images with the OpenStreetMap or
COTS
navigation data,
FTS
maps, and topographic maps as auxiliary data (GMES, 2011).
Sertel and Akay (2015) aimed to create accurate, detailed
LCLU
map especially in urban areas of Gaziantep City from
SPOT
-
5 images. The Object based classification method was used to
determine
LCLU
types and besides
SPOT
-5 images, Normalized
Difference Vegetation Index (
NDVI
) and Normalized Difference
Water Index (
NDWI
) maps, cadastral maps, Openstreet maps, road
maps and Land Cover maps were integrated to the classification
process. Urban Atlas nomenclature was used to identify
LCLU
classes. The results showed that integration of satellite images
with different spatial data increase the accuracy of urban
LCLU
map with suggested approach in the study, urban areas repre-
sented with seventeen classes and mapped with 94% or higher
overall accuracy. Akay and Sertel (2016), determined urban
LCLU
changes in Gaziantep City Centre between 2010 and 2015 using
SPOT
-5 and
SPOT6
images, respectively. Object based classifica-
tion applied with integration of various ancillary data namely
Normalized Difference Vegetation Index (
NDVI
), Difference Water
Index (
NDWI
) maps, cadastral maps, OpenStreetMaps, road maps
and Land Cover maps. The European Urban Atlas project
LCLU
maps legend used for determination of
LCLU
class definitions
in the study. The decision-tree based object oriented classifica-
tion method suggested in this study fulfilled the geometric and
thematic requirements of 1:10 000 scale Urban Atlas project.
This study showed that these maps could be used for support
master planning, monitor transportation infrastructure, generate
location based services and support street navigation.
Landscape metrics have been used for defining relation-
ships between the structural landscape features and the
ecological functions (Leitao
et al
., 2006). Spatial metrics
first used for landscape ecology studies with describing the
land covers’ composition and spatial arrangement and then
now they are used for urban areas for defining the urban
patterns and structures (Mesev, 2007). Different researchers
used metrics for defining urban sprawl pattern relation with
ecological features (Herold
et al
., 2002; Ji
et al
., 2006; Sudhira
et al
., 2004; Uuemaa
et al
., 2009). There are also studies more
specific about urban features with using metrics and high
resolution satellite images (Ghafouri
et al
., 2016; Prastacos
and Kochilakis, 2012; Stan
et al
., 2016; Uuemaa
et al
., 2011;
Zheng
et al
., 2016).
Prastacos and Kochilakis (2012) calculated six landscape
metrics using Urban Atlas land cover information for nine
of Greece’s provinces. The calculated metrics are;
CONTAG
,
Patch Density (
PD
), Edge Density (
ED
), Largest Patch Index
(
LPI
), Euclidean Nearest Neighbor Distance (
ENN_MN
), Fractal
Dimension (FRAC) indices. By interpreting the calculated
metrics, they made inferences and comparisons about the
structures and forms of these cities. Herold
et al
. (2002) used
landscape metrics to identify and quantify urban growth in
two test sites in the Santa Barbara area. In this study, aerial
photo data were used for 1978, 1979, and 1998. Hierarchical
US Geological Survey scheme is used as the classification
scheme. Uuemaa
et al.
(2009) calculated 15 landscape metrics
for 35 regions in Estonia and analyzed factors and main com-
ponents to determine which landscape metrics could work
better for the Estonian map. Their results showed that there
are four main components that define the landscape structure
which are dominance, contrast, shape complexity, and com-
position. Landscape metrics proposed according to the study
results and defining these components respectively; (
ED
) or
Simson’s Diversity Index (
SIDI
), Total Edge Contrast Index
(
TECI
) or Mean Edge Contrast Index (
ECON_MN
), Mean Contrast
Index Shape Index-SHAPE_MN) and Patch Richness Density
(
PRD
). Ghafouri
et al
. (2016) aimed to explore the relation-
ship between human-derived factors and landscape metrics
as a landscape pattern indicator, and to identify appropriate
metrics for modeling this relationship. Landscape metrics for
class level were calculated by using
LCLU
maps for 32 districts
belonging to Mazandaran and Guilan provinces located to the
south of Hazar Sea.
Although urban atlas maps are available for several Euro-
pean cities, there is not such data for a huge amount of cities
outside the Europe. Therefore, it is important to discover ap-
propriate remote sensing data and methods to create up-to-date
urban atlas-like maps of cities other than selected
EU
cities.
This research aims to create up-to-date high resolution
detailed urban atlas-like
LCLU
maps of two different districts
with different landscape characteristics and conduct land-
scape metrics comparison between them. The study contains
two districts of Karabük, Turkey namely Safranbolu and
Center which are neighborhood districts but have very dif-
ferent landscape characteristics. Iron and steel factories are
located in Karabük Center district, and it is well known with
mining industry. Whereas, Safranbolu district has urban sites
and archeological city residences which is declared as World
Cultural Heritage. 1.5 m
SPOT
6/7 satellite images obtained in
2016 were used in this study as main remote sensing data in
conjunction with Zoning plans and OpenStreetMap road data
for determining road and artificial surface classes, forest maps
for forest areas and
NDVI
images. Object based classification
method was applied by using this data set in order to create
accurate urban
LCLU
maps. Then, different landscape metrics
namely Patch Density (
PD
), Edge Density (
ED
), Largest Patch
Index (
LPI
), Euclidean Nearest Neighbor Distance (
ENN_MN
),
Area-Weighted Mean Fractal Dimension Index (
FRAC_AM
) and
Contagion (
CONTAG
) metrics were calculated for two districts
both urban area and district area scales and results of differ-
ent metrics were compared. Analyzes of urban areas with
landscape metrics allows objective evaluations to define the
structure of urban areas and representation of differences or
similarities in the spatial structure of different urban regions.
Study Area
Karabük City located on Black Sea part of Turkey lies between
40° 5
′
and 40° 15
′
north latitudes and between 32° 15
′
and
32° 20
′
east longitudes (Figure 1a) covering an area of 4.145
km² (Karabük Special Provincial Administration, 2015). In
this research, we selected Center and Safranbolu districts of
Karabük City as two pilot areas (Figure 1).
The iron and steel industry is the main economical income
of Karabük City. Additionally, the city has extensive forest
areas, a variety of vegetation and historical assets (Republic
of Turkey Ministry of Environment and Urbanization, 2009.)
The mean elevation of Karabük is 350 m and the city cov-
ers hilly areas with steep slopes specifically outside the city
center. Center district is located at the center of the city and
has neighborhood with all other districts of Karabük City.
The surface area of the district is 70.400 ha and population is
134,406 according to 2016 Address-based population registra-
tion. Iron and steel industrial facilities are established at the
Center district including 420 ha of organized industrial zone.
On the contrary, Safranbolu district is an important historical
site including architectural structures from Ottoman times,
played a key role in the caravan trade over many centuries
and is in UNESCO’ s World Heritage list (UNESCO, n.d). With
the National Conservation Law, No 2863, Safranbolu was
declared as urban and natural site. Safranbolu is an attractive
touristic center due to its historical and cultural structure and
architecture. The population of Safranbolu is 63,965 accord-
ing to 2016 Address-based population registration. Table 1
shows total area values, population values and numbers of
settlements for each district.
712
November 2018
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
