See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/329216831

COMPARISON OF INTERPOLATION METHODS FOR METEOROLOGICAL DATA

Thesis · May 2011

CITATIONS READS

2 565

3 authors, including:

Merve Keskin
Finnish Geospatial Research Institute FGI, National Land Survey of Finland
35 PUBLICATIONS 79 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Exploring the Cognitive Processes of Map Users Employing Eye Tracking and EEG View project

EnviroGRIDS - 11. Building Capacity for a Black Sea Basin Observation and Assessment System Supporting Sustainable Development View project

All content following this page was uploaded by Merve Keskin on 29 May 2019.

The user has requested enhancement of the downloaded file.

 İSTANBUL TECHNICAL UNIVERSITY

CIVIL ENGINEERING FACULTY

GEOMATICS ENGINEERING

COMPARISON OF INTERPOLATION METHODS FOR

METEOROLOGICAL DATA

Prepared by
MERVE KESKIN
010060315
KORAY ÖZDOĞU
010080701

Consulted by
Assist. Prof. Dr. Çiğdem Göksel
Prof. Dr.Necla Uluğtekin
Assist. Prof. Dr. Ahmet Özgür DoğruÖ

ISTANBUL, MAY, 2011

PREFACE

This thesis is based upon studies conducted during February 2011 to May 2011 at
the Department of Geomatics Engineering in Istanbul Technical University,
Turkey.

We would like to express my sincere gratitude to our leading supervisor Assist.

Prof. Dr. Ahmet Özgür Doğru. Without his advise and unique support this thesis
would never had become a reality. Also we would like thank the staff at the
Department of Geomatics Engineering and creators and coordinaters of
EnviroGRIDS project.

Finally, we wish to express our greatest thanks to our family, friends and
colleagues, who have supported us during this time period.

Koray Özdoğu & Merve Keskin, May 2011.

2
CONTENTS

Page

PREFACE ………………………………………………………………...….…......ii

CONTENT ..........………………………………………………………...…..….....iii

TABLE LIST .........……………………………………………………………...... v

EQUATION LIST ................................................................................................... v

FIGURE LIST ......…………………………………………………………….........vi

1. INTRODUCTION.................................................................................................. 7

1.1. A brief Information about EnviroGRIDS at The Black Sea Catchment

project ..................................................................................................................... 9

2. METHODOLOGY .............................................................................................. 10

2.1. Interpolation .................................................................................................. 10

2.1.1 Spatial Interpolation ............................................................................... 10

2.1.2 Spatial Interpolation in GIS .................................................................. 11

2.1.2.1 Inverse Distance Weighted Interpolation.......................................11

2.1.2.2 Spline Interpolation .........................................................................13

2.1.2.3 Kriging Interpolation.......................................................................15

2.1.2.4 Natural Neighborhood Interpolation..............................................18

2.2. Comparison of Spatial Interpolation Methods ........................................... 19

2.3. Extrapolation ................................................................................................. 21

2.4.Accuracy Assesment ...................................................................................... 21

3. CASE STUDY ..................................................................................................... 22

3.1. Case Study of Bulgaria ................................................................................ 22

3.1.1. Data of the Study .................................................................................... 23

3.1.2. GIS Implementation .............................................................................. 25

3.1.3. Determination of Spatial Interpolation Method ................................. 26

3
 3.1.4. Accuracy Assessments ........................................................................... 30

3.2. Case Study of Turkey ................................................................................... 34

3.2.1 Data of the Study ..................................................................................... 35

3.2.2. GIS Implementation .............................................................................. 37

3.2.3. Determination of Spatial Interpolation Method ................................. 38

3.2.4. Accuracy Assessments ........................................................................... 41

3.3. Mapping......................................................................................................... 42

4. CONCLUSION..................................................................................................... 45

REFERENCES ......................................................................................................... 46

CURRICULUM VITAE .......................................................................................... 49

4
 TABLE LIST

Table 3.1: Database view of wind energy data..................................................... 23

Table 3.2: Control Stations…………………………………………………………27

Table 3.3: The actual Wind Data at10 m for Control Stations…………………….31

Table 3.4: Errors and RMSEs for each method……………………………………31

Table 3.5: The actual Wind Data at 100 m for Control Stations………………..….32

Table 3.6: Errors and RMSEs for each method…………………………………….32

Table 3.7: The actual Wind Data at10 m for Second Control Stations……………..33

Table 3.8: The actual Wind Data at100 m for Second Control Stations……………33

Table 3.9: Errors and RMSEs for each method……………………………………34

Table 3.10: Database view of wind speed and temperature.....................................35

Table 3.11: Control Stations………………………………………………………...38

Table 3.12: The actual Wind Data at 10 m for Control …………………………….41

Table 3.13: Errors and RMSEs for each method…………………………………..41

EQUATION LIST

(2.1)

(2.4)

5
 FIGURE LIST

Figure 2. 1: How Spline Interpolation Works………………………………………14

Figure 2.2: Semivariance…………………………………………………………..17

Figure 3.1: Location of the stations in Bulgaria...................................................... 23

Figure 3.2: A map of monitoring stations in Bulgaria…………………………… 26

Figure 3.2: Result of IDW interpolation method for wind speed data at 10 m……27

Figure 3.3: Result of IDW interpolation method for wind speed data at 100 m…..28

Figure 3.4: Result of Kriging interpolation method for wind speed data at 10 m…28

Figure 3.5: Result of Kriging interpolation method for wind speed data at 100 m..29

Figure 3.6: Result of Natural Neighborhood interpolation method for wind speed
data at 10 m………………………………………………………………………… 29

Figure 3.7: Result of Natural Neighborhood interpolation method for wind speed
data at 100 m……………………………………………………………………….. 30

Figure 3.8: Location of the stations in Turkey..........................................................35

Figure 3.9: Result of IDW interpolation method for wind speed data at 10 m........39

Figure 3.10: Result of Kriging interpolation method for wind speed data at 10 m..39

Figure 3.11: Result of Natural Neighborhood interpolation method for wind speed
data at 10 m………………………………………………………………………….40

Figure 3.12: Result of Spline interpolation method for wind speed data at 10 m…40

Figure 3.13: Wind Speeds at 10m. Map of Bulgaria……………………………….42

Figure 3.14: Wind Speeds at 100m. Map of Bulgaria………………...……………43

Figure 3.15: Wind Speeds at 10m. Map of Turkey…………………………………44

Figure 3.16: Average Temperature Map of Turkey…………………………………44

6
INTRODUCTION

In recent decades, energy became one of the most crucial topics in the world.
Producing energy, finding energy sources and making decisions about which type of
energy will be used are vital concerns of energy management. Finding out the energy
potentials and bringing out the current state of renewable energy sources will help us
to make realistic assumptions and further suggestions about useful utilizations of
energy. The visualization of the status and managing the existing data will be much
more convenient and easier by mapping process and establishing a geographic
information system (GIS).

Temperature, humidity, wind speed, and etc. are considered as measured values
forming the data used to monitor and determine meteorological conditions in a
region. However, these are point source data since they are collected as a result of the
measurements at a specific point. Spatial Interpolation methods are used to represent
these data by surfaces. In this context point based vector data is interpolated to the
raster surface, so that each raster cell represents a measurement such as cell’s
relationships to a fixed point or specific concentration level. Raster data derived by
using suitable spatial interpolation method is the main source of the meteorological
maps. So that spatial interpolation is the most important step of the mapping process
of the meteorological data.

The most important thing in mapping the meteorological data is determining the
suitable methodology to execute. Acquiring the data set to use, selecting the
optimum amount data to map, deciding which attributes related to entities will be
needed, finding out the ways to generate additional data out of the data in hand-if
neccessary, determining the technique of providing new data (for instance; deciding
an interpolation method to apply), setting out all the other possible techniques and
comparison of them, investigating the accuracy of the work and making the final
product- which is a map- qualified and look visually nice and user friendly are the
steps of producing map.

Additionally, we may not have actual information about every single point in our
study area and most of the time; it is what we have to deal with. Interpolation is a

7
method to represent the point source data as surfaces. Geographic Information
Systems (GIS) offer a chance to evaluate attribute and spatial data together and to
make statistical analysis. . Many of the GIS software are currently provide spatial
interpolation tools for several methods such as Kriging, Inverse Distance Weighted
(IDW), Natural Neighbour, Spline, and etc. However the problem is determining the
interpolation method which gives the best calculation result for each specific case
and study area, because the accuracy of the spatial interpolation is quite related the
total number of point sources, their locations and distribution within the study area.

This study, which is developed and being executed as a subtask of the EnviroGRIDS
Project supported by European Union within the 7th Framework Programme, aims at
to determine optimum spatial Interpolation Method for mapping meteorological data
in Bulgaria and along the Black Sea cost of Turkey. Performances of the spatial
interpolation methods used in ArcGIS software were assessed regarding the
applications executed by using meteorological data of Bulgaria and Turkey.

8
1.1. A brief Information about EnviroGRIDS at The Black Sea Catchment
project

A brief InformatIon about The Black Sea Catchment project:

The Black Sea Catchment is world-wide known to be one of the unsustainable

ecological development and insufficient resource management, leading to serious
environmental, social and economic issues (URL 7).

The core environmental problem of the Danube River Catchment can be described as
“ecologically unsustainable development and inadequate water resources
management”. The problems are caused by different factors, such as: inadequate
management of wastewater/solid waste, ecological unsustainable industrial activities,
and inadequate land management and improper agricultural practices. They generate
several direct consequences: pollution of surface/groundwater, eutrophication, and
accelerated runoff /erosion. These consequences have, on the other hand, the
following main effects: decline in quality of life, human health risks, degradation of
biodiversity, economic decline, and reduced availability of water (URL 7).

The Black Sea itself is also affected by severe environmental degradation. In 1995, it
was rated with the highest concerns in five out of seven environmental categories,
making it the worst of any of the European seas (Stanners and Boudreau 1995).
Some signs of recovery have been observed in the last years, but eutrophication
remains a severe problem. The Black Sea Hydrological Catchment represents a very
interesting case study to test the capacity of integrating large data sets to assess
vulnerability and sustainability issues related to freshwater resources as various
scales. The project is therefore essentially concentrating on the terrestrial part of the
Black Sea Hydrological Catchment, not to be confused with the marine part often
considered when using the Black Sea Basin terminology (URL 7).

Istanbul Technical University is one of the partners of this multi-disciplined project.

Our part is to do mapping -based on the data and researches- in order to set out wind
energy potentials, introducing different interpolation techniques and determining the
best method related to given data set while mapping.

9
2. METHODOLGY

2.1. Interpolation

Interpolation is basically an estimation method and a process of acquiring new mid-

data values between existing data values. In other words, as Moler described,
interpolation is the process of defining a function that takes on specified values at
specified points (1989). It is different to "smoothing", the techniques have the
characteristic that the estimated curve passes through all the given points and the idea
is that the points are in some sense correct and lie on an underlying but unknown
curve, the problem is to be able to estimate the values of the curve at any position
between the known points (Bourke, 1999).

In numerical analysis, interpolation is a kind of a method of building new data points

within the domain of a different set of known points.

In science and engineering, curve fitting or regression analysis is used to construct a

function fitting the data points in hand which have been achieved by sampling or
various experiments. Thus, interpolation is a certain condition of curve fitting, where
the function has to pass exactly through the data points.

2.1.1 Spatial Interpolation

According to Nigel M. Waters, spatial interpolation is the procedure of estimating

the value of properties at unsampled sites within the area covered by existing
observations and in almost all cases, the property must be interval or ratio scaled.
This method may be considered as the reverse process of selecting the some points
from a Digital Elevation Model (DEM)-representing the area exactly and precisely.
The logic lying behind spatial interpolation is to observe the points which lie close
together in space, tend to be more likely to have similar values than points far apart
(1988).

There are several interpolation methods such as piecewise constant interpolation

(nearest neighbor), linear interpolation, polynomial interpolation, spline
interpolation, linear regression, curve fit through arbitrary points, interpolation via
Gaussian processes, interpolation in digital signal processing. In this study spatial
interpolation methods widely used within the GIS software were considered as
special cases.

10
2.1.2 Spatial Interpolation in GIS

Spatial interpolation is a very significant and commonly used feature in many GIS
softwares. As declared by UC Santa Barbara Department of Geography, spatial
interpolation can be used in GISs:

o to provide contours for displaying data graphically,

o to calculate some property of the surface at a given point,

o to change the unit of comparison when using different data structures

in different layers,

o frequently is used as an aid in the spatial decision making process

both in physical and human geography and in related disciplines such
as mineral prospecting and hydrocarbon exploration.

It can be realized that many of the spatial interpolation techniques are two-
dimensional developments of the one dimensional methods previously developed for
time series analysis. One unit looks at areal procedures and some applications when
the other unit introduces spatial interpolation and examines point based interpolation
simultaneously (URL 1).

In this study, spatial interpolation methods of inverse distance weighted, spline,

kriging and natural neighborhood are explained in detail since they are considered as
methods used for comparisons.

2.1.2.1 Inverse Distance Weighted Interpolation

Based on ESRI Dictionary of terms, Inverse Distance Weighted interpolation (IDW)

is an interpolation technique that estimates cell values in a raster from a set of sample
points that have been weighted so that the farther a sampled point is from the cell
being evaluated, the less weight it has in the calculation of the cell's value.

This interpolation method calculates cell values by averaging sample data point
values in the neighborhood of every cells. When the point gets closer to the center of
the cell trying to be estimated, it gains more influence in the averaging calculations.
In other words, IDW assumes that the influence of the variable to be mapped gets
lower as the distance from its sampled location gets higher (URL 2). Let’s examine
some basic parameters in IDW inierpolation in ArcGIS software.

11
Power parameter’s properties:

 To be able to control the importance of known points on the interpolated

values based on their distance from the output point.

 By defining a higher power, more emphasis is placed on the nearest points,

and the resulting surface will have more detail (be less smooth).

 Specifying a lower power will give more influence to the points that are
farther away, resulting in a smoother surface. A power of two is most
commonly used with IDW and is the default (ESRI, ArcGIS Resource
Center).

Nyugen’s research also supports the idea that the high value setting for Power
parameter will emphasize more on the nearest points and will produce a more detail
and less smooth surface; whereas, the lower power will give more power to the far
way points and producing smoother surface. The default power value- commonly
used with IDW- is 2 and based on several trials, Nyugen found out that the default
value stabilized the power between far-away and the nearest points (Nyugen, 2009).

By applying a Search Radius, the characteristics of the interpolated surface can be

controlled. Calculation of each interpolated cells is obtained by a fixed or variable
search radius. Processing speed can be enhanced by limiting the number of points. It
is also possible to limit the number of points when the points have no spatial
correlation because of their cell location (URL 2). Again in Nyugen’s application
(2009), he chose to limit the number of point, in order not to take the far away points
into consideration, because there will not be any correlation between them. By
observing the spiky distribution histogram of the topographic map he produced, he
discovered that the point data set has unusual height of elevation; if the data doesn’t
fit to normal distribution.

A Fixed Search Radius requires a neighborhood distance- dictating the radius of the
neighborhood- and a minimum number of points (indicating the minimum number of
measured points to use within the neighborhood). For each interpolated cell, the
distance of the radius is constant. In calculation of each interpolated cell, the
measured points falling within the radius are used. The search radius will ascend
until it can encircle the minimum number of points, if there are less measured

12
neighboring points than the specified minimum. With a Variable Search Radius,
specifying the number of points used when calculating the interpolated cell values
makes the radius distance change for each interpolated cell. By this way, depending
on the density of the measured points near the interpolated cell, some neighborhoods
will be large and some will be small. Generally, smaller neighborhoods or a
minimum number of points are used while the phenomenon has a great amount of
variation (URL 2).

There are some cons and pros of IDW. The advantages are:

 It is a simply understandable principle.

 The calculation speed is fairly fast.

 The algorithm is not as complicated as the Kriging method.

 It produces reasonable results for many type of data set.

 This method is better fit for evenly distributed normal distribution data set.

The disadvantages are:

 When using different setting of parameters’ values, it is hard to get a decent

interpolation raster.

 For the best outcome, one needs to have a DEM raster layer as a source of
verification, or the geography knowledge of the study area, which may not be
realistic in a real life application.

 When the characteristics of study area are not known, the interpolation can be
effected by rough distribution of observed points, since an equal weight
would be assigned to each of the data points even if it is in a cluster (Nyugen,
2009).

2.1.2.2 Spline Interpolation

The Spline method is an interpolation method that estimates values using a

mathematical function that minimizes overall surface curvature, resulting in a smooth
surface that passes exactly through the input points. It fits a mathematical function to
a specified number of nearest input points while passing through the sample points.
This method is best for generating gently varying surfaces such as elevation, water
table heights, or pollution concentrations (URL 2).

13
It can be named as zero-order interpolation, thiessen polygons, voronoi polygons or
drichlet cells. The idea is simple and based on determining the value at a location
with an unmeasured attribute by assigning to it the value of the closest measured
value out of known sample point values (URL 1) How spline interpolation works is
shown in Figure 2.1.

Figure 2.1: How Spline Interpolation Works

There are two Spline methods:

The Regularized method

 It creates a smooth, gradually changing surface with values that may lie
outside the sample data range.

 The weight parameter defines the weight of the third derivatives of the
surface in the curvature minimization expression.

 The values entered for this parameter must be equal to or greater than zero.
The typical values that may be used are 0, 0.001, 0.01, 0.1, and 0.5.

The Tension method

 It controls the stiffness of the surface according to the character of the

modeled phenomenon.

 It creates a less smooth surface with values more closely constrained by the
sample data range.

 The weight parameter defines the weight of tension. The higher the weight,
the coarser the output surface.

 The values entered must be equal to or greater than zero. The typical values
are 0, 1, 5, and 10.

14
Number of points identifies the number of points used in the calculation of each
interpolated cell. The more input points you specify, the more each cell is influenced
by distant points and the smoother the output surface. The larger the number of
points, the longer it will take to process the output raster (URL 3).

2.1.2.3. Kriging Interpolation

Natural data are difficult to model using smooth functions because normally random
fluctuations and measurement error combine to cause irregularities in sampled data
values. Kriging was developed to model those stochastic concepts (URL 3).

Kriging is a geostatistical method of spatial data interpolation that can be used to

limit the number of samples in eco-risk assessments. The method gets its name from
D.G. Krige, a South African mining engineer, who used the technique to more
accurately predict the extent of gold deposits in unsampled areas in 1963. Kriging is
an interpolation method that optimally predicts data values by using data taken at
known nearby locations. It can be both two-dimensional and three-dimensional.
Kriging is based on the assumption that the parameter being interpolated at a site is a
regionalized variable. A regionalized variable varies in a continuous manner spatially
so that data values from points nearer each other are more correlated. Data values
from widely separated points are statistically independent in kriging (Leadon, 2000).

Interpolating a surface from points using kriging in GIS software

 Kriging is a processor-intensive process. The speed of execution is dependent

on the number of points in the input dataset and the size of the search
window.

 The Semivariogram Properties dialog box has several models to choose from.
When the Kriging method is set to Ordinary, the available models are
Spherical, Circular, Exponential, Gaussian, and Linear. When the Kriging
method is set to Universal, the available models are Linear with linear drift
and Linear with quadratic drift.

 The optional output variance of prediction raster contains the kriging variance
at each output raster cell.

15
  Some input datasets may have several points with the same X,Y coordinates.
If the values of the points at the common location are the same, they are
considered duplicates and have no affect on the output. If the values are
different, they are considered 'coincident' points. Interpolation functions may
handle this data condition differently, in some cases using the first coincident
point encountered for the calculation, or sometimes the last. This may cause
some locations in the output raster to have different values than what you
might expect. The solution is to prepare your data by removing these
coincident points.

 Low values within the output variance of prediction raster indicate a high
degree of confidence in the predicted value. High values may indicate a need
for more data points (URL 3).

How Krige and Variogram work

Kriging is an advanced geostatistical procedure that generates an estimated surface

from a scattered set of points with z-values. Kriging in ArcObjects has two methods:
Krige and Variogram. The difference between them is they allow different levels of
control over the operation.

Kriging is a complex procedure that requires greater knowledge about spatial

statistics than can be conveyed in this command reference. Before using the kriging,
one should have an understanding of the fundamentals of kriging and have assessed
the appropriateness of your data for modeling with this technique.

Kriging is based on the regionalized variable theory that assumes that the spatial
variation in the phenomenon represented by the z-values is statistically homogeneous
throughout the surface.

The spatial variation is quantified by the semivariogram. The semivariogram is

estimated by the sample semivariogram, which is computed from the input point
dataset. The value of the sample semivariogram for a separation distance of h
(referred to as the lag) is the average squared difference in z-value between pairs of
input sample points separated by h. The sample semivariogram is calculated from the
sample data with the equation 2.1:

16
 (2.1)

where n is the number of pairs of sample points separated by the distance h.

Below is an example of a typical semivariogram (Figure 2.2):

Figure 2.2: Semivariance

Kriging offers two types of surface estimators: Ordinary Kriging and Universal
Kriging. Ordinary Kriging is represented by the Spherical, Circular, Exponential,
Gaussian, and Linear types. With these options, Kriging uses the mathematical
function specified with the semiVariogramType argument to fit a line or curve to the
semivariance data in the semivariogram.

Ordinary Kriging assumes that the variation in z-values is free of any structural
component. These five models are provided to ensure that the necessary conditions
of the variogram model are satisfied. These methods and conditions are discussed in
McBratney and Webster (1986). The variance is calculated based on the average
variance of all point pairs within each interval of the cell size. The variogram is then
fitted to the variance points using the Levenberg–Marquardt Method of nonlinear
least squares approximation. A minimum of three points (a variance value at three
distances) is required for the fit. By increasing the cell size, the number of sample
points per cell size interval will be increased, thereby providing enough data points
to estimate the semivariogram. Once the semivariogram is estimated, a smaller cell
size can be used in creating the actual output raster.

17
In some cases, one may suspect that the spatial variation in z-values in the data
contains local trends. Universal Kriging assumes that the spatial variation across the
surface also has a structural component referred to as drift. Drift is a systematic
change in the z-values in a particular direction. Universal Kriging assumes that the
spatial variation in z-values is the sum of three components: a structural component
(drift), a random but spatially correlated component, and random noise representing
the residual error. The structural component represents a constant trend over the
surface. The random noise is assumed to be spatially independent and have a normal
distribution. Once the structural effects have been accounted for, the remaining
variation is spatially homogeneous such that the z-value difference between input
sample points is merely a function of the distance between them as with Ordinary
Kriging (URL 2).

Here are some general issues when performing kriging:

 Make an assessment of the appropriateness of your data for use with the
kriging technique.

 Use with several different surface estimators that approximate your data.

 Examine the outSemiVariance in ArcMap. Click the ArcGIS Spatial Analyst

drop-down menu, point to Surface Analysis, and click Contour to create
contours of the variance. Look for areas where the selected model has high
variance values. More sample points may be needed in these locations.

 If you think your data has a local trend, use the Universal1 or Universal2
estimator and generate contours on the variance output. Repeat the process
using differently sized neighborhoods with both methods and observe the
changes in the variance. Choose the parameters that yield the best results
(URL 3).

2.1.2.4. Natural Neighborhood Interpolation

Natural neighbor interpolation finds the closest subset of input samples to a query
point and applies weights to them based on proportionate areas in order to interpolate
a value (Sibson, 1981). One of its names is Sibson or “area-stealing” interpolation.
Its basic properties are that it's local, using only a subset of samples that surround a

18
query point, and that interpolated heights are guaranteed to be within the range of the
samples used. It does not infer trends and will not produce peaks, pits, ridges or
valleys that are not already represented by the input samples. The surface passes
through the input samples and is smooth everywhere except at locations of the input
samples. Breaklines may be used, in the case of TIN to Raster interpolation, to
augment the surface, creating linear discontinuities where appropriate such as along
roadsides and water bodies. It adapts locally to the structure of the input data,
requiring no input from the user pertaining to search radius, sample count, or shape.
It works equally well with regularly and irregularly distributed data (Watson, 1992).

Interpolating a surface from points using Natural Neighbor in GIS

software

 The Natural Neighbor tool can efficiently handle large numbers of input
points. Other interpolators may have difficulty with large point datasets.

 Some input datasets may have several points with the same X,Y coordinates.
If the values of the points at the common location are the same, they are
considered duplicates and have no affect on the output. If the values are
different, they are considered 'coincident' points. Interpolation functions may
handle this data condition differently, in some cases using the first coincident
point encountered for the calculation, or sometimes the last. This may cause
some locations in the output raster to have different values than what you
might expect. The solution is to prepare your data by removing these
coincident points.

 If the cell center of the perimeter cells of the output raster fall outside
the convex hull (defined by the input points) then those cells will be assigned
NoData values. If an input point falls within one of these perimeter cells and
the cell center falls outside the convex hull, the cell will still be assigned a
value of NoData (URL 3).

2.2. Comparison of Spatial Interpolation Methods

The question is “Which interpolation is the best?”. There are several interpolation
methods used in spatial interpolation. It is needed to choose the optimal interpolation
method for our subject. Here are some former studies which are going to help to
determine our method:

19
Research 1:

In 2006, according to the research called “Comparison of interpolation methods for

mapping climatic and bioclimatic variables at regional scale” carried out in
Universit`a di Roma ‘La Sapienza’ it is pointed out that universal kriging with
external drift obtained the best performances for seventeen variables of the twenty-
one analysed, neural network interpolator has proven to be more efficient for three
variables and de-trended inverse distance weighting (D-IDW) for only one. Terrain
variables and geographical location have been used as predictors of the climate
variables: longitude, latitude, elevation, aspect, slope, continentality and estimated
solar radiation. Based on these results, the universal kriging estimates are used to
produce the climatic and bioclimatic maps aimed at defining the bioclimatic
envelope of species. (Attorre et al.)

Research 2:

Borga and Vizzaccaro (1996) focused on the ties of Kriging with a deterministic
interpolation procedure, known as multiquadratic surface fitting. When those two are
compared, it has been realized that kriging equations with a linear variogram model
are identical in form to equations of multiquadratic surface fitting with cone surfaces.
The issue of the accuracy of both estimators is discussed through a case study where
hourly rainfall maps of real storm events collected by radar provided the reference
rainfall. It is shown that kriging performs better at lower gauge density, while at
higher gauge density the accuracy of both estimators is similar.

Research 3:

Based on a research done in Arizona States Universit by Anderson, three methods of

interpolation (Spline, Inverse Distance Weighting (IDW) and Kriging) to be used on
air temperature data from the Phoenix Metropolitan area. The study identified the
best spatial interpolation method to use for the creation of continuous data for air
temperatures and Kriging performed better overall.

Research 4:

Pierre Goovaerts published a paper named “Performance comparison of

geostatistical algorithms for incorporating elevation into the mapping of
precipitation” which presents presents three geostatistical algorithms for
incorporating a digital elevation model into the spatial prediction of rainfall: simple

20
kriging with varying local means, kriging with an external drift, and colocated
cokriging. As a result, it has been obtained that ordinary kriging yields more accurate
predictions than linear regression when the correlation between rainfall and elevation
is moderate (less than 0.75 in the case study).

As mentioned above -based on former studies and researchs-, kriging method

performs beter overall while interpolating climatic, bioclimatic and hydrologic
variables.

2.3. Extrapolation

Extrapolation is basically estimating a value of a variable outside- a known range

from values- within a known range by assuming that the estimated value follows
logically from the known values. Using known or observed data to infer or
calculate values for unobserved times, locations or other variables outside a
sampled area. In the absence of data, extrapolation is a common method for
making predictions, but it is not always accurate. For example, based on observed
economic indicators, an economist can make predictions about the state of the
economy at a future time. These predictions may not be accurate because they
cannot take into account seemingly random events like natural disasters (URL 5).

To predict the values for locations outside the area

Geostatistical Analysts interpolates the values of the selected variable at all locations
that lie within the minimum spatial bounding box. The minimum spatial bounding
box is the smallest box that can be created to encompass all the input sample points.
However, this bounding box may not produce a map that exceeds the bounding box
is called extrapolation. The resulting geostatistical layer from an extrapolation will
cover the area that is already identified.

2.4.Accuracy Assesment

Spatial interpolation inevitably introduces errors. Predictions at unobserved points

are based on limited information and will only by shear chance be equal to the true
value. Interpolation errors affect the quality of the interpolated map and will
propagate through subsequent analyses, possibly leading to poor decisions. It is

21
therefore important to quantify the accuracy of the interpolated maps. Interpolation
accuracy assessment is routinely done in geostatistics, where the predicted value at
an unobserved point is accompanied by a probability distribution of the associated
interpolation error (Heuvelink et al., 2003).

As Price (2000) explained that the standard deviation is the average amount by which
scores in a distribution differ from the mean, ignoring the sign of the difference
and sometimes, it is defined as the average distance between any score in a
distribution and the mean of the distribution.

The standard deviation measures the spread of the data about the mean value. It is
useful in comparing sets of data which may have the same mean but a different
range. The Root Mean Square (RMS) error in GIS represents the difference between
the original control points and the new control point locations calculated by
the transformation process (URL 6).

Here is the formula of root mean square error shown in equation 2.4:

(2.4)

3. CASE STUDY

3.1. Case Study of Bulgaria

The data of Bulgaria, which includes the wind speed values measured at 10 and 100
meters (m/s) and energy measurements (W/m²), were collected at 39 meteorological
monitoring stations distributed in Bulgaria (See Figure 3.1). Stated data were
provided by The Black Sea Regional Energy Centre (BSREC), the Bulgarian partner
of EnvriroGRIDS project.

22
 Figure 3.1: Location of the stations in Bulgaria

3.1.1. Data of the Study

The existing data was vector maps of Bulgaria from ESRI world vector dataset and
tabular data of wind and solar measurement values. First of all relational database of
attribute data were formed by using Microsoft Access.

Table 3.1 : Database view of wind energy data

Location Wind speed (10m/s) Wind speed (100 m/s) Energy (W/m²)

AHTOPOL 4.63 8.60 39.78

BLAGOEVGRAD 5.39 9.51 18.25

BOTEV 9.61 13.28 90.01

BURGAS 6.46 10.49 63.75

CHERNIVRYH 9.41 13.37 90.43

23
CHIRPAN 4.58 7.68 24.96

DOBRICH 4.38 7.05 24.86

DRAGOMAN 4.98 9.36 44.58

ELHOVO 4.90 8.73 25.69

EMINE 7.44 12.13 78.44

KALIAKRA 6.88 11.77 90.22

KARNOBAT 4.79 7.51 16.58

KAZANLYK 4.98 9.15 25.33

KJUSTENDIL 4.54 7.52 12.13

KNEJA 5.26 8.46 26.47

KYRDJALI 6.14 9.56 35.00

LOM 5.30 8.58 31.31

LOVECH 4.84 7.08 21.60

MONTANA 5.04 8.40 28.96

MURGASH 8.99 14.85 106.53

MUSALA 7.96 12.36 92.33

NOVOSELO 5.43 8.93 18.56

ORJAHOVO 5.74 9.29 18.42

PAZARDJIK 5.40 7.62 19.73

PLEVEN 5.18 8.13 29.87

PLOVDIV 5.09 8.33 26.23

24
 RAZGRAD 4.74 7.67 32.90

ROJEN 4.36 7.21 27.97

RUSE 5.80 9.40 47.46

SANDANSKI 4.93 7.42 21.41

SHABLA 6.09 10.91 89.56

SILISTRA 5.26 8.45 30.53

SLIVEN 5.69 7.97 25.97

SOFIA 4.04 6.06 11.40

SVISHTOV 5.96 9.68 35.25

VARNA 5.75 8.46 40.56

VELIKO
TARNOVO 4.48 6.74 13.79

VIDIN 4.11 6.64 12.91

VRACA 4.92 7.20 19.19

As we can infer from the wind energy values of the station data almost fits to normal
distribution curve.

3.1.2. GIS Implementation

Both geometric and attribute data were imported ArcGIS 9.3 software and the
relations between geometric and attribute data were provided by using join tool
depending on relational database basis. Related attribute tables were joined by
matching “Location” coloumns of each table. Spatial Interpolation Methods were
then applied by using Spatial Analyst Extension of ArcGIS.

25
 Figure 3.2: A map of monitoring stations in Bulgaria

3.1.3. Determination of Spatial Interpolation Method

Three interpolation methods that are Inverse Distance Weighted, Kriging and Natural
Neighborhood were applied by using existing Bulgarian data. In this process, four of
the monitoring stations were determined as control stations and they were taken out
form the source data for testing the efficiency of interpolation methods. Data of the
remaining 35 monitoring stations were interpolated by using three spatial
interpolation methods of Spatial Analyst extension of ArcGIS software. Data of 35
stations were both interpolated and then extrapolated for each method. Finally the
interpolated raster surfaces were extracted depending on the country boundaries of
Bulgaria. Table 3.2 indicates the geographic coordinates of selected control stations.

26
 Table 3.2. Control Stations

Latitude Longitude Height

SOFIA 42,65 23,38 568

LOVECH 43,13 24,73 220

PLOVDIV 42,07 24,85 154

SLIVEN 42,67 26,32 259

Results of the interpolation methods were indicated in the following figures: Figure
3.2- Figure 3.7. The maps for accuracy assessment were obtained from 35 stations
with IDW, Kriging and Natural Neighbor interpolation.

Figure 3.2: Result of IDW interpolation method for wind speed data at 10 m.

27
Figure 3.3: Result of IDW interpolation method for wind speed data at 100 m.

Figure 3.4: Result of Kriging interpolation method for wind speed data at 10 m.

28
Figure 3.5: Result of Kriging interpolation method for wind speed data at 100 m.

Figure 3.6: Result of Natural Neighborhood interpolation method for wind speed
data at 10 m.

29
 Figure 3.7: Result of Natural Neighborhood interpolation method for wind speed
data at 100 m.

3.1.4. Accuracy Assessments

For assessing the accuracy of each interpolation method, values of control station
calculated using data of 35 stations were sampled from each interpolated raster
surfaces depending on the geographic coordinates of each station.

Accuracy assessment was done by comparing the measured values of each station
with estimated values. In this context errors at each station were first calculated and
then root mean square error of each method were determined depending on
calculated errors. Table 3.3 indicates measured and estimated wind speed values at
each control station.

Interpolation for 10m/s wind data

Following table shows the real values of 10 m/s wind data for selected stations.

30
 Table 3.3. The actual Wind Data at10 m for Control Stations

Estimated Values
Wind Speed
at 10m IDW Kriging Nat. Neigh.

SOFIA 4.040687 8.61485 6.010405 9.182572

LOVECH 4.844892 6.085177 5.789081 6.966969

PLOVDIV 5.093465 5.151019 6.075618 5.146211

SLIVEN 5.692505 5.261918 5.823363 4.764056

Errors and RMSEs at 10 m.were calculated for each method as indicated in Table 3.4

Table 3.4. Errors and RMSEs for each method

Natural
IDW Kriging Neighborhood

SOFIA -4.5742 -1.9697 -5.1419

LOVECH -1.2403 -0.9442 -2.1221

PLOVDIV -0.0576 -0.9822 -0.0527

SLIVEN 0.4306 -0.1309 0.9284

RMS 2.37960 1.19927 2.81988

Same processes were also applied by using wind speed data measured at 100m.
Results were presented in Table 3.5 and Table 3.6.

31
 Table 3.5. The actual Wind Data at 100 m for Control Stations

Wind Speed
Latitude Longitude Height 100m/s

SOFIA 42,65 23,38 568 6.062710

LOVECH 43,13 24,73 220 7.080118

PLOVDIV 42,07 24,85 154 8.331278

SLIVEN 42,67 26,32 259 7.968214

Errors and RMSEs were calculated for each method as indicated in Table 3.4;

Table 3.6. Errors and RMSEs for each method

Natural
IDW Kriging Neighborhood

SOFIA -6.59741 -3.490193 -7.50438

LOVECH -2.25282 -2.028003 -3.15488

PLOVDIV 0.13010 -1.228566 0.21617

SLIVEN -0.65680 -1.367887 0.02118

RMSE 3.501760 2.217811 4.07174

As seen on both comparison tables, Kriging perfomed better over all, with RMSE of
2.217811. However, the deviations are too much. This results come from the position
of points because some points wind’s values are much bigger than average.
Especially choosing Sofia as a control station was not a good idea when its location
is considered. Because Sofia is very close to Chernivryh station and this point was
also affecting the interpolation results. So, another solution idea which was to take
Sofia out and put another station came out. To ensure the homogeneous distribution

32
and having the average wind energy values, Razgrad was chosen. Here are the new
control stations:

Table 3.7 The actual Wind Data at10 m for Second Control Stations

Wind Speed
Latitude Longitude Height 10m/s

LOVECH 43,13 24,73 220 4.844892

PLOVDIV 42,07 24,85 154 5.093465

RAZGRAD 43,52 26,52 344 4.740668

SLIVEN 42,67 26,32 259 5.692505

Following table shows the real values of 100 m/s wind data for selected stations.

Table 3.8 The actual Wind Data at 100 m for Second Control Stations

Estimated Values
Wind Speed
at 10m IDW Kriging Nat. Neigh.

LOVECH 4.844892 6.085177 5.789081 6.966969

PLOVDIV 5.093465 5.151019 6.075618 5.146211

RAZGRAD 4.740668 4.740738 5.370565 4.763105

SLIVEN 5.692505 5.261918 5.823363 4.764056

Errors and RMSEs were calculated for each method as indicated in Table 3.4;

33
 Table 3.9 Errors and RMSEs for each method

Error in Error in Natural

Station Name Error in IDW Kriging Neighborhood

LOVECH -1.24029 -0.94419 -2.12208

PLOVDIV -0.05755 -0.98215 -0.05275

RAZGRAD -0.00007 -0.62990 -0.02244

SLIVEN 0.43059 -0.13086 0.92845

RMSE 0.65708 0.75333 1.15850

Final Comparison of three methods,

As seen on Table 3.9, RMSEs of each three methods are substantially decreased
when compared to previously selected control stations and changing one control
station affected the results. The least RMSE is 0.65708 which belongs to IDW
method. The errors of Kriging and IDW are pretty close to each other. As a result,
IDW performed better than the other two interpolation methods. This outcome can be
connected to a cause that IDW gives better results, if the data fits to normal
distribution.

3.2. Case Study of Turkey

The data of Turkey was collected from 41 meteorological monitoring stations and the
spatial distribution of the monitoring stations and general view of the study area are
depicted in the Figure 3.8.

34
 Figure 3.8: Location of the monitoring stations in Turkey

3.2.1 Data of the Study

The existing data was vector maps of Turkey from ESRI world vector dataset and
tabular data of wind measurements values. The tabular data was derived from the
raw data obtained from General Directorate of Meteorology. The derived data
includes average wind speed and average temperature of 15th day of each month in
the year of 2003.

Table 3.10: Database view of wind speed and temperature

Location Wind speed (10m/s) Average_Temperature (Celsius)

Afyon 2,158333 0,566667

Aksaray 1,700000 -4,000000

Amasya 1,183333 1,266667

Ankara 2,140000 -0,120000

Ardahan 1,691667 0,391667

Artvin 1,866667 0,891667

35
 Bartin 1,275000 1,016667

Bayburt 2,041667 1,575000

Bilecik 2,457143 0,742857

Bolu 1,383333 0,075000

Bursa 2,291667 1,950000

Cankiri 1,008333 3,608333

Corum 1,833333 1,958333

Duzce 1,016667 4,475000

Erzincan 1,908333 1,191667

Erzurum 2,175000 -1,625000

Eskisehir 2,500000 -2,200000

Giresun 1,300000 1,733333

Istanbul 2,340000 0,420000

Gumushane 2,191667 1,325000

Karabuk 1,058333 4,025000

Kastamonu 1,508333 -0,258333

Kayseri 1,108333 1,116667

Kirikkale 2,166667 1,433333

Kirklareli 0,958333 4,491667

Kirsehir 2,250000 0,383333

Kocaeli 1,083333 3,800000

36
 Konya 2,155556 0,022222

Kutahya 1,175000 1,175000

Nevsehir 2,108333 0,033333

Ordu 1,850000 1,525000

Rize 1,225000 1,133333

Sakarya 1,050000 2,083333

Samsun 2,100000 0,816667

Sinop 1,066667 1,475000

Sivas 1,108333 1,433333

Tokat 2,108333 1,000000

Trabzon 2,266667 -0,500000

Usak 2,209091 0,400000

Yozgat 2,350000 0,275000

Zonguldak 2,066667 1,850000

3.2.2. GIS Implementation

As the first step of the implementation a relational database of attribute data were
formed by using Microsoft Access. Both geometric and attribute data were imported
ArcGIS 9.3 software and the relations between geometric and attribute data were
provided by using join tool depending on relational database basis. Related attribute
tables were joined by matching “Location” columns of each table. Spatial
Interpolation Methods were then applied by using Spatial Analyst Extension of
ArcGIS.

37
3.2.3. Determination of Spatial Interpolation Method

Four interpolation methods that are Inverse Distance Weighted, Kriging, Spline and
Natural Neighborhood were applied by using existing Turkey data. In this process,
four of the monitoring stations were determined as control stations and they were
taken out form the source data for testing the efficiency of interpolation methods.
Data of the remaining 37 monitoring stations were interpolated by using four spatial
interpolation methods of Spatial Analyst extension of ArcGIS software. Data of 37
stations were interpolated. Finally the interpolated raster surfaces were extracted
depending on the country boundaries of Turkey. Table 3.11 indicates the geographic
coordinates of selected control stations.

Table 3.11 : Control Stations

Latitude Longitude Height

CORUM 40,55 34,94 776

TOKAT 40,33 36,56 611

BAYBURT 40,25 40,23 1584

BILECIK 40,14 29,98 539

Results of the interpolation methods were indicated in the following figures: Figure
3.9- Figure 3.12. The maps obtained by 37 stations for accuracy assessment with four
different interpolation.

38
 Figure 3.9: Result of IDW interpolation method for wind speed data at 10 m.

Figure 3.10: Result of Kriging interpolation method for wind speed data at 10 m.

39
Figure 3.11: Result of Natural Neighborhood interpolation method for wind speed
data at 10 m

Figure 3.12: Result of Spline interpolation method for wind speed data at 10 m

40
3.2.4. Accuracy Assessments

For assessing the accuracy of each interpolation method, values of control station
calculated using data of 37 stations were sampled from each interpolated raster
surfaces depending on the geographic coordinates of each station.

Accuracy assessment was done by comparing the measured values of each station
with estimated values. In this context errors at each station were first calculated and
then root mean square error of each method was determined depending on calculated
errors. Table 3.12 indicates measured and estimated wind speed values at each
control station.

Table 3.12: The actual Wind Data at 10 m for Control Stations

Estimated Values
Wind Speed
at 10m
IDW Kriging Nat. Neigh. Spline

CORUM 1,833333 1,747548 1,720682 1,663268 1,671279

TOKAT 2,108333 1,454606 1,564339 1,31356 1,324309

BAYBURT 2,041667 1,894895 1,740354 1,912005 2,124031

BILECIK 2,457143 1,640164 1,705422 1,589864 1,750214

Errors and RMSEs were calculated for each method as indicated in Table 3.13;

Table 3.13: Errors and RMSEs for each method

IDW Kriging Natural Neigh. Spline

CORUM 0,085785 0,112651 0,170065 0,162054

TOKAT 0,653727 0,543994 0,794773 0,784024

BAYBURT 0,146772 0,301313 0,129662 -0,08236

41
 BILECIK 0,816979 0,751721 0,867279 0,706929

RMSE 0,530027 0,491043 0,597823 0,535604

Final Comparison of four methods,

As it seen from Table 3.13 it is clear that Kriging is the best interpolation method for
Turkey. Accuracy of those four interpolation methods in an order as below.
KRIGING>IDW>SPLINE>N.N.

3.3. Mapping

Mapping process was done for both Bulgaria and Turkey by using ArcGIS software.
Convenience of colour alternatives and arranging the legend of maps were taken into
consideration in order to increase the comprehensibility of data.

Two different maps were created for Bulgaria. Maps for wind speeds at 10 m. and
wind speeds at 100m. are shown in Figure 3.13 and Figure 3.14. IDW method was
chosen for mapping because of its accuracy and visiual quality.

Figure 3.13: Wind Speeds at 10 m. Map of Bulgaria

42
 Figure 3.14: Wind Speeds at 100 m. Map of Bulgaria

There are two maps which were created for Turkey. Wind speeds at 10 m. and
average temperature maps are shown in Figure 3.15-3.16. Despite the fact that
Kriging interpolation was found the most accurate method, IDW interpolation
method was used while creating maps because of visual preference.

43
Figure 3.15: Wind Speeds at 10 m. Map of Turkey

Figure 3.16: Average Temperature Map of Turkey

44
 4. CONCLUSION

Based on the researches done, the idea was that kriging performs better over all the
interpolation methods. For Turkey data, four different interpolation methods (IDW,
Kriging, Natural Neighbor, Spline) are applied and as it explained by the given
tables and data results, it is clear that Kriging is the most accurate method for Turkey.
However for Bulgaria data, we found out that Inverse Distance Weighted
Interpolation gave us the best and more accurate results while mapping the wind
energy data. We can refer from here that interpolation methods may differ from data
to data and the data distribution related to study area.

Kriging requires a technical knowledge before it is applied because it is much more

complicated than the other interpolation methods. Some parameters- like variogram
type or interpolation type- have to be determined before the interpolation process is
carried out. The result changes if the values are taken as default values or pre-
determined. To take the default values of ordinary kriging in this study may be the
reason why kriging was not the best method as it was expected to be.

It has already known that Inverse Distance Weighted method works better for normal
distributed data and it has been tested and confirmed in Bulgaria data.

For the accuracy assessment, changing the control stations affects the accuracy of the
data produced and it may also change the accuracy order of the interpolation
methods. However, testing the interpolation methods with the same stations at
different heights does not change the accuracy order (see Table 3.4 and 3.6).

Mapping has been done based on the data and researches in order to set out wind-
energy potentials and to introduce different interpolation methods. Considering the
importance of energy, Turkey and Bulgaria wind maps will create a big advantage for
new investments to produce energy.

45
REFERENCES

Anderson, S., n.d., “An Evaluation of Spatial Interpolation Methods on Air

Temperature in Phoenix, AZ”, Departmant of Geography, Arizona State
University, Tempe, Arizona, USA. Retrieved from:

Attorre, F., Alfo, M., De Sanctis, M. (and others), 2007. “Comparison of

Interpolation Methodsfor mapping clomatic and bioclimatic variables at
regional scale”, International Journal of Climatology (DOI: 10.1002/joc.
1495).

Borga M., Vizzaccaro A., 1997. “On the interpolation of hydrologic variables:
formal equivalence of multiquadratic surface fitting and kriging”,
Dipartimento Territorio e Sistemi Agro-Forestali, Agripolis, Università di
Padova, 35020 Legnaro, Padova, Italy.

Bourke, P., 1999. Interpolation Methods, University of Western Australia,

“Interpolation Methods”.

Burrough, P.A., 1986. Principles of Geographical Information Systems for land

Resources Assessment, Clarendon, Oxford. See Chapter 8.

Davis, J.C., 1986. Statistics and Data Analysis in Geology, 2nd edition, Wiley, New
York. (Also see the first, 1973, edition for program listings.)

Goovaerts, P., 1999. “Performance comparison of geostatistical algorithms for

incorporating elevation into the mapping of precipitation”, The University of
Michigan, Department of Civil and Environmental Engineering, Ann Arbor,
MI, U.S.A.

Heuvelink, G., B., M., Edited by Caetano, M., Painho, M., 2003. “Incorporating
Process knowledge in spatial interpolation of environmental variables”,
Environmental Sciences Group, Wageningen University and Research
Centre, 6700 AA Wageningen, The Netherlands.

Johnston, K., Ver Hoef, V.F., Krivoruchko, K., Lucas, N., 2003. “Using ArcGIS
Geostatistical Analyst”, 7th International Symposium on Spatial Accuracy
Assessment in Natural Resources and Environmental Sciences. United States
of America.

46
Kahaner, D., Moler, C. and Nash, S., 1989. “Numerical Methods and Software”,

Prentice Hall, Englewood Cliffs, NJ, USA.

Lang C. n.d., “Krigining Interpolation”, Dept. of Computer Science, Cornell

University, Ithaca, New York, United States.

Leadon, J. C., 2000, “Kriging in Eco-Risk Assessments”, Naval Facilities

Engineering Command San Diego, CA.,92101-3327.

McBratney, A.B., and R. Webster, 1986. Choosing Functions for Semi-variograms

of Soil Properties and Fitting Them to Sampling Estimates. Journal of Soil
Science 37: 617-639.

Nguyen, T., 2009. “Project 5 Interpolation From Simple to Advanced”, Penn State

University, World Campus – GEOG 586.

Price, I., Dr., 2000. “Chapter 4: Analysing the Data Part II : Descriptive

Statistics”, University of New England, Armidale, NSW, 2351.

Sibson, R., 1981. "A Brief Description of Natural Neighbor Interpolation," Chapter

2 in Interpolating multivariate data, John Wiley & Sons, New York, pp. 21-

36.
Waters, N.M., 1988. "Expert Systems and Systems of Experts," Chapter 12 in W.J.

Coffey, ed., Geographical Systems and Systems of Geography, University of

Western Ontario, London, Ontario.

Watson, D., 1992. “Contouring: A Guide to the Analysis and Display of Spatial

Data”. Pergamon Press, London.

URL 1 : http://www.geog.ubc.ca/courses/klink/gis.notes/ncgia/u40.html#SEC40.1
Offical website of UC Santa Barbara Department of Geography.
Accessed on 18.05.2011.

URL 2 : http://help.arcgis.com/en/arcgisdesktop/10.0/help/ Offical website of ESRI

ArcGIS Resource Center. Accessed on 17.05.2011.

URL 3 : http://support.esri.com/en/ Offical website of ESRI ArcGIS Desktop 9.3

Online Help. Accessed on 18.05.2011.

47
URL 4 : http://www.uiowa.edu/~geog/health/index9.html Offical website of
Geography Department of the University of IOWA. Accessed on
19.05.2011.

URL 5 : http://support.esri.com/en/knowledgebase/Gisdictionary/term/extrapolation
Offical website of GIS Dictionary, ESRI. Accessed on 20.05.2011.

URL 6:
HTTP://ADS.AHDS.AC.UK/PROJECT/GOODGUİDES/GİS/SECT72.HTML
Offical website of GIS Guide to Good Practice. Accessed on 20.05.2011.

URL 7: http://www.envirogrids.net/index.php?option=com_content&view=article
&id=1&Itemid=3. Offical website of EnviroGRIDS. Accessed on
20.05.2011.

48
 CURRICULUM VITAE

Merve KESKIN was born in Istanbul in 1988. She completed her elementary
education at Goztepe Pansiyonlu Elementary School. She achieved an mid-
intermediate English language degree in Marmara Cagdas Bilimler Vakfi in 2000.
Her high school education was at Istek Semiha Sakir Private High School.

In 2006-2007 academic year, she started to her undergraduate education at the

department of Geomatics Engineering in Istanbul Technical University. She achieved
a certificate of A2 level German language in Goethe Institute in 2010. She studied
one semester in Hannover, Germany with ERASMUS programme. She has very well
communication skills in English and German in written and oral manner.

Also, she is a student member of HKMO (Chamber of Survey and Cadastre

Engineers). She attended IGSM (International Geodesy Students Meeting) in
Valencia, Spain in 2008; in Zurich, Switzerland in 2009; in Zagreb, Croatia in 2010
and in Newcastle, United Kingdom in 2011. She has good computer skills and she is
keen on literature.

Koray ÖZDOĞU was born in İzmit in 1988. He completed his elementary school at
Şehit Namık Tümer Elementary School as an honour student. Then he started to
İzmir Anatolian High School. At 2006 he was graduated and went to United States
for language course for one year. He succeeded to get TOEFL IBT score and passed
the preparation year of university. Thus he is very well with using English.

In 2008, he started undergraduate education at the department of Geomatics

Engineering in Istanbul Technical University. He studied in Greece one semester
with Erasmus Program.

As a social manner he is a member of HKMO (Chamber of Survey and Cadastre

Engineers). He is also member of JFK (Geodesy and Fotogrammetry Club).

49

View publication stats