iran territories (Fig. 1). It stretches out
between (35° 10´ 00´´ - 36° 55´ 00´´) latitude and (43° 25´ 00´´ - 46° 20´ 00´´) longitude, encompasses an
area of 19,860.6 km2 and comprises a wide variety of landscapes. The Little Zab River (LZR), which is
also known as Lower Zab or sometimes Lesser Zab River is the largest tributary of Tigris River with
about 71 % of its basin situated in Iraq and the rest in Iran. The LZR is an 8th order river that flows from
northeast to southwest. It feeds Dokan Lake Reservoir and flows further into Tigris River. The river is fed
by rainfall, snowmelt and many springs resulting in a high discharge in spring and low discharge in
summer. The average annual discharge of the LZR is 7.17 km3, of which 5.07 km3 are retarded every year
since the construction of Dokan Dam (Frenken, 2009). The climate of the basin varies from semi-arid in
the north and northeastern parts to arid in the south and southwestern part. Areas of higher elevation in
the north and northeast of the basin receive substantial rainfall quantities with a mean annual precipitation
of up to 850 mm. This decreases notably to less than 315 mm/a in the area of lower elevation in the south
and southwest of Dokan Lake until the confluence with the Tigris River. Many different geological units
are exposed in the Little Zab River basin with ages ranging from Jurassic to Quaternary. The upper part of
the basin area is located within the highly folded zone and Zagros Suture Zone (Jassim and Goff, 2006). It
is characterized mainly by alternation of different rock types with different lithological properties, while
the lower part of the basin is located within the foot hills zone, which is characterized by clastic
unresisting rocks (Aziz 1983).
2 Materials and methods
2.1 Remote sensing imagery data
Satellite imagery was used from the Landsat Multispectral Scanner System (Landsat2-MSS L1T)
acquired in September 1976, from Landsat 5 Thematic Mapper (TML1T) acquired in September 2000,
1990 and 1984 and from Landsat 8 OLI acquired in April 2014 and September 2014. All this data was
obtained from the USGS archive (http://earthexplorer.usgs.gov/). The downloaded bands of Landsat
8OLI, TM and MSS scenes were superimposed (excluding the thermal band) to form multispectral
images using ENVI5.1 software. After the acquisition of the images seven bands of OLI8, five bands of
TM, four bands of MSS except the thermal bands of each image scene were superimposed to form a
single multispectral image dataset using the "layer stack" function. From the acquired data, if possible,
only those images without cloud cover were selected. The Universal Traverse Mercator (UTM) system
3
Freiberg Online Geoscience Vol 43, 2015
with the datum of WGS84 and Zone 38 was selected as projection for all data. The QuickBird satellite
collects image data with a pixel resolution of 0.65 m. Images from this satellite are an excellent basis for
identifying LULC classes. QuickBird images were used in order to effectively identify the spatial
distribution characteristics of urban and built-up land and other unspecified objects in Landsat images.
This data was also used to investigate and provide data (e.g. ground truth points) for areas that are
inaccessible, to refine the classification of the satellite images and as reference image for geo-referencing.
2.2 Radiometric calibration and atmospheric correction
The radiometric calibration, atmospheric correction, and other preprocessing steps were done to prepare
the satellite images for LULC classification and change detection. The satellite image scenes were first
converted to radiances; then each image scene was converted to satellite reflectance using the Landsat
calibration tool in ENVI 5.1. The necessary information such as acquisition date and sun elevation was
obtained from the Landsat header files while the average elevation of each image scene was obtained
from digital elevation data (DEM). Multi-temporal multispectral digital images collected by Landsat
satellites for more than 38 years were utilized for preparing the LULC maps and change detection maps.
Figure 1: Location map of the study area
4
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
The correction itself involves the conversion of digital number (DN) values to spectral radiance (at the
sensor), the conversion of spectral radiance to apparent reflectance (at the sensor) and finally the removal
of atmospheric effects, which result from absorption and scattering (atmospheric correction) (Lillesand et
al. 2014; Cui et al. 2014; Chander and Markham 2003). The spectral radiance is usually converted to
reflectance because spectral radiance depends on the degree of illumination of the feature (irradiance),
which is affected by many variables such as time of the day, season, latitude, etc. Whereas reflectance
provides a standardized measure, which directly allows comparing images. Thus all images were
radiometrically corrected.
Atmospheric corrections are unnecessary for maximum likelihood image classification of a single date
image (Singh 1989; Song et al. 2001). It is, however, necessary for minimizing reflectance variations
within the individual LULC classes before image classification and the detection of changes between
images from different dates. Jeffries-Matusita distance is one of the spectral separability measures
commonly used in remote sensing applications. It is used to quantify the spectral separability during
target detection (Santillan et al. 2011). In this study, radiometric calibration and atmospheric correction
have been necessary because multi-temporal data was utilized. These corrections are used to remove
atmospheric interferences and to put multi-temporal data to the same radiometric scale in order to detect
and monitor terrestrial surface cover changes (Chen et al. 2005; Song et al. 2001; Serra et al. 2003).
2.3 Geo-referencing of images
From comparing the results of the satellite data with the ground truth points and QuickBird data, it was
observed that all data is acceptable and compatible except for the Multispectral Scanner (MSS) sensor
scenes, from which selected points did not exactly fit the corresponding ground control points and ground
features. Thus MSS scenes were corrected by using the image-to-image geo-referencing operation tool of
ENVI based on ground control points. The Landsat MSS scenes were geo-referenced with a root mean
square error of less than 0.5 pixels. Several control points were selected for geo-referencing each MSS
scene with the reference image. Subsequently, the images of each year were mosaicked to generate new
images covering the entire study area. For this purpose, the mosaicking tool based on geo-referenced
images was used. Then the MSS images were reprojected to the Universal Transverse Mercator grid using
the nearest neighbor resembling method. Finally, subsets of the images of the LZRB were created using
the region of interest (ROI) layer tool of ENVI.
2.4 Field data (ground truth points) and auxiliary data
Field surveys of LZRB were conducted during wet season (April) and dry season (September) 2014 to
identify the LULC and detect changes in water and vegetation cover between spring and summer using
Landsat 8 OLI imagery. The global position system (GPS) Garmin and Nikon Camera were used during
the collection of ground truth points. In addition to this data the available ground control points and
photos from previous field surveys are also used in this work.
Existing auxiliary data comprising topographical maps, QuickBird satellite images, digital spatial
geological data, laboratory reports and thematic layers were prepared as well. This data was employed to
facilitate fieldwork, help with the interpretation of remote sensing data and to validate the results of
LULC mapping and change detection assessments. There are 507 ground truth points from which 288
points were collected during the field survey of the study period and from previous field surveys on the
study area. The rest of the ground control points were collected based on the auxiliary data (Figure 2).
The ground truth points were subdivided into two groups, one group for selecting training sites for the
supervised classification and the second group for the assessment of the accuracy of the image index
results.
5
Freiberg Online Geoscience Vol 43, 2015
2.5 Images processing and interpretation
Image pre-processing, supervised classification, accuracy assessment and change detection using image
indices was performed.
2.5.1 Supervised classification
In this study a supervised classification using maximum likelihood was applied based on the spectral
differences between different classes. These differences were used to subdivide the LULC of the LZRB
into separate classes. Prior knowledge is utilized to execute supervised classification, depending on
previously collected training sites from certain areas of known LULC. In order to execute a supervised
classification, it is necessary to collect spectral signatures from training areas, which are then used to
“train” the classification algorithm (Chen and Stow 2002; Jusoff et al. 2009). Several training site were
collected from different places of the study area depending on the collected ground control points, field
observations and the auxiliary data. Training sites for each LULC class were collected by selecting many
training sites for the same class considering their spatial distribution. Based on the statistics of these
training sites, each pixel in the classified image of the LULC map was then assigned to these training
sites.
The LULC map for the LZRB was generated based on the pixel-by-pixel supervised classification using
Landsat 8OLI images from September 2014. Classes have been separated by selecting the spectral
signatures of each class from different image sites, using the seed properties method that is provided in
ERDAS Imagine V.11 software. The results were manually corrected by ArcGIS, using mainly OLI8
images of 30 meter spatial resolution. The image was improved and enhanced using digital image
enhancement processing techniques to highlight some LULC classes before interpretation. Image
enhancement processes alter the impression of the image on the viewer. In Erdas software different
enhancement techniques are available: contrast enhancement, linear and nonlinear contrast stretching,
density slicing, Gaussian stretching and so forth. Because the enhancement distorts digital pixel values,
supervised classification was carried out on the original images.
6
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
Figure 2: Ground control points used for verifying the accuracy of the LULC map
Integrating the supervised classification results of the Landsat 8 OLI and QuickBird data offered the most
satisfactory results. A QuickBird image (2006) was used when detailed information was required;
otherwise Landsat data was used. Several steps were used in order to improve the classification and to
resolve interference problems between classes depending on visual interpretation with the help of
auxiliary data and QuickBird images. The main interference existed between the urban and the built-up
land class with the barren land subclasses particularly in small residential areas such as villages and
towns. The misclassification in residential areas is due to the heterogeneous composition of buildings,
roads, and wooded and non-wooded areas (Ward et al. 2000). After supervised classification, all images
were vectorized into polygons to modify the classified image results for more accuracy using several
methods such as band ratio, and visual digitizing. The spatial extents of LULC were calculated in Arc
Map for each subbasin based on their derived extents and the produced LULC map. The classification of
the LULC was produced according to a modified USGS classification (Anderson et al. 1976) to be more
compatible with the goal of the study and the environmental conditions of the LZRB area. Finally, a set of
13 LULC classes and subclasses were recognized in the LZRB area.
2.5.1.1 Accuracy assessment
The increasing use of satellite data in different scientific fields in the last decades is because of its low
costs, and also because it provides faster and more powerful results. However, results obtained from
satellite data are accompanied by a certain error probability. Accuracy is considered to be the degree of
closeness of the results to values accepted as true. To evaluate the quality of thematic maps produced
7
Freiberg Online Geoscience Vol 43, 2015
from raster images, examining the accuracy of the classification and evaluating the agreement of the
resulting map for the particular purpose is the main approach (Foody, 2008). The accuracy of the
classification is assessed by comparing the classification results with known information. The minimum
level of interpretation accuracy in the identification of LULC categories from remote sensor data should
be at least 85 % (Wright and Morrice 1997; Anderson et al. 1976). According to Thomlinson et al. (1999)
the target should be an overall accuracy of 85 % with no class less than 70 % accuracy. Despite the
attractiveness of having a target value and a standard method for accuracy assessment, it appears the
remote sensing researchers are not achieving the typically specified targets yet (Foody, 2002).
There are many factors that can affect the accuracy value such as the characteristics of the satellite data,
the scale of the study area, and the details in the LULC classes. Thematic maps extracted from remote
sensing data should undergo a statistically rigorous accuracy assessment before being used for scientific
investigations (Stehman and Czaplewski 1998). Accuracy in this study was evaluated with an error matrix
or so-called confusion matrix. An error or confusion matrix is among the most common approaches used
for calculating the accuracy of thematic maps derived from multispectral imagery (Smits et al. 1999;
Congalton and Green 2002; Liu et al. 2007). An error matrix displays the results from the comparison of
reference class labels of the LULC categories with the real results (Stehman and Czaplewski 1998).
For accuracy assessment, the overall accuracy, the producer’s accuracy, the user’s accuracy and the
overall kappa coefficient was calculated. The user’s accuracy was calculated by dividing the number of
correctly classified pixels for each LULC class by the total number of pixels in the classified image. The
producer's accuracy refers to the number of correct pixels in the classified image for each LULC class
divided by the total number of pixels in the reference data. Accordingly, the user's accuracy reflects that a
given pixel can be identified on the earth surface as it is in the classified image, whereas producer’s
accuracy refers to the percentage of a given class that is correctly identified on the map. The overall
accuracy is computed by dividing the total number of correctly classified pixels by the total number of
reference pixels (Congalton 2001; Rogan et al. 2002). In this study, the accuracy assessment of the final
classified image was validated against earth surface features using randomly distributed ground truth
points. These points were considered as references for examining the accuracy of the LULC map of
LZRB and the change detections results. The accuracy assessment was conducted for each classification
result. Agreements and disagreements of the classification have been assessed by using an error matrix
and simple descriptive statistics (Table 1).
The Kappa index was also calculated for the classified image to measure the accuracy of the results. The
Kappa statistics of the LULC map result after classification is 0.811. The Kappa coefficient expresses the
proportionate reduction of the errors generated by a classification process compared with the error of a
completely random classification (Forghani et al. 2007). A Kappa value of 1 indicates perfect agreement,
and a value of 0.811 expresses that the produced LULC map avoided 81.1 % of the errors. The accuracy
achieved for the LULC map of the LZRB is well accepted amongst remote sensing specialists and GIS
practitioners (Forghani et al. 2007; Zhang et al. 2002; Harris and Ventura 1995).
2.5.2 Change detection (image indices)
One of the most important applications of satellite data is to detect the changes for the same object
between different seasons and different years. The process by which variation in an object at different
times or changes that occur in LULC are monitored and identified over a certain number of years is
known as change detection (Singh 1989,Tewolde and Cabral 2011). Changes in LULC result in changes
in radiance values and these changes will be largely related to radiance changes but also to other factors
(Ingram et al. 1981). The impact of these factors such as differences in atmospheric conditions, Sun
angle and soil moisture can partially be reduced by selecting suitable data (Singh 1989; Jensen, 1983).
8
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
Table 1: Accuracy assessment of the LULC map of the LZRB, reference totals is the total number of
tested pixels in the reference image, classified totals is the number of classified pixels in the classified
image, number correct is the number of correctly classified pixels, which agree with the pixels in the
reference image. The producer’s, the user’s and the overall accuracy are explained in paragraph 2.5.1.1.
LULC units Reference
totals (true)
Classified
totals
(predicted)
Number
correct
Producer’s
accuracy
Users
accuracy
Water 22 22 20 90.9 90.9
Urban and built-up Land 41 29 29 70.7 100
Natural vegetation 26 24 19 73.1 79.2
Cropland and pasture 40 36 35 87.5 97.2
Cultivated land 37 37 26 94.6 74.3
Harvested land 40 33 29 72.5 87.9
Burned land 17 6 5 29.4 83.3
Igneous and/or
metamorphic rocks 36 32 27 75 84.4
Carbonate rocks 26 36 24 92.3 66.7
Clastic rocks 47 54 46 97.9 85.2
Conglomerate 21 24 21 100 87.5
Bare soil 22 23 21 95.5 91.3
Mixed barren land 132 151 121 91.7 80.1
Totals 507 507 423
Overall classification accuracy = 83.4 %, Overall Kappa statistics = 0.81
Change detection analysis is used to determine the nature; extent and rate of land cover changes through
spatial and temporal variations (Rogan and Miller 2006; Zubair 2006). Information from change detection
can be used for land management and future planning by linking many important variables such as
urbanization, water management, deforestation and land degradation (Mancino et al. 2014). Image
differencing is executed by subtracting images from two different time periods of the same location or
scene (Fox et al. 2007). Spectral radiance values of satellite data can be analyzed individually for each
band or for a combination of many bands (Singh, 1989). The difference image for two or more dates
shows either an increase or a reduction of particular LULC categories and displays also the variation in
their spatial distribution. Many indices have been developed and utilized to detect changes in different
locations and under different environmental conditions around the world. The Normalized Difference
Water Index (NDWI) and the Normalized Difference Vegetation Index (NDVI) have been widely used to
produce useful time series results. Multi-temporal satellite images are commonly used to identify and
monitor spatial and temporal changes in water and vegetation cover. In this study, the threshold of the
reflectance of each index was determined by manually checking the index values and comparing the
results with false-color images. All index equations were run using the ‘Model Maker’ in Erdas Image
software. The size of the all LULC classes and the image index areas in the different years have been
computed in ArcGIS. Integrating remote sensing and GIS technologies was proven to be very useful for
detecting changes.
3 Results and discussion
3.1 Land Use Land Cover map classes
Thirteen LULC classes and subclasses have been identified in the LZRB; these are: urban and built-up
land, barren land, which involves 6 subclasses (igneous and/or metamorphic rocks, carbonate rocks,
clastic rocks, mixed barren land, bare soil and conglomerate), agricultural land, which involves three
categories (cropland and pasture, cultivated land and harvested land), natural vegetation, burned land and
water. The spatial distribution of the LULC classes is illustrated in Figure 3. The covered area and
9
Freiberg Online Geoscience Vol 43, 2015
percentage that each LULC class covers are shown in Table 2. Figure 4 shows which percentage of the
LZRB each class covers.
Figure 3: LULC map of the LZRB (Landsat 8OLI, September 2014)
Table 2: Area and percentage that each LULC class covers in the classified image of the LZRB
(Landsat 8 OLI, September 2014)
LULC class and subclass LZRB (km
2
)
% of the LZRB
Igneous and/or
metamorphic rocks
Barren
land*
1293.3
6.5
Carbonate rocks 1207.1
6.1
Clastic rocks 2103.1
10.6
Conglomerate 566.3
2.9
Bare soil 2036.3
10.3
Mixed barren land 6841.1
34.5
Cropland and pasture Agricultural
land**
783.3
3.9
Cultivated land 980.2
4.9
Harvested land 1322.9
6.7
Urban and built-up land 113.2
0.6
Natural vegetation 2278.1
11.5
Burned land 187.1
0.9
Water 148.6
0.8
percent 19860.7
100
* Total barren land 14047.3 km
2
with 70.7 % of the total area,
**Total agriculture land 3086.45 km2 with 15.5 % of the total area
10
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
Figure 4: Pie chart of the LULC classes in percent of the total area of the LZRB
3.1.1 Urban and built-up land
Within the last years, urban and also rural areas in the LZRB have grown rapidly due to the significant
economic growth in Kurdistan Region in Iraq as a whole and the higher demand for living space due to
the population increase. This trend has been most pronounced in cities, towns and villages that are located
in relatively flat areas. There are different approaches for mapping urban and built-up land such as visual
interpretation of aerial and satellite data, digitizing topographic map and digital interpretation using
satellite image processing techniques (Wang et al., 2010). The huge urbanization in the map area is
represented by many cities and towns like Koisanjaq, Raniyah, Qal`at Dizah, Taq Taq, Altun Kupri,
Hawija, Chuwaarta and Penjween within the territory of Iraq and Piranshar, Pasveh, Baneh and Sardasht
in the Iranian territory. Additionally, villages and scattered houses are distributed all over the map area.
The built-up land extracted using Landsat 8OLI 2014 occupies the smallest class with only 113.2 km2, i.e.
not more than 0.57 % of the total area (Table 2). It is difficult to determine urban areas in the LZRB
solely by digital image processing due to the use of various kinds of building materials. This results in the
confusion with other LULC classes as most of the villages are made of rocks and clay derived from
surrounding areas. In addition, the low spatial and spectral resolution of the images and the uncertainty of
spectral characteristics add to the obstacles in distinguishing this class. The spectral signature of villages
in the area is very difficult to separate from fallow land. Thus the classification was complemented by the
visual interpretation of QuickBird data using ArcGIS. Combining visual interpretation and digital image
processing leads to the best results (Jensen, 2000). The determined urban and built-up land is highlighted
in reddish blue to bluish pink color in 753 RGB in the Landsat 8 OLI image (Figure 5, A and B).
11
Freiberg Online Geoscience Vol 43, 2015
3.1.2 Natural vegetation
Natural Vegetation includes all vegetation cover that grows naturally, like forests, shrubs, perennial herbs,
and grasses. Orchards that produce various fruits and nuts are also included in this class because of their
limited spatial extension. It is distributed mainly in the high-mountains area of the basin and has the same
spectral signature as sparse trees and forest patches. Consequently, the LZRB provides valuable
information for assessing the relation between topography, morphometry, climate type and soil
characteristics and their effect on the diversity and composition of the vegetation. The vegetation cover in
arid and semiarid regions more strongly depends on the environmental conditions than in humid regions
(Böer & Sargeant, 1998). The environmental factors that change the composition of the vegetation in arid
and semi-arid regions include the geological and topographical situation, the climate conditions, the soil
type and grazing. They all lead to a variation in spatial and temporal changes in the composition of the
vegetation. The morphometric characteristics of the river basin have a direct effect on the spatial
distribution of plant communities (Al-Rowaily et al. 2012). The northern and north eastern parts of the
main basin receive the highest annual rainfall with an average of about 783 mm/a. The rainfall in the
lowland (west and northwest) gradually decreases; the average annual rainfall there is about 308 mm/a.
Generally, native vegetation and orchards are distributed in the high mountains in the north- and
northeastern parts of the main basin north of and around Dokan Lake (Figure 5, C and B). There are
different types of hazel filbert, nut trees, and fruits in the mountainous area. Shrubs mostly comprise
herbaceous plant species that mostly do not grow higher than 1 m and are distributed as scattered plants
or small communities throughout the LZRB. Several springs exist in the high-relief areas; they are
surrounded by different types of vegetation cover. The present study also showed that most mixed forests
exist in the northern and north eastern part of the LZRB and are associated with higher precipitation than
in the lower part of the main basin. The natural vegetation of the LZRB covers an area equal to
2278.07 km2, i.e. 11 % (Table 2) of the LULC map area.
3.1.3 Agricultural land
Agricultural land includes all types of agricultural land that can be distinguished at the acquisition date of
satellite image. The availability of water is the main factor controlling the distribution of species in arid
and semiarid regions as a result of the limited amount of precipitation and frequency of droughts (Zheng
et al. 2013). Agricultural lands are permanently changing at various spatial and temporal scales in
response to human activity and environmental factors.
The agricultural landscapes of the LZRB area are dominated by the farming of cereal crops. The
agricultural land in the low-relief area is based mainly on the production of cereals like barley and wheat.
The agricultural lands of the LZRB are mainly distributed in the flat terrain, low-relief land and hilly
terrain area but also along the flood plain of the main river course and its tributaries and alluvial fans
around Dokan Lake. Thus the variation of this land use along the main river from high-altitude
mountainous area to flat and hilly terrain has a direct influence on the vegetation’s composition and
diversity. The types of the agricultural land are:
3.1.3.1 Cropland and pasture
The cropland component of the agricultural land must be given significant attention to because its
frequently undergoes spatial and temporal changes which directly influence the biogeochemical and
hydrologic cycles of the land cover patterns (Wardlow et al. 2007). Cropland and pasture involve many
components in agricultural studies such as harvested cropland, cropland used only for pasture, the
alternation between crops and pasture, and land more or less permanently used for one of the purposes,
cropland or pasture (Anderson et al. 1976). The alluvial fans that surround Dokan Lake are fertile with
dense agricultural activities, where different crops grow on the large wheat farms in the summer season.
12
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
Most of the croplands that are mainly irrigated by precipitation are distributed in the northern and
northeastern parts of the main basin area, because the precipitation there is higher than in the southern
part, which is almost dry throughout the year. The LULC in the mountainous areas is characterized by
scattered cropland patches along narrow flood plains and Piedmont plains.
Farmland can be distinguished from natural vegetation by distinctive geometric features and lighter color
than natural vegetation (Figure 5, E and F). The red/green ratio was used to distinguish forests from
croplands. As the red band (0.64–0.67 µm) is the red-color absorption band of the chlorophyll of healthy
green vegetation and the green band (0.53–0.59 µm) represents the radiation reflected from leaf surfaces
this ratio can be used to discriminate broad classes of vegetation. Croplands appeared as lighter (brighter)
tone and forests appeared as darker tone and also the SWIR 2 /green ratio could be used to distinguish
forests from croplands. But it could not distinguish forests from water bodies. Consequently, different
attempts have been made in this research to discriminate natural vegetation from agriculture land; band-
ratio visual interpretation by manual class identification using ground survey data and mask creation
around agriculture land have been performed. The main obstacles in the separation of cropland from
pasture are that both classes have the same reflectance and both are growing close to each other or as
mixture on the same agriculture land. As it was apparent that the spectral signatures of these two classes
cannot be distinguished from each other they were merged.
The lower part of the LZRB is fertile and most wheat farms are found in this plain particularly in Al-
Haweja District in the lowest part of the basin. Cropland and pasture in the LZRB cover an area of about
786.73 km2 which represents 3.96 % of the LULC classes of the basin (Table 2 and Figure 4).
3.1.3.2 Cultivated land
Cultivated lands are prepared and improved lands for agricultural purposes to grow crops for food
production purposes. Leaving land unplanted for most of or for an entire growing season to enhance the
soil’s fertility for crop production in the following year is known as fallowing (Havlin et al. 1995). Fallow
land is included in the cultivated land class because it is difficult to discriminate between them. This can
also be justified by the fact that fallow land occurs only at a limited spatial extent in the basin area
because of the high fertility of the soil there.
The cultivated lands have different spectral signatures in different parts of the LZRB, depending on the
type and source of the soil material, which impacts the color and also some other properties of soil, like
its moisture content and texture. Agricultural lands are characterized by their regular geometric shape and
their location along road networks. These properties help to identify this land cover type when looking at
it from the aerial perspective.
Cultivated lands cover approximately the same area as cropland and pasture; it amounts 980.21 km2, i.e.
4.94 % of the map area (Table 2 and Figure 4). The spatial extent of cultivated lands is shown in Figure 3.
In some locations, there are large areas in which barren land and cultivated land are significantly mixed.
This is because they are composed of the same materials. So, they had to be separated based on the visual
interpretation of the filed reconnaissance survey. Cultivated land has a dark brown to brown color with
dark maroon color in the multispectral arrangement RGB753 of Landsat 8OLI (Figure 5 G and H).
13
Freiberg Online Geoscience Vol 43, 2015
Figure 5: Built-up land (A and B), natural vegetation (C and D), cropland (E and F) and cultivated land
(G and H), The images on the left show photos of the LULC type and the right image represent false-
color composites RGB753 of the Landsat image OLI 8.
14
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
3.1.3.3 Harvested land
The satellite images acquired stem from the end of the dry season (after harvesting season). The lands that
are classified and named as harvested land reflect the non-vegetated spectral response of the soil and crop
residues from the fields of wheat and barley in the LZRB area. Wheat and barley are usually planted in
November and December and are harvested in May or June. Therefore, they cover the land until the
harvesting time.
In the basin area dry herbaceous are widely distributed. As they have the same spectral reflection as
harvested land, these two land cover types can easily be confused (Figure 6A). Therefore, most of the
agricultural lands and dry herbaceous in the low relief land appeared almost the same as bare land with
the typical bright cyanic color on the RGB754 false color composite (Figure 6B). The results of the field
work showed that the bright cyanic color belongs to stubbles or other remnants of the harvested crops.
Harvested land covered about 1322.93 km2 of the LZRB area, i.e. 6.66 % of the map area (Table 2).
Figure 6: Harvested land (A and B), burned land (C and D) and water (E and F), The left image shows photos
of the LULC type and the right image represents false-color composites of the Landsat 8 OLI, RGB753 image.
15
Freiberg Online Geoscience Vol 43, 2015
3.1.4 Burned land
Burning land is a traditional way that aims to prepare the agricultural land before cultivation by removing
stubble residues or the remnants of harvested crops from the previous crop season (Figure 6C). It is an
annual practice for easily clearing the land for cultivation. Furthermore, burning harvested fields is done
to get rid of the rodents from farmlands before the next crop season (Eiumnoh and Shrestha 2000).
Burn agriculture is widely used in the territory of Iraq and particularly in the Kurdistan region in the north
of Iraq. Harvested land and pastures are burnt regularly to improve their quality (Lindell 2011). Although
the chemical changes resulting from burning can contribute to an improvement of the soil’s fertility, this
effect is only temporal (Hölscher et al. 1997). Fires in some cases get out of control, expand to areas of
natural vegetation and consequently affect areas that the farmer actually did not want to burn. The adverse
effects of the burn agriculture, like land degradation and erosion, need to be understood by the people in
order to prevent this bad practice. Repeated burning on slopes caused an increase in the erosion rate
(Scotter 1972) and subsequently, surface runoff. The increase in the spatial extent of bare rock and soil
can be better explained by the burning of bushes on slopes. The geochemical effect of burning organic
material in the topsoil is a lack of important nutrients such as C, N and S by volatilization processes
(Giardina et al. 2000; Sommer et al. 2004; Ewel et al. 1981). Several studies report the resulting ash from
burn land to reduce soil acidity as burning also increases the amount of one or several exchangeable base
cations such as magnesium and calcium. Moreover, burn agriculture impacts the C cycle and
consequently affects global climate. Information on the occurrence of burned land can help agricultural
scientists and local authorities to monitor this problem and its effect on the ecosystem. Stone (1971) states
that after twelve years of burning there is no difference in grain productivity on the site where wheat
straw has been burned. These burned lands can be identified by their distinct maroon color in the Landsat
8 OLI, RGB 753 image (Figure 6D). Burned lands are distributed in many parts of the basin (Figure 3).
They cover an area of about 187.13 km2 in LZRB, i.e. 0.94 % of the total map area (Table 2, Figure 4).
3.1.5 Water
The water category comprises water bodies, streams, canals, and other linear water bodies, lakes,
reservoirs, bays and estuaries (Anderson et al. 1976). Water bodies in the LZRB are represented by
Dokan Lake and LZR which appear as linear features surrounded by strips of agricultural lands,
particularly in low-relief areas. The delineation of water areas depends on the scale of the data
presentation and the resolution characteristics of the satellite data used for preparing the LULC map
(Anderson et al. 1976). There are also some permanent tributaries and many intermittent and ephemeral
streams, which are usually active during rainy season. During heavy rain most of the valleys are flooded.
Some of them are flash-flooded during strong rain storms and supply huge amounts of water to the LZR.
Dokan Lake is an artificial reservoir mainly supplied by water from the LZR. It represents an important
reservoir for the storage of water for agricultural land irrigation, drinking water supply, power generation,
and flood control. Additionally, it is surrounded by fertile agricultural land extending over the surface of
alluvial fans (Figure 6E and F.). Although the variation of the spatial extent of water bodies and the water
level of the lake are artificially controlled by the dam, the climatic conditions, particularly the fluctuation
of the rainfall rate is also crucial for the assessment of the amount of water in the lake. The fluctuation of
the water level is directly reflected by the surface water body’s spatial extension. Consequently, the
artificially controlled Dokan dam at least partly affected the statistical calculation of the water class in the
LULC map.
There are many other water resources in the LZRB such as natural springs that are distributed in different
places particularly in the mountainous area. Wetlands in the LZRB are also included in the water class. It
is mainly represented by bogs in high mountain area, and small backswamps nearby the main river and
around some springs. Wetlands occur only to a very limited extent and are hard to be depicted in the
LULC map, due to scale limitations. The area covered by water in the LZRB accounts for about 148.6
km2 which is 0.75 % of the total map area (Table 2 and Figure 4).
16
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
3.1.6 Barren land
The land that is not used for any purpose is commonly known as barren land. Barren land categories are
used to classify lands with a limited capacity to support life and having less than 5 percent vegetative
cover (scattered vegetation) (Lillesand et al. 2014). Generally, the surface of barren land is covered by
salt flats, sand dunes, mud flats, beaches, bare exposed rock, bare soil, or by salt-affected soils (Anderson
et al. 1976).
Barren land is more vulnerable to erosion than vegetated land which is more resistant to and protected
against soil erosion (Iqbal and Khan 2014). Barren land occurred everywhere in the study area. The data
reveals that the total area of barren land covers 14047.3 km2 which corresponds to 70.73 % of the total
map area (Table 2 and Figure 4). The subcategories of barren land in the LZRB include mixed barren
land, igneous and/or metamorphic rocks, clastic rocks, carbonate rocks, conglomerates and bare soil.
Remote sensing data and techniques supply means for successfully distinguishing between and mapping
of exposed rock classes and other associated weathering products. They provide information that is very
useful in addition to geological maps of the bedrocks (Leverington and Moon 2012). Landsat 8 OLI
images have been successfully used to map the exposed rocks within the LZRB. Bare exposed rocks in
the LZRB are subdivided into igneous and/or metamorphic rocks, carbonate rocks, and clastic rocks,
which are described hereinafter.
3.1.6.1 Igneous and/or metamorphic rocks
The application of Landsat 8 OLI satellite data to identify the types of igneous and metamorphic rocks
was successful. Mafic and ultramafic rocks can easily be discriminated because fresh mafic rocks have a
higher reflectance than fresh ultramafic rocks (Macias 1995). There are many factors that pose obstacles
to the mapping of rocks like extensive vegetation covers or dry herbaceous, soil and weathering products.
The spectral reflectance of geological materials can be controlled by surface alteration rather than by the
material’s fresh internal mineralogy (McSween et al. 2009). Thus, the characteristics of the rock’s
spectral reflectance are strongly impacted by the physical and chemical conditions of the rock outcrops.
Landsat 8 OLI images allow the effective discrimination between ultramafic rocks and mafic rocks. Mafic
rocks have lighter colors than ultramafic rocks due to their relatively high silica content, which has a high
spectral reflectance. Mafic rocks have magenta color in the mutispectral band composite image 753 RGB
while ultramafic rocks have dark violet to bluish color in that image image (Al-Rubaiay et al. 2010). The
same results are obtained from the Landsat 8 OLI image (Figure 7B). Igneous and metamorphic rocks are
characterized by the distinct and multispectral reflections, which are merge to one class after
classification.
Igneous and metamorphic rocks occupy 1293.3 km2, i.e. 6.51 % (Table 2 and Figure 4) of the LZRB.
They are located in the complex mountainous area in the north and north east of the LZRB (Figure 3).
This class comprises different types of igneous and metamorphic rocks with some sedimentary rocks that
belong to the Qandil Metamorphosed Series (Cretaceous), Shalair Series (Early-Late Cretaceous), Katar
Rash (Volcanic) Group (Late Cretaceous), Intrusive Complex (Earl-Late Cretaceous) and the Walash
Volcanic Rock Group (Paleocene/Eocene -Oligocene). An example for igneous and metamorphic rocks
within the basin are mafic or basaltic lava, laminated and andalusite schist, gneiss, phyllites, serpentinite,
quartzite, recrystallized and massive metamorphosed limestone, andesites, diorite, granodiorite, syenite,
nepheline syenite (Sissakian 1993; Ma’ala 2007). The Mawat Ophiolite succession from the bottom
upwards consists of ultramafic rocks (Figure 7A), gabbro, spilitic basalt and metabasalt (Al-Mehaidi
1974). There are also different igneous and metamorphic rocks mapped and reported in geological maps
of Mirwan and Mahabad in Iran territory such as pyroxene hornfelse facies, homogenous phyllite, gneiss,
acidic volcanic rocks, andesitic to basaltic volcanic rocks with pillow structures, spotted slate, andalusite
and chiastolite schist, ophiolite, marble, serpentinite, gabbro, diorite and so forth.
17
Freiberg Online Geoscience Vol 43, 2015
Figure 7: Mafic and ultramafic rocks (A and B), carbonate rocks (C and D), conglomerates (E and F) and
mixed barren land (G and H), the images on the left show photos of the LULC type and the right images
represent false-color composites of the Landsat 8 OLI, RGB753 image.
18
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
3.1.6.2 Sedimentary rocks
The LZRB area is mainly covered by sedimentary rocks. They primarily comprise clastic rocks
(sandstones, siltstones and shales) and carbonates. Sedimentary rocks are exposed in different parts of the
LZRB area. In the central and upper part of the basin, mainly carbonate rocks are present; south and
southwest of Dokan Lake clastic rocks are most common. The lithological classification based on satellite
image interpretation was combined with the digitized results from the GEOSURV-Iraq geological maps.
Three types of sedimentary rocks were differentiated in the map area; they are described hereinafter.
- Carbonate rocks
Carbonate minerals have diagnostic absorption features in their reflectance spectra in the SWIR and TIR
band due to electronic and vibrational processes, so that the spectra can be used to discriminate carbonate
minerals from other minerals (Van Der Meer and de Jong 2001; Clark 1999; Russ 2011).
The exposed carbonate rocks of different geological formations appear in a light pinkish color in the band
combination of 753 RGB in the Landsat 8 OLI image (Figure 7D). The high spectral reflection of these
rocks is due to their light color (Figure 7C). The carbonate rocks that appear in a dark color are grouped
together with the mixed barren land subclass, because they are covered by thin weathered horizons, soil,
vegetation, or affected by different coloring constituents. Carbonate rocks cover an area of about 1207.1
km2, ie. 6.08 % of the total LZRB area (Table 2 and Figure 4). The carbonate rocks identified from
Landsat 8 OLI are mainly present around Dokan Lake and in the northeastern part of the main basin
(Figure 3).
- Clastic sedimentary rocks
Clastic rocks such as sandstones, siltstones and claystones or other unconsolidated materials belong to
different formations such as the Fatha, Injana, Mukdadiyah and the Bai Hassan Formations in Iraq and the
Ruteh, Parut and the Lulan Formations in Iran (Figure 3). These sediments are mainly distributed in the
central and southwestern parts of the LZRB and to a limited extent also in the north eastern part of the
main basin. Clastic rocks in the central and southwestern parts are exposed in the anticline and syncline
on both sides of the LZR. These rocks can be discriminated by their spectral signatures depending on the
spectral response of the contained materials. They exhibit different color variations ranging from violet,
pink to light pinkish for sandstones, siltstones and claystones, respectively, on Landsat 8 OLI RGB753
color composite images (Figure 7F). Clastic rocks that are predominantly composed of sandstone are
characterized by darker color shadings than those composed of siltstone and claystone. In the LZRB,
clastic sedimentary rocks cover an area of about 2103.1 km2 which corresponds to 10.59 % of the map
area (Table 2 and Figure 4).
- Conglomerates
The conglomerate subclass has been separated from the clastic rocks because they are widely distributed
and easily distinguishable in the Landsat data. The coarse-grained sedimentary rocks (conglomerates)
generally show dark magenta color shades in the Landsat 8 OLI, RGB 753 color composite images
(Figure 7F). They also exhibit spectacular dendritic drainage patterns in highly dissected badland terrain
morphology. Moreover, the hills within badland areas consist of conglomerates characterized by smooth
surfaces (Figure 7E). Conglomerates are separated from other clastic rocks because they can easily be
distinguished using the Landsat data. Conglomerates are distributed mainly in the south and south west of
Dokan Lake on the flanks of some anticlines and along the main river course.
The exposed conglomerates in the LZRB mainly belong to the Bai Hassan Formation, recent Quaternary
sediments of river terraces and recent valley fill sediments. The river terraces consist of different
19
Freiberg Online Geoscience Vol 43, 2015
lithologies eroded from high-slope landscapes. They are generally composed of unconsolidated coarse-
grained sediments. The valley fill sediments consist of alluvial material from mass wasting, surface runoff
from highlands and sediment load of rivers and valleys. The size of the gravels generally varies from few
mm up to boulders and locally blocks of up to 2 m were noticed in the upper reaches of some streams. In
the lower reaches of the valleys the size of gravel range from fine pebbles to up to 20 cm, often mixed
with a sand matrix. This class occupies 566.28 km2, i.e. 2.85 % of the LZRB (Table 2 and Figure 4).
3.1.7 Bare soil
Bare soil is restricted mainly to the lower part of the LZRB in the foot hill zone, which appears at the
surface of piedmont, sheet run off and polygenetic sediments. Bare soil stretches out on both sides of the
LZR’s course and is almost utilized for agricultural purposes. The main agricultural areas are widely
distributed within this part of the region due to gentle slopes and fertile soils. Furthermore, irrigation
canals are distributed within this part particularly in the south and southeast of LZR in the surrounding of
Al-Haweja Town. In the areas of bare soil that are located on the southwestern side the agricultural
activity is lower than on the opposite side because of limited or abandoned irrigation canals (Figure 3).
Generally, bare soils consist of eroded materials (silty and clayey materials with sandy and gypsiferous
admixtures) that cover the pre-Quaternary rocks. Bare soil is distinguished from other barren land because
it has a different spectral signature that makes it possible to discriminate it from the surrounding areas. In
many areas the bare soils can be recognized on the basis of their light color shade caused by its soft
texture. The high spectral reflectance of some areas is caused by the fine sediments that make up the
soil’s surface or dry bare soil with sparce vegetation cover. The variation of the physical properties (such
as moisture content) of the bare soils is the reason for the variation of the reflectance of this class. Bare
soils cover an area of about 2036.3 km2 which corresponds to 10.25 % of the LZRB (Table 2, Figure 4).
3.1.8 Mixed barren land
The class mixed barren land is used when a mixture of different barren land features occur and the most
dominant land use occupies less than two thirds of the area (Anderson et al. 1976). The mixed barren land
in the LZRB is represented mainly by the products of denudation and is largely composed of
unconsolidated or semi-consolidated recent alluvial, argillic, sandy alluvial and fluvial sediments of
weathered friable loose formations (Figure 7G). Mixed barren land is the largest LULC class in the
LZRB, representing 34.44 % and occupying 6841.1 km2 of the total LULC map (Table 2 and Figure 4).
The mixed barren land is classified on the basis of its spectral reflection and field checking. Different
shades of this class on the Landsat 8 OLI, RGB 753 color composite images are present due to big
differences between the source materials (Figure 7H). These differences comprise the sediments’
environment, variations in texture, the type of lithology, the degree of consolidation and so forth.
Moreover, many formations are grouped together in this category because they are composed of different
types of materials.
3.2 Change detection using image indices
In this research, the NDWI and NDVI were calculated from Landsat 8OLI 2014 images in wet (April)
and dry (September) season while from TM5 (2000), TM5 (1990), TM5 (1984) and MSS (1976) the
indices were calculated only for the dry season. Satellite images were utilized to determine the
spatiotemporal changes in natural vegetation, cropland, surface water, and urban and built-up land.
Changes in NDBI were identified using visual interpretation. The wet season in the LZRB extends from
20
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
October to April and the dry season extends from May to September. Therefore, the selected months for
the analysis of seasonal changes in the LZRB were April (wet or high-runoff season) and September (dry
season). Water features were manually identified and NDVI thresholds were manually applied to classify
thematic images into two classes, land and water. The best fit for the water features and the NDVI
thresholds for each index were determined using visual interpretation of the color composites of the
satellite data and also using the field work data.
3.2.1 Calculation of magnitude and percentage of changes
The magnitude and percentage of the changes detected represents the degree of spatial expansion or
reduction of the area occupied by the respective LULC class. Positive values indicate increases while
negative values indicate decreases of the spatial extent of the regarded LULC class. The magnitude of
change is calculated by subtracting the area of the respective class for each recent satellite image date
from that of the first image date. The value of change is calculated by subtraction between the recent and
first image and also between each successive date (Table 3). The percent of change (P) is then calculated
according to the following equations:
= R– F(. 1)
SD = R − O(eq. 2)
P =
∗ 100%(eq. 3) (Abubakar and Anjide 2012)
Where: RF is the recent reference, R is the recent date (2014, 2000, 1990, 1984), F is the reference (First)
date (1976), SD is the Successive Date, O is the Old date, P is the percentage of change relative to the
reference date (1976).
The NDWI, NDVI and NDBI indices are extracted from the surrounding area depending on the defined
thresholds and masks. In this study, changes in spatial extent of water and vegetation cover as well as
urban and built-up land between the earliest and latest image data and also between successive image
periods were calculated. An increase in the spatial extent is marked as positive (+).Change detection
identifies changes in terms of location and extent which are then utilized to generate spatial distribution
maps of the image indices.
21
Freiberg Online Geoscience Vol 43, 2015
Table 3: Change detection in the LZRB using image indices: A: area in km2 for each index, SD: the
difference in km2 for each index between successive dates, i.e. recent date (R) minus old date (O), RF: the
difference in km2 for each index between an image date and the reference date (1976), P: percent of
change relative to the reference date.
Year A SD
RF P Year A
SD
RF P
NDBI (urban and built-up land) NDVI (natural vegetation)
2014 April
3283.2
2014 Sept. 113.2 52 102.3 9.4 2014 Sept. 2278.1
-726.9
-368.5
-0.14
2000 61.2 20 50.3 4.6 2000 3005 -69.7 358.4 0.14
1990 41.2 15 30.3 2.8 1990 3074.6
699.7 428 0.16
1984 26.2 15.4 15.4 1.4 1984 2375 -271.6
-271.6
-0.1
1976 10.9 0 0 0 1976 2646.6
0 0 0
NDWI (surface water bodies with Dokan Lake)
NDVI (cropland and pasture)
2014 April 230.7 2014 April
5134.1
2014 Sept. 148.6 -12.9 -155.2 -162.1 2014 Sept. 783.3 65.1 204.5 0.35
2000 161.5 -55.5 -149.2 -0.5 2000 718.3 -21.5 139.5 0.24
1990 217 25.1 -93.7 -0.3 1990 739.8 119.1 161 0.28
1984 191.9 -118.8 -118.8 -0.9 1984 620.7 41.9 41.9 0.07
1976 310.7 0 0 0 1976 578.8 0 0 0
NDWI (Dokan Lake only)
SD = R - O
RF: R-F
P = (R-F)/F*100%
(-): decreasing trend
2014 April 156.31
2014 Sept. 110.6 -27.6 -131.8 -0.5
2000 138.1 -53.1 -104.2 -0.4
1990 191.3 57.6 -51.1 -0.2
1984 133.7 -108.7 -108.7 -0.5
1976 242.4 0 0 0
3.2.2 NDBI
The fast expansion and changing pattern of NDBI reflects the changing economic and social conditions
and the increasing population. Several NDBI indices have been applied but they only gave results of low
accuracy. Therefore the NDBI was extracted mainly based on the visual interpretation of images by on-
screen digitizing of the urban areas for 1976, 1984, 1990, 2000 and 2014. The delineation of urban and
built-up land was performed using ArcGIS software. In the LZRB there are mainly urban settlements, but
also many rural settlements like towns and villages along the main river course. The total built-up land
area in 2014 was 113.2 km2, which is only 0.57 % of the total LZRB area. It has increased to 102.3 km2
from 1976 to 2014 i.e. by 9.39 %. The largest increase in built-up land area occurred from 2000 to 2014
(Table 1 and Figure 8). The distributions of urban and built-up land in the LZRB vary strongly. The
increase in their spatial extent goes step by step, which can be seen when comparing images from
successive dates (Figure 9).
22
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
Figure 8: Changes in image indices within the time period 1976-2014 and seasonal change between April
and September in 2014
3.2.3 NDWI
Remote sensing satellite data is widely used in water resources monitoring and management. It supplies
more accurate and convenient information for mapping surface water bodies and for monitoring the
dynamics of water (Ji et al.2009) .In this study the NDWI was applied to identify surface water using the
method of extracting thematic images from band ratios [equation 4 (Ji et al. 2009)] and [equation 5 (El-
Asmar et al. 2013)].
NDWIforLandsat8andTM5 = ()
(). 4 4 (Ji et al. 2009)
NDWIforMSS = (()
((). 5(El-Asmar et al. 2013)
23
Freiberg Online Geoscience Vol 43, 2015
Figure 9: Change detection map for September 1976, 1984, 1990 and 2000, April 2014 and September
2014
24
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
There are different image indices that have been calculated and examined to delineate surface water; the
aforementioned equations were found to be the most applicable and most accurate for determining the
spatial and temporal changes of the water features.
The NDWI has successfully been used to extract surface water features from Landsat 8 OLI and TM5 and
MSS images. Several image indices have been applied to extract water bodies from the MSS satellite
sensor but none of them produced convincing results because MSS images lack a shortwave (SWIR)
band. Therefore, different NDWI have been proposed to determine the best performing index and to
establish appropriate thresholds for the MSS for clearly identifying water features. The resulting NDWI
images highlight water bodies that are present in the study area. the NDWI index ranges from -1 to +1
with water bodies represented by high values (El-Asmar et al. 2013).
The reflectance of clear water is high in the green band and low in the near infrared because water is
characterized by strong absorption in the near infrared wavelengths range and beyond. Therefore, for
calculating the NDWI two bands are used: green (0.5-0.6 m) and near infrared (0.8-1.1 m). In this
study it was found that the above-mentioned equations are best for calculating the NDWI. Water features
in MSS images show positive NDWI values; a threshold of ≥ 0.13 was found to be the most suitable and
most accurate. Modeling spectral reflectance curves is not enough for water extraction because there are
many spectral mixtures that have the same spectral reflectance as water. The modeling was done using the
model maker in Erdas imagine software. Mixtures of water pixels with some other objects’ pixels have
been identified and processed depending on the matching between NDWI results with ground truth
points, field surveys, QuickBird image data and other axillary data using GIS.
There are also interferences between water identified with the NDWI applied on MSS images and the
ultramafic rocks exposed in the north eastern part of the basin. These rock exposures have the same
spectral characteristics as water and the same threshold value. By examining the NDWI in detail it was
observed that shadows are also mixed with extracted water features in some places. The interface areas
have been removed using GIS by converting the NDWI result to polygons and manually removing it
based on visual interpretation. Finally, the thresholds for finding water features were identified by visual
analyzing the band composition of Landsat 8 OLI, TM5 and MSS. The extracted results were validated
with ground-truth observations throughout the LZRB by field surveys and collected training sites. The
NDWI in the LZRB is mainly represented mainly by the linear features of the LZR, its main permanent
tributaries and Dokan Lake. The spatial variations of the NDWI during the (1976-2014) period are
illustrated in Figure 9. There is a dramatic decrease in the extent of the surface water in the LZRB (Table
3 and Figure 8). The change in NDWI during the period of 1976-2014 can be attributed mainly to the
variation in the storage of Dokan Lake which depends on the climatological conditions and the
management of Dokan Dam Reservoir (Table 3 and Figure 10). In this study, the surface area changes of
Dokan Lake during the period of 1976–2014 were investigated. Dokan Lake is one of the largest water
storage lakes in Iraq. The water storage of Dokan Lake is of economic importance for the power
generation and also provides a suitable place for recreation. There are huge variations in the spatial extent
of Dokan Lake between April and September that reflect the difference between wet (rainy) and dry
season (Figure 10). Times series of the spatial extension of Dokan Lake are displayed in Figure 10.
Dokan Lake is in a critical situation now as a result of decreasing surface water area with time and
increased water demand for agricultural and domestic uses. It is necessary to monitor the decrease of the
Dokan lake surface area regularly to understanding the main factors influence on the lake and to provide
the best management.
3.2.4 NDVI
The NDVI is the basic index for measuring the vegetation cover and for distinguishing vegetation from
other earth surface covers. NDVI calculations are based on the fact that the chlorophyll in living plants
strongly absorbs radiation in the visible range of red light and strongly reflects radiation in the near-
infrared range of the spectrum (Tucker 1979; Mather and Koch 2011). Although, there are many new
indices that have been developed to distinguish vegetation from non-vegetated lands and some of them
also take into account the soil’s behavior, the NDVI is still the most commonly used index (Forkuo and
25
Freiberg Online Geoscience Vol 43, 2015
Frimpong 2012). Changes in the vegetation cover (natural vegetation, cropland and pasture) of the LZRB
were detected using the NDVI in five observation times within the period of 1976-2014. Different sensors
are sensitive to different wavelengths in the bands that were used to determine NDVI values. Therefore,
there are systematic differences in the NDVI values derived from different sensors (Ouyang et al. 2012).
Figure 10: Change in the spatial extension of Dokan Lake during the period of 1976-2014 and the
seasonal change between April and September in 2014 based on the NDWI index results.
26
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
Hence, there are some slight differences in the NDVI thresholds between the different Landsat sensors.
Changes in vegetation cover during the analysis period (1976-2014) were detected using the NDVI
differencing images technique. First, the NDVI indices were calculated from the Landsat images during
the analysis period (1976-2014) using NDVI depending on the general normalized difference between
near infrared (NIR) and visible red (RED) (eq. 6). The resulting NDVI images were subtracted between
each successive date (i.e. 1976-1984, 1984-1990, 1990-2000 and 2000-2014 (Table 3)). NDVI images
with positive values indicate increasing vegetation while negative NDVI values indicate decreasing
vegetation cover. The use of the NIR and the red band was shown to be significantly correlated with the
amount of green leaf area or green leaf biomass (Tucker 1979; Singh 1989).
=ρNIR − ρRed
ρNIR + ρRed. 6
The results from applying equation 6 are NDVI images in which areas of sparse to none vegetation
correspond to lower values and vegetation to high values (Myneni et al. 1995). In the present study, the
best fit threshold values for vegetation cover identification were based on comparing image indices with
visual analyses of satellite data of RGB color composite images and field surveys. Determining the spatial
extension of cropland areas and separating them from natural vegetation was done by clipping the NDVI
image using a mask polygon for the cropland areas in the LZRB which represented the actual cropland
area extension. This can be considered as a possible representation of the cropland’s extension during the
period of 1976-2014. Several masks have been prepared for each date’s image depending on the
geometric shape of the farmland’s physiographic character, its spatial characteristics, the color shade of
the image, the vegetation pattern and field surveys. Finally, the natural vegetation has been separated
from cropland for each date using GIS according to the aforementioned indicators. The reflectance of
surface vegetation covers depends on their spectral response characteristics, structural properties of the
vegetation and the underlying soil (Myneni et al. 1995). There was no visible difference between the
vegetation extracted using the NDVI and that extracted based on the supervised classification of the
Landsat 8 OLI image from September 2014.
The NDVI of the natural vegetation displays an oscillating pattern during the period of 1976-2014 with a
maximum decrease in the image from 2014, while the NDVI of the cropland and pasture displayed only
slight variations during the same period (Table 3 and Figure 8). The variation in the spatial extent of the
NDVI of natural vegetation contributes mainly to the fluctuations of the climatological factors such as the
intensity and the frequency of the precipitation (Figure 9). The pasture land has a limited extent and is
mainly restricted to the seasonal river courses and the bank of the main river because the satellite images
were acquired in the end of the dry season. There are huge differences in the vegetation cover between
dry and wet seasons due to climatological and anthropogenic effects. During wet season there are large
farmland areas of wheat and barley even in hilly terrains that grow together with grass and pastures in
spring. The summer season is characterized by summer croplands of limited spatial extent that mainly
depends on the irrigation with water from the LZR and groundwater wells.
4 Conclusions
LULC change detection studies have to be performed and updated regularly to support a variety of future
planning activities and to resolve environmental issues. This research combined remote sensing, GIS, and
detailed ground information to produce LULC maps and to evaluate and detect changes in the LZRB
between 1976, 1984, 1990, 2000 and two dates in 2014 – one in April and one in September to compare
seasonal changes. From the LULC map of the LZRB thirteen classes and subclasses have been extracted
based on supervised classification. Barren land is the dominated class covering an area of about 14,047.26
km2 with a percentage of 70.7 %. The mixed barren land is the largest category of the barren land class
occupying 34.4 % of the main basin; the natural vegetation is the second largest class occupying 11.5 %.
The change detection results for the (1976-2014) period show that the area of the urban- and built-up
lands has increased from 10.9 km2 in 1976 to 113.2 km2 in 2014 due to the population increase and the
27
Freiberg Online Geoscience Vol 43, 2015
economic growth in the last decade. The NDWI shows a negative trend, it has decreased from 310.7 km2
in 1976 to 148.6 km2 in 2014. The negative trend of the NDWI is correlated mainly with the shrinkage in
the Dokan Lake area. The vegetation cover of the basin has been classified into two groups; NDVI
(cropland) and NDVI (natural vegetation). The NDVI (cropland) shows a positive trend during (1976-
2014) while the NDVI (natural vegetation) shows no clear trend during the same period. The increasing
of the agricultural land area is due to the increasing agricultural practices to meet the growing local
demand for food. In 2014, there has been a significant difference in the NDVI and the NDWI between the
wet and the dry season. The increase of the natural vegetation, cropland and water area in the wet season
2014 is mainly correlated to climatic conditions. Generally, the results indicate that during the dry
(summer) season from 1976 to 2014, the area of the natural vegetation has undergone negligible or
limited changes.
5 Acknowledgments
The research was supported by the Ministry of Higher Education and Scientific Research of Iraq
(MoHESR), and the TU Bergakademie Freiberg. We are grateful to the Geological Survey of Iraq for
providing the necessary data and support during field work. The authors are grateful to M.Sc. Hussein A.
Jassas, M.Sc. Ahmed T. Al-Rubaiay, and M.Sc. Arslan A. Othman for their valuable comments. We also
gratefully thank M.Sc. Mandy Hoyer for her very valuable and constructive comments and suggestions,
which helped us to improve the paper.
6 References
Abubakar M and Anjide A (2012) Analysis of Land Use/Land Cover Changes to Monitor Urban Sprawl
in Keffi-Nigeria. Environmental Research Journal, 6, 130–135.
Al-Mehaidi HM (1974) Geologicl Investigation of Mawat-Chuwarta Area, NE Iraq, GEOSURV,
Baghdad, Iraq. Int. Rep. No. 609.
Al-Rowaily SL, El-Bana MI, Al-Dujain, FA (2012) Changes in vegetation composition and diversity in
relation to morphometry, soil and grazing on a hyper-arid watershed in the central Saudi Arabia.
CATENA, 97, 41–49. doi:10.1016/j.catena.2012.05.004
Al-Rubaiay TA, Al-Ma′amar AA, Hattab, AA (2010) Integration of remotely sensed data and GIS
techniques to study Lesser Zab River Basin. Internalreport No. 3328, Geosurv, Iraq, Baghdad, 73 P.
Baghdad.
Anderson BJ, Hardy EE, Roach J T, Witmer RE (1976) A Land Use And Land Cover Classification
System For Use With Remote Sensor Data. U.S. Geological Survey Professional Paper 964 (Vol.
2001).
Aziz. MT, Ibraheem FA, Sebesta J, Hassan AR (1983) The Lesser Zab River basin project photo
engineering geological and geomorphological mapping. Internal report. no. 1405, GEOSURV-
IRAQ.
Barrios A (2000) Agriculture and Water Quality. American Farmland Trust, Center for Agriculture in the
Environment.
Böer B, Sargeant D (1998) Desert perennials as plant and soil indicators in Eastern Arabia. Plant and Soil,
199(2), 261–266. doi:10.1023/A:1004318610230
Chander G, Markham B (2003) Revised Landsat-5 TM radiometric calibration procedures and
postcalibration dynamic ranges. Geoscience and Remote Sensing, IEEE Transactions on, 41(11),
2674–2677.
Chen D, Stow D (2002) The effect of training strategies on supervised classification at different spatial
resolutions. Photogrammetric Engineering and Remote Sensing, 68(11), 1155–1162.
28
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
Chen X, Vierling L, Deering D (2005) A simple and effective radiometric correction method to improve
landscape change detection across sensors and across time. Remote Sensing of Environment, 98(1),
63–79. doi:10.1016/j.rse.2005.05.021
Clark RN (1999) Spectroscopy of rocks and minerals, and principles of spectroscopy. Manual of Remote
Sensing, 3, 3–58.
Congalton RG (2001) Accuracy assessment and validation of remotely sensed and other spatial
information. International Journal of Wildland Fire, 10(4), 321–328.
Congalton RG, Green K (2002) Assessing the Accuracy of Remotely Sensed Data: Principles and
Practices. Taylor and Francis.
Cooper CM (1993) Biological effects of agriculturally derived surface water pollutants on aquatic
systems - A review. Journal of Environmental Quality.
Cui L, Li G, Ren H, He L, Liao H, Ouyang N, Zhang Y (2014) Assessment of atmospheric correction
methods for historical Landsat TM images in the coastal zone: A case study in Jiangsu, China.
European Journal of Remote Sensing, 47, 701–716.
Del Valle HF, Elissalde NO, Gagliardini DA, Milovich J (1998) Status of desertification in the
Patagonian region: Assessment and mapping from satellite imagery. Arid Land Research and
Management, 12(2), 95–121.
Di Gregorio A, Jansen LJ (2000). Land cover classification system: LCCS: classification Nations,
concepts and user manual. Food and Agriculture Organization of the United Rome. Rome, Italy.:
FAO.
Eiumnoh A, Shrestha RP (2000) Application of DEM data to Landsat image classification: Evaluation in
a tropical wet-dry landscape of Thailand. Photogrammetric Engineering and Remote Sensing, 66(3),
297–304.
El-Asmar HM, Hereher ME, El Kafrawy SB (2013) Surface area change detection of the Burullus
Lagoon, North of the Nile Delta, Egypt, using water indices: A remote sensing approach. The
Egyptian Journal of Remote Sensing and Space Science, 16(1), 119–123.
doi:10.1016/j.ejrs.2013.04.004
Environment Canada (2001) Threats to Sources of Drinking Water and Aquatic Ecosystem Health in
Canada. National Water Research Institute, Burlington, Ontario. NWRI Scientific Assessment
Report Series No. 1. 72p. Page 69 – 15. Burlington, Ontario.
Ewel J, Berish C, Brown B, Price N, Raich J (1981) Slash and Burn Impacts on a Costa Rican Wet Forest
Site. Ecology, 62(3), 816–829. doi:10.2307/1937748
Foody GM. (2002) Status of land cover classification accuracy assessment. Remote Sensing of
Environment, 80(1), 185–201.
Foody GM (2008) Harshness in image classification accuracy assessment. International Journal of
Remote Sensing, 29(11), 3137–3158. doi:10.1080/01431160701442120
Forghani A, Cechet B, Nadimpalli K (2007) Object-based classification of multi-sensor optical imagery
to generate terrain surface roughness information for input to wind risk simulation. In 2007 IEEE
International Geoscience and Remote Sensing Symposium (pp. 3090–3095). IEEE.
doi:10.1109/IGARSS.2007.4423498
Forkuo EK, Frimpong A (2012) Analysis of forest cover change detection. International Journal of
Remote Sensing Applicaions, 2(4), 82–92.
Fox J, Rindfuss RR, Walsh SJ, Mishra V (2007) People and the Environment: Approaches for Linking
Household and Community Surveys to Remote Sensing and GIS. Springer US.
Frenken K (2009) Irrigation in the Middle East region in figures AQUASTAT Survey-2008. Water
Reports, (34).
29
Freiberg Online Geoscience Vol 43, 2015
Giardina C, Sanford R, Døckersmith I, Jaramillo V (2000) The effects of slash burning on ecosystem
nutrients during the land preparation phase of shifting cultivation. Plant and Soil, 220(1-2), 247–
260. doi:10.1023/A:1004741125636
Griffith JA, Martinko EA, Whistler JL, Price KP (2002) Interrelationships among landscapes, ndvi, and
stream water quality in the U.S. central plains. Ecological Applications, 12(6), 1702–1718.
doi:10.1890/1051-0761(2002)012[1702:IALNAS]2.0.CO;2
Harris PM, Ventura SJ (1995) The integration of geographic data with remotely sensed imagery to
improve classification in an urban area. Photogrammetric Engineering and Remote Sensing.
Havlin J, Schlegel A, Dhuyvetter KC, Shroyer JP, Kok H, Peterson D (1995) Great plains dryland
conservation technologies Publication , Manhattan, KS: Kansas State University, Vol. S-81, 7–11.
Hölscher D, Ludwig B, Möller RF, Fölster H (1997) Dynamic of soil chemical parameters in shifting
agriculture in the Eastern Amazon. Agriculture, Ecosystems and Environment, 66(2), 153–163.
doi:http://dx.doi.org/10.1016/S0167-8809(97)00077-7
Ingram K, Knapp E, Robinson JW (1981) Change detection technique development for improved
urbanized area delineation. Technical Memorandum CSC/TM-81, 6087.
Iqbal MF, Khan IA (2014) Spatiotemporal Land Use Land Cover change analysis and erosion risk
mapping of Azad Jammu and Kashmir, Pakistan. The Egyptian Journal of Remote Sensing and
Space Science, 17(2), 209–229. doi:10.1016/j.ejrs.2014.09.004
Jassim SZ, Goff JC (2006) Geology of Iraq. Geological Society of London.
Jensen JR (2000) Remote sensing of the environment: an earth resource perspective. Prentice Hall.
Jensen JR (1983) Urban/suburban land use analysis. Manual of Remote Sensing, 2, 1571–1666.
Ji L, Zhang L, Wylie B (2009) Analysis of dynamic thresholds for the normalized difference water index.
Photogrammetric Engineering and Remote Sensing, 75(11), 1307–1317.
Jusoff K, Ismail MH, Ali NH (2009) Spectral separability of tropical forest tree species using airborne
hyperspectral imager. Journal of Environmental Science and Engineering, 3(1), 37–41.
Leverington DW, Moon WM (2012) Landsat-TM-Based Discrimination of Lithological Units Associated
with the Purtuniq Ophiolite, Quebec, Canada. Remote Sensing, 4(12), 1208–1231.
doi:10.3390/rs4051208
Lillesand T, Kiefer RW, Chipman J (2014) Remote sensing and image interpretation. John Wiley and
Sons.
Lindell L (2011) Environmental Effects of Agricultural Expansion in the Upper Amazon [Elektronisk
resurs]: A study of river basin geochemistry and hydrochemistry, and farmers’ perceptions. Växjö,
Kalmar: Linnaeus University Press.
Liu C, Frazier P, Kumar L (2007) Comparative assessment of the measures of thematic classification
accuracy. Remote Sensing of Environment, 107(4), 606–616.
Ma’ala KA (2007) The geology of Sulaimaniyah quadrangle sheet NI-38-3, GEOSURV, Baghdad, Iraq.
Int. Rep. No.3095.
Macias LF (1995) Remote sensing of mafic-ultramafic rocks: examples from Australian Precambrian
terranes. Australian Geological Survey Organisation.
Mancino G, Nolè, A, Ripullone F, Ferrara A (2014) Landsat TM imagery and NDVI differencing to
detect vegetation change: assessing natural forest expansion in Basilicata, southern Italy. iForest -
Biogeosciences and Forestry, 7(2), 75–84. doi:10.3832/ifor0909-007
Mather P, Koch M (2011) Computer processing of remotely-sensed images: an introduction. John Wiley
and Sons.
30
Al-Saady et al. Land use and land cover mapping and change detection in the Little Zab River Basin (LZRB)
McSween HY, Taylor GJ, Wyatt MB (2009) Elemental composition of the Martian crust. Science (New
York, N.Y.), 324(5928), 736–9. doi:10.1126/science.1165871
Myneni RB, Hall FG, Sellers PJ, Marshak AL (1995) The interpretation of spectral vegetation indexes.
IEEE Transactions on Geoscience and Remote Sensing, 33(2), 481–486.
Nian Y, Li X, Zhou J, Hu X (2014) Impact of land use change on water resource allocation in the middle
reaches of the Heihe River Basin in northwestern China. Journal of Arid Land, 6(3), 273–286.
doi:10.1007/s40333-013-0209-4
Ouyang W, Hao F, Skidmore AK, Groen TA, Toxopeus A G, Wang T (2012) Integration of multi-sensor
data to assess grassland dynamics in a Yellow River sub-watershed. Ecological Indicators, 18(0),
163–170. doi:http://dx.doi.org/10.1016/j.ecolind.2011.11.013
Reed BC, Loveland TR, Tieszen LL (1996) An approach for using AVHRR data to monitor U.S. great
plains grasslands. Geocarto International, 11(3), 13–22.
Rogan J, Franklin J, Roberts DA (2002) A comparison of methods for monitoring multitemporal
vegetation change using Thematic Mapper imagery. Remote Sensing of Environment, 80(1), 143–
156. doi:10.1016/S0034-4257(01)00296-6
Rogan J, Miller J (2006) Integrating GIS and Remotely Sensed Data for Mapping Forest Disturbance and
Change. In Understanding Forest Disturbance and Spatial Pattern (pp. 133–171). CRC Press.
doi:doi:10.1201/9781420005189.ch6
Russ JC (2011) The Image Processing Handbook, Sixth Edition. CRC Press.
Santillan J, Makinano M, Paringit E (2011). Integrated Landsat image analysis and hydrologic modeling
to detect impacts of 25-year land-cover change on surface runoff in a Philippine watershed. Remote
Sensing, 3(6), 1067–1087.
Scotter GW (1972) Fire as an ecological factor in boreal forest ecosystems of Canada.Scotter, G. W.
1972. Fire in the Environment. USDA Forest Service p. 15-24. (U. S. F. Service, Ed.). Dept. of
Agriculture, Forest Service.
Serra P, Pons, X, Sauri D (2003) Post-classification change detection with data from different sensors:
some accuracy considerations. International Journal of Remote Sensing, 24(16), 3311–3340.
Singh A (1989) Digital change detection techniques using remotely-sensed data. International Journal of
Remote Sensing.
Sissakian VK (1993) The geology of Kirkuk Quadrangle sheet NJ-38- 2, GEOSURV, Baghdad, Iraq. Int.
Rep. No. 2229.
Smits PC, Dellepiane, SG, Schowengerdt RA (1999) Quality assessment of image classification
algorithms for land-cover mapping: A review and a proposal for a cost-based approach.
International Journal of Remote Sensing, 20(8), 1461–1486. doi:10.1080/014311699212560
Sommer R, Vlek PG, Deane de Abreu Sá T, Vielhauer K, de Fátima Rodrigues Coelho R, and Fölster H.
(2004) Nutrient balance of shifting cultivation by burning or mulching in the Eastern Amazon –
evidence for subsoil nutrient accumulation. Nutrient Cycling in Agroecosystems, 68(3), 257–271.
doi:10.1023/B:FRES.0000019470.93637.54
Song C, Woodcock CE, Seto, KC, Lenney MP, Macomber SA (2001) Classification and Change
Detection Using Landsat TM Data. Remote Sensing of Environment, 75(2), 230–244.
doi:10.1016/S0034-4257(00)00169-3
Stehman SV, Czaplewski RL (1998) Design and Analysis for Thematic Map Accuracy Assessment.
Remote Sensing of Environment, 64(3), 331–344. doi:10.1016/S0034-4257(98)000108
Stone EL (1971) Effect of prescribed burning on long term productivity of Coastal Plain soils.Proc.
USDA Forest Service, Southeastern Forest Experiment Station. Prescribed burning Symposium
proceedings. 160 pp. Asheville, N.C. p. 115-138.
31
Freiberg Online Geoscience Vol 43, 2015
Tewolde MG, Cabral P (2011) Urban Sprawl Analysis and Modeling in Asmara, Eritrea. Remote
Sensing, 3(12), 2148–2165. doi:10.3390/rs3102148
Thomlinson JR, Bolstad PV, Cohen WB (1999) Coordinating Methodologies for Scaling Landcover
Classifications from Site-Specific to Global. Remote Sensing of Environment, 70(1), 16–28.
doi:10.1016/S0034-4257(99)00055-3
Tucker CJ (1979) Red and photographic infrared linear combinations for monitoring vegetation. Remote
Sensing of Environment, 8(2), 127–150. doi:10.1016/0034-4257(79)90013-0
Ulbricht KA, Teotia HS, Civco DL (1993) Supervised Classification to Land Cover Mapping in Semi-
Arid Environment of NE Brazil Using Landsat-TM and SPOT Data. INTERNATIONAL
ARCHIVES OF PHOTOGRAMMETRY AND REMOTE SENSING, 29, 821.
Van Der Meer F, de Jong SM (2001) Imaging Spectrometry:: Basic Principles and Prospective
Applications (Vol. 1). Springer Science and Business Media.
Wang, L, Gong P, Ying Q, Yang Z, Cheng X, Ran Q (2010) Settlement extraction in the North China
Plain using Landsat and Beijing-1 multispectral data with an improved watershed segmentation
algorithm. International Journal of Remote Sensing, 31(6), 1411–1426.
doi:10.1080/01431160903475332
Ward D, Phinn SR, Murray AT (2000) Monitoring growth in rapidly urbanizing areas using remotely
sensed data. Professional Geographer, 52(3), 371–386.
Wardlow B, Egbert S, Kastens J (2007) Analysis of time-series MODIS 250 m vegetation index data for
crop classification in the U.S. Central Great Plains. Remote Sensing of Environment, 108(3), 290–
310. doi:10.1016/j.rse.2006.11.021
Wright GG, Morrice JG (1997) Landsat TM spectral information to enhance the land cover of Scotland
1988 dataset. International Journal of Remote Sensing, 18(18), 3811–3834.
doi:10.1080/014311697216630
Zhang Q, Wang J, Peng X, Gong P, Shi P (2002) Urban built-up land change detection with road density
and spectral information from multi-temporal Landsat TM data. International Journal of Remote
Sensing, 23(15), 3057–3078. doi:10.1080/01431160110104728
Zheng JG, Chen YW, Wu GX (2013) Association of Vegetation Patterns and Environmental Factors on
the Arid Western Slopes of the Helan Mountains, China. Mountain Research and Development,
33(3), 323–331. doi:10.1659/MRD-JOURNAL-D-12-00088.1
Zubair AO (2006) Change detection In land use and land cover using remote sensing data and GIS.
Unpublished Master Thesis, University of Ibadan.
32
Land use and land cover dynamics in the dry Lake
Haramaya Watershed in eastern Ethiopia using
remote sensing
Shimelis, B. Gebere Department of Hydrogeology, Institute of Geology, Technische
Universität Bergakademie Freiberg, Gustav-Zeuner Str. 12,
09599 Freiberg, Germany. Email: shimelisberhanu@yahoo.com
Merkel, Broder Department of Hydrogeology, Institute of Geology, Technische
Universität Bergakademie Freiberg, Gustav-Zeuner Str. 12,
09599 Freiberg, Germany. Email: merkel@geo.tu-freiberg
Agumassie, Tena A. Water and Land Resource Center, Addis Ababa, Ethiopia.
Email: tena.a@wlrc-eth.org
Abstract: The knowledge of land use and land cover is important for properly managing, planning
and monitoring natural resources. The aim of this study was to generate land use maps for the
study area and to understand the land use and land cover changes using remotely sensed satellite
imageries from 1985 to 2011. The images were geometrically corrected to a common map
projection followed by image processing operations. In ERDAS, supervised classification based
on the maximum likelihood algorithm was applied to the Landsat images acquired in 1985, 1995,
2006 and 2011. To check the accuracy of the classification, ground truth data was also collected.
Post classification change detection was applied in order to assess changes in land use and land
cover over time using IDRISI software. Rapid population growth has created land use
misbalances. The results showed that dramatic changes in land use and land cover have occurred
with shrinkage of water bodies, cultivated land, forests and grassland and at the same time
expansion of Catha edulis (chat)/shrubs as well as settlement areas. The result revealed that an
absolute shrinkage and loss of water bodies has occurred due to an extensive and massive
clearance of forests and grasslands. Between 1985 and 2011, forests became smaller and more
fragmented and declined from 202.6 ha to 101.6 ha. Only patches of mature forests are left. They
are under threat from expansion of land for chat production and settlements. In the mentioned
period, the total area of water bodies decreased from 3.5 % to 1.0 % while the area of grassland
and cultivated land decreased from 29.2 % to 19.6 % and 42.4 % to 32.6 %, respectively. For a
sustainable development of the watershed resources, proper land use planning is essential.
Keywords: Land use land cover changes, remote sensing, image analysis
1. Introduction
Land use and land cover changes are among the main causes for climate changes at local, regional and
global scales (Etter et al. 2006; Manandhar et al. 2009). Decreases in soil productivity, biodiversity
losses, shortages in water resources and global warming are often related to land use and land cover
changes (Dwivedi et al. 2005; Mas et al. 2004; Zhao et al. 2004). The changes affect both human and
33
Freiberg Online Geoscience Vol 43, 2015
natural systems and are recognized as the key factor for global change (Bockstael 1996; Reis 2008;
Yadav et al. 2012).
There is a variety of driving forces of land use and land cover changes such as urbanization,
population growth and economic booming (Gong et al. 2013). The growing human population and its
associated problems such as the rise in demand for land, water and forests along with poor resource
management practices, are the major driving forces behind the land use and land cover changes
(Hobbs et al. 1991; Lin and Ho 2003; Tanrivermis 2003). Hence, to meet the increasing demands of
basic human needs and welfare, land use and land cover information and potentials for their best use is
important for the selection, planning and implementation of land use systems (Fan et al. 2007;
Manonmani and Suganya 2010; Pandey 2012; Yadav et al. 2012).
Over the last two or three decades, remote sensing techniques have been widely acknowledged as
powerful tools to study land use and land cover changes at global to regional and local scales (Gong et
al. 2011; Jensen 2007; Lambin et al. 2003; Manandhar et al. 2009; Mayaux et al. 2008). Remotely
sensed imagery data provides reliable data at a more frequent interval and at relatively low costs. It
makes it possible to obtain land use and land cover information and generates results that are more
eye-catching than those obtained with traditional methods (Jat et al. 2008; Lu et al. 2004;
Mahmoodzadeh 2007; Wang et al. 2010). Using satellite images is attractive due to their regular
repetitive coverage, the recording of data from the same geographic area at the same time of the day,
and the consistent scale and look-angle (Jensen 1986).
Valuable information on land use and land cover change can efficiently be extracted from remote
sensing imageries. In this study, topographic maps, aerial photos, TM and ETM + Landsat images of
the studied watershed and other supporting documents were used to prepare land use and land cover
maps for years 1985, 1995, 2006 and 2011. Image classification was done using supervised image
classification with a maximum likelihood classifier. Supervised classification is usually known as
knowledge-based expert classification technique that depends on reference data to improve the
accuracy of the classified image (Berberoglu et al. 2007; Chen et al. 2007). In order to see the nature
of the changes of land use and land cover, maps were derived from satellite imageries, and post
classification change detection was performed. Accuracy assessment was also done using an error or
confusion matrix. An error matrix is the most commonly used method to present the accuracy of the
classified images (Fan et al. 2007).
A rapid change in land use and land cover is one of the challenges, which seriously contribute to
environmental degradation in Ethiopia in general and in the Haramaya catchment in particular. It is
witnessed that extensive shifting and intensive cultivation and the continuous increase in population
put significant stress not only on the land but also on water resources of the watershed. And also poor
practices of water use, over-exploitation of surface water and groundwater, clearing of forests and
siltation of lakes is continuing due to the increasing demand for water and biomass caused by the
population growth. Maps drawn from a series of satellite images of the study area show the Lake
Haramaya has disappeared completely.
Remote sensing data has been used for land use land cover dynamics analysis by (Assen 2011) in the
study area. On his paper, Catha edulis (chat) is categorized under agricultural crops; moreover his
findings reveal an increase in cultivated land but according to (Lemessa 2001) and (CESAR), Catha
edulis (chat) is not an agricultural crop but a shrub, which requires high quantities of irrigation water
compared with agricultural crops such as cereals or vegetables. In their study, (Taffesse et al. 2012)
also stated that in 2003/04 Catha edulis (chat) production in Ethiopia used only 1.2 % of the total area
cultivated; however, over five years, until 2008/09, its production increased by 6.1 % per year.
Different from his findings, this paper also confirms a decrease in cultivated land from time to time
34
Gebere et al. Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern Ethiopia
while Catha edulis (chat) farming increased significantly. Hence, based on intensive field visits and
the collection of ground truth data, a more realistic land use land cover dynamics analysis is provided
in this work.
Remotely sensed data in the form of Landsat images and aerial photographs is important for land use
and land cover mapping and is cost-effective (Ali et al. 2012; Solaimani et al. 2010). However, little
attention was paid to it in Ethiopia. Most of the land use and land cover maps developed in the country
are based on traditional procedures. With this fact in mind, this study aimed at understanding land use
and land cover changes and consequences for human activities in the dry Lake Haramaya watershed
over the past 26 years.
2. Study area and dataset
2.1. Description of the study area
The study area is located in the eastern highlands of Ethiopia (9o22ꞌ58ꞌꞌ to 9o31ꞌ22ꞌꞌ N and 41o53ꞌ44ꞌꞌ to
42o12ꞌ43ꞌꞌ E) covering an area of 15,319.4 ha as shown in Fig. 1. The watershed consists of a vast
depression area bounded by adjacent highlands. The elevation of the area ranges from 2460 m above
sea level in the northeastern parts to 2006 m.a.s.l. close to the central depression area.
Fig. 1: Map of the study area
The climate in the study area is semi-arid with a mean annual rainfall of 743 mm. Rainfall is bimodal;
its maxima occur in April and August with the highest peak in August (139.5 mm) while the smaller
peak is in April (97.5 mm). January is the driest month with a rainfall amount of 8.5 mm. The
elevation of the catchment causes is the reason why temperatures are moderate with an average
temperature of 16.4°C. It varies from 12.6°C in December to 19.0°C in June. Fig. 2 shows the annual
climatic variation in temperature and precipitation in the study area. The mean annual temperature
between 1970 and 2010 ranged from 12.4°C to 19.8°C while the annual rainfall during same period
ranged from 469 mm to 1104 mm.
35
Freiberg Online Geoscience Vol 43, 2015
Fig. 2: Mean climate data of the study area in the period from 1970 to 2010
The watershed generally consists of rocks ranging in age from Precambrian to recent depositions.
Stratigraphically, from the bottom to the top, the following order can be found: granite (Precambrian),
sandstone and limestone (Mesozoic sedimentary rocks) and recent sediments (Quaternary). The
predominant soil types of the watershed are characterized by medium to fine-textured materials.
Agriculture is the economic basis of the communities in the watershed (crop, vegetables, livestock and
cash crops such as Catha edulis (chat)). According to (Mesfin et al. 2011), on average each household
in the area owns about 0.74 ha of land. In general, activities other than agriculture seem to be very
limited. Cattle, sheep, goats and poultry are the most common domestic animals raised in the rural
area whereas the people in urban areas are mainly government, non-government, and private business
employees or traders.
The increasing population size can contribute to many types of environmental stress (Grimm et al.
2008). Rapid population growth and unplanned human settlements are the major driving forces to the
need of increased food production and stress for natural resources like soil, forests, water and fish. The
Haramaya watershed is the main source of water for the towns of Harar and Haramaya and nearby
communities. Rapid population growth in the area puts great stress on the limited resources available –
water, land, forests, etc. According to the Central Statistical Agency of Ethiopia 1994 census report,
the population living around the watershed was estimated to have been 166,597 while in 2007 the
number increased to 271,018; i.e. it increased by 63 %. According to the report, 49,711 people lived in
the watershed and the number increased by 74 % to 86,703 in 2007.
2.2. Cultivation of chat
Chat is an evergreen perennial shrub plant that belongs to the Celastraceae family. The plant is known
under different vernacular names: Khat is used in English and in Arabic, and chat in Ethiopia
(Lemessa 2001). In Ethiopia, chat is one of the main cash crops and is the major sources of country’s
foreign currency. There is a strong demand for it on local and foreign markets. It is cultivated for
chewing its fresh leaves and young tender twigs (Belwal and Teshome 2011). Moreover, the other
parts of the plant are used for various purposes such as fencing, as construction material for houses, as
firewood, etc. It is planted in rows with row spacing of ½ m to 1 m.
36
Gebere et al. Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern Ethiopia
The suckers or branches are used for vegetative propagation of chat. Due to the poor germination rate
of the seeds, they are not used for propagation. Chat farms are being irrigated when there is water
scarcity during dry season. In rainy season there is no need for irrigation since most parts of Ethiopia
get adequate amounts of rainwater. Usually chat is intercropped with cereals and vegetables. Chat can
be harvested all year round at any time of the day, but it is often harvested in the early morning or late
afternoon. Generally, well-established, irrigated mature chat plants can possibly be harvested up to 2-3
times a year depending on the age, management practices and the fertility of the soil.
2.3. Preparation of datasets
Landsat scenes were acquired from the USGS (U.S. Geological Survey) GLOVIS (Global
Visualization Viewer) website (http://glovis.usgs.gov/). While summer in Ethiopia is manly cloudy,
winter days are usually relatively cloud-free. Hence, four predominantly cloud-free winter Landsat
scenes between 1985 and 2011 were used. Three of these are Landsat TM scenes acquired in 1985,
1995 and 2011, while the last one is a Landsat ETM+ scene acquired in 2006. The dates of the images
were chosen to be as closely to each other as possible, but at least in the same vegetation season.
Other materials used are aerial photos (1986, 1996 and 2005) and topographic maps (1986 and 1999)
at a scale of 1:50,000 published by the Ethiopian Mapping Authority. ASTER 30 m resolution Digital
Elevation Models (DEM) were used to delineate the catchment of the study area. The DEM data had
some missing data points where no data was recorded by the mission due to some problems.
Therefore, 3DEM Terrain Visualization Software was used to fill the gaps. Arc Hydro Tools 9, which
is an extension for ArcGIS, was used to characterize the watershed of the study area. The results were
compared with those obtained from SRTM 90 m resolution DEM using both Arc Hydro and
Rivertools 2.4.
3. Methods
3.1. Image processing
In this study, supervised land use and land cover image classification was done. For image analysis, all
visible and infrared bands (bands 1–5 and 7) were included. While the eye provides the best
interpretation, computer algorithms are quicker and allow more consistent classifications. According
to Eastman (Eastman 2006) and Mayaux et al. (Mayaux et al. 2008), the human eye is often used to
guide digital classification or in conjunction with digital classification. For classification of images as
well as for accuracy assessment, training areas (i.e. reference points) were collected.
A number of independent training areas were randomly collected during field survey using a GPS
device. This survey was made to accurately determine the locations of the land use and land cover
class references. As much as possible an effort was made to make the randomly selected points to be
representative for the entire watershed. Also sufficient data was taken for land features such as water
bodies, forests and settlements during the field visit. Initially visual interpretation of images including
aerial photographs, land use maps, information from Google Earth was done and interviews were
made with the elder community members, who are familiar with the land use and land cover history of
the study area over long time periods. Features such as settlements and water bodies from topographic
maps were also used for selecting the training areas.
Land use and land cover classification was performed in ERDAS software using supervised
classification with the maximum likelihood algorithm. Compared with other classification algorithms,
the supervised maximum likelihood classification method gives higher classification accuracies
(Bolstad and Lillesand 1991; Lillesand and Kiefer 1994; Mengistu and Salami 2008; Reis 2008;
Solaimani et al. 2010). It is the most commonly and most widely used technique (Dewan and
37
Freiberg Online Geoscience Vol 43, 2015
Yamaguchi 2009; Serra et al. 2003; Yacouba et al. 2009). This technique is based on the decision rule
that unknown class pixels belong to the class with the highest likelihood among several classes (Foody
2002; Foody et al. 1992; Franklin et al. 2003; Solaimani et al. 2010).
Eight signature classes were created using the signature editor of ERDAS software. This process was
performed for all four Landsat images. The spectral profile for the land use and land cover classes of
the watershed is shown in Fig. 3. The following land use and land cover classes were identified:
settlements, forest plantations, grassland, cultivated land, Catha edulis (chat)/shrubs, bare land,
shallow water and water bodies. The descriptions of the land cover classes are presented in Tab. 1.
Fig. 3: Spectral profile of the land use and land cover classes
Tab. 1: Description of thesland cover classes
Land Cover Class Description
Bare land Area of exposed soils and bare exposed rocks
Shallow water Swampy area where water is on the surface
Cultivated Land Areas used for annual crop cultivation and fallow lands, which are
permanently or not irrigated.
Forest Plantation Represents both natural and enhanced plantation of forest areas that are
stocked with trees capable of producing timber or other wood products that
mainly are eucalyptus and conifers.
Water Body All bodies of water are in this class. It is the area that remains water-logged
throughout the year.
Settlement Areas where there is a permanent concentration of people building and other
man-made structures. Houses, roads, buildings, and bare soil.
Catha edulis (chat)/Shrub Catha edulis (chat), bushes, scrubs and tall herb communities.
Grass land Area covered with grass that is used for grazing and that covered for a
considerable period of the year.
38
Gebere et al. Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern Ethiopia
3.2. Accuracy assessment
Accuracy assessment of the final image is important once the classifications are completed. It is
important for post classification change detection analysis. Accuracy assessment of remotely sensed
images involves the comparison of the classified map with reference data (Congalton 1991). It can
give information on the quality of the map produced. In order to compare two classified images of
different dates, an error matrix or confusion matrix can be used (Congalton 1991; Congalton and
Green 2008). An error matrix is a c x c square array (c is the number of rows and columns), which is
used for comparing information from the classified image (row data) with reference data (column
data). Its most common application is land use and land cover change analysis using images of the
same target land features from two dates (Congalton and Green 2008; Ismail and Jusoff 2008).
To assess the accuracy of the independently classified images, the produced land cover maps and
reference data were compared as stated by (Congalton 1991). To check the accuracy of the produced
land use and land cover maps, error matrices were created. More than 40 reference points (training
areas) for each classified images were used for the accuracy assessment. The standard accuracy
assessment criteria such as producer’s accuracy, user’s accuracy, overall classification accuracy and
the Kappa statistics were used to show the accuracy of classification images.
The Kappa statistics were computed using the following equation:
( )
( )
)1(
1
2
1 1
∑
∑ ∑
=
++
= =
++
−
−
r
i
ii
r
i
r
i
iiii
xxN
xxxN
where r is number of rows in the error matrix, xii is number of observations in row i and column i, xi+
is the total number of observations in row i, x+i is the total number of observations in column i and N is
total number of observations included in the matrix.
3.3. Change analysis
Change detection is one of the most important applications of remote sensing. Post classification
change detection analysis is ideally one of the most simple change detection methods (Xu et al. 2009).
It is a comparative analysis of independently classified images of different dates. For the post
classification change detection in this study, each image was classified independently followed by a
thematic overlay of the classified images. The method results in a “from-to” change matrix showing
the transitions between the individual classes in a certain time period (Fichera et al. 2012).
In this study, post classification change detection was carried out to quantitatively show from-to
changes. In order to detect and quantify the extent of changes from one land cover category to another
at a later date, in IDRISI software the cross tabulation technique was used. For cross tabulation
operations, IDRISI was found to be better than ERDAS Imagine. Microsoft Excel was used to
organize some of the data for change detection analysis and to quantitatively determine the change
dynamics. The general workflow of the supervised classification is shown in Fig. 4.
39
Freiberg Online Geoscience Vol 43, 2015
Fig. 4: General workflow for a supervised land use land cover classification
4. Results and discussion
4.1. Results
This study gives a better understanding of land use and land cover changes in the study area of the
watershed for a period of more than two decades due to rise in number of population. To understand
the changes in the watershed, the land use and land cover categories were extracted for the year 1985,
1995, 2006 and 2011 using supervised maximum likelihood classification technique as a means of
image classification. All the images were classified into eight thematic classes: forest plantation, water
body, settlement, cultivated land, grassland, bare land, shallow water and chat/shrubs. The results from
the study indicate that the land use and land cover of the watershed has considerably changed.
As shown in Tab. 2, the majority of the area was covered by cultivated land in 1985 but from time to
time, much of the cultivated land was converted to Catha edulis (chat)/shrubs and settlements. For
example, from 1985 to 2011, the cultivated land cover decreased by 1,503.2 ha whereas the
Chat/shrubs cover increased by 3,144.0 ha. In 1985, 42.4 % of the watershed were covered by
cultivated land. This number decreased to 32.6 % in 2011. The area covered with Catha edulis
(chat)/shrubs increased from 20.7 % in 1985 to 41.2 % in 2011. Settlement areas also increased from
0.7 % in 1985 to 1.6 % in 2011. Bare land increased from 2.1 % in 1985 to 2.4 % in 2011. Water
bodies covered 3.5 % of the study area in 1985 but they decreased to 2.3 % in 1995; it further
decreased to 1.5 % in 2006, and in 2011 it covered only 1.0 %. It is obvious that the majority of the
cultivated land was converted to Catha edulis (chat)/shrubs as Fig. 5 shows.
40
Gebere et al. Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern Ethiopia
Fig. 5: Land use and land cover maps of the study area in the period from 1985 until 2011
Tab. 2: Extent and percentage of land use and land cover between 1985 and 2011
LULC
Categories
1985 1995 2006 2011
ha % ha % ha % ha %
Bare land 314.1 2.1 323.2 2.1 329.8 2.2 367.0 2.4
Chat/Shrubs 3170.9 20.7 4377.3 28.6 5145.6 33.6 6314.9 41.2
Cultivated land 6499.6 42.4 6231.3 40.7 5809.5 37.9 4996.4 32.6
Forest plantations 202.6 1.3 172.0 1.1 139.8 0.9 101.6 0.7
Grassland 4472.8 29.2 3626.2 23.7 3381.3 22.1 3005.1 19.6
Settlements 103.9 0.7 129.9 0.8 186.6 1.2 252.3 1.6
Shallow water 14.8 0.1 102.3 0.7 96.0 0.6 124.1 0.8
Water bodies 540.7 3.5 357.1 2.3 230.9 1.5 158.1 1.0
Total 15319.4 100 15319.4 100 15319.4 100 15319.4 100
Tab. 3 shows the extent of changes in land use and land cover between 1985 and 1995. A significant
amount of the cultivated land was converted to Chat/shrubs. For example, in 1985 cultivated land
covered 6,499.6 ha of the watershed and about 3,197.0 ha of the land remained unchanged in 1995.
This means that the cultivated land was partly converted to different land use and land cover classes
41
Freiberg Online Geoscience Vol 43, 2015
such as 1.5 ha to water bodies, 1,251.7 ha to grassland, 1,849.3 ha to Catha edulis/shrubs, 43.1 ha to
settlements, 122.3 ha to bare land, 3.6 ha to shallow water and 31.2 ha to forest plantations. A
significant amount of the cultivated land was converted to Catha eduli/shrubs.
Tab. 3: Change matrix of land use and land cover (ha) from 1985 (columns) to 1995 (rows)
Classes 1985
BL CS CL FP GL ST SW WB Total
1995
BL 90.5 3.2 122.3 1.7 100.5 5.0 0.0 0.1 323.2
CS 0.5 2014.0 1849.3 34.9 438.3 11.3 0.9 28.1 4377.3
CL 88.2 595.9 3197.0 40.4 2259.0 37.9 0.1 12.8 6231.3
FP 0.9 50.6 31.2 45.3 43.9 0.0 0.0 0.2 172.0
GL 130.7 474.5 1251.7 76.1 1568.1 21.9 6.3 97.0 3626.2
ST 3.2 10.4 43.1 1.4 43.8 27.8 0.0 0.1 129.9
SW 0.0 21.2 3.6 2.7 9.8 0.0 7.5 57.5 102.3
WB 0.0 1.2 1.5 0.0 9.5 0.0 0.0 345.0 357.1
Total 314.1 3170.9 6499.6 202.6 4472.8 103.9 14.8 540.7 15319.4
Overall Kappa Index of Agreement = 69.8 %
Note: WB = water bodies, SW = shallow water, BL = bare land, GL = grassland, CS = Catha edulis
(chat)/shrubs, CL = cultivated land, ST = settlements and FP = forest plantations
The land use and land cover changes from 1995 to 2006 can be derived from Tab. 4. For instance, of
6,231.3 ha which were covered by cultivated land in 1995, 2,785.3 ha remained unchanged in 2006.
Hence, 3,446.0 ha of the cultivated land were converted to other land use and land cover classes such
as 3.7 ha to water bodies, 1,353.5 ha to grassland, 1,895.8 ha to Chat/shrubs, 84.1 ha to settlements,
84.8 ha to bare land, 5.9 ha to shallow water and 18.3 ha to forest plantations. In fact, a significant
amount of the cultivated land was converted to Chat/shrubs. In the same time period, out of
4,377.3 ha, 2,287.9 ha of Chat/shrubs remained unchanged.
Tab. 4: Change matrix of land use and land cover (ha) from 1995 (columns) to 2006 (rows)
Classes 1995
BL CS CL FP GL ST SW WB Total
2006
BL 91.0 2.9 84.8 3.8 140.7 4.1 0.0 2.5 329.8
CS 12.8 2287.9 1895.8 54.1 830.8 19.1 19.8 25.3 5145.6
CL 129.3 1406.9 2785.3 49.9 1355.9 47.2 5.9 29.3 5809.5
FP 0.5 48.2 18.3 29.3 40.7 2.7 0.0 0.2 139.8
GL 78.3 618.0 1353.5 34.5 1112.1 23.0 44.0 117.8 3381.3
ST 11.3 7.7 84.1 0.5 48.6 33.9 0.0 0.5 186.6
SW 0.0 1.4 5.9 0.0 25.5 0.0 11.3 51.8 96.0
WB 0.0 4.2 3.7 0.0 71.9 0.0 21.2 129.8 230.9
Total 323.2 4377.3 6231.3 172.0 3626.2 129.9 102.3 357.1 15319.4
Overall Kappa Index of Agreement = 66.7 %
The changes in land use and land cover in the study area between 2006 and 2011 can be obtained from
Tab. 5. For example, Catha edulis (chat)/shrubs covered 5,145.6 ha in 2006 and 3,120.8 ha remained
unchanged in 2011. However, 2,024.8 ha were converted to different land use and land cover classes
like 20.3 ha to settlement areas, 846.1 ha to grasslands, 1,127.4 ha to cultivated land, 3.8 ha to bare
land, 3.7 ha to shallow water and 20.3 ha to forests. A significant increase in the Catha edulis
42
Gebere et al. Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern Ethiopia
(chat)/shrubs cover from 2006 to 2011 was noticed, from 5,145.6 ha to 6,314.9 ha. Again, the
shrinkage in cultivated land had the largest contribution to the expansion of Catha edulis (chat) farm.
Tab. 5: Change matrix of land use and land cover (ha) from 2006 (columns) to 2011 (rows)
Classes 2006
BL CS CL FP GL ST SW WB Total
2011
BL 132.3 3.8 65.5 0.0 152.2 12.6 0.4 0.1 367.0
CS 2.3 3120.8 2139.0 57.7 975.9 10.0 8.1 1.1 6314.9
CL 149.5 1127.4 2340.5 18.3 1255.0 81.8 18.7 5.0 4996.4
FP 0.3 20.7 21.6 42.6 16.5 0.0 0.0 0.0 101.6
GL 32.1 846.1 1133.2 18.5 901.3 15.5 32.3 26.1 3005.1
ST 13.3 20.3 106.0 2.3 43.7 66.7 0.0 0.0 252.3
SW 0.0 3.7 1.4 0.0 35.5 0.0 27.9 55.5 124.1
WB 0.0 2.8 2.2 0.4 1.2 0.0 8.6 143.1 158.1
Total 329.8 5145.6 5809.5 139.8 3381.3 186.6 96.0 230.9 15319.4
Overall Kappa Index of Agreement = 67.8 %
In 1985, the watershed was dominated by cultivated land (6,499.6 ha) but later in 2011, 2,629.5 ha of
the cultivated land had been converted to Catha edulis (chat)/shrubs. The result in Tab. 6 shows an
increase in Catha edulis (chat)/shrubs from 1985 to 2011. In 1985 Catha edulis (chat)/shrubs covered
about 3,170.9 ha of the land; it increased to 6,314.9 ha in 2011. The increase in Catha edulis
(chat)/shrubs is, recorded in 2011, due to the transformation of other land use and land cover classes to
it. For instance, water bodies covered 98.9 ha of the land, grassland contributed 1,121.1 ha, cultivated
land contributed 2,629.5 ha and forest plantations occupied 76.3 ha of the entire study area. The
settlement area also increased from 103.9 ha to 252.3 ha during the same time period. It was noticed
that the shrinkage in cultivated land to a large extent contributed to the increase in Catha edulis (chat).
Tab. 6: Change matrix of land use and land cover (ha) from 1985 (columns) to 2011 (rows)
Classes 1985
BL CS CL FP GL ST SW WB Total
2011
BL 83.3 4.6 82.3 0.6 129.9 6.0 0.0 60.2 367.0
CS 2.6 2365.1 2629.5 76.3 1121.1 17.9 3.4 98.9 6314.9
CL 190.5 365.6 2354.0 31.6 1945.2 38.0 5.5 66.0 4996.4
FP 0.3 27.8 19.4 34.1 19.7 0.2 0.0 0.1 101.6
GL 32.3 390.9 1316.9 49.2
1083.0
15.2 0.8 116.9 3005.1
ST 5.1 15.5 90.8 10.4 100.9 26.6 0.0 2.9 252.3
SW 0.0 0.3 0.3 0.0 1.0 0.0 5.0 117.5 124.1
WB 0.0 1.2 6.4 0.3 71.9 0.0 0.0 78.3 158.1
Total 314.1 3170.9 6499.6 202.6 4472.8 103.9 14.8 540.7 15319.4
Overall Kappa Index of Agreement = 65.0 %
Post classification of land use and land cover changes was performed to see the change from one land
use and land cover type to another land use and land cover type during the time periods 1985–1995,
1995–2006, 2006–2011 and 1985–2011 as compiled in Tables 3, 4, 5 and 6, respectively. All possible
changes and no changes were identified, with no changes shaded in grey color along the diagonals. For
example, between 1985 and 2011 (Table 6), more than 100 ha of forest plantations were cleared.
During the time period 1985 – 2011, the majority of the cultivated land (2,629.5 ha), water bodies
(98.9 ha), bare land (2.6 ha) and grassland (1,121.1 ha) was converted to Catha edulis (chat)/shrubs.
Bare land and settlements have never been transformed to shallow water or water bodies. Most of the
43
Freiberg Online Geoscience Vol 43, 2015
land in the watershed was converted to Catha edulis (chat)/shrubs. The overall Kappa index
agreement of the change detections ranges from 65.0 % to 69.8 %.
The assessment of the results for the classification accuracy for the four sets of Landsat images is
presented in Tab. 7. The overall classification accuracies for 1985, 1995, 2006 and 2011 are 86.0 %,
88.3 %, 82.8 % and 88.2 % respectively, with overall Kappa statistics of 0.798, 0.827, 0.753 and
0.811. The producer’s and user's accuracies also showed consistently high results, accounting for 66 %
to 100 % and 77 % to 100 %, respectively.
Tab. 7: Accuracy assessment for the classified images
Year Type of
Landsat Date Path/Row Resolution
Overall
classification
accuracy
Overall
Kappa
statistics
1985 TM 21 Feb, 1985 166 – 54 30 m 86.0 % 0.798
1995 TM 17 Feb, 1995 166 – 54 30 m 88.3 % 0.827
2006 ETM+ 06 Nov, 2006 166 – 54 30 m 82.8 % 0.753
2011 TM 12 Jan, 2011 166 – 54 30 m 88.2 % 0.811
Data source: USGS Global Visualization Viewer http://glovis.usgs.gov/
4.2. Discussion
In this study, four land use and land cover maps were produced using remotely sensed data through
supervised image classification with maximum likelihood classification. The main idea of image
classification is to automatically assign all pixels in an image to specific land use and land cover
classes (Lillesand and Kiefer 1994). The results of the supervised classification into eight land use and
land cover classes are shown in Fig. 6. Post classification change detection was also employed in order
to compare the four independently classified images. Eight land use and land cover classes were
identified in the studied watershed: forest plantations, water bodies, settlements, cultivated land,
grassland, bare land, shallow water and chat/shrubs.
Fig. 6: Land use and land cover changes in percent
44
Gebere et al. Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern Ethiopia
The rising population number and increasing socioeconomic demands create pressure on land use and
land cover which results in unplanned and uncontrolled changes in land use and land cover (Ikiel et al.
2013). The study result reveals that shifting large shares of the agriculture to Catha edulis
(chat)/shrubs cultivation and clearing forests for settlements is one of the causes of deforestation in the
watershed. The transformation of forest areas is related to human activities that affect the environment
(Gong et al. 2013; GUO et al. 2006). Forest lands are exploited extremely for housing and firewood
production as well as for satisfying the land demand for Catha edulis (chat) production. If the present
trend of deforestation in the study area keeps continuing, it is just a matter of time when the whole
forest cover will have been converted.
The study area is well known for extensive Catha edulis (chat) production due to availability of basic
infrastructure such as roads and due to the proximity of markets (Lemessa 2001). The rapid decline of
cultivated land may be largely attributed to land use conversions to Catha edulis (chat) production and
to the expansion of settlement areas. Presently Catha edulis (chat) is more and more being produced
by the local people and sold on local markets as well as exported to neighboring countries. This
creates the opportunity for local farmers to be able to improve their standard of living by earning
significant amounts of benefit instead of growing cereal crops or other vegetables.
All in all, this study revealed an increase in Chat/shrubs, bare land, settlements and shallow water
whereas cultivated land, grassland, water bodies and forest plantations decreased. A large share of the
water bodies were converted to shallow water and grasslands. Agriculture is the economic mainstay in
the study area and multiple cropping is practiced in most parts of it. Most of the farmers in the
watershed are small-scale farmers engaged in mixed farming, i.e. cropping and livestock production.
All marginal and grazing lands were brought under cultivation of Chat. Currently, Chat production
now occupies a significant amount of the watershed area. The production of this plant requires
substantial amounts of water mostly using surface irrigation.
Continuous population growth with an increasing individual demand for water used for irrigation,
municipal and industrial uses result in strong competition in the allocation of the scarce water
resources of the watershed. Between 1994 and 2007, the number of people living in the watershed
grew by 74 %. Changes in land use and land cover and improper use of the watershed’s resources have
also caused a substantial decline of the wetland resources of the study area. These may be major
causes for the complete demise of the lake. Lake Haramaya does no longer exist (the storage capacity
decreased to zero) for subsistence irrigation, domestic and industrial uses.
5. Conclusions
This study provides land use and land cover changes that occurred in Lake Haramaya watershed
between 1985 and 2011 using remotely sensed images. Remote sensing techniques are useful and
detailed ways for producing land use and land cover maps. Post classification change detection was
employed to analyze the land use and land cover changes in the study area. Information of land use
and land cover and its changing patterns over time is very important especially for the management of
natural resources of the watershed such as forests, water and soil. The main objective of the study was
to detect spatial and temporal land use and land cover changes using satellite imagery.
It was shown that the Catha edulis (chat)/shrubs cover increased, as well as the area occupied by bare
land, shallow water and settlements while the cultivated land, grassland, water body and forest
plantation cover decreased. The most important reason for the expansion of Catha edulis (chat)
production is that it is an attractive plant for increasing the income of the local farmers, who sell it to
local market and neighboring countries, and that it creates employment opportunities for the local
communities. Moreover, the climate pattern of the region is very suitable for growing Catha edulis
45
Freiberg Online Geoscience Vol 43, 2015
(chat). Although the plant is an important economic resource in the watershed, its high water demand
remained unknown and unexplored.
Rapid population growth along with the scarce natural resources and unmanaged cultivation practices
cause environmental changes in the watershed. The extent of cultivated land, water bodies, forest
plantations and grasslands are shrinking in favor of settlements and Catha edulis (chat)/shrubs farms.
According to the Central Statistics Agency reports, the population of the watershed annually increased
at a rate of 4 % between 1994 and 2007, a very high figure, compared with the national average being
2.7 %. The population in the towns of Harar, Awoday and Haramaya showed an annual growth rate of
about 3.2 %, which is still higher than the national growth rate.
Currently, uncontrolled and unplanned construction activities along with a high rate of population
growth cause an extensive depletion of water resources, which may lead to serious food scarcity and
misery for the rising population in the area in the near future unless immediate measures are taken.
These land use maps could be used as valuable means for monitoring changes in land use and land
cover patterns, for providing water resource information, for evaluating the extent of deforestation and
for dealing with lake rehabilitation in the study area as well. Information on land use and land cover
pattern changes is therefore critically important for land resources management and for future planning
activities. So, remote sensing is a powerful tool that can be used to extract information on land use for
monitoring, assessing, managing and planning natural resources.
Acknowledgements
This study was funded by the Engineering Capacity Building Program of Ethiopia (ECBP). The
authors gratefully acknowledge the support rendered by Haramaya University in providing office
facilities, materials helpful for this study and a vehicle for field data collection.
46
Gebere et al. Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern Ethiopia
References
Ali Z, Tuladhar A, Zevenbergen J (2012) An integrated approach for updating cadastral maps in
Pakistan using satellite remote sensing data International Journal of Applied Earth Observation
and Geoinformation 18(0):386-398 doi:http://dx.doi.org/10.1016/j.jag.2012.03.008
Assen M (2011) Land use/cover dynamics and its implications in the dried Lake Alemaya watershed,
eastern Ethiopia J Sustain Dev Africa 13(267-284
Belwal R, Teshome H (2011) Chat exports and Ethiopian economy: opportunities, dilemmas and
constraints Afr J Bus Manage 5(9):3635-3648
Berberoglu S, Evrendilek F, Ozkan C, Donmez C (2007) Modeling forest productivity using Envisat
MERIS data Sensors 7(10):2115-2127
Bockstael NE (1996) Modeling economics and ecology: the importance of a spatial perspective
American Journal of Agricultural Economics:1168-1180
Bolstad PV, Lillesand TM (1991) Rapid maximum likelihood classification vol 57. vol 1. American
Society for Photogrammetry and Remote Sensing, Bethesda, MD, ETATS-UNIS
CESAR Khat. Center for Substance Abuse Research. http://www.cesar.umd.edu/cesar/drugs/khat.asp.
Accessed June 23, 2015 2015
Chen X, Cai X, Li H (2007) Expert classification method based on patch-based neighborhood
searching algorithm Geo-spat Inf Sc 10(1):37-43 doi:10.1007/s11806-006-0145-y
Congalton RG (1991) A review of assessing the accuracy of classifications of remotely sensed data
Remote sensing of environment 37(1):35-46
Congalton RG, Green K (2008) Assessing the accuracy of remotely sensed data: principles and
practices. CRC press,
Dewan AM, Yamaguchi Y (2009) Land use and land cover change in Greater Dhaka, Bangladesh:
Using remote sensing to promote sustainable urbanization Applied Geography 29(3):390-401
Dwivedi R, Sreenivas K, Ramana K (2005) Cover: Land‐use/land‐cover change analysis in part of
Ethiopia using Landsat Thematic Mapper data International Journal of Remote Sensing
26(7):1285-1287
Eastman JR (2006) Idrisi Andes Guide to GIS and image processing:87-131
Etter A, McAlpine C, Pullar D, Possingham H (2006) Modelling the conversion of Colombian
lowland ecosystems since 1940: Drivers, patterns and rates Journal of environmental
management 79(1):74-87
Fan F, Weng Q, Wang Y (2007) Land use and land cover change in Guangzhou, China, from 1998 to
2003, based on Landsat TM/ETM+ imagery Sensors 7(7):1323-1342
Fichera CR, Modica G, Pollino M (2012) Land cover classification and change-detection analysis
using multi-temporal remote sensed imagery and landscape metrics European Journal of Remote
Sensing 45(1):1-18
Foody GM (2002) Status of land cover classification accuracy assessment Remote sensing of
environment 80(1):185-201
Foody GM, Campbell NA, Trodd NM, Wood TF (1992) Derivation and applications of probabilistic
measures of class membership from the maximum-likelihood classification LLLTLL 58(9):1335-
1341
47
Freiberg Online Geoscience Vol 43, 2015
Franklin J, Rogan J, Phinn SR, Woodcock CE (2003) Rationale and conceptual framework for
classification approaches to assess forest resources and properties. In: Remote Sensing of Forest
Environments. Springer, pp 279-300
Gong C, Chen J, Yu S (2011) Spatiotemporal dynamics of urban forest conversion through model
urbanization in Shenzhen, China International journal of remote sensing 32(24):9071-9092
Gong C, Yu S, Joesting H, Chen J (2013) Determining socioeconomic drivers of urban forest
fragmentation with historical remote sensing images Landscape and urban planning 117(57-65
Grimm NB, Faeth SH, Golubiewski NE, Redman CL, Wu J, Bai X, Briggs JM (2008) Global change
and the ecology of cities Science 319(5864):756-760
GUO L, XIA B-C, YU S-X, GONG C-F (2006) Effect of anthropogenic disturbances on the temporal-
spatial changes of landscape patterns at Taishan Mountain Chinese Journal of Eco-Agriculture
4(060
Hobbs NT, Schimel DS, Owensby CE, Ojima DS (1991) Fire and grazing in the tallgrass prairie:
contingent effects on nitrogen budgets Ecology 72(4):1374-1382
Ikiel C, Ustaoglu B, Dutucu AA, Kilic DE (2013) Remote sensing and GIS-based integrated analysis
of land cover change in Duzce plain and its surroundings (north western Turkey) Environmental
monitoring and assessment 185(2):1699-1709
Ismail MH, Jusoff K (2008) Satellite data classification accuracy assessment based from reference
dataset International Journal of Computer and Information Science and Engineering 2(2):96-102
Jat MK, Garg PK, Khare D (2008) Monitoring and modelling of urban sprawl using remote sensing
and GIS techniques International journal of Applied earth Observation and Geoinformation
10(1):26-43
Jensen JR (1986) Introductory digital image processing: a remote sensing perspective. Pearson
Prentice-Hall,
Jensen JR (2007) Remote Sensing of the Environment: An Earth Resource Perspective. Second edn.
Pearson Prentice Hall,
Lambin EF, Geist HJ, Lepers E (2003) Dynamics of land-use and land-cover change in tropical
regions Annual review of environment and resources 28(1):205-241
Lemessa D (2001) Khat (Catha edulis): Botany, distribution, cultivation, usage and economics in
Ethiopia Addis Ababa: United Nations Development Programme, Emergencies Unit for Ethiopia
(UNDP-EUE)
Lillesand TM, Kiefer RW (1994) Remote Sensing and Image Interpretation. John Willey & Sons Inc.,
United States of America
Lin GCS, Ho SPS (2003) China's land resources and land-use change: insights from the 1996 land
survey Land use policy 20(2):87-107
Lu D, Mausel P, Brondizio E, Moran E (2004) Change detection techniques International journal of
remote sensing 25(12):2365-2401
Mahmoodzadeh H (2007) Digital change detection using remotely sensed data for monitoring green
space destruction in Tabriz International Journal of Environmental Research 1(1):35-41
Manandhar R, Odeh IOA, Ancev T (2009) Improving the accuracy of land use and land cover
classification of Landsat data using post-classification enhancement Remote Sensing 1(3):330-
344
48
Gebere et al. Land use and land cover dynamics in the dry Lake Haramaya Watershed in eastern Ethiopia
Manonmani R, Suganya GMD (2010) Remote sensing and GIS application in change detection study
in urban zone using multi temporal satellite International Journal of Geomatics and Geosciences
1(1):60-65
Mas J-F et al. (2004) Assessing land use/cover changes: a nationwide multidate spatial database for
Mexico International Journal of Applied Earth Observation and Geoinformation 5(4):249-261
doi:http://dx.doi.org/10.1016/j.jag.2004.06.002
Mayaux P, Eva H, Brink A, Achard F, Belward A (2008) Remote sensing of land-cover and land-use
dynamics. In: Earth Observation of Global Change. Springer, pp 85-108
Mengistu DA, Salami AT (2008) Application of remote sensing and GIS inland use/land cover
mapping and change detection in a part of south western Nigeria African Journal of
Environmental Science and Technology 1(99-109
Mesfin W, Fufa B, Haji J (2011) Pattern, trend and determinants of crop diversification: empirical
evidence from smallholders in eastern Ethiopia Journal of Economics and Sustainable
Development 2(8):78-89
Pandey D (2012) Land Use and Land Cover Planning of Gondia Municipal City, Maharashtra State,
India Using Remote Sensing & GIS Techniques Int J Life Sc Bt & Pharm Res 1(4):46-64
Reis S (2008) Analyzing land use/land cover changes using remote sensing and GIS in Rize, North-
East Turkey Sensors 8(10):6188-6202
Serra P, Pons X, Sauri D (2003) Post-classification change detection with data from different sensors:
some accuracy considerations International Journal of Remote Sensing 24(16):3311-3340
Solaimani K, Arekhi M, Tamartash R, Miryaghobzadeh M (2010) Land use/cover change detection
based on remote sensing data (A case study; Neka Basin) Agric Biol JN Amer 1(1148-1157
Taffesse AS, Dorosh P, Gemessa SA (2012) Crop Production in Ethiopia: Regional Patterns and
Trends Food and agriculture in Ethiopia: Progress and policy challenges 74(53
Tanrivermis H (2003) Agricultural land use change and sustainable use of land resources in the
Mediterranean region of Turkey Journal of Arid Environments 54(3):553-564
Wang K, Franklin SE, Guo X, Cattet M (2010) Remote sensing of ecology, biodiversity and
conservation: A review from the perspective of remote sensing specialists Sensors 10(11):9647-
9667
Xu L, Zhang S, He Z, Guo Y The comparative study of three methods of remote sensing image change
detection. In: Geoinformatics, 2009 17th International Conference on, 2009. IEEE, pp 1-4
Yacouba D, Guangdao H, Xingping W (2009) Assessment of land use cover changes using ndvi and
dem in Puer and Simao counties, Yunnan Province, China World Rural Observations 1(2):1-11
Yadav PK, Kapoor M, Sarma K (2012) Land use land cover mapping, change detection and conflict
analysis of Nagzira-Navegaon corridor, Central India using geospatial technology International
Journal of Remote Sensing and GIS 1(2):90-98
Zhao GX, Lin G, Warner T (2004) Using Thematic Mapper data for change detection and sustainable
use of cultivated land: a case study in the Yellow River delta, China International Journal of
Remote Sensing 25(13):2509-2522
49
The influence of volatile organic components on the
stable isotopic composition of the groundwater in
Tanjero area - Kurdistan region, Iraq
Kareem, Aras Department of Hydrogeology, Institute of Geology, Technische
Universität Bergakademie Freiberg, Gustav-Zeuner Str. 12,
09599 Freiberg, Germany. Email: arasomer@yahoo.com
Merkel, Broder Department of Hydrogeology, Institute of Geology, Technische
Universität Bergakademie Freiberg, Gustav-Zeuner Str. 12,
09599 Freiberg, Germany. Email: merkel@geo.tu-freiberg
Abstract: Landfills in urban areas are among the major sources for organic contamination. Seasonal
weather changes accelerate the disintegration of organic matter and precipitation increases the risk of
leakage through the subsurface into the groundwater. BTEX compounds are among the most common
organic contaminants in the environment, particularly on industrial sites. In this study, they were
found in Tanjero’s industrial area southeast of Sulaimani city near Tanjero River. BTEX are present
there as a result of random dumping of volatile solvents by petroleum refineries on the waste dump
site as well as dumping of municipal waste rich in organics from Sulaimani city. The stable isotopes of
water in the wells around the waste site show an evaporation effect as compared to wells that are more
than 1 kilometer away from the landfill site. The seasonal variations of BTEX concentrations in
groundwater are correlated with the stable isotopes of water (δ18O, δ2H) as a result of increasing bio-
degradation, the lack of rainfall and dropping groundwater table in summer. Due to heat developed by
the microbial degradation of BTEX and other organic components, seepage water evaporates and leads
to a shift of the signature of the stable isotopes of water. This changed isotope signature can also be
traced in the groundwater in the downstream plume of the dump site.
Keywords: Landfill; petroleum refinery; stable isotopes; BTEX; groundwater contamination
1 Introduction
The growing population in Sulaimani city (Kurdistan region, northern Iraq) led to an increase in solid
waste generation that reached up to 1000 tons per day (Rashid 2010) (Fig. 1). In the last decade the
industrial zone in Tanjero area was characterized by an increasing number of petroleum refineries. The
area is located to the southwest of Sulaimani city (12 km from the city center). Waste, warehouses,
and fuel tanks are randomly spread without any control measures in place or any concept for environ-
mental protection. The industrial remains are dumped uncontrolledly on a fenced but non-supervised
and not constructed landfill site. The area of the landfill is about 50 hectares allocated for the munici-
pal waste dumping. The wells around the waste dump area are contaminated with benzene, toluene,
ethylbenzene and xylenes (BTEX) that caused changes in the quality of the pumped water and the
suitability for human consumption.
The seasonal weather changes are significant in Kurdistan region of Iraq. The climate is characterized
by cold and snowy winters and long, warm dry summers. Autumn and spring are short. During sum-
mer, the day temperature may reach 45°C. The main characteristic of the seasonal rainfall distribution
is the absence of precipitation in summer (June-September). Rainy season usually starts in mid-
October and ends in the beginning of May. The month with the greatest amount of precipitation in the
50
Freiberg Online Geoscience Vol 43, 2015
whole region is January (Stevanovic and Markovic 2003) (Tab. 1). The area of interest consists of
clastic rocks (Tanjero formation) with a thickness approximately 140 m and clastic alluvium sedi-
ments that contain gravel, sand, and clay (Budy et al. 1980; Ali 2007). Tanjero is a river flowing per-
manent along the southern margin of Sulaimani city and feeding Darbadikhan Lake in the east-
southeast (Habib 2003). This small river is polluted by the sewage of the city and some villages
around (NI 2008).
Fig. 1: Map of the study area
Tab. 1: Average monthly climate parameters of the study area over a period of 25 years (Mustafa 2006).
Month Air tempera-
ture (°C)
Relative hu-
midity (%)
Water vapor
pressure (mbar)
Rainfall
(mm)
Sunshine
(h)
Evapora-
tion (mm)
January 5.7 68.4 6.89 125.4 4.9 51.8
February 6.7 64.9 6.7 107.7 5.7 57.5
March 10.8 58.7 7.8 112.1 6.9 108.5
April 16.6 54.3 9.9 77.6 6.3 143.6
May 22.4 39.4 10.9 37.2 9.5 250.7
June 29 24.6 7.1 4.6 11.6 342.1
July 33.1 22.6 11.8 0.7 11.5 416.9
August 32.1 22.9 11.1 0 11.8 368.8
September 28.3 25 9.7 3.9 10.5 263.5
October 22 37.4 9.5 26 7.9 165.8
November 13.7 59.2 9 96.6 6.2 78.6
Vegetation area
51
Kareem, A.; Merkel, B. Influence of VOCs on the stable isotopic composition of groundwater in Tanjero area
The groundwater contamination in the study area is caused by the dumping of solid and liquid waste
and contaminants spread by surface runoff, river drainage systems, wastewater discharges, eutrophica-
tion and littering (Patrick et al. 2002). The increased amount of waste and contaminants endangers
local environments, natural resources, public health, local economies, and proper living conditions.
Various diseases like cancer result from the exposure to hazardous emissions mainly from waste in-
cineration without proper technologies. Especially the worker faces high health risks (EPA 2007).
Waste water discharges damage the soil and irrigation system (Green, Jammenjad 1997). The relative
share different components contribute to the total sum of municipal garbage dumped on the landfill
site is showed in Tab. 2 (Ray 1958).
Tab. 2: Main components of the municipal garbage (modified after (Ray 1958)).
Category %
Paper 30
Putrescible 25
Fines 10
Glass 8
Metal 8
Plastics 8
Textile 3
Rubber 3
Miscellaneous 5
The oil refinery industry produces several types of hydrochloric and sulfochloric acid leachate waste.
The refining and processing of crude oil requires water; one gallon of crude oil generates 30 gallons of
wastewater (Ray 1958).
Benzene, toluene, ethylbenzene and xylene (BTEX) are often found together and all have their origin
in the petroleum production and leaks from underground fuel storages and landfills. These compounds
are relatively mobile and can easily dissolve in water and hence contaminate groundwater (Davis et al.
1999; Kennedy 1992; Batlle-Aguilar 2008). BTEX compounds have been reported to be toxic com-
pounds having hazardous effects on terrestrial biota as well as aquatic organisms (An 2004). Some
regional environmental isotope studies have been done around the study area. Results from Haji Oma-
ran (Mawlood 2003) and Rania area (Al Manmi 2008) yielded a LMWL (Local meteoric water line)
defined by the equation:
δ2H=7.7δ18O+14.4 (Hamamin, Saeed 2012)
While the GMWL (Global meteoric water line) is defined by the equation:
δ2H = 8 δ18O +10 (Craig 1961).
Data from stable isotope analysis can be used to determine the origin of contaminants, the combination
of different sources contributing to a multi-source plume, but also to characterize their complex
transport processes and to assess measures for contaminated site remediation. Quantitatively assessing
isotopic fractionation can be done for studying the progress of environmental processes like biodegra-
dation. (North et al. 2006).
December 8.2 68.5 6.3 111.6 5.8 51
52
Freiberg Online Geoscience Vol 43, 2015
2 Materials and methods
15 water samples were collected from several wells around the landfill site (Fig.1 and Tab. 3) accord-
ing to the standards for water sampling for BTEX and stable isotope analyses (Borden et al. 1996;
Headley and Rae 1992). Sample A13 was taken southeast of the waste dump site and refineries and
sample A14 was taken from Tanjero River northwest of the waste dump site and refineries. A15 was
collected from Sarchnar spring 10 km away from the dump site. The rest of the samples were collected
from water wells in the industrial area at various distances from the waste dump site. For the GC-MS
determination, a 50 ml glass tube was filled with 25 ml of sample and sealed with caps using a crimp-
er. The volume of the stable isotopes analysis glass tube was 30 ml. The sample glass tubes filled
completely and firmly closed. All samples were with vials containing 1.5 ml in the auto sampler and
the average of the results calculated from 10 runs.
Tab. 3: Names and locations of the water samples.
Sample
No. Name of the location Latitude Longitude
A1 Yellow house well 35o
30´ 12.7´´ 45o 26´
11.5´´
A2 Metal melting factory well 35o
29´ 30.1´´ 45o 26´
2.1´´
A3 Livestock warehouse (well depth 42 m) 35o
28´ 50.6´´ 45o 26´
38.6´´
A4 Livestock warehouse (well depth 15 m) 35o
28´ 32´´ 45o 26´
38.6´´
A5 A house south of scrape location (well depth 75 m) 35o
30´ 10´´ 45o 25´
36´´
A6 The well of the School 35o
30´ 12.2´´ 45o 25´
39´´
A7 The well of the pebble factory 35o
28´ 49.4´´ 45o 25´
26.4´´
A8 The well of the pebble crushing factory 35o
29´ 5.7´´ 45o 25´
53.3´´
A9 The well beside the stream at the west bank 35o
28´ 41.7´´ 45o 25´
38.1´´
A10 The well of the oil factory 35o
28´ 35.6´´ 45o 25´
36.9´´
A11 The well of the building material warehouse 35o
28´ 4´´ 45o 25´
59.8´´
A12 Qaywan oil warehouse 35o
27´ 51.7´´ 45o 28´
23.9´´
A13 Tanjero Stream south the waste site 35o
28´ 15.9´´ 45o 26´
59.3´´
A14 Tanjero Stream above the industrial area 35o
30´ 30.9´´ 45o 22´
38.8´´
A15 Sarchnar Spring 35o
35´ 14´´ 45o 23´
0.9´´
BTEX was determined by headspace technique with gas chromatography and detection by mass spec-
trometry (Thermo Scientific Ultra-ISQ GC-MS) in the laboratory of the Department of Hydrogeology
of TU Bergakademie Freiberg according to the procedures described by (Butler et al. 2010). The sta-
ble isotopes δ18O and δ2H were measured by a liquid-water isotope analyzer based on cavity ring down
spectroscopy with a reproducibility of 0.3 ‰ for δ2H and 0.1 ‰ for δ18O (LGS, Los Gatos Research)
in the laboratory of the Department of Hydrogeology of TU Bergakademie Freiberg.
53
Kareem, A.; Merkel, B. Influence of VOCs on the stable isotopic composition of groundwater in Tanjero area
3 Results and discussion
The stable isotope analysis data for winter and summer is shown in Fig. 2 and Fig. 3. The standard
deviation and error bars are included as well for indicating the uncertainty of the stable isotopes data
whose importance (Donald et al. 2001) underlined.
Fig. 2: Scatter-gram of δ18O (‰VSMOW) versus δ2H with error bars for winter samples
Fig. 3: Scatter-gram of δ18O (‰VSMOW) versus δ2H with error bars for summer samples
A-1
A-2
A-3
A-4
A-5
A-6
A-7
A-8
A-9
A-10
A-11
A-12
A-13
A-14
A-15
-45
-43
-41
-39
-37
-35
-33
-31
-29
-8 -7.5 -7 -6.5 -6 -5.5 -5
δ2H (‰VSMOW)
δ18O (‰VSMOW)
A-1
A-2
A-4
A-3
A-5
A-6
A-7
A-8
A-9
A-10
A-11
A-12
A-13
A-14
A-15
-45
-40
-35
-30
-25
-7 -6.5 -6 -5.5 -5 -4.5 -4
δ2H (‰VSMOW)
δ18O (‰VSMOW)
54
Freiberg Online Geoscience Vol 43, 2015
In summer, two leachate samples were collected from two interflow drainages on the waste dump site
and analyzed for BTEX as shown in Tab. 4.
Tab. 4: BTEX concentrations in the leachate of the drainages.
3.1 BTEX and stable isotopes results for the end of December 2012 sam-
ples
BTEX components revealed rather low concentrations (< 0.55 ppb) during winter (Fig. 4) and elevated
concentrations of up to 3.4 ppb during summer (Fig. 6). During winter, the sampling point A4 showed
some higher BTEX values while in summer A3, A4, A9, A10, and A12 showed significantly higher
BTEX concentrations. During winter season, the xylene contents in the water samples A3 and A4
were elevated with 0.54 and 0.11 ppb, respectively, as compared to the other well and surface water
samples. The wells from which sample A3 and A4 were taken are located at 15 m distance from the
eastern fence of the landfill site. In the other groundwater samples xylene concentrations were equal to
or less than 0.11 ppb and the concentrations of the other BTEX components (benzene, toluene, and
ethylbenzene) were less than 0.1 ppb except for sample A4.
The average rainfall was 119.1 mm and the air temperature was 7.8 °C from November 2012 until the
end of January 2013. The stable isotopes plot around the LMWL except for those of sample A3 and