ArticlePDF Available

Detection and mapping of green-cover and landuse changes by advanced satellite image processing techniques (a case study:

Authors:

Abstract

The long-term occupation of Azerbaijan territories by Armenian forces had extreme forms of socioeconomic shocks, destructive geo-environmental, ecological, and evident Green-Cover (GC) and Landuse (LU) unenthusiastic changes. During the investigations multi-spectral and high-resolution Sentinel-2 satellite images sampled from 2016 to 2021 and normalized Diffe-rence Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI) were used. An Object-Based Image Analysis (OBIA) was adjusted in the segmentation and knowledge-based classification processes using the eCognition Developer and a Cellular Automata Markov-Chain (CA-MC) model with probabilistic rules and has been introduced within the TerrSet IDRISI Selva software. Indexing methods indicated that significant changes in GC are taken place in the sampled rayons of Qubadli, Zangilan, and Jabrayil (located in the southern Eastern Zangezur Economic Region) over the past years. Subsequently, OBIA methods approved that majority of the negative changes are detected in LU types, predominantly on the forests (-4.7%) and pastures canopies (-4.6%) considering the last six years. In addition, a reliable CA-MA prediction map designated that there would be recognizable upbeat progressive increases in barren (+4.8%) and abandoned (+5%) lands, particularly the emergence of vegetation vulnerability in the region. Consequently, accurate image processing and mapping of the current situation of the Azerbaijan liberated lands have to be the most urgent tasks of the geographers, ecosystem scientists, and remote sensing specialists before the government starts reconstruction and rehabilitation projects.
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
27
ANAS Transactions Earth Sciences 2 / 2022
http://www.journalesgia.com
DETECTION AND MAPPING OF GREEN-COVER AND LANDUSE CHANGES BY
ADVANCED SATELLITE IMAGE PROCESSING TECHNIQUES
(A case study: Azerbaijan Eastern Zangezur Economic Region)
Rasouli A.A.1, Safarov S.H.2, Asgarova M.M.3, Safarov E.S.2, Milani M.4
1Department of Environmental Sciences, Macquarie University, Sydney, Australia
Level 4, 12 Wally's Walk, Macquarie University, North Ryde,
Sydney, Australia:aarasuly@yahoo.com
2 Ministry of Science and Education of the Republic of Azerbaijan Institute
of Geography, Baku, Azerbaijan
115, H.Javid ave., Baku, Azerbaijan, AZ1143: safarov53@mail.ru (corresponding author)
3Azerbaijan State Pedagogical University, Baku, Azerbaijan
matanat_askerova@mail.ru
4Bandirma Onyedi Eylul University, Faculty of Engineering and
Natural Sciences, Bandirma, Turkey
Yeni Mahalle, Shehit Astsubay Mustafa Soner Varlık Caddesi, 77, 10200,
Bandirma, Turkey: mmilani@bandirma.edu.tr
Keywords: GC/LU changes,
Sentinel-2 satellite imagery,
dynamic and threshold spectral
indexing, knowledge-based
OBIA classification,
prediction CA-MC maps
Summary. The long-term occupation of Azerbaijan territories by Armenian forces had extreme
forms of socioeconomic shocks, destructive geo-environmental, ecological, and evident Green-Cover
(GC) and Landuse (LU) unenthusiastic changes. During the investigations multi-spectral and high-
resolution Sentinel-2 satellite images sampled from 2016 to 2021 and normalized Diffe-rence Vegeta-
tion Index (NDVI) and Normalized Difference Water Index (NDWI) were used. An Object-Based
Image Analysis (OBIA) was adjusted in the segmentation and knowledge-based classification pro-
cesses using the eCognition Developer and a Cellular Automata Markov-Chain (CA-MC) model with
probabilistic rules and has been introduced within the TerrSet IDRISI Selva software.
Indexing methods indicated that significant changes in GC are taken place in the sampled ray-
ons of Qubadli, Zangilan, and Jabrayil (located in the southern Eastern Zangezur Economic Re-
gion) over the past years. Subsequently, OBIA methods approved that majority of the negative
changes are detected in LU types, predominantly on the forests (-4.7%) and pastures canopies
(-4.6%) considering the last six years. In addition, a reliable CA-MA prediction map designated
that there would be recognizable upbeat progressive increases in barren (+4.8%) and abandoned
(+5%) lands, particularly the emergence of vegetation vulnerability in the region. Consequently,
accurate image processing and mapping of the current situation of the Azerbaijan liberated
lands have to be the most urgent tasks of the geographers, ecosystem scientists, and remote sensing
specialists before the government starts reconstruction and rehabilitation projects.
© 2022 Earth Science Division, Azerbaijan National Academy of Sciences. All rights reserved.
Introduction
Occupying the territory of neighboring countries
can have serious catastrophic social, economic, and, of
course, geographical, and ecological consequences
(General Assembly Security Council, 2009). One of the
vital consequences of the occupation of the Karabakh
and neighboring rayons by Armenia have been signifi-
cant changes in Green-Cover (GC)/Landuse (LU)
(GC/LU) types; and other environmental features of the
region (Salan, 2007; Scheffer, 2010). Our understand-
ing of how armed conflicts affect GC/LU change was
limited meanwhile after the liberation of Azerbaijan oc-
cupied rayons in November 2020 accommodating fun-
damental and complex changes and is a very essential
task formation about LU conditions and related changes
during the time (Baumann et al., 2015).
Azerbaijan geographers, ecologists, environ-
mentalists and urban planners are well aware more
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
28
than any scientist else in the world what damage has
been done by the occupation of the territories by
Armenian forces during the last 30 years (Giovarelli
and Bledsoe, 2001; Rasouli et al., 2018c). The fun-
damental driving force of GC/LU change is related
to temporal-spatial processes caused by war activi-
ties such as the compulsory seizure of agricultural
and livestock activities, along with the destruction of
rare forests (Hasanov et al., 2017). Moreover, we
should consider the uncontrolled erosion of soil, pas-
ture, and unique historical monument buildings de-
struction caused by unprincipled mining activities,
multiple bombings, and land-mine landings within
the occupied territories of Azerbaijan (Report: In-
vestigating the environmental dimensions…, 2021).
Eventually, the improper landuse and lack of proper
care of human and natural resources led to wide-
spread destruction and even extinction of many
plants and wildlife species in the region. Therefore,
to detect the current situation of the liberated lands,
few classic and advanced image-processing methods
could be applied to analyze post-war changes, focus-
ing on the southern rayons.
To date, several methods have been proposed in
the process of producing GC/LU maps through satel-
lite image processing, as it could be regarded as the
science and art of acquiring information about the
Earth's surface without actually being in contact with
it (Franklin and Wulder, 2002). Remote sensing
makes it possible to collect data by sensing and re-
cording reflected or emitted energy and processing,
analyzing, and production of a variety of practical
information in various fields (Lillesand, Kiefer,
2004). Through this modern technology, large
amounts of row data and information in the form of
digital satellite images are prepared and made avail-
able to researchers (Khandelwal et al., 2014; Coper-
nicus, Sentinel-2, 2020). Up to now, many different
methods have been developed in the processing data
obtained from remote sensing technology, each with
its advantages and limitations. Accordingly, in the
last few decades, a wide range of image processing
procedures, especially focusing on image indexing
and classifying, methods are provided, each with its
own goals, strengths, and weaknesses (Nelson and
Khorram, 2018). By reviewing the image processing
literature, it can be found that one of the well-known
methods of satellite image processing is information
extraction through their indexing; which usually
leads to the production of different styles of GC
maps (Platt and Rapoza, 2008). In addition, the liter-
ature surveys different classification methodologies
adopted for the effective LU productions by the pro-
cessing of satellite images. In an overview, this
could be simply divided into the traditional classifi-
cation methods, Object-Based Image Analysis,
Deep-Learning, and Hybrid/Integrated approaches
(Hussain et al., 2013; Rehman and Hussain, 2018).
Classic methods are the initial techniques for im-
age interpretation and information extraction. Satellite
image classification groups together the pixels of the
image into several different defined classes. The pix-
els are grouped based on the digital values extracted
from the satellite images (Bahadur, 2009). The prima-
ry areas of interest are in vegetation indexing, unsu-
pervised and supervised classifications, accuracy as-
sessment by the processing of Landsat and Sentinel
imagery (Sentinel Online, 2018). There are many
ways in which these techniques are classified and cat-
egorizing these techniques into supervised and unsu-
pervised is the most common way. Meanwhile, hu-
man image analysts have to play crucial roles in both
unsupervised and supervised image classification
procedures (Khatami et al., 2016). In reviewing the
subject literature, the advantages and disadvantages of
classic image processing based on their efficiency
and depends on the type of data have been exten-
sively stated (Hay and Castilla, 2006; Khandelwal et
al., 2014).
Methods of extracting information through the
indexing of satellite images have been common for
many years, some of them may refer to Normalized
Difference Vegetation Index (NDVI) and Normal-
ized Difference Water Index (NDWI) indicators
with their advantages and limitations (Thenkabail et
al., 2018). In recent years, intending to overcome the
limitations of traditional methods in the processing
of satellite images, new hybrid methods have been
introduced to high-light land surfaces such as vege-
tation cover and water bodies, by addressing a com-
bination of fuzzy and thresholding methods (Rasouli
et al., 2018a). Traditionally, two indexes of NDVI
and NDWI are well-known and widely used land-
cover indicators. NDVI is a simple, but effective
index for quantifying green vegetation (Pettorelli,
2013). It normalizes green leaf scattering in Near-
Infra-red wavelengths with chlorophyll absorption in
red wavelengths. The value range of the NDVI is 1
to 1. Negative values of NDVI (values approaching
1) correspond to water. Values close to zero (0.1
to 0.1) generally correspond to barren areas of rock,
sand, or snow. Low, positive values represent shrub
and grassland (approximately 0.2 to 0.4), while high
values indicate temperate and tropical rainforests
(values approaching 1). In return, NDWI makes use
of reflected NIR radiation and visible green light to
enhance the presence of such features while elimi-
nating the presence of soil and terrestrial vegetation
features. It is suggested that the NDWI could be a
fine proxy for the water indexing and provide re-
searchers with better estimations if applied by
threshold and legal methods.
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
29
In a more precise step of satellite image pro-
cessing, we can mention the OBIA method that is
one of the most modern approaches (Hay and Cas-
tilla, 2008). In a simple view, an OBIA classification
involves segmenting an image into objects (groups
of pixels) that have geographical features such as
shape, length, and topological entities such as dis-
tance and adjacency parameters. These attributes
make a knowledge base for the sample objects,
which can be called upon in the image indexing and
classification process, with a greater possibility for
detecting changes in higher resolution imagery (Te-
odoro and Araújo, 2014). OBIA is one of several
approaches developed to overcome the limitations of
the pixel-based approaches to incorporate spectral,
textural and contextual information to identify the-
matic classes in an image. Such kind of classifica-
tion is producing meaningful material distribution
maps via the identification of individual pixels or
groups of pixels with similar spectral responses to
incoming radiation (Tiede et al., 2017). While clas-
sic unsupervised and supervised classification is
pixel-based, OBIA groups pixels into representative
vector shapes with size, shape, texture, spectral, ge-
ographic context and geometry, though the use of
OBIA and related decision functions has resulted in
significant advances in the accuracy of GC/LU maps
(Rasouli et al., 2018b). In consequence, such advance
and suitable techniques would be an initial addition to
the country literature by mapping of Azerbaijan libe-
rated lands (Rasouli and Mammadov, 2020).
After the liberation of the occupied territories,
GC and LU change would be perhaps the most im-
portant concern in many rayons of EZER and
Karabakh as well. It is recognized such dramatic
changes can significantly impact regional climate,
ecosystem stability, water balance, stream silt up,
biodiversity, and socioeconomic practices, thereby
impinging on the overall quality of life in the com-
ing years. Consequently, the prediction of LU using
time serious data is very important for future man-
agement plans. In reviewing the background of the
subject, various methods of predicting changes in
GC/LU have been proposed, and it is regularly em-
ployed for a diverse suitability measure as a proxy of
human influence on land change processes. A CA-
MC model is one in which the future state of a sys-
tem can be predicted purely based on the proximate-
ly preceding state (Yagoub and Bizreh, 2014). Pre-
dicting future change is achieved by creating a tran-
sition probability matrix of GC/LU change from pe-
riod one to period two. Multispectral satellite images
and the CA-MC model were used by the researcher
to predict the GC/LU change in different regions. It
also computed states between different GC and
quantified the transition rate between different LU.
The factors in which driving forces of changes were
combined are selected to provide the estimation of
future scenarios (Mammadov and Rasouli, 2021).
Scientists are now able to apply a CA-MC to simu-
late future LU changes, quantitatively to predict the
dynamic changes of landscape patterns, with some
limitations in the prediction of the spatial pattern of
LU changes (Kamusoko, 2019). Intending to extract
accurate information from the resulting changes in
the research topic, the basic objectives of the current
study are: (a) applying rule-based and thresholding
indexing methods to understand the initial status of
GC during the last period of the occupation stage;
(b) creating more accurate methods of OBIA to ac-
cess more details of LU status and resulting changes;
(c) predicting future LU changes based on CA-MC
model inside the sampled rayons of the EZER.
These aims could be accompanied by accurate in-
formation that facilitates future rehabilitations plan-
ning (Rasouli et al., 2021a).
The Study Area
In this research, three rayons of Qubadli, Zangi-
lan and Jabrayil were arbitrary selected as the study
area that is entirely located inside EZER of the Re-
public of Azerbaijan. Administratively, this region is
one of the 14 economic regions of Azerbaijan, that
borders Armenia to the west, Iran to the south and con-
sists of two more Kalbajar and Lachin rayons (Fig. 1).
On July 7, 2021, the President of Azerbaijan Ilham
Aliyev signed Decree "on the new division of econo-
mic regions in the Republic of Azerbaijan (Decree of
the President of the Republic of Azerbaijan, 2021).
The study area has its weather conditions of
climate that would be affected by region complex
relief and enduring water bodies. In a territory along
the Aras River with semi-desert and dry steppes,
winter passes drily, and in higher elevated areas, the
climate is mildly humid. The precipitation (250 to
400 mm) in the study region varies from south to the
north and is mostly determined by the mountain
range; indicating a highly continental and mountain-
ous climate (Sylven et al., 2009). There are a few
summits (up to 2270 meters) in the study area and
this mountain range passes Aras's ravine lower in the
direction of south-east and makes flatlands. The
most recognized hydrographic network is the Aras
River that after forming the border between Azerbai-
jan, Iran and Armenia, flows into Azerbaijan in the
east of the country. Nearly about 100 km down-
stream of the border it flows into the Kura River,
which continues to flow southeast towards the Cas-
pian Sea. Forests spread in the mountainous parts of
the study area, mostly at the heights of 1800-2000
meters, gradually become lower and create subalpine
and alpine meadows.
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
30
Fig. 1. The geographic location of EZER in Azerbaijan, indicating the study area boundary
The occupation of the EZER rayons (and other
Azerbaijan territories) by Armenia caused significant
damage to the forests, especially those located in the
frontal lines and southern areas, with a very rich flora
of more than 4.500 species of higher plants
(Valigholizadeh and Karimi, 2016). During the occu-
pation period, agriculture became abounded and less
economically important for both countries, though the
overall production of agricultural products increased
in the last years of occupation periods (FAOSTAT,
2014). During the occupation stage, the fundamental
driving force of GC/LU changes has been related to
war activities such as the compulsory seizure of agri-
cultural and livestock activities, along with the de-
struction of rare forests (Nagorno-Karabakh con-
flict… IDMC, 2021). Moreover, we should consider
the uncontrolled erosion of soil, pasture and unique
historical monument buildings destruction caused by
unprincipled mining activities, multiple bombings,
and land-mine landings within the occupied territo-
ries. Eventually, the improper landuse and lack of
proper care of human and natural resources led to
widespread destruction and even extinction of many
plants and wildlife species in the region.
Data processed and techniques applied
a) Image collection
To acquire GC and LU maps with high-resolution
optical imagery, the European Commission and the
ESA jointly developed the Sentinel-2 satellite mission
under the Copernicus program (Berger et al., 2012;
Meer et al., 2014). The Sentinel-2 mission is composed
of two satellite constellations (Sentinel-2A and 2B)
running in a sun-synchronous orbit. The orbit height is
786 km, the inclination is 98.5 degrees, the revisit pe-
riod is 5 days, and the design life is 7 years (Drusch et
al., 2012). The Sentinel-2 satellite is a push-broom
multispectral imager (MSI) and the basic information
is shown in Table 1. It has 13 channels, the spectral
range is from 0.4 to 2.4 µm, and the spatial resolutions
are 10 m, 20 m, and 60 m (Caballero et al., 2019). The
Sentinel-2 data product is Level-1C (L1C) top-of-
atmosphere (TOA) reflectance data processed by geo-
metric correction. The download address for the data
source is https: //scihub.copernicus.eu/dhus/#/home.
Due to the balanced vegetation cover and the absence
of cloud cover in the summer months, the satellite ima-
ges are limited related to July were processed.
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
31
Table 1
Basic information of the Sentinel-2 satellite
Band
Number
Basic
Descriptions
Range (µm)
Central
Wavelength (µm)
Spatial
Resolution (m)
1
Coastal Aerosol
0.443
60
2
Blue
0.490
10
3
Green
0.560
10
4
Red
0.665
10
5
Vegetation Red Edge 1 (VRE1)
0.705
20
6
Vegetation Red Edge 2 (VRE2)
0.740
20
7
Vegetation Red Edge 3 (VRE3)
0.783
20
8
Near-Infrared (NIR)
0.842
10
8A
Vegetation Red Edge 4
0.865
20
9
Water Vapor
0.945
60
10
Shortwave Infrared Cirrus
1.375
60
11
Shortwave Infrared a
1.610
20
12
Shortwave Infrared b
2.190
20
Sentinel-2 MSI covering 13 spectral bands
(443-2190 nm), with a swath width of 290 km and a
spatial resolution of 10 m (four visible and NIR
bands), 20 m (six red edge and shortwave infrared
bands) and 60 m, three atmospheric correction bands
(Sentinel Online, 2018). In the study area, the Senti-
nel-2 images were used for investigating the changes
of GC/LU over 6 years from 2016 to 2021. To ob-
tain cloud-free imagery for the July months in the
region the best comparability time-series were
searched when the sky is usually clear. The selection
of study areas required the consideration that the
experimental areas are representative and, thus, the
vegetation and non-vegetation types in the study ar-
eas should be abundant and diverse. Water bodies
include Aras River and Khudafarin reservoir, and
non-vegetation types include bare-land and other
objects.
Fig. 2 shows two samples of Sentinel-2 images
RGB combinations for years 2016 and 2021, with a
highest spatial resolution (10*10 m). The reduction
of vegetation is completely detectable, particularly
the rapid decline of green-cover is visible across the
selected rayons (Sentinel-2 MSI, 2020). Even, the
soil salinization process can visually be observed in
the region marked with white, reddish, and orange
colores in the satellite imagery.
b) Image Processing
To achieve the main objectives of the current
research, several methods of image processing were
step-wisely applied to the satellite datasets. First,
based on the adapted GC/LU, Sentinel-2 images
were provided and in the data preprocessing step the
digital number (DN) values of each image were
converted to the top of atmosphere (TOA) reflec-
tance inside the ERDAS Imagine software (Rasouli
et al., 2020). The conversion of the raw pixel DN to
the TOA reflectance is a basic step for satellite im-
age preprocessing that represents only the brightness
characteristics of ground objects in the image pro-
cessing process (Chander et al., 2009). The details of
conversion methods are given in the relevant works
of literature (Lillesand, Kiefer, 2004).
Then each corrected multispectral satellite im-
agery was imported to the eCognition software set-
ting, and accordingly, mixed RGB layers were creat-
ed. To obtain functional information from the satel-
lite images, we propose a tool, called estimation of
scale parameter (ESP), that builds on the idea of lo-
cal variance (LV) of object heterogeneity within a
scene (Ikokou and Smit, 2013). When segmentation
settings were modified based on the ESP depend-
ing on the image quality (bands available, and image
spatial-resolution), pixels of any satellite image were
grouped into image objects before object-based clas-
sification can be performed. The degree of heteroge-
neity within an image object was controlled by a
subjective measure called the 'scale parameter', as
implemented in the mentioned software. Variation in
heterogeneity is explored by evaluating LV plotted
against the corresponding scale. The thresholds in
rates of change of LV (ROC-LV) indicate the scale
levels at which the image can be segmented most
appropriately, relative to the data properties at the
scene level. It should be also noted that in the seg-
mentation stage of satellite images, the default value
for scale parameter is taken about 45 and the shape
and compactness indexes were suggested as 0.3 and
0.7 respectively.
Along with implementing a few dynamic rule-
based indexing procedures, the following equations
were step-wisely applied to create the basic NDVI and
NDWI indexes: it will be considered as vegetation.
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
32
Fig. 2. Samples of Sentinel-2 imagery reflecting green cover destruction inside the EZER (2016 and 2021)
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
33
󰇛 󰇜
"
if Mean NDVI ≥ 0.30 and Brightness ≤ 38 and 45 [shape:0.3 compct.:0.7] it will be considered as vegetation. (1)
and 󰇛 󰇜
if Water <= 1 < unclassified <= 0 < unclassified on NDWI (2)
The NDVI is a standardized vegetation index
that allows us to generate an image showing the rela-
tive biomass. The chlorophyll absorption in Red
band and relatively high reflectance of vegetation in
the NIR band are using for calculating NDVI. In
turn, NDWI makes use of reflected NIR radiation
and visible green light to enhance the presence of
water bodies while eliminating the presence of soil
and terrestrial vegetation features. In the dynamic
thresholding stage, each index is regarded as an in-
dicator or measure of specified land surfaces, which
typically refers to a spectral measure of the change
in the satellite image band's reflections (Dozier,
1989). Inside the eCognition Developer, an index
layer calculation algorithm inserts a new image layer
by calculation of a spectral index as differentiating
between NDVI and NDWI indexes (Rasouli and
Mammadov, 2020).
Thereafter, a multi-resolution segmenta-
tion algorithm based on the OBIA algorithms
was applied to split out each image into unclassified
“object primitives” that form the basis for the image
objects and the rest of the image analysis (Baatz and
Schape, 2000). Segmentations, and the resulting
characteristics of object primitives and eventual im-
age objects, were set based on shape, size, color, and
pixel context parameters. The values of the parame-
ters define how much they influence spectral and
spatial characteristics of each image layer (Blaschke,
2010). At a simple glance, inside the eCognition
software, the segmentation process is based on pre-
defined parameters derived from real-world
knowledge of the features that researchers wish to
identify. In GC mapping, seemingly, this will re-
quire an understanding of image layer weights in the
rule-based indexing and segmentation stages. By
applying more advanced algorithms, in the classifi-
cation step, inside the Class Hierarchy, the basic
classes LU types such as water, forest, pasture, agri-
culture, barren lands, abounded lands and likely vul-
nerable areas could be assigned (eCognition Refe-
rence Book, 2019). Fig. 3 provides an overview of
the extracting of GC/LU based on image process
methods.
Before the final analysis of LU maps, for each
class, a sampling procedure was applied and a TTA
mask was created in the assessment of the classifica-
tion accuracy (Foody, 2002). Then, with sufficient
satisfaction from the results, all recognized LU clas-
ses were exported to the TerrSet IDRISI Selva soft-
ware for later analysis.
At the final stage, to understand the quantity,
location, and nature of future LU changes a Markov
analysis was applied inside the TerrSet software
(Clark Labs, 2017). For this job, all maps classified
by the OBIA method were pairwise compared to
detect patterns of changes between successful years.
This kind of predictive LU change modeling is ap-
propriate when the past trend of LU changing pat-
tern is known (Eastman, 2009). In practice, a CA-
MC is a stochastic process (based on probabilities)
with discrete state space and discrete or continuous
parameter space. In such a random process, the state
of a system "s" at the time (t+1) depends only on the
state of the system at a time "t" not on the previous
states.
The CA-MC model not only explains the quan-
tification of conversion states between the LU types
but can also reveal the transfer rate among them.
This technique is commonly used in the prediction
of geographical characteristics with no after-effect
event, which has now become an important predict-
ing method in geographic research (Rasouli et al.,
2018c). Based on the conditional probability formu-
la-Bayes, the prediction of LU changes is calculated
by equation 3: 󰇛󰇜󰇛󰇜󰇛󰇜 (3)
Where󰇛󰇜,󰇛󰇜 are the system status at the
time of (t) or (t + 1); 󰇛󰇜 is the transition probabil-
ity matrix in a state which is calculated as equation 4
(Benz et al., 2004).



 (4)
Where 󰇡
 󰇛
󰇜󰇢
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
34
Fig. 3. A simplified workflow in the extracting GC/LU maps
According to the matrix of the initial 󰇛󰇜 and
the transition probability of the nth stage󰇛󰇜, the
LU distribution in the Karabakh region in the future
can be calculated by using a computer simulation.
The CA-MC simulation model 󰇛󰇜 is as equation 5:
󰇛󰇜󰇛󰇜󰇛󰇜󰇛󰇜󰇛󰇜 (5)
Consequently, applying a CA model it made
possible to simulate a more accurate model where
space and time are discrete variables and interactions
assigned are local variables. In a CA model, the
transition of a cell from one landcover to another
depends on the state of the neighborhood cells. This
is based on the idea that a cell will have a higher
probability to change to land-cover class ‘a’ than to
a land-cover class ‘b’ if the cell is in closer proximi-
ty to land-cover class a’. Thus, the CA model not
only uses the information of the previous state of a
land-cover as done by a CA-MC model but also uses
the state of neighborhood cells for its transition rules
(Adhikari and Southworth, 2012). The CA-MC can
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
35
quantitatively predict the dynamic changes of land-
scape patterns, while it is not good at dealing with
the spatial pattern of landscape change (Mammadov
et al., 2018). On the other hand, Cellular Automata
(CA) can predict any transition among any number
of categories (Pontius and Malanson, 2005). Com-
bining the advantages of the CA model and the
space layout forecast of the CA-MC, a model with a
better LU change in both time and spatial dimension
was recognized to simulate maps for the year 2023.
To ensure that the model is reliable in predicting an
LU for a specific projected time-series, it was vali-
dated using existing datasets by comparing maps,
using Kappa statistics nearly about 94 present (Ra-
souli et al. 2020).
Results Obtained
a) Comparison of the GC Indices
To compare the changes of the water surface,
vegetation, and non-vegetation covers inside the
study area, the mean values of NDWI and NDVI are
shown in Fig. 4. Examination of GC during the last
6 years shows the fact that significant negative
changes have been made in the study area. Although
the decrease in vegetation can be traced in all three
rayons of the EZER region, the rate of decrease in
Jabrail rayon is very higher than others.
To illustrate the changes in GC changes from
2016 to 2021 in the study area, Fig. 5 is given.
Linear equations indicate that there is a conside-
rable decrease in vegetation cover and a contrasting
meaningful increase in non-vegetated classes. R2 values
of more than 0.92 percent approved a significant dec-
rease in vegetated areas and an increase in the non-
vegetated covers inside the study area from 2016 to 2021.
b) Development of LU Maps
By introducing OBIA segmentation and classi-
fication procedures, LU maps, with much more de-
tails were produced for the study area. Careful ex-
amination of the resulting maps indicated that the
reduction of green cover areas, particularly in forest
and pasture classes, all around the sampled rayons is
quite considerable. As it is noticeable in exampled
maps of 2018 and 2021, changes are evident, and
significant changes can be visually seen in the
southern rayons of EZER by emerging more barren
lands and changes in agricultural activities (Fig. 6).
To ensure that the classification result is reliable,
visual interpretation and a quantitative accuracy index
are used to conduct accuracy assessments. All maps
were validated using existing datasets by comparing
maps, using Kappa statistics nearly about 94 present.
At the final stage, applying the CA-MC model,
an LU map for 2023 was produced inside TerrSet
software that makes it possible to apply a variety of
tools available and predict a set of reliable LU pre-
dictions for the future time series. An example of a
prediction LU map for the sampled areas is shown in
Fig. 7. As you can notice, changes in the LU classes
are quite evident, by emerging more barren and
abandoned lands and considerable reduction in for-
est and pasture classes inside the sampled rayons.
Validated accuracy results confirm reliable pre-
diction LU map was very high, with an overall accu-
racy of 0.95 and Kappa statistics nearly 94% for the
2023 prediction model. For comparison, LU values
from 2016 to 2021 are illustrated in Fig. 8, along
with the results of the CA-MC model of 2023. Three
linear equations approve the vegetation covers
(mostly forest and pasture, with value 0.97) are
being continuously reducing and almost the same
proportion is added to the barren and abandoned
lands with value of 0.81 and 0.74 respectively.
The values in parentheses (in %) indicate a de-
crease (-) or increase (+) concerning any of LU clas-
ses. Accordingly, the linear regression (R²) equations
approve significant changes in the Barren-Lands
(with α = 0.05 at 95%) and Forest (α = 0.01 at 99%)
levels and comparably less significant in Abandoned-
Lands class with α = 0.2 at 80%. Regarding the criti-
cal values and sample sizes, no significant variations
were statistically detected in other LU classes.
Discussion
Our main goal was to assess GC and LU chang-
es inside the EZER, with special care in the southern
rayons of Qubadli, Zangilan, and Jabrayil that se-
verely were affected by the military occupation of
Armenia over the past 30 years. Primary, to under-
stand the extent of changes in basic GC types, we
processed multi-temporal Sentinel-2 imagery and
introduced two rule-based indexing techniques. It
was found that inside the study area, particularly
Zangilan and Jabrayil rayons, reduction of GC is
accelerated during the recent years at least, and is
continuously being destroyed. Hence, dynamic
thresholding indexing methods could be effective
methods by enclosing the spectral, spatial, mul-
titemporal, and multisensor information; and ancil-
lary data into indexing procedures, including NDVI
and NDWI indexes (Mammadov and Rasouli, 2021).
Furthermore, the OBIA segmentation and classifica-
tion results indicated that the vegetation trend (par-
ticularly forests and pastures) is negative in the de-
creasing mode and the LU changes are signs all over
the EZER rayons. Nevertheless, accuracy assess-
ment is an integral part of any indexing procedure,
and uncertainty and error propagation in the classic
image-processing chain is still an important factor
influencing the final GC map's accuracy. Therefore,
identifying the weakest links in the chain and then
reducing the uncertainties is critical for the im-
provement of digital image processing accuracy for
many investigators (Lobo and Chick, 1996).
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
36
Fig. 4. Classified GC maps produced by indexing techniques adjusted for the study area
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
37
Fig. 5. Changes in the GC values for sampled years inside the southern rayons of EZER
Additionally, it is believed that OBIA methods
could be regarded as a sub-discipline of geo-
information science devoted to partitioning remote
sensing imagery into meaningful image objects, and
assessing their characteristics through spatial, spec-
tral, and temporal scales (Blaschke, 2010). Although
the OBIA is found to be a very advanced image pro-
cessing procedure, nevertheless we think that it
could be compared in future research along with
quite professional programs as SVM, and Machine-
Learning / Deep-Learning approaches to detect high-
ly LU changes with high confidence (Langat et al.,
2019). In this way, by processing high-resolution
imagery we may carry out much more detail by ex-
panding the study site to other liberated rayons with
an emphasis on the Karabakh region to visualize the
most likely reduc-tion in forest covers, even evaluat-
ing the landuse changes in the liberated from the
enemy urban areas.
In addition, final prediction maps created by a
CA-MC model confirmed that the reduction in LU
will be continued according to certain probabilistic
rules (Roy et al., 2015). Such models can be cited as
suitable tools for modeling landuse change, especial-
ly in geographical areas that are not accessible by
land operations (Keshtkar and Voigt, 2016). The
overall accuracy values of 0.95 found for the Kappa
index (with 94%) indicated a high degree of simi-
larity between the actual maps and the maps result-
ing from simulations, which supports the accuracy
of the model and the possibility of dependence in
predicting the direction of the green covers reduction
for future years, as the result showed that the forests
would be diminished to the northern and northwest-
ern parts in 2023 toward the other rayons. The past
negative trends in LU may likely continue in that
direction as the Probability Matrix of the Markov
model shows. During future investigation plans, we
could consider several mixed OBIA, Fuzzy, and
Conventional Neural Network (CNN), Deep-
Learning methods, as much as advanced artificial
intelligence approaches (Song et al., 2012; Liu et al.,
2019). It is also possible to refer to the multi-
functional solutions: such as the region 3D elevation
height, morphologic features (aspect, slope parame-
ters) integrated with the OBIA-based analysis (Kato,
2020).
24 23 21 22 19 17
666
593 596 557 535
460
1644
1717 1713 1753 1776
1850
y = -35.514x + 692.13
R² = 0.924
y = 35.629x + 1617.5
R² = 0.925
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2016 2017 2018 2019 2020 2021
Area, km2
Year
Water Vegetated Non-Vegetated
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
38
Fig. 6. Sentinel-2 LU maps of 2018 and 2021 classified by OBIA techniques
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
39
Fig. 7. An example of CA-MC predicated map for 2023 inside the sampled rayons
Our basic suggestions are establishing a real-time
satellite monitoring system, connecting to the ground-
based observation stations to improve the end-user
GC/LU productions and plans. We strongly believe
that these suggestions are the logical way of accessing
the up-to-date observation, which could be run for the
country's current and future needs (Duarte et al.,
2018). There is a need to have long-term GC and
LU planning to highlight the role of advanced image
processing in obtaining much more accurate and us-
able information from the land charters and behav-
iors. It will let scientists think about designing intel-
ligent-based rural and town infrastructures; based on
smart economic needs and sustainable development
strategies (Rasouli et al., 2021b). As it is evident, the
long-term occupation of the EZER and Karabakh
rayons of course inside and around the front-line
there are uncountable destructions in the region that
caused by Armenian forces with significant damages
to the forests, especially those were targeted by the
illegal logging processes and deliberated bushfires
(General Assembly Security Council, 2009). Wheth-
er you are monitoring, mapping green cover or eco-
logical damages, combating loss of natural re-
sources, or just helping countries meet their sustain-
able development goals due to the climate change
impacts, chances are high that we will need accurate
and spatially detailed information on GC and LU pro-
cedures. Undoubtedly, shortly, environmental chan-
ges in the regional scale and even Azerbaijan extent
with of natural or human origin will continue and
the resulting scenarios may be more catastrophic.
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
40
Fig. 8. Changes in the LU types during the last six years inside the sampled rayons.
As a result, based on undeniable research expecta-
tions, the Earth observation satellites, like those
from Landsats, Copernicus, Azercosmos high-
resolution imagery, and definitely through advanced
image processing procedures are key to providing
highly accurate information in dealing with envi-
ronmental issues in the country.
Concluding Remarks
There is no doubt that the long-term occupation
of EZER rayons has seriously affected the region of
both GC and LU types. During the last 30 years,
Azerbaijan infrastructures, the environment, and
ecosystems have also been seriously affected in
many ways. The current opening research results
confirmed that both rule-based indexing and ad-
vanced OBIA methods could be very effective ways
of detection in the trustable GC and LU maps. To
conclude:
Modern image processing has many better ways
to reach practical knowledge of geospatial mat-
ters and regional environmental behaviors. In-
tegration of OBIA, Fuzzy, and ML/DL methods
could be seen as a particular vision in the image
processing procedures.
Most of the negative changes are detected in the
southern parts of the EZER, including forests
(-4.7%) and pastures (-4.6%); subsequently a
meaningful progressive increase in barren lands
(+4.8 %), and the emergence of abandoned
lands (+5%) in LU types. All rayons of
Karabakh may have the greatest damages; par-
ticularly those that are imposed on the green
cover of the region. They require careful and
immediate multipurpose geo-environmental in-
vestigations.
Azerbaijan particularly the EZER and
Karabakh geo-environments must be watch-
fully monitored by employing advanced remote
sensing technology. This will lead to the con-
struction of relevant multi-purpose geo-spatial
LU databases that are essential actions in the
current and future land-reclamation plans.
The CA-MC model has proven to be flexible
and applicable, when it using available data, al-
so it has proven to be efficient in dealing with
yf o= -20.605x + 306.58
R² = 0.97
yba= 24.679x + 975.29
R² = 0.81
yab = 19.881x + 367.23
R² = 0.74
0
200
400
600
800
1000
1200
1400
July-
12/2016
July-
28/2017
July-
03/2018
July-
03/2019
July-
07/2020
July-
12/2021
CA-MC-July
2023
A
r
e
a
K
m
2
Landuse Classes
Water (-0.2) Forest (-4.7) Pasture (-4.6)
Agriculture (+0.5) Barren Lands (+4.8) Abandoned Lands (+5)
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
41
LU through monitoring, forecasting, and reduc-
tion trend if the lands are left as before.
The methods applied in the present study can be
adjusted to other liberated areas and along the
front lines over a longer time.
A multi-platform and multi-purpose real-time
monitoring system is almost immediately re-
quired if it is not yet too late. Such reclamation
projects could be managed by Azerbaijani staff
to do this critical nationwide advanced interest.
REFERENCES
Adhikari S., Southworth J. Simulating forest cover changes of
Bannerghatta National Park based on a CA-Markov model:
a remote sensing approach. Remote Sens., Vol. 4, No. 10,
2012, pp. 3215-3243.
Baatz M., Schape A. Multi-resolution segmentation an optimi-
zation approach for high quality multi-scale segmentation. In:
(Strobl J. et al. eds.) Angewandte Geographische Informations
verarbeitung XII, Beiträgezum AGIT Symposium Salsburg
2000, Karlruhe, Herbert Wichmann Verlag, pp. 12-23.
Bahadur K. Improving Landsat and IRS image classification:
evaluation of unsupervised and supervised classification
through band ratios and DEM in a mountainous landscape in
Nepal. Remote Sensing, Vol. 1, No. 4, 2009, pp. 1257-1272.
Baumann M., Radeloff V., Avedian V., Kuemmerle T. Land-use
change in the Caucasus during and after the Nagorno-Karabakh
conflict. Reg. Environ. Change, Vol. 15, 2015, pp. 1703-1716.
Benz U., Hofmann P., Willhauck G., Lingenfelder I., Heynen
M. Multi-resolution, object-oriented fuzzy analysis of re-
mote sensing data for GIS-ready information. ISPRS Journal
of Photogrammetry and Remote Sensing, Vol. 58, No. 3,
2004, pp. 239-258.
Berger M., Moreno J., Johannessen J.A., Levelt P.F., Hanssen
R.F. ESA’s Sentinel missions in support of Earth system
science. Remote Sens. Environ., Vol. 120, 2012, pp. 84-90.
Blaschke T. Object-based image analysis for remote sensing.
ISPRS International Journal of Photogrammetry and Remote
Sensing, Vol. 65, No. 1, 2010, pp. 2-16.
Caballero I., Ruiz J., Navarro G. Sentinel-2 satellites provide
near-real time evaluation of catastrophic floods in the West
Mediterranean. Water, Vol. 11, No. 12, 2019, p. 2499.
Chander G., Markham B.L., Helder D.L. Summary of current
radiometric calibration coefficients for Landsat MSS, TM,
ETM+, and EO-1, ALI sensors. Remote sensing of envi-
ronment, Vol. 113, No. 5, 2009, pp. 893-903.
Clark Labs, Clark University. “TerrSet Geospatial Monitoring
and Modeling System”. Clark University, USA, Accessed
21 February 2017, https://clarklabs.org/wp-content/uploads/
2016/03/TerrSet18-2_Brochure_WEB.pdf.
Copernicus, Sentinel-2, Satellite Missions eoPortal Directory.
Directory.eoportal.org. Retrieved 5 March 2020,
https://eoportal.org/web/eoportal/satellite-missions/c-
missions/copernicus-sentinel-2-2020.
Decree of the President of the Republic of Azerbaijan of July 7,
2021 No. 1386 "On the new division of economic regions in the
Republic of Azerbaijan", 2021, https://president.az/articles/
52389 (in Azerbaijani).
Dozier J. Spectral signature of Alpine snow cover from LAND-
SAT Thematic Mapper. Remote Sensing of Environ-
ment, Vol. 45, 1989, pp. 9-22.
Drusch M., Del Bello U., Carlier S., Colin O., Fernandez V., Gascon
F., Hoersch B., Isola C., Laberinti P., Martimort P. Sentinel-2:
ESA’s optical high-resolution mission for GMES operational ser-
vices. Remote Sens. Environ., Vol. 120, 2012, pp. 25-36.
Duarte D., Nex F., Kerle N. Vosselman G. Satellite image clas-
sification of building damages using airborne and satellite
image samples in a deep learning approach. ISPRS Annals
of Photogrammetry, Remote Sensing & Spatial Information
Sciences, Vol. 4-2, 2018, pp. 89-96.
ЛИТЕРАТУРА
Adhikari S., Southworth J. Simulating forest cover changes of
Bannerghatta National Park based on a CA-Markov model:
a remote sensing approach. Remote Sens., Vol. 4, No. 10,
2012, pp. 3215-3243.
Baatz M., Schape A. Multi-resolution segmentation an optimi-
zation approach for high quality multi-scale segmentation. In:
(Strobl J. et al. eds.) Angewandte Geographische Informations
verarbeitung XII, Beiträgezum AGIT Symposium Salsburg
2000, Karlruhe, Herbert Wichmann Verlag, 2000, pp. 12-23.
Bahadur K. Improving Landsat and IRS image classification:
evaluation of unsupervised and supervised classification
through band ratios and DEM in a mountainous landscape in
Nepal. Remote Sensing, Vol. 1, No. 4, 2009, pp. 1257-1272.
Baumann M., Radeloff V., Avedian V., Kuemmerle T. Land-use
change in the Caucasus during and after the Nagorno-Karabakh
conflict. Reg. Environ. Change, Vol. 15, 2015, pp. 1703-1716.
Benz U., Hofmann P., Willhauck G., Lingenfelder I., Heynen M.
Multi-resolution, object-oriented fuzzy analysis of remote sensing
data for GIS-ready information. ISPRS Journal of Photogram-
metry and Remote Sensing, Vol. 58, No. 3, 2004, pp. 239-258.
Berger M., Moreno J., Johannessen J.A., Levelt P.F., Hanssen
R.F. ESA’s Sentinel missions in support of Earth system
science. Remote Sens. Environ., Vol. 120, 2012, pp. 84-90.
Blaschke T. Object-based image analysis for remote sensing.
ISPRS International Journal of Photogrammetry and Remote
Sensing, Vol. 65, No. 1, 2010, pp. 2-16.
Caballero I., Ruiz J., Navarro G. Sentinel-2 satellites provide
near-real time evaluation of catastrophic floods in the West
Mediterranean. Water, Vol. 11, No. 12, 2019, p. 2499.
Chander G., Markham B.L., Helder D.L. Summary of current
radiometric calibration coefficients for Landsat MSS, TM,
ETM+, and EO-1, ALI sensors. Remote sensing of envi-
ronment, Vol. 113, No. 5, 2009, pp. 893-903.
Clark Labs, Clark University. “TerrSet Geospatial Monitoring
and Modeling System”. Clark University, USA, Accessed
21 February 2017, https://clarklabs.org/wp-content/uploads/
2016/03/TerrSet18-2_Brochure_WEB.pdf.
Copernicus, Sentinel-2, Satellite Missions eoPortal Directory.
Directory.eoportal.org. Retrieved 5 March 2020,
https://eoportal.org/web/eoportal/satellite-missions/c-
missions/copernicus-sentinel-2-2020.
Dozier J. Spectral signature of Alpine snow cover from LAND-
SAT Thematic Mapper. Remote Sensing of Environ-
ment, Vol. 45, 1989, pp. 9-22.
Drusch M., Del Bello U., Carlier S., Colin O., Fernandez V., Gascon
F., Hoersch B., Isola C., Laberinti P., Martimort P. Sentinel-2:
ESA’s optical high-resolution mission for GMES operational ser-
vices. Remote Sens. Environ., Vol. 120, 2012, pp. 25-36.
Duarte D., Nex F., Kerle N. Vosselman G. Satellite image clas-
sification of building damages using airborne and satellite
image samples in a deep learning approach. ISPRS Annals
of Photogrammetry, Remote Sensing & Spatial Information
Sciences, Vol. 4-2, 2018, pp. 89-96.
Eastman J.R. IDRISI taiga guide to GIS and image processing.
Clark Labs for Cartographic Technology and Geographic Anal-
ysis, Clark University, Worcester, MA, USA, 2009, 141 p.
eCognition Reference Book. eCognition Developer for Win-
dows operating system, Version 9.5.1, Trimble, 2019,
https://ecognition.blog/whats-new-in-ecognition-9-5-1/.
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
42
Eastman J.R. IDRISI taiga guide to GIS and image processing.
Clark Labs for Cartographic Technology and Geographic Anal-
ysis, Clark University, Worcester, MA, USA, 2009, 141 p.
eCognition Reference Book. eCognition Developer for Win-
dows operating system, Version 9.5.1, Trimble, 2019,
https://ecognition.blog/whats-new-in-ecognition-9-5-1/.
FAOSTAT. Statistical database of the Food and Agricultural
Organization. 2014, Available at: http://faostat.fao.org/.
Foody G.M. Status of land cover classification accuracy assess-
ment. Remote Sensing of Environment, Vol. 80, No.
1, 2002, pp. 185-201.
Franklin S.E., Wulder M.A. Remote sensing methods in medi-
um spatial resolution satellite data land cover classification
of large areas. Progress in Physical Geography, Vol. 26, No.
2, 2002, pp.173-205.
General Assembly Security Council. Letter dated 30 September
2009 from the Permanent Representative of Azerbaijan to
the United Nations addressed to the Secretary-General, Pro-
tracted conflicts in the GUAM area and their implications
for international peace, security and development. The situa-
tion in the occupied territories of Azerbaijan. Security
Council Sixty-fourth session, 2009.
Giovarelli R., Bledsoe D. Land reform in Eastern Europe: West-
ern CIS, Transcaucuses, Balkans, and EU Accession Coun-
tries. Prepared for FAO (Food and Agriculture Organization)
by the Rural Development Institute (RDI), Seattle, Washing-
ton, 2001, http://www.fao.org/3/AD878E/AD878E00.htm.
Hasanov Z.M., Ibrahimov Z.A., Nabiyev V.R. Beech forests of
Azerbaijan: The modern condition, age structure and regen-
eration. Annals of Agrarian Science, Vol. 15, No. 4, 2017,
pp. 453-457.
Hay G.J., Castilla G. Object-based image analysis, strengths,
weaknesses, opportunities, and threats (SWOTs). From
OBIA 2006 International Archives of Photogrammetry, Re-
mote sensing, and Spatial Information Science, 2006.
Hay G.J., Castilla G. Geographic object-based image analysis
(GEOBIA): a new name for a new discipline. In: (Blaschke
T., Lang S., Hay G.J. eds.) Object-based image analysis:
spatial concepts for knowledge-driven remote sensing appli-
cations. Springer. Berlin, Germany, 2008, pp. 75-89.
Hussain M., Chen D., Cheng A., Wel H., Stanley D. Change
detection from remotely sensed images: from pixel-based to
object-based approaches. ISPRS J. Photogramm. Remote
Sens., Vol. 80, 2013, pp. 91-106.
Ikokou G., Smit J. A technique for optimal selection of segmen-
tation scale parameters for object-oriented classification of
urban scenes. South Afr. J. of Geomatics, Vol. 2, No. 4,
2013, pp. 358-369.
Kamusoko C. Remote sensing image classification in R.
(Springer Geography) 1st ed. Springer . 2019, 237 p.
Kato L. Integrating open-street map data: in object-based land-
cover and land-use classification for disaster recovery. LAP
LAMBERT Academic Publishing. 2020, 76 p.
Keshtkar H., Voigt W. A spatiotemporal analysis of landscape
change using an integrated Markov chain and cellular automata
models. Model. Earth Syst. Environ., Vol. 2, No. 10, 2016.
Khandelwal P., Singh K., Mehrotra A. Unsupervised change
detection from satellite images using KCN. LAP LAM-
BERT Academic Publishing. 2014, 100 p.
Khatami R., Mountrakis G., Stehman S.V. A meta-analysis of
remote sensing research on supervised pixel-based land cov-
er image classification processes: general guidelines for
practitioners and future research. Remote Sens. Environ.,
Vol. 177, 2016, pp. 89-100.
Langat P.K., Kumar L., Koech R., Ghosh M.K. Monitoring of
land use/land-cover dynamics using remote sensing: A case
of Tana River Basin, Kenya. Geocarto Intern., Vol. 36, No.
13, 2019, pp. 1470-1488.
FAOSTAT. Statistical database of the Food and Agricultural
Organization. 2014, Available at: http://faostat.fao.org/.
Foody G.M. Status of land cover classification accuracy assess-
ment. Remote Sensing of Environment, Vol. 80, No.
1, 2002, pp. 185-201.
Franklin S.E., Wulder M.A. Remote sensing methods in medi-
um spatial resolution satellite data land cover classification
of large areas. Progress in Physical Geography, Vol. 26, No.
2, 2002, pp.173-205.
General Assembly Security Council. Letter dated 30 September
2009 from the Permanent Representative of Azerbaijan to
the United Nations addressed to the Secretary-General, Pro-
tracted conflicts in the GUAM area and their implications
for international peace, security and development. The situa-
tion in the occupied territories of Azerbaijan. Security
Council Sixty-fourth session, 2009.
Giovarelli R., Bledsoe D. Land reform in Eastern Europe: West-
ern CIS, Transcaucuses, Balkans, and EU Accession Coun-
tries. Prepared for FAO (Food and Agriculture Organization)
by the Rural Development Institute (RDI), Seattle, Washing-
ton, 2001, http://www.fao.org/3/AD878E/AD878E00.htm.
Hasanov Z.M., Ibrahimov Z.A., Nabiyev V.R. Beech forests of
Azerbaijan: The modern condition, age structure and regen-
eration. Annals of Agrarian Science, Vol. 15, No. 4, 2017,
pp. 453-457.
Hay G.J., Castilla G. Object-based image analysis, strengths,
weaknesses, opportunities, and threats (SWOTs). From
OBIA 2006 International Archives of Photogrammetry, Re-
mote sensing, and Spatial Information Science, 2006.
Hay G.J., Castilla G. Geographic object-based image analysis
(GEOBIA): a new name for a new discipline. In: (Blaschke
T., Lang S., Hay G.J. eds.) Object-based image analysis:
spatial concepts for knowledge-driven remote sensing appli-
cations. Springer. Berlin, Germany, 2008, pp. 75-89.
Hussain M., Chen D., Cheng A., Wel H., Stanley D. Change de-
tection from remotely sensed images: from pixel-based to
object-based approaches. ISPRS J. Photogramm. Remote
Sens., Vol. 80, 2013, pp. 91-106.
Ikokou G., Smit J. A technique for optimal selection of segmen-
tation scale parameters for object-oriented classification of
urban scenes. South Afr. J. of Geomatics, Vol. 2, No. 4,
2013, pp. 358-369.
Kamusoko C. Remote sensing image classification in R.
(Springer Geography) 1st ed. Springer . 2019, 237 p.
Kato L. Integrating open-street map data: in object-based land-
cover and land-use classification for disaster recovery. LAP
LAMBERT Academic Publishing. 2020, 76 p.
Keshtkar H., Voigt W. A spatiotemporal analysis of landscape
change using an integrated Markov chain and cellular au-
tomata models. Model. Earth Syst. Environ., Vol. 2, No. 10,
2016.
Khandelwal P., Singh K., Mehrotra A. Unsupervised change de-
tection from satellite images using KCN. LAP LAMBERT
Academic Publishing. 2014, 100 p.
Khatami R., Mountrakis G., Stehman S.V. A meta-analysis of
remote sensing research on supervised pixel-based land cov-
er image classification processes: general guidelines for
practitioners and future research. Remote Sens. Environ.,
Vol. 177, 2016, pp. 89-100.
Langat P.K., Kumar L., Koech R., Ghosh M.K. Monitoring of
land use/land-cover dynamics using remote sensing: A case
of Tana River Basin, Kenya. Geocarto Intern., Vol. 36, No.
13, 2019, pp. 1470-1488.
Lillesand T., Kiefer R. Chipman J. Remote sensing and image
interpretation. 5th ed. John Wiley and Sons Inc. New York,
2004, 763 p.
Liu S. Qi Z. Li X. Gar-On A. Integration of convolutional neural
networks and object-based post-classification refinement for
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
43
Lillesand T., Kiefer R. Chipman J. Remote sensing and image
interpretation. 5th ed. John Wiley and Sons Inc. New York,
2004, 763 p.
Liu S. Qi Z. Li X. Gar-On A. Integration of convolutional neural
networks and object-based post-classification refinement for
land use and land cover mapping with optical and SAR data.
Journal of Remote Sensing, Vol. 11, No. 6, 2019, pp. 690;
https://doi.org/10.3390/rs11060690.
Lobo A., Chick O.A., Casterad A. Classification of Mediterranean
crops with multisensor data: Per-pixel versus per-object statis-
tics and image segmentation. International Journal of Remote
Sensing, Vol. 17, No.12, 1996, pp. 2385-2400.
Mammadov R., Rasouli A.A. Practical satellite image pro-
cessing inside the ERDAS imagine. Provided for the Fast
Training Program. Publisher “The Institute of Geography of
Azerbaijan National Academy of Sciences”. Baku, Azerbai-
jan, 2021, 350 p.
Mammadov R., Rasouli A.A., Mobasher H. Applying an object-
based classification approach through a cellular automata-
Markov method in land-cover/landuse change detection proce-
dure, case of the Urmia Lake. Proceeding of the Eurasian GIS
2018 Congress 04-07 September 2018 Baku, Azerbaijan.
Published by Nobel Akademik Yayıncılık Eğitim Danışmanlık
Tic. Ltd. Şti Baku/Azerbaijan. 2018, pp. 272-284.
Meer F.D., Werff H.M.A., Ruitenbeek F.J.A. Potential of ESA’s
Sentinel-2 for geological applications. Remote Sens. Envi-
ron., Vol. 148, 2014, pp.124-133.
Nagorno-Karabakh conflict: finding common ground in respect
of the dead. International Displacement Monitoring Centre
(IDMC). News and Press Release ICRC, 2021, 16 April.
Nelson S., Khorram S. Image processing and data analysis with
ERDAS IMAGINE. 1st ed. CRC Press. 2018, 350 p.
Pettorelli N. The normalized difference vegetation index. 1st
Edition. Kindle Edition. OUP Oxford. 2013, 224 p.
Platt R.V., Rapoza L. An evaluation of an object-oriented para-
digm for land use/land cover classification. The Professional
Geographer, Vol. 60, No.1, 2008, pp. 87-100.
Pontius R.G., Malanson J. Comparison of the structure and accuracy
of two land change models. International Journal of Geograph-
ical Information Science, Vol.19, No. 2, 2005, pp. 745-748.
Rasouli A.A., Mammadov G.Sh., Asgarova M.M. Mastering
spatial data analysis inside the GIS setting. Elm. Baku,
2021a, 396 p.
Rasouli A.A., Mammadov R., Asgarova M.M. Application of
satellite image processing methods in mapping of landcov-
er/landuse changes inside the Karabakh liberated territo-
ries. Materials of the On-line Conference of the Azerbaijan
National Academy of sciences Biodiversity of Karabakh,
land and water resources: past, present and future”, 2021b,
p.132.
Rasouli A., Mammadov R. Preliminary satellite image analysis
processing the Landsat imagery inside the ArcGIS setting.
Lambert Academy Publishing. Germany, 2020, 264 p.
Rasouli A.A., Mammadov R., Aliyev V. Detection of Caspian Sea
coastline changes by Fuzzy-based object-oriented image analy-
sis. Abbstracts of The Second Eurasian RISK-2020 Conference
and Symposium, 12-19 April, 2020 Tbilisi / GEORGIA,
2020, https://books.aijr.org/index.php/press/catalog/book/93.
Rasouli A.A., Mammadov R., Mobasher H. Object-based water
bodies extraction method by processing of Sentinel satellite im-
agery case study: Baku City, Azerbaijan. Proceeding of Eura-
sian GIS 2018 Congress 04-07 September 2018 Baku, Azer-
baijan. Published by Nobel Akademik Yayıncılık Eğitim
Danışmanlık Tic. Ltd. Şti Baku/Azerbaijan. 2018a, pp. 284-290.
Rasouli A.A., Mammadov R., Pishnamaz M., Hushmand A., Sa-
farov E. Assessment of forest cover changes by applying ob-
ject-oriented procedures inside the Qarabağ Occupied Region.
Eurasian GIS 2018 Congress 04-07 September 2018 Baku,
land use and land cover mapping with optical and SAR data.
Journal of Remote Sensing, Vol. 11, No. 6, 2019, pp. 690,
https://doi.org/10.3390/rs11060690.
Lobo A., Chick O.A., Casterad A. Classification of Mediterra-
nean crops with multisensor data: Per-pixel versus per-
object statistics and image segmentation. International Jour-
nal of Remote Sensing, Vol. 17, No.12, 1996, pp. 2385-
2400.
Mammadov R., Rasouli A.A. Practical satellite image processing
inside the ERDAS imagine. Provided for the Fast Training Pro-
gram. Publisher: “The Institute of Geography of Azerbaijan Na-
tional Academy of Sciences”. Baku, Azerbaijan, 2021, 350 p.
Mammadov R., Rasouli A.A., Mobasher H. Applying an object-
based classification approach through a cellular automata-
Markov method in land-cover/landuse change detection proce-
dure, case of the Urmia Lake. Proceeding of the Eurasian GIS
2018 Congress 04-07 September 2018 Baku, Azerbaijan.
Published by Nobel Akademik Yayıncılık Eğitim Danışmanlık
Tic. Ltd. Şti Baku/Azerbaijan. 2018, pp. 272-284.
Meer F.D., Werff H.M.A., Ruitenbeek F.J.A. Potential of ESA’s
Sentinel-2 for geological applications. Remote Sens. Envi-
ron., Vol. 148, 2014, pp.124-133.
Nagorno-Karabakh conflict: finding common ground in respect
of the dead. International Displacement Monitoring Centre
(IDMC). News and Press Release ICRC, 2021, 16 April.
Nelson S., Khorram S. Image processing and data analysis with
ERDAS IMAGINE. 1st ed. CRC Press. 2018, 350 p.
Pettorelli N. The normalized difference vegetation index. 1st
Edition, Kindle Edition. OUP Oxford. 2013, 224 p.
Platt R.V., Rapoza L. An evaluation of an object-oriented para-
digm for land use/land cover classification. The Professional
Geographer, Vol. 60, No.1, 2008, pp. 87-100.
Pontius R.G., Malanson J. Comparison of the structure and accuracy
of two land change models. International Journal of Geograph-
ical Information Science, Vol.19, No. 2, 2005, pp. 745-748.
Rasouli A.A., Mammadov G.Sh., Asgarova M.M. Mastering
spatial data analysis inside the GIS setting. Elm. Baku,
2021a, 396 p.
Rasouli A.A., Mammadov R., Asgarova M.M. Application of sat-
ellite image processing methods in mapping of landcov-
er/landuse changes inside the Karabakh liberated territories.
Materials of the On-line Conference of the Azerbaijan Na-
tional Academy of sciences Biodiversity of Karabakh, land
and water resources: past, present and future”, 2021b, p.132.
Rasouli A., Mammadov R. Preliminary satellite image analysis
processing the Landsat imagery inside the ArcGIS setting.
Lambert Academy Publishing. Germany, 2020, 264 p.
Rasouli A.A., Mammadov R., Aliyev V. Detection of Caspian Sea
coastline changes by Fuzzy-based object-oriented image analy-
sis. Abbstracts of The Second Eurasian RISK-2020 Conference
and Symposium, 12-19 April, 2020 Tbilisi / GEORGIA,
2020, https://books.aijr.org/index.php/press/catalog/book/93.
Rasouli A.A., Mammadov R., Mobasher H. Object-based wa-
ter bodies extraction method by processing of Sentinel sat-
ellite imagery case study: Baku City, Azerbaijan. Proceed-
ing of Eurasian GIS 2018 Congress 04-07 September 2018
Baku, Azerbaijan. Published by Nobel Akademik Yayın-
cılık Eğitim Danışmanlık Tic. Ltd. Şti Baku/Azerbaijan.
2018a, pp. 284-290.
Rasouli A.A., Mammadov R., Pishnamaz M., Hushmand A., Sa-
farov E. Assessment of forest cover changes by applying ob-
ject-oriented procedures inside the Qarabağ Occupied Region.
Eurasian GIS 2018 Congress 04-07 September 2018 Baku,
Azerbaijan. Published by Nobel Akademik Yayıncılık Eğitim
Danışmanlık Tic. Ltd. Şti Baku/Azerbaijan. 2018b, p. 74.
Rasouli A.A., Mammadov R., Safarov E., Mohammadzadeh K.
Fuzzy object-based landcover/use mapping of The Karabagh
region by processing of Sentinel satellite imageries. Eura-
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
44
Azerbaijan. Published by Nobel Akademik Yayıncılık Eğitim
Danışmanlık Tic. Ltd. Şti Baku/Azerbaijan. 2018b, p. 74.
Rasouli A.A., Mammadov R., Safarov E., Mohammadzadeh K.
Fuzzy object-based landcover/use mapping of The Karabagh
region by processing of Sentinel satellite imageries. Eura-
sian GIS 2018 Congress 04-07 September 2018 Baku,
Azerbaijan. Published by Nobel Akademik Yayıncılık
Eğitim Danışmanlık Tic. Ltd. Şti Baku/Azerbaijan. 2018c,
p. 73.
Rehman S., Hussain M. Fuzzy C-means algorithm based satellite
image segmentation. Indonesian Journal of electrical engineer-
ing and computer science, Vol. 9, No. 2, 2018, pp. 332-334.
Report: Investigating the environmental dimensions of the 2020
Nagorno-Karabakh conflict. Conflict and Environment Ob-
servatory, Published: February 2021, Categories: Publica-
tions, Law and Policy.
Roy S., Farzana K., Papia M., Hasan M. Monitoring and predic-
tion of land use/land cover change using the integration of
Markov chain model and cellular automation in the South-
eastern Tertiary Hilly Area of Bangladesh. Int. J. Sci. Basic
Appl. Res., Vol. 24, No.4, 2015, pp. 125-148.
Sayilan M.O. Karabakh conflict between 1988-95. Master Pro-
ject. Ankara University, Ankara, 2007, 54 p. (in Turkish).
Scheffer M. Foreseeing tipping points. Nature, Vol. 467, No.
7314, 2010, pp. 411-412.
Sentinel Online. MultiSpectral Instrument (MSI) Overview,
European Space Agency. Retrieved 3 December 2018,
https://www.esa.int/.
Sentinel-2 MSI User Guides. Radiometric Resolutions, Sentinel-2
MSI, Sentinel Online. Sentinel.ESA.int. Retrieved 5 March
2020, https://sentinels.copernicus.eu/web/sentinel/user-guides/
sentinel-2-msi/resolutions/radiometric.
Song X., Duan Z., Jiang X. Comparison of artificial neural net-
works and support vector machine classifiers for land cover
classification in Northern China using a SPOT-5 HRG im-
age. International Journal of Remote Sensing, Vol. 33, No.
10, 2012, pp. 3301-3320.
Sylven M., Reinvang R., Andersone-Lilley Z. Climate change in
southern Caucasus: impacts on nature, people and society.
Report WWF Norway. 2009, pp. 1-35.
Teodoro A., Araujo R. Exploration of the OBIA methods avail-
able in SPRING non-commercial software to UAV data
processing. In: Proceedings of the Earth Resources and En-
vironmental Remote Sensing/GIS Applications V, Amster-
dam, the Netherlands, Vol. 9245, 2014.
Thenkabail P.S., Lyon J.G., Huete A. Fundamentals, sensor
systems, spectral libraries, and data mining for vegetation
(hyperspectral remote sensing of vegetation). The 2nd ed.,
CRC Press. 2018, 489 p.
Tiede D., Krafft P., Fureder, P., Lang S. Stratified template
matching to support refugee camp analysis in OBIA-
Workflows. Remote Sens., Vol. 9, No. 4, 2017, 326 p.
Valigholizadeh A., Karimi M. Geographical explanation of the
factors disputed in the Karabakh geopolitical crisis. Journal
of Eurasian Studies, Vol. 7, 2016, pp. 172-180.
Yagoub M., Bizreh A. Prediction of land cover change using
Markov and cellular automata models: the case of Al-Ain,
UAE, 1992-2030. Journal of the Indian Society of Remote
Sensing, Vol. 42, No. 3, 2014, pp. 665-671.
sian GIS 2018 Congress 04-07 September 2018 Baku,
Azerbaijan. Published by Nobel Akademik Yayıncılık
Eğitim Danışmanlık Tic. Ltd. Şti Baku/Azerbaijan. 2018c,
p.73.
Rehman S., Hussain M. Fuzzy C-means algorithm based satellite
image segmentation. Indonesian Journal of electrical engineer-
ing and computer science, Vol. 9, No. 2, 2018, pp. 332-334.
Report: Investigating the environmental dimensions of the 2020
Nagorno-Karabakh conflict. Conflict and Environment Ob-
servatory, Published: February 2021, Categories: Publica-
tions, Law and Policy.
Roy S., Farzana K., Papia M., Hasan M. Monitoring and predic-
tion of land use/land cover change using the integration of
Markov chain model and cellular automation in the South-
eastern Tertiary Hilly Area of Bangladesh. Int. J. Sci. Basic
Appl. Res., Vol. 24, No. 4, 2015, pp. 125-148.
Sayilan M.O. 1988-95 Arası dağlık Karabağ Sorunu. Master
Project. Ankara University, Ankara, 2007, 54 p.
Scheffer M. Foreseeing tipping points. Nature, Vol. 467, No.
7314, 2010, pp. 411-412.
Sentinel Online. MultiSpectral Instrument (MSI) Overview, Eu-
ropean Space Agency. Retrieved 3 December 2018,
https://www.esa.int/.
Sentinel-2 MSI User Guides. Radiometric Resolutions, Sentinel-2
MSI, Sentinel Online. Sentinel.ESA.int. Retrieved 5 March
2020, https://sentinels.copernicus.eu/web/sentinel/user-guides/
sentinel-2-msi/resolutions/radiometric.
Song X., Duan Z., Jiang X. Comparison of artificial neural net-
works and support vector machine classifiers for land cover
classification in Northern China using a SPOT-5 HRG im-
age. International Journal of Remote Sensing, Vol. 33, No.
10, 2012, pp. 3301-3320.
Sylven M., Reinvang R., Andersone-Lilley Z. Climate change in
southern Caucasus: impacts on nature, people and society.
Report WWF Norway. 2009, pp. 1-35.
Teodoro A., Araujo R. Exploration of the OBIA methods avail-
able in SPRING non-commercial software to UAV data
processing. In: Proceedings of the Earth Resources and En-
vironmental Remote Sensing/GIS Applications V, Amster-
dam, the Netherlands, Vol. 9245, 2014.
Thenkabail P.S., Lyon J.G., Huete A. Fundamentals, sensor
systems, spectral libraries, and data mining for vegetation
(hyperspectral remote sensing of vegetation). The 2nd ed.,
CRC Press. 2018, 489 p.
Tiede D., Krafft P., Fureder, P., Lang S. Stratified template
matching to support refugee camp analysis in OBIA-
Workflows. Remote Sens., Vol. 9, No. 4, 2017, 326 .p
Valigholizadeh A., Karimi M. Geographical explanation of the
factors disputed in the Karabakh geopolitical crisis. Journal
of Eurasian Studies, Vol.7, 2016, pp. 172-180.
Yagoub M., Bizreh A. Prediction of land cover change using
Markov and cellular automata models: the case of Al-Ain,
UAE, 1992-2030. Journal of the Indian Society of Remote
Sensing, Vol.42, No. 3, 2014, pp. 665-671.
“Azərbaycan Respublikasında iqtisadi rayonların yeni bölgüsü
haqqında” Azərbaycan Respublikası Prezidentinin 2021-ci il
7 iyul tarixli 1386 nömrəli Fərmanı, 2021,
https://president.az/articles/52389.
A.A.Rasouli et al. / ANAS Transactions, Earth Sciences 2 / 2022, 27-45; DOI: 10.33677/ggianas20220200080
45
ОБНАРУЖЕНИЕ И КАРТИРОВАНИЕ ИЗМЕНЕНИЙ РАСТИТЕЛЬНОГО ПОКРОВА И ЗЕМЛЕПОЛЬЗОВАНИЯ
С ПОМОЩЬЮ ПЕРЕДОВЫХ МЕТОДОВ ОБРАБОТКИ СПУТНИКОВЫХ ИЗОБРАЖЕНИЙ
(На примере Восточно-Зангезурского экономического района Азербайджанской Республики)
Расули A.A.1, Сафаров С.Г.2, Аскерова M.M.3, Сафаров Э.С.2, Милани М.4
1Департамент наук об окружающей среде, Университет Маккуари, Сидней, Австралия
Австралия, Сидней, Нот Райд, Воллиз Вок,12, Уровень 4: aarasuly@yahoo.com
2Министерство науки и образования Азербайджанской Республики, Институт Географии
им. акад. Г.Алиева, Баку, Азербайджан
AZ1070, г.Баку, просп. Г.Джавидa, 115: safarov53@mail.ru
3Азербайджанский Государственный Педагогический Университет, Баку, Азербайджан
AZ1000, Баку, ул.У.Гаджибейли, 68: matanat_askerova@mail.ru
4 Университет Бандырма Оньеди Эйлул, Бандырма, Турция
Yeni Mahalle, Shehit Astsubay Mustafa Soner Varlık Caddesi, 77, 10200, Бандырма, Турция: mmilani@bandirma.edu.tr
Резюме. При исследованиях были использованы методы обработки многоспектральных спутниковых изображений вы-
сокого разрешения Sentinel-2, снятых с 2016 по 2021 год, установлены Нормализованный разностный индекс растительно-
сти (NDVI) и Нормализованный разностный водный индекс (NDWI) для карт водной поверхности, растительности и без
растительности путем применения правил динамического определения пороговых значений, также был скорректирован
объектно-ориентированный анализ изображений (OBIA) в процессах сегментации и классификации с использованием
eCognition Developer для определения пространственных изменений в землепользовании. Наконец, для прогнозирования
типов землепользования в программном обеспечении TerrSet IDRISI Selva была введена модель Марковской цепи клеточ-
ных автоматов (CA-MC) с вероятностными правилами. Методы индексации показали, что за последние годы (2016-2021 гг.)
в районах Губадлы, Зангилан и Джебраил, расположенных на юге Восточно-Зангезурского Экономического Региона про-
изошли значительные изменения в растительном покрове. Впоследствии методы OBIA подтвердили, что в последние годы
оккупации данной территорий армянскими войсками большинство негативных изменений обнаруживается в типах земле-
пользования, преимущественно в лесах (-4,7%) и на пастбищных покровах (-4,6%). Кроме того, надежная карта прогнозов
CA-MA указывает на то, что в ближайшие годы будет наблюдаться заметный рост площадей бесплодных (+ 4,8%) и забро-
шенных (+ 5%) земель. Следовательно, точная обработка изображений и картографирование текущей ситуации на осво-
божденных от оккупации землях Азербайджана должны стать самой неотложной задачей географов, ученых-
экосистемологов и специалистов по дистанционному зондированию до того, как правительство приступит к проектам ре-
конструкции и восстановления.
Ключевые слова: изменения в растительном покрове и землепользовании, спутниковые изображения Sentinel-2, динамическое
и пороговое спектральные индексирования, научно-обоснованная OBIA классификация, карты прогнозирования CA-MC
PEYK TƏSVİRLƏRİNİN MÜTƏRƏQQİ İŞLƏNMƏ ÜSULLARININ TƏTBİQİ İLƏ BİTKİ ÖRTÜYÜ VƏ
TORPAQDAN İSTİFADƏDƏ BAŞ VERMİŞ DƏYİŞİKLƏRİN AŞKARLAMASI VƏ XƏRİTƏLƏŞDİRİLMƏSİ
(Azərbaycan Respublikasının Şərqi Zəngəzur iqtisadi rayonun timsalında)
Rəsuli A.A.1, Səfərov S.H.2, Əsgərova M.M.3, Səfərov E.S.2, Milani M.4
1Ətraf mühitə dair elmlər departamenti, Makkuari Universiteti, Avstraliya, Sidney
Sidney, North Ryde, Wally's Walk, 12, səviyyə 4: aarasuly@yahoo.com
2Azərbaycan Respublikası Elm və Təhsil Nazirliyi, akad. H.Əliyev ad. Coğrafiya İnstitutu, Bakı, Azərbaycan
AZ 1143, Bakı ş., H.Cavid prosp., 115: safarov53@mail.ru
3Azərbaycan Dövlət Pedaqoji Universiteti, Bakı, Azərbaycan
AZ1000, Bakı, Ü.Hacıbəyli, 68
4Bandırma Onyedi Eylul Universiteti, Mühəndislik və təbiət elmləri fakultəsi, Bandırma, Türkiyə
Yeni Mahalle, Shehit Astsubay Mustafa Soner Varlık Caddesi, 77, 10200, Bandirma, Turkiyə
Xülasə. Azərbaycan ərazilərinin uzun müddət ərzində erməni işğalı altında qalması, sosial-iqtisadi sarsıntıların ddindən artıq
ağrılı formaları, mçinin bitki örtüyü və torpaqdan istifadə də daxil olmaqla, ətraf mühitdə kar dağıdıcı geoekoloji dəyişikliklərlə
nəticələndi. Buna görə də, 2016-cı ildən 2021-ci ilədək çəkilmiş yüksək həlletmə qabiliyyətinə malik “Sentinel-2” çoxspektral peyk
təsvirlərinin, həmçinin bitkilərin (NDVI) və suların (NDWI) normallaşdırılmış müxtəliflik indekslərinin işlənilmə metodlarından isti-
fadə edilmişdir. Seqmentasiya prosesləri və eCognition Developer təsnifatları və TerrSet IDRISI Selva proqram təminatı tətbiq olun-
maqla, (OBIA) təsvirlərin obyektlər üzrə səmtlənmiş təhlili korrektə edilmiş, ehtimal olunan qaydalarla, qəfəs avtomatlarının Markov
zəncirinin (CA-MC) modeli yaradılmışdı. Daha sonra OBİA metodları təsdiqlədi ki, bu ərazinin erməni qoşunları tərəfindən işğalının
son illərində (2016-2020) mənfi dəyişikliklərin əksəriyyəti əsasən meşələrdə (-4.7%) və otlaq örtüklərində (-4.6%) aşkar edilmişdir).
Bundan əlavə, etibarlı CA-MA proqnostik xəritəsi, yaxın illərdə həm istifadəyə yarasız torpaqlarda (+ 4.8%), həm də tərk edilmiş
(+ 5%) ərazilərdə nəzərəçarpacaq artımların olacağı göstərir. Nəticə etibarilə, hökumətin yenidənqurma və bərpa layihələrinə
başlaması ilə əlaqədar coğrafiyaçıların, ekosistem alimlərinin və uzaqdan zondlama mütəxəssislərinin ən aktual vəzifəsi Azərbay-
canın ğaldan azad edilmiş torpaqları üzrə mövcud peyk görüntülərinin dəqiq lənməsi və müvafiq xəritələrin hazırlanması ol-
malıdır.
Açar sözlər: torpaq örtüyü torpaq istifadəsindəki dəyişikliklər, Sentinel-2 peyk görüntüləri, dinamik və hüdud spektral indeks-
ləməsi, elmi əsaslı OBIA təsnifatı, CA-MC proqnoz xəritələri
... Meanwhile, other optical sensors, such as Sentinel-2, may be considered to improve the SCA maps' accuracy. Sentinel-2 has a lower temporal resolution than the Landsat OLI, and we may combine their high-spatial resolution bands to achieve better results in future studies [84,85]. It is also worthwhile to test the sensitivity of ALOS-Palsar DEM data to reconstruct fine SCA datasets and associated products such as snow depth and snow water equivalence (SWE) [86][87][88][89]. ...
Article
Full-text available
This study detected the spatial changes in Snow Cover Area (SCA) over the Snowy Mountains in New South Wales, Australia. We applied a combination of Object-Based Image Analysis (OBIA) algorithms by segmentation, classification, and thresholding rules to extract the snow, water, vegetation, and non-vegetation land covers. For validation, the Maximum Snow Depths (MSDs) were collected at three local snow observation sites (namely Three Mile Dam, Spencer Creek, and Deep Creek) from 1984 to 2020. Multiple Landsat 5, 7, and 8 imageries extracted daily MSDs. The process was followed by applying an Estimation Scale Parameter (ESP) tool to build the local variance (LV) of object heterogeneity for each satellite scene. By matching the required segmentation parameters, the optimal separation step of the image objects was weighted for each of the image bands and the Digital Elevation Model (DEM). In the classification stage, a few land cover classes were initially assigned, and three different indices—Normalized Differential Vegetation Index (NDVI), Surface Water Index (SWI), and a Normalized Differential Snow Index (NDSI)—were created. These indices were used to adjust a few classification thresholds and ruleset functions. The resulting MSDs in all snow observation sites proves noticeable reduction trends during the study period. The SCA classified maps, with an overall accuracy of nearly 0.96, reveal non-significant trends, although with considerable fluctuations over the past 37 years. The variations concentrate in the north and south-east directions, to some extent with a similar pattern each year. Although the long-term changes in SCA are not significant, since 2006, the pattern of maximum values has decreased, with fewer fluctuations in wet and dry episodes. A preliminary analysis of climate drivers’ influences on MSD and SCA variability has also been performed. A dynamic indexing OBIA indicated that continuous processing of satellite images is an effective method of obtaining accurate spatial–temporal SCA information, which is critical for managing water resources and other geo-environmental investigations.
Article
Full-text available
Object-based image analysis (OBIA) has been widely used for land use and land cover (LULC) mapping using optical and synthetic aperture radar (SAR) images because it can utilize spatial information, reduce the effect of salt and pepper, and delineate LULC boundaries. With recent advances in machine learning, convolutional neural networks (CNNs) have become state-of-the-art algorithms. However, CNNs cannot be easily integrated with OBIA because the processing unit of CNNs is a rectangular image, whereas that of OBIA is an irregular image object. To obtain object-based thematic maps, this study developed a new method that integrates object-based post-classification refinement (OBPR) and CNNs for LULC mapping using Sentinel optical and SAR data. After producing the classification map by CNN, each image object was labeled with the most frequent land cover category of its pixels. The proposed method was tested on the optical-SAR Sentinel Guangzhou dataset with 10 m spatial resolution, the optical-SAR Zhuhai-Macau local climate zones (LCZ) dataset with 100 m spatial resolution, and a hyperspectral benchmark the University of Pavia with 1.3 m spatial resolution. It outperformed OBIA support vector machine (SVM) and random forest (RF). SVM and RF could benefit more from the combined use of optical and SAR data compared with CNN, whereas spatial information learned by CNN was very effective for classification. With the ability to extract spatial features and maintain object boundaries, the proposed method considerably improved the classification accuracy of urban ground targets. It achieved overall accuracy (OA) of 95.33% for the Sentinel Guangzhou dataset, OA of 77.64% for the Zhuhai-Macau LCZ dataset, and OA of 95.70% for the University of Pavia dataset with only 10 labeled samples per class.
Article
Full-text available
Accurate and reliable information about the situation in refugee or internally displaced person camps is very important for planning any kind of help like health care, infrastructure, or vaccination campaigns. The number and spatial distribution of single dwellings extracted semi-automatically from very high-resolution (VHR) satellite imagery as an indicator for population estimations can provide such important information. The accuracy of the extracted dwellings can vary quite a lot depending on various factors. To enhance established single dwelling extraction approaches, we have tested the integration of stratified template matching methods in object-based image analysis (OBIA) workflows. A template library for various dwelling types (template samples are taken from ten different sites using 16 satellite images), incorporating the shadow effect of dwellings, was established. Altogether, 18 template classes were created covering typically occurring dwellings and their cast shadows. The created template library aims to be generally applicable in similar conditions. Compared to pre-existing OBIA classifications, the approach could increase the producer's accuracy by 11.7 percentage points on average and slightly increase the user's accuracy. These results show that the stratified integration of template matching approaches in OBIA workflows is a possibility to further improve the results of semi-automated dwelling extraction, especially in complex situations.
Article
Full-text available
Geopolitical crises are among the international crises that have geographic nature and root. Therefore, despite international political or economic crises that may have more negative consequences than those of geopolitical ones, no crisis threatens the nations-states' factors as do geopolitical crises because the specialty of geographic basis is rooted in national and prestigious values. Thus, the compromise about them is quite hard and impossible, and these crises often last so long. In this regard, if a geopolitical crisis, like the case of Karabakh, is rooted in bilateral disputes over geographic factors, including geographic-political, geographic-cultural and geographic-economic factors, it would have special complexities.
Article
Full-text available
Spatially land-cover models are necessary for sustainable land-cover planning. The expansion of human-built land involves the destruction of forests, meadows and farmlands as well as conversion of these areas to urban and industrial areas which will result in significant effects on ecosystems. Monitoring the process of these changes and planning for sustainable use of land can be successfully achieved by using the remote sensing multi-temporal data, spatial criteria and predictor models. In this study, land-cover change analysis and modeling was performed for our study area in central Germany. An integrated Cellular Automata–Markov Chain land change model was carried out to simulate the future landscape change during the period of 2020–2050. The predictive power of the model was successfully evaluated using Kappa indices. As a consequence, land change model predicts very well a continuing downward trend in grassland, farmland and forest areas, as well as a growing tendency in built-up areas. Hence, if the current trends of change continue regardless of the actions of sustainable development, drastic natural area decline will ensue. The results of this study can help local authorities to better understanding the current situation and possible future conditions as well as adopt appropriate strategies for management of land-cover. In this case, they can create a balance between urban development and environmental protection.
Article
Full-text available
Optical satellite imagery has been used for analyzing the spatial distribution; temporal changes and prediction of land use/land cover (LULC) categories in the southeastern hill tracts of Bangladesh. This is the primary work in this study area where multi-temporal Landsat imagery has been used to generate LULC maps for the years of 1989, 2003 and 2014. In the first step, a total seven LULC categories were derived using the integration of NDVI and conventional supervise classification techniques which were further evaluated using the error matrix table and kappa statistics. The result obtained from the accuracy assessment process presented an overall accuracy of more than 90% for all three years with Kappa statistics 1989 (88.67), 2000 (92.33) and 2014 (89.67) respectively. A rigorous change analysis was carried out, using Land Change Modeler (LCM) where dramatic changes were noticed in hilly forest, shrub land and crop land categories for 1989-2014 time span. From 1989 to 2014 hilly forest decreased 3062 sq. km, shrub land and cropland increased 2990.82 sq.km and 258.74 sq.km area respectively. In the second stage, Markov chain-cellular automata was used to model and predict LULC in the study area. At first Markov chain was used to generate transition probability matrices between LULC categories and then cellular automata was used to predict 2014 LULC map. After successful validation of observed and predicted LULC maps of 2014 (Overall accuracy 91.43% and Kappa statistics 90.00), the combined procedure was employed to simulate land-use/land-cover for 2028 and 2042. The simulated result of 2014-2042 shown a dramatic decrease in hilly forest (1831 sq.km.) and water reservoirs (148 sq.km.) whereas a significant increase was noticed in the shrub land (1889 sq.km.), crop land (91 sq.km.) and settlement area (7.5 sq.km). All these findings from the research could offer an opportunity for the more skillful management and policy making on biodiversity, forest, land and other environmental resources in the study region. URL: http://gssrr.org/index.php?journal=JournalOfBasicAndApplied&page=article&op=view&path[]=4539
Article
Full-text available
The UAE has witnessed rapid urban development and economic growth in recent years. With its ambitious vision to become one of the advanced nations by 2021, planners and policy-makers need to know the most likely direction of future urban development. In this study, remotely sensed imagery coupled with cellular automata models were used to predict land cover in Al Ain, the second largest city in the Emirate of Abu Dhabi. Markov and cellular automata models were used for 1992 and 2006 to predict land cover in 2012. Land Use and Land Cover maps for the study area were derived from 1992, 2006, and 2012 Landsat satellite images (TM, ETM+). The models achieved an overall accuracy of approximately 80 %. A Markov model was applied for 2006 and 2012 to predict land cover in 2030. The results conformed to the general trend of the Al Ain Master Plan 2030. This study demonstrates that remote sensing, with the availability of free Landsat data, is a viable technology that could be used to help in the prediction process especially in developing countries, where data availability is a problem.
Article
Full-text available
The appetite for up-to-date information about earth's surface is ever increasing, as such information provides a base for a large number of applications, including local, regional and global resources monitoring, land-cover and land-use change monitoring, and environmental studies. The data from remote sensing satellites provide opportunities to acquire information about land at varying resolutions and has been widely used for change detection studies. A large number of change detection methodologies and techniques, utilizing remotely sensed data, have been developed, and newer techniques are still emerging. This paper begins with a discussion of the traditionally pixel-based and (mostly) statistics-oriented change detection techniques which focus mainly on the spectral values and mostly ignore the spatial context. This is succeeded by a review of object-based change detection techniques. Finally there is a brief discussion of spatial data mining techniques in image processing and change detection from remote sensing data. The merits and issues of different techniques are compared. The importance of the exponential increase in the image data volume and multiple sensors and associated challenges on the development of change detection techniques are highlighted. With the wide use of very-high-resolution (VHR) remotely sensed images, object-based methods and data mining techniques may have more potential in change detection.
Article
Full-text available
Remote sensing of land cover in Mediterranean regions is complicated by the high landscape diversity, which is typical of both natural and agricultural lands. This spatial complexity reduces the accuracy of the common per-pixel classification of multi-spectral remotely sensed imagery. In this paper we show that per-field statistics derived from multi-spectral imagery enhances separability between different crops and terrain categories. We also found that Synthetic Aperture Radar (SAR) imagery processed with Thematic Mapper (TM) imagery in a synergistic context produces a higher enhancement of discrimination if per-field statistics are used. Finally, we show that image segmentation is a convenient way to apply this approach avoiding field digitizing by computing per-segment statistics of training fields and classifying the segmented image through Linear Canonical Discriminant Analysis.
Article
There has been a recent surge of interest in remote sensing and its use in ecology and conservation. This book focuses explicitly on the Normalized Difference Vegetation Index (NDVI), a simple numerical indicator and powerful tool that can be used to assess spatio-temporal changes in green vegetation. The NDVI opens the possibility of addressing questions on scales inaccessible to ground-based methods alone; it is mostly freely available with global coverage over several decades. This text provides an authoritative overview of the principles and possible applications of the NDVI in ecology, environmental and wildlife management, and conservation. NDVI data can provide valuable information about temporal and spatial changes in vegetation distribution, productivity, and dynamics; allowing monitoring of habitat degradation and fragmentation, or assessment of the ecological effects of climatic disasters such as drought or fire. The NDVI has also provided ecologists with a promising way to couple vegetation with animal distribution, abundance, movement, survival and reproductive parameters. Over the last few decades, numerous studies have highlighted the potential key role of satellite data and the NDVI in macroecology, plant ecology, animal population dynamics, environmental monitoring, habitat selection and habitat use studies, and paleoecology. The chapters are organized around two sections: the first detailing vegetation indices and the NDVI, the principles behind the NDVI, its correlation with climate, the available NDVI datasets, and the possible complications and errors associated with the use of this satellite-based vegetation index. The second section discusses the possible applications of the NDVI in ecology, environmental and wildlife management, and conservation.