**2.1 Systemic methodology**

278 Urban Development

In the majority of the cases, when a coastal residential unit reaches the critical population threshold, the coastal urban concentration leads to the formation of harbors (Polyzos, 2011; Sekovski et al., 2011). Since harbors exist in most coastal cities, maritime transport and the related economic accessory facilities are particularly significant factors for their development. The harbor-based economies of coastal cities, suggest an engine of economic growth, which can be expressed in variable ways, such as composition of employment centers, attraction of investment and trade and creation of production and market places for commodities and consumption (Li, 2003). Of course, the seaside location privilege diachronically entailed the risk of attacks in the history, such as piracies (Birnie, 1987; Anderson, 1995). Today coastal cities suggest a vital source of income for national economies (Li, 2003), fact that is obviously observed even in the case of Greece (Polyzos et

The environmental facile of coastal cities, suggests also a powerful developmental axis. Many coastal cities are located and being developed at places where additional (to seaside) physical facilities, energy source and raw material deposits exist (Decker et al., 2002), factors that are supposed to be essential for the reinforcement of the coastal city's economy. For example, the existence of water supplies and arable lands consist considerable criteria for the location of cities in general. Many coastal cities were either located (when possible) near to river endings, where potable and irrigation water supplies are plentiful so as their hydroenergy exploitation is more accessible. The deltaic end of the Greek "Evros" river represents a characteristic paradigm for the coastal city of "Alexandroupolis" environmental vantage. Regarding raw material deposits, the coastal city of "Kavala" in Greece, which is located in a geographic position plenty of marble deposits, suggest an indicant example. Environmental amenities of coastal cities also operate as an axis of cultural development and education, as an absorbent mechanism of surplus rural population, as well as the nursery for civic spirit and social harmony (Li, 2003). History of Greece can indicate various paradigms of coastal cities (suggestively at regions of "Attica", "Ionian Islands", "Crete", "Dodekanisa" etc.) that

The touristic facility constitutes a fundamental developmental factor to coastal cities and this statement seems to be more significant in the case of Greece, by the time that the country posses a coastline greater than 10.000 km (Cori, 1999). Nowadays, the coastal cities dynamics are mainly directed by the developmental axis of touristic utilization (Cori, 1999; Miller & Auyong, 1991), revealing a considerable amount of profitable potentials for the economy of a country. The touristic development, considered as a respond variable, on depends on other economic variables, such as the increase of peoples' leisure time (at least in developed countries), the communication and transportation improvement (Sekovski et al., 2011), the environmental, cultural and coastal attraction etc., which suggest motivations and render coastal areas to comprise obvious spatial destinations for the conduction of the

The concept of coastal tourism includes, in its definition, the full range of tourism, leisure, and recreationally oriented activities that take place in the coastal zone and in the offshore coastal waters (Hall, 2001). Coastal tourism is considered one of the faster evolving forms of contemporary tourism, and, in the case of Greece, it seems to be the most promising developmental potential that the country should exploit, in order to overpass its modern economic problems. Moreover, considering the above in a larger scale framework, Greece

consisted cultural centers and cores of science and arts development.

al., 2011).

touristic procedure.

The initial approach, into the analysis of coastal cities developmental performance, was based on a systemic point of view. The term "systemic" has a wide conceptual range and refers to this class of methodologies that aspires to model, on a theoretical basis, the structural components of the study object and to describe its operations, under the perspective that it suggests a subsystem of a wider systemic environment. In the case of this paper, the study object refers to a manifold of chosen Greek coastal cities, which are distributed over the geographic area of Greece and suggest a spatial system. The total of coastal cities is grouped, by a geographical division, into administrative clusters, which suggest the Greek prefectures, as defined before the act of "Kallikratis" (Act 3852/2010). Consequently, the spatial units that are studied in this paper refer, on one hand, to each coastal city (spatial monad) and, on the other hand, to the pre-"Kallikrateian" Greek prefectures (spatial units consisting groups of spatial monads).

The theoretic framework of the systemic approach was inspired from the Drivers-Pressures-State-Impacts-Responses (DPSIR) model (Sekovski et al., 2011). The DPSIR model suggests a chain framework of (the five title) concepts that describe, on a step (discrete) way, the inputoutput mechanism of a socio-economic phenomenon with its geographic dimensions. Given a short description, the Drivers (motivation of acts) lead to pressures on the geographical system, pressures that result to a state (balance status). The conducted operations, targeting to achieve the balance, may have environmental and economic impacts on the wider including system. Finally, the Responses present a set of societal and policy makers' prioritizations, in order to reduce the undesired impacts and, in general, to improve the performance of the system. The schematic framework of the DPSIR model, adopted to the case of Greek coastal cities and to the needs of this study, is presented at figure 1.

### **2.2 Analytic approach**

The analytic approach suggests a supplementary or further analysis of the systemic modeling, which was presented above. It targets to describe, under a quantitative perspective, the parts that synthesize the study system and to quantify the systemic mechanisms. Regarding the case of Greek coastal cities, the quantitative tools that are used for the research derive from the scientific sectors of Spatial Analysis, Regional Economics

The Evolution and Spatial Dynamics of Coastal Cities in Greece 281

*CF c c c dc i n*

The Theil index constitutes a statistic function for inequality measurement analysis. The index appertains in the entropy measures family and has a comparative utility. The index's

> 1 <sup>1</sup> ln *<sup>n</sup> i i*

whereas *xi* stands for the arithmetic value of the observation *i* for the characteristic (variable) *Χ* (*i*=1,…,*n*), *lnx* is the napierian logarithmic function and *μ* the mean of the *n* observations *xi* (i=1,…,n). Theil index ranges into the interval 0 < *Τ* < ln*n* (Polyzos, 2011). The Theil index is considered useful in order to reveal the cities that presented the greatest variations in their

At the following analysis the statistical methods of bivariate correlations and one-sample Ttest are used. The bivariate correlations procedure (Norusis, 2004) computes Pearson's correlation coefficient with their significance levels. The linear correlation formula is a measure of linear association that measures the level, in which two variables are related. The

> (,) () () *Cov X Y*

Where, *Cov*(*X*,*Y*) stands for the covariance of the variables *X*, *Y* and *Var*(*X*), *Var*(*Y*) for the

The one-sample T-test procedure (Blalock, 1972; Hays, 1981; Norusis, 2004) tests whether the mean of a single variable differs from a specified theoretic value. The procedure tests the difference between the sample mean and the known or hypothesized value and specifies the level of confidence for the difference. This test is hence used to compare the means between two samples, the coastal and the terrestrial, by assuming that the hypothesized value of the

The purpose of this test in the case of Greek coastal cities is to compare some characteristics that appear in the coastal and the terrestrial prefectures and to draw conclusions. The

*DX X* 1 2 is the difference of the two means (coastal and terrestrial) and *SD* is the

mathematic formula of the t-test statistical function is expressed by the ratio

mathematic formula of Pearson's Bivariate Correlations is shown at relation (4),

*r*

test regards the mean value of the second sample.

*i x x <sup>T</sup> n* 

*n i*

mathematical formula is shown at equation (3) (Tsiotas and Polyzos, 2011).

**2.2.2 Regional econometric tools** 

chronological evolution.

**2.2.3 Statistical tools** 

respective variations.

1 1, <sup>1</sup> { ,..., } min , 1,...,

(3)

*Var X Var Y* (4)

*i ij i j ij*

(2)

*D <sup>D</sup> <sup>t</sup>*

*<sup>S</sup>* where

*i*

*w* 

*w dc*

*n n*

and Statistics. The analytic tools, which are used in this study, regard some common Spatial Analytic (Mitchell, 2005), Regional Econometric (Polyzos, 2011; Tsiotas and Polyzos, 2011) and Statistical measures (Blalock, 1972; Hays, 1981; Norusis, 2004) and are described briefly below.

Fig. 1. The DPSIR framework scheme for the case of Greek coastal cities

#### **2.2.1 Spatial analytic tools**

The Weighted Mean Center (WMC) (Mitchell, 2005) suggests a plane point that has coordinates (*x*, *y*) the average calculation of the coordinates of all the features in the study area, as shown at relation (1), where *xi*, *yi* is the symbolism of the coordinates and *wi* of the weights of a point *I* (Polyzos et al., 2011).

$$\text{VMCC} = (\overline{X}, \overline{Y}) \left| \overline{X} = \frac{\sum\_{i=1}^{n} w\_i \cdot x\_i}{\sum\_{i=1}^{n} w\_i}, \overline{Y} = \frac{\sum\_{i=1}^{n} w\_i \cdot y\_i}{\sum\_{i=1}^{n} w\_i} \right. \tag{1}$$

WMC is a useful measure for tracking changes in a spatial distribution or for comparing the distributions of different types of features. In this study WMC is used to compare the spatial distributions of the Greek coastal cities appearance for the decades 1961 up to 2001. The WMC's shifting suggests a five step trail (WMC1961, WMC1971, WMC1981, WMC1991, WMC2001,) and is expected to reveal the demographic transformation and the potentials of the coastal cities in Greece.

Central Feature (CF) tool identifies the most centrally located feature in a point, line, or polygon feature class. Distances from each feature to every other feature in the dataset are calculated (weighted) and summed (Mitchell, 2005). The mathematical formula of the central feature is related to the closeness centrality formula (Wang et al., 2011) and is shown at relation (2). The use of this tool is expected to disclose the most vital Greek coastal city in terms of population and centrality (Polyzos et al., 2011).

$$\text{CF} = c \in \{c\_1, \dots, c\_n\} \left| \min \left( dc\_i = \frac{\sum\_{i=1}^n w\_i \cdot \sum\_{j=1, i \neq j}^n dc\_{ij}}{w\_i} \right) \right|, i = 1, \dots, n \tag{2}$$

#### **2.2.2 Regional econometric tools**

280 Urban Development

and Statistics. The analytic tools, which are used in this study, regard some common Spatial Analytic (Mitchell, 2005), Regional Econometric (Polyzos, 2011; Tsiotas and Polyzos, 2011) and Statistical measures (Blalock, 1972; Hays, 1981; Norusis, 2004) and are described briefly

Fig. 1. The DPSIR framework scheme for the case of Greek coastal cities

The Weighted Mean Center (WMC) (Mitchell, 2005) suggests a plane point that has coordinates (*x*, *y*) the average calculation of the coordinates of all the features in the study area, as shown at relation (1), where *xi*, *yi* is the symbolism of the coordinates and *wi* of the

(,) ,

WMC is a useful measure for tracking changes in a spatial distribution or for comparing the distributions of different types of features. In this study WMC is used to compare the spatial distributions of the Greek coastal cities appearance for the decades 1961 up to 2001. The WMC's shifting suggests a five step trail (WMC1961, WMC1971, WMC1981, WMC1991, WMC2001,) and is expected to reveal the demographic transformation and the potentials of

Central Feature (CF) tool identifies the most centrally located feature in a point, line, or polygon feature class. Distances from each feature to every other feature in the dataset are calculated (weighted) and summed (Mitchell, 2005). The mathematical formula of the central feature is related to the closeness centrality formula (Wang et al., 2011) and is shown at relation (2). The use of this tool is expected to disclose the most vital Greek coastal city in

*WMC X Y X Y*

1 1

*i i n n i i i i*

*n n*

1 1

*ii ii*

(1)

*wx wy*

*w w*

below.

**2.2.1 Spatial analytic tools** 

the coastal cities in Greece.

terms of population and centrality (Polyzos et al., 2011).

weights of a point *I* (Polyzos et al., 2011).

The Theil index constitutes a statistic function for inequality measurement analysis. The index appertains in the entropy measures family and has a comparative utility. The index's mathematical formula is shown at equation (3) (Tsiotas and Polyzos, 2011).

$$T = \frac{1}{n} \sum\_{i=1}^{n} \frac{\mathbf{x}\_i}{\mu} \cdot \ln \frac{\mathbf{x}\_i}{\mu} \tag{3}$$

whereas *xi* stands for the arithmetic value of the observation *i* for the characteristic (variable) *Χ* (*i*=1,…,*n*), *lnx* is the napierian logarithmic function and *μ* the mean of the *n* observations *xi* (i=1,…,n). Theil index ranges into the interval 0 < *Τ* < ln*n* (Polyzos, 2011). The Theil index is considered useful in order to reveal the cities that presented the greatest variations in their chronological evolution.

#### **2.2.3 Statistical tools**

At the following analysis the statistical methods of bivariate correlations and one-sample Ttest are used. The bivariate correlations procedure (Norusis, 2004) computes Pearson's correlation coefficient with their significance levels. The linear correlation formula is a measure of linear association that measures the level, in which two variables are related. The mathematic formula of Pearson's Bivariate Correlations is shown at relation (4),

$$r = \frac{Cov(X, Y)}{\sqrt{Var(X) \cdot Var(Y)}} \tag{4}$$

Where, *Cov*(*X*,*Y*) stands for the covariance of the variables *X*, *Y* and *Var*(*X*), *Var*(*Y*) for the respective variations.

The one-sample T-test procedure (Blalock, 1972; Hays, 1981; Norusis, 2004) tests whether the mean of a single variable differs from a specified theoretic value. The procedure tests the difference between the sample mean and the known or hypothesized value and specifies the level of confidence for the difference. This test is hence used to compare the means between two samples, the coastal and the terrestrial, by assuming that the hypothesized value of the test regards the mean value of the second sample.

The purpose of this test in the case of Greek coastal cities is to compare some characteristics that appear in the coastal and the terrestrial prefectures and to draw conclusions. The mathematic formula of the t-test statistical function is expressed by the ratio *D <sup>D</sup> <sup>t</sup> <sup>S</sup>* where

*DX X* 1 2 is the difference of the two means (coastal and terrestrial) and *SD* is the

The Evolution and Spatial Dynamics of Coastal Cities in Greece 283

coastal cities manifold, the available decennial sub-manifolds of the Greek coastal cities are presented as pie charts, for the corresponding coastal cities, to the map of figure 2. During the study period, the Greek urban population coverage changed 29,5%, from 56,21% in 1961 to 72,79% in 2001. This fact implies, firstly, the structural changes that the Greek economic model was subjected (the meanwhile period) to, which suggests the country's transposition from the agricultural-based economy model (primary sector) to the services provision

Secondly, the meanwhile urban growth can be related to the phenomenon of agglomeration that benefited the two Hellenic metropolitan cities, "Athens" and "Thessaloniki", with respect to the trends observed in the wider Mediterranean level (Cori, 1999). The growing Greek coastal cities coverage of the total country's population in 1961 reaches a coverage of 16,42%, whereas in 2001 this percentage extended to 19,43%. The map of figure 2 also depicts the geographical distribution of the Greek coastal cities population growth. This map is considered useful in the level that it illustrates some geographical formations (or clusters) of this population growth. It seems worth telling that the greatest coastal concentrations in the Greek territory appear into places that do not present morphology of open (non-curved) coast lines, but in those which are attributed with physical protection.

This fact comes to an agreement with the previous historical placement.

Fig. 2. Comparative pie charts of the coastal cities population distribution for the period

Figure 3 maps the spatial locations of the Greek Geographical Mean Center (GMC) and the group of Weighted Mean Centers (WMC), which were calculated for the period 1961-2001. This map also depicts the geographical locations of the two Greek metropolitan cities

economy model (tertiary sector).

1961-2001.

standard error of the difference. The one-sample T-test process was considered easier to apply than the corresponding ANOVA, which is also commonly used to compare sample means since the samples are compared in pairs. The constraint of the T-test method and the rest common linear ones assumes that the data is normally distributed, especially with respect to skewness. Consequently, outlying values should be carefully checked and the use of boxplots is applied in order to manage these cases.
