Tipping the scales G U

(1)

1

U

NIVERSITY OF

G

RONINGEN

M

ASTER’S

T

HESIS

Tipping the scales

A multi-scalar approach on settlement location modelling

Author:

Alexandra Katevaini

Supervisor:

dr P.M. van Leusen

A thesis submitted in fulfilment of the requirements for the degree of Master of Arts

in the

Research Master’s Programme in Archaeology Faculty of Arts

(2)

2

Acknowledgements

I would like to express my deep gratitude to dr. P.M. van Leusen, my research supervisor, for his patient guidance, encouragement, and number of hours he has spent teaching me the technical skills of GIS applications and helping me with the completion of the thesis.

I would also like to extend my thanks to everyone at the Groningen Institute of Archaeology for their understanding to have me working in the computer facilities during the pandemic of 2020.

My grateful thanks are also extended to prof. dr. P.D. Jordan, who is the second reader of this thesis.

I wish to acknowledge the advice and help provided by prof. dr. Sofia Voutsaki, prof. dr. Peter Attema and dr. C. G. Williamson during the entirety of my studies in Groningen.

Finally, I wish to thank my family, my tutors in the University of Athens and my friends for their support and encouragement throughout my study.

(5)

5

1. Introduction

The term ‘scale’ can describe tangible objects, abstract ideas and size relations in space and time. Apart from everyday life uses, the word has different meanings within different sciences or dissimilar meanings within the same science. In Geography, scale as a term and its effects on the research are examined very closely as altering, facilitating, or troublesome factors (Smith 2000, 724; Gregory 1994, 544; Rogers et al. 2013; Gregory et al. 2009; Watson 1978). With the increase in research that produces spatial data, not only in Geography but also in Biology, Archaeology, and Epidemiology to name a few, collective research attempts encounter the problem of incorporating data with different resolution without losing or altering the information (Gotway and Young 2002). Lately, the importance of scales and their potential problems are realised in archaeology (Haggis 2006). Due to the nature of archaeology as a study field and the data available, the research interest or the process of collecting data usually defines the spatial scale. The spatial scale, meaning the size of the research area, can potentially affect the methodological tools and the results they produce. As a result, many types of research that examine spatial relations and use tools such as settlement location models are criticised on the size of the sample they used. The exact effects of scale sometimes become more profound than the research interest or understanding of archaeologists. After identifying the problem in that scales case on the analysis of spatial pattern, specific statistic tools can counteract the effect (Bevan and Conolly 2006; Palmisano 2014). However, there is no systematic research on the direct effect of the scales in other tools used for interpreting spatial relations. This type of research would be the first step in understanding the limitations and, hopefully, the utilisation of tools to overcome problems caused by spatial scale.

A standard method to study the correlations between settlement behaviour and environmental and social variables is Settlement Location Modelling (SLM). SLMs analyses point patterns of settlements recognised and dated by archaeological survey and excavation based on the correlation of variables imposed by exogenous environmental influences (referred as ‘first-order effects’), and social variables created in local interaction between settlements (known as ‘second-order effects’). Those variables are examined by multivariate analysis, allowing the researcher to

(6)

6

discuss the prevailing environmental variables and landscape affordances. Additionally, SLM is at the core of ‘predictive’ modelling. The analysis of the input sites properties is used to indicate the probability of encountering similar sites in similar landscapes settings. Lastly, archaeologists use SLMs, as supporting methods to material culture studies, to identify, interpret, and compare patterns in different chronological periods. To sum up, SLMs are commonly used in archaeology, and their dependency on the spatial scale is recognised but not researched in depth.

The chosen approach to investigate the influence of the spatial scale on the models was comparing three interlocked models with a gradually augmented number of observations and spatial scale. The models have the same geographical, temporal and research restraints. The choice of the geographical, temporal and research range allows a backdrop to review the effect of scale on the models. The validation is based on the use of existing archaeological theories as benchmarks for the results of the models. Bearing those criteria in mind the archaeology of the island of Crete was chosen as the restraint for the models. Crete has a long history of habitation and archaeological research and is a field the author is in depth familiar with. The temporal range of the models covers the period from the end of Bronze Age (1200 BC) to Classical Period (323 BC). At the end of the Bronze Age, settlements were moved to more defensive and mountainous locations as opposed to the rural and coastal areas inhabited in the previous periods. Over the years, the number of settlements is gradually reduced, and only a few new settlements are created. By the time Crete enters the Classical Period, only a handful of settlements are left, and at this point are defined as independent city-states. Researchers that examine this period have connected the earlier change in settlement patterns with the creation of the city-states (Wallace 2010).

Previous settlement pattern analysis on the island is used as the roadmap for selecting the variables for the models (Spencer and Bevan 2018). The variables were rendered by computer application of Geographical Information Systems (GIS) and analysed using maps, graphs and tests. Throughout the computation process, the effect of the scale was observed and accounted for.

(7)

7

The goal of this research is to analyse the effect of spatial scales on the SLMs. The leading question is: how can spatial scale affect the Settlement Location Models? In Chapter 2, the geographical, temporal and research context of the models is analysed. Initially Crete is presented as the research context. Moreover, the temporal and geographical range are defined and described in relation to the archaeology of Crete. Lastly there is an analysis of the dataset of the models. Chapter 3 is centred on the methodology. Initially, details regarding the type of SLMs of this research are provided. Furthermore, the computation processes of the variables that constitute the first and second order effects are presented. Lastly, the tools that used for the analysis are discussed. The results of the SLMs are presented in Chapter 4. The results are presented per chronological period and then the results of each SLM are presented separately. The effect of spatial scales on the models is discussed in Chapter 5. The effect is calculated based on the analysis of the results of the models in two fields. The first one is the interpretation of the results of the models when those are compared per period and cross-examined with the existing archaeological theories presented in Chapter 2. Their similarities and differences are combined to explore how the spatial scale affected the models. The second field is focused on the computation process of the models, as highlighted in Chapter 3, and the effect of the spatial scales is discussed from a technical perspective.

2. The contextualisation of the SLM

The primary focus of this Chapter is the contextualisation of the SLMs in the archaeology of Crete. The archaeological context of the models is established for the geographical, temporal and research range of the models. The structure of the chapter is as follows. Initially, Crete is presented as the backdrop for the SLMs and its selection is justified based on the history of the island, research, and the author's familiarity. Furthermore, the chronological range of the models and the leading theories are discussed per period. Furthermore, the geographical range of the models is defined and described by local toponyms. Lastly, the dataset of the models is analysed.

(8)

8

2.1. Background on Crete

Crete is located in the Eastern Mediterranean Sea. It is the biggest and most populated Greek Island and the fifth-largest island in the Mediterranean Sea (total area 8,336 sp. km). Humans have inhabited the island since the Palaeolithic Period (130,000 BC). During the Bronze Age, Crete was the centre of the Minoan Civilization (3250- 1050 BC). Archaeological research of this period connects the construction of large buildings with central courts, described as ‘Palaces’, and the intensification of external relationships with centralised administration, a stable economy and complex social organisation. By the end of the Bronze Age, after a gradual decline in wealth and power, the island inhabitants adjusted their lifestyle (1200- 1050 BC). The collapse of Crete follows a similar pattern with the rest of the Eastern Mediterranean. New settlements, rituals and mortuary practices begin emerging in the landscape, in a different setting than previously preferred (Watrous 1999). Those new foundations will be the base of the Classical and Hellenistic cities. From the Hellenistic period onward, the island seized to be independent and was ruled by other powers in the Mediterranean Basin like the Romans, the Byzantine Empire, Andalusian Arabs, the Venetian Republic and lastly the Ottoman Empire. In 1898, Cretans became independent of the Ottoman Empire and joined the Greek State officially by 1913.

The intensification of interest in the human past in the 19th century made Crete the centre of archaeological interest. The focus of early research in Crete was the Bronze Age remains and the epigraphic evidence of the historic period. This long-lasting interest on the Bronze Age and Classical history of the island led researchers to attempt to fill the gap between them. Some of the Late Bronze Age settlements cover that gap. Their later strata of occupation are dated in the Classical Period. Based on epigraphic evidence, they were independent city-states. Archaeological campaigns, focused on Geometric and Archaic periods, have shed light on the structure, architecture, ceramic production and ritual practices connected with the settlements. Some of these sites were abandoned, and others continued in later periods, adding more research information. Another result of these campaigns was the rise in the number of collective publications that attempt to summarise and present contemporary dated archaeological researches in a specific region or the whole of the

(9)

9

island. Those publications are either specialised in a particular type of sites, like cemeteries, settlements and sanctuaries, or to a specific period such as Geometric to Archaic and include all the types of sites that are dated within the predefined chronological range (Kanta 1980; Nowicki 2000; Sjögren 2003; Prent 2010).

Collective, overarching and long dureé researches observed that the settlement patterns changed from coastal and rural areas to more defensible locations in the end of the Bronze Age (Nowicki 2011). This phenomenon is common in the Aegean and SLMs are used to analyse the extent and intensity in each case. The effect of the environmental factors in the location and longevity of the settlements during transition periods have been used to examine regional patterns (Mirabello Bay and Mouliana region) or temporal patterns (Late Minoan IIIC to Early Iron Age) (Spencer and Bevan 2018; Koh and Clinton 2015; Wallace 1997; Wallace 2003). For this research, the settlement location model variables will follow the patterns already created by previous research and examine them in the long dureé (from the end of the Bronze Age till Classical Period).

Lastly, the archaeology of Crete was chosen as the backdrop for the SLMs, because it is my academic interest. An understanding of the history of the island and the research are valuable in better comprehending the bibliographical evidence and the biases of research. Secondly, the familiarisation with the Cretan landscape allowed a critical input on the results of the computation processes and the differentiation of GIS generated landscape and the actual one. Finally, personal experience of the landscape allowed me to create the dataset with more accuracy.

2.2. The temporal range of the SLMs

The temporal range extends from 1200 to 323 BC effectively covering seven archaeologically defined chronological periods: Late Minoan IIIC (LMIIIC), Subminoan (SM), Protogeometric (PG), Geometric (G), Orientalising (O), Archaic (A), and Classical (C) (Table 1). The definition and the grouping of the settlements are based on chronological sequences. Those are established by the typology and sequence of shapes of excavated ceramic assemblages around the island. The main

(10)

10

ceramic assemblage sequences are established based on the fine-ware pottery from Knossos and surrounding cemeteries or votive offerings from Idaeon and Diktaeon caves (Desborough 1948; Warren 2007; Sjorgen 2008, 51- 55; Tsipopoulou et al. 2003). The rest of the settlements are dated based on synchronisms with Knossos, Mainland and Cyprus (Popham 1970; Hallager 2007; Kanta 2003a; Kanta 2003b). Lately, there is a new interest in publishing pottery assemblages from other sites and the creation of new chronological sequences (Erickson 2000; Kotsonas 2006).

Period Ceramic Phase Estimated Absolute Dates

Final Neolithic 4500-3000 cal BC

Prepalatial EM I-II 3000- 2200 cal BC

Late Prepalatial EM III- MM IA 2200- 1950 cal BC

Protopalatial MMIB- MMII 1950-1800 cal BC

Neopalatial MMIII- LMIB 1800/1700- 1490/1430 cal BC Third Palace/ Postpalatial LMII 1490/1430- 1430/1390 cal BC LMIIIA- LMIIIB 1430/1390- 1200/1190 cal BC

LMIIIC 1200/1190- 1100 cal BC Sub-Minoan 1100- 1000 cal BC Proto-geometric 1000- 800 cal BC Geometric 800-700 cal BC Orientalising 700- 600 cal BC Archaic 600- 500 cal BC Classical 500- 323 cal BC Hellenistic 323- 66 cal BC

Early Roman 66 BC- 400 calAD

Roman 400- 700 calAD

Table1 Chronological table, the phases examined in this research are highlighted (Manning 1999)

Apart from the chronological periods, researchers identify two transitional periods within this temporal range. The first is the turn from Bronze to Iron Age, with the introduction of a new metalworking method. The second transition represents the changes in the society in the 8th century and is considered the final push for the creation of the city-states in Greece (Sjorgen 2008, 56- 58). The settlements of the end of the Bronze Age evolved organically into the city-states. (Wallace 2010). This continuation has been observed and approached from different perspectives and research standpoints such as religion, settlement layouts, architectural features, and multi-disciplinary methods to describe and explain the change and continuation in features (Prent 2005; Haggis and Mook 2007; Coldstream 2006; Nowicki 2002; Wallace 2010). The patterns are analysed within their chronological period and in

(11)

11

comparison to the other periods allowing interpretations. In the descriptions of the chronological periods that follows, the existing archaeological theories on settlement patterns are the main focus.

2.2.1. Late Minoan IIIC (1200/1190- 1100 cal BC)

Late Minoan IIIC is the final phase of the Minoan period. It synchronises with the disturbances in the East Mediterranean attributed to the ‘Sea People’. Researchers have described the change in the settlement patterns during this period as preferring more defensive locations, with steep slopes and avoiding the coastal zone. They highlight the location of settlements on ridges over the coast, with high visibility to the sea and the general area. The settlements have been classified as ‘coastal defensive sites’, ‘fortified citadels’ and ‘refugee sites’ (Nowicki 2000, 12-14). Researchers argue that there was a danger from the sea and the settlements were created to be protected from this danger (Nowicki 2007, 438- 442). The dating of fortification walls in this period supports this interpretation (Hayden 1988, 19). Even though researchers recognise a change in the settlements during LMIIIC, the exact character and classification of the settlements has been the centre of debate (Haggis 1999). A result of this debate was a list with the characteristic of the LMIIIC settlements was created. The settlements are described as defensive with the focus on hinterland and an economy based on livestock. At the same time, the population increased causing the creation of clusters of settlements in close proximity (Borgona 2003). The coastal zones are uncertain and avoided. In regards to the social structure of the settlements, it is defined as local and nucleated, and there is evidence of the emerging of military elites due to the general instability and the lack of palatial control. Lastly, based on the study of the material culture from excavations, there are intense interregional economic interactions.

2.2.2. Subminoan (1100- 1000 cal BC)

The Subminoan period is not broadly accepted (Hallager 2010). Even though it is recognised in Central Crete, in East Crete. Researchers argue that the SM material from Central Crete is contemporary with the local late LMIIIC material (Nowicki

(12)

12

2000, 243). Some do not recognise it as a period and use the term Early Iron Age (EIA). In regards to the settlement patterns, the LMIIIC ones remain the prevailing type of settlement. (Wallace 2003, 257; Haggis 1993, 327- 331; Haggis 2001).

2.2.3. Protogeometric (1000- 800 cal BC)

The Protogeometric period comes as a change to the stability of the previous ones. The settlement pattern from LMIIIC to PG changes as there is a general pattern of aggregative movement of regional populations towards central settlements (Judson 2019, 4- 5; Nowicki 2002). A great number of the sites are abandoned, mainly the more defensive ones. The surviving settlements expand in size, and new ones are created (Wallace 2003, 257). The new settlements have small territories located in regions with proximity to arable areas, highlighting the significance of hinterlands and subsistence. (Wallace 1997; 2006; 2013). Within this period, there is intense nucleation. There are phenomena of reuse of defensive settlements. Researchers attribute this to memory. The sites that continue are highly visible within their landscape, potentially acting as monuments. Lastly, from this point onwards, most of the settlements are positioned near communications routes (Wallace 2003, 259).

2.2.4. Geometric (800- 700 cal BC)

According to researchers, the period between the Geometric and Archaic is the point that ‘poleis’ are starting to form on Crete, consistent with the second transition. Moreover, the identity of the habitants of Eastern Crete is crystallised as separate from the rest of the island (Coldstream 2013). Homer (Od. xix. 176) includes the inhabitants of the East Crete with the name ‘Eteocretans’, this differentiation to the rest of the population. This can be considered ancestry and a stepping stone for political definition. The Eteocretan characteristic is also used to differentiate the geometric pottery in Eastern Crete (Tsipopoulou 2013; Whitley 1998). During this period, there is a tendency of returning to sites that were abandoned in the previous but for specialised symbolic activities or construction of tombs connected with the abandoned buildings (Wallace 2003, 262). Lastly, the phenomenon of nucleation

(13)

13

observed in the previous period is considered the reason for inter-settlement friction that will continue until the Archaic period (Wallace 2003, 257).

2.2.5. Orientalising (700- 600 cal BC)

Within Orientalising period, researchers observe a reduction in material culture as contemporary craftsmanship and production of objects such as bronze armour, votive plaques and intricate pottery vessels declines (Erickson 2010). This drop re-enforces the description of the 6th and 5th century as a period of economic recession and isolation of the island from the Greek World. In this period, more sites are abandoned. Abandonment of urban states indicates a more significant change in the social structure of Crete (Haggis and Mook, 2007; Haggis 2012).

2.2.6. Archaic (600- 500 cal BC)

In the Archaic period, the idea of the poleis is defined (Guizzi 2013). By this period, the sites that serve as memory places due to their high visibility points are abandoned (Wallace 2003, 260). Based on the results of survey campaigns, there is a pattern of shifting habitation within settlements in close range (Haggis 1993, 294- 319). Lastly, during the A period, there is an intensification in the constructions of public buildings connected with the urban functions of ‘poleis’. However, not all the settlements with such constructions survived into the Classical period (Mook and Haggis, 2013).

2.2.7. Classical (500- 400 cal BC)

Little is known about the classical cities of the island. There is a small number of systematic researches focused solemnly on the period. As a result, there are only a few researches on the pottery and the in-situ inscriptions (Erickson 2010; Wallace 2010, p 20-21). However, the extensive systematic survey campaigns in Eastern Crete record finds from all the periods. Based on the evidence from those surveys, by the Classical period clusters of settlements seize to exist, the rural areas are abandoned, and the population moves to other locations (Haggis 1993, 294- 319). In general,

(14)

14

Classical city-states are considered the result of the nucleation that started in the PG period (Wallace 2010, 251).

To conclude, in LMIIIC the settlement pattern of Crete was completely altered, and till the C few new settlements were created. The majority of the sites that started at the end of Bronze Age did not continue to become city-states. For this reason, each SLM includes models produces per chronological period. In this way, the results of the models can indicate contemporary environmental preferences and networks as well as differences between periods. Lastly, the results of the models are comparable with the aforementioned theories on settlement patterns.

2.3. The geographical range of the SLMs

Due to the size of the island and the extensive archaeological research, the geographical range was also defined. Examining the archaeology of Crete in parts is a broadly accepted way of classifying research. Crete is divided in West, Central and East based on the Mount Ida (or Mount Psiloritis). Central Crete is North and South of the mountain, including the Messara plain. East and West of the Mount are called Eastern and Western Crete, respectively (Figure 1). For this research, East of Crete was chosen as the testing ground for the SLMs.

Figure 1 Topographic map of Crete island, Greece (accessed from

(15)

15

Eastern Crete has been the centre of several intensive archaeological campaigns. Surveys and excavations cover a large portion of the land, allowing a better interpretation of the urban and rural economy. Due to the extensive area covered by researches, the settlements in East Crete allow a better-rounded understanding of the displacement in the landscape and social changes.

Within Eastern Crete there are further regional classifications. From West to East, it consists of the sub-regions of Lasithi plateau and the Dikti Mountains, the Mirabello Bay and the Isthmus of Ierapetra, the Thrypiti and Orno mountains, and the Siteia region. The Lasithi plateau is excluded from the research area. Researchers describe the region as a specialised area compared to the rest of the East Crete (Fernandes et al. 2012; Nowicki 1991; 1995; van Effenterre 1980). As a result, the research area covers from the E side of the Dikti mountain range to the E coast of the island. The Dikti Mountains divide the research area from the rest of Crete. However, crossing is possible from the N, close to the area of the modern city of Agios Nikolaos. Another possible crossing is to the S, the modern South road connecting Ierapetra with Messara follows this passage. Mirabello Bay includes the Isthmus of Ierapetra area and the region between Orno and Thrypiti, called Kavousi. Lastly, the Siteia region is consisting of smaller sub-regions, the Siteia bay, where the harbour of the modern city is located, and a valley to its S and the Siteiaka Mountains. To the N of the mountains is an area called Cave Sidero. The valley of Siteia is the crossing that the modern road traverses and connects the modern cities of Ierapetra and Siteia. The road crosses between the mountain range of Thrypiti and the Sitiaka Mountains. Its coastal areas primarily to the South and the East currently have little to no vegetation and are characterised by gorges and cliffs.

Even though Eastern Crete is divided into smaller groups that share similar topographical features. In regards to the N to S axis, the borders are coasts. The northern coast, except for the Mirabello and Siteia bay, is steep. On the other hand, the southern coast, except the area covered by Siteiaka Mountains is flat. The settlements are divided to N and S by the mountains and E and W based on their position in relation to the Isthmus and the Siteia valley.

(16)

16

The topography of this area has been considered to be the reason for the fragmentation of the local population. The similarity of the topographic features and the fragmentation of the populations create a proper environment to test the effect of spatial scales on SLMs.

2.4. The dataset of the SLMs

Having decided on the defined geographical area and temporal range , the dataset was based on collective publications as the primary source. The first publication used was the monograph “Defensible Sites in Crete c. 1200- 800 BC (LMIIIB/ IIIC through Early Geometric)” by Nowicki (2000). The author presents the settlements in the whole island that were inhabited between 1200- 800 BC. The settlements were located during archaeological campaigns, surveys and excavations. He also includes the results of his personal survey that covered the mountainous areas of the island and was focused on locating the Late Minoan defensive sites. The author provided information on the location, the condition of preservation, previous research and the suspected phases of occupation based on pottery assemblages. The second publication that was used was “Cretan Locations. Discerning Site Variations in Iron Age and Archaic Crete” the monograph by Sjögren was published in 2003 and was focused on settlement, ritual and mortuary sites from Early Iron Age to Archaic period. This publication was used to cross-reference the dating for the settlements that are in both publications and to add in the dataset new sites dating from the Early Iron Age onwards. There is no collective publication about the Classical period on the island. However, both monographs included the Classical period in their chronological tables when contemporary material culture was located. This fragmented information was used to substantiate the continuation of those settlements in the Classical period.

Both publications used data that were produced during different type of archaeological campaigns. This creates unevenness on the type of information available. The sites that are recognised during the process of excavation provide more information, especially regarding their size, architecture, duration of occupation and type of settlement. Moreover, even within the sites that are excavated the data they provide are not always comparable as they represent only the latest phase of

(17)

17

occupation. In the cases, that sites survived longer, the earlier occupation phases were destroyed or covered by the later strata. For the sites located during survey only the information regarding their location, estimation of the size and duration of occupation are available.

The coarse ware pottery is more common in surveys than fine ware one. The chronology of the surveys is created based on coarse ware assemblages to create a rough estimation of the duration of occupation of the settlement (Haggis 1993, 37- 53). The variations in the coarse-ware typology are based on different fabric composition over time due to the use of different materials. Eastern Crete is a diverse region concerning raw materials (Erickson 2010, 24). This form of chronology is usual and recognised in Eastern Crete; however, the coarse pottery aseemblages sequences are not directly comparable with the fine ware ones (Sjorgen 2008). The study and establishment of fine ware pottery sequences from excavations can only be used and compared with other excavated context. The study and publication of new pottery assemblages and the establishment of better-defined sequences have allowed the researchers to question the chronologies of the settlements (Erickson 2000). Alongside the new chronological sequences, past chronologies are debated (Hallager 2010; Mook 2004, 169).

Considering these biases, the dataset had to accommodate sites with different amount of information available. For this reason, the only data regarding the settlements used to analyse their relations are their coordinates and the proposed period of use. Based on those, the sites were divided into three primary groups based on their location, in relation to the research area of the models, and seven subgroups by their chronology (Table 2).

Group LMIIIC SM PG G O A C Total

SLM1 20 5 13 7 7 5 2 21 SLM2 31 13 23 20 12 13 4 37 SLM3 56 32 40 37 26 26 9 66

Table 2 Number of settlements per chronological period, and the total number in each group.

The groups like the SLMs are successive to each other. The settlements in SLM1 are included in SLM2. In the same manner, the settlements of SLM2 are part of SLM3 (Table 3, 4, 5).

(18)

18

No Site LMIIIC SM PG G O A C

54 Agios Ioannis Katalimata/ Kastrolakkos Y

56 Agios Ioannis Psychro Y y y

35 Asari Kefala Y

25 Chrysopigi Korakia Y

60 Kato Chorio Profitis Elias Y

28 Kavousi Azoria Y 29 Kavousi Kastro Y 30 Kavousi Vronda Y 40 Kephala Vasilikis Y 64 Koutsounari Karfi Y 24 Lastros Vigla Y 42 Monastiraki Chalasmeno Y 43 Monastiraki Katalimata Y

14 Myrsini Kastello/ Ellinika y 45 Oreino Epano Ellinika

46 Oreino Kastri

47 Oreino Kato Ellinika (Petrokopia) 53 Stauvrochori Skalia 23 Tourloti Kastri 66 Vainia Skouros (StoSkouro) 67 Vainia Stavromenos

Table 3 SLM1 settlements and period of use.

54 Agios Ioannis Katalimata/ Kastrolakkos 55 Agios Ioannis Plagia

56 Agios Ioannis Psychro 34 Agios Stefanos Kastello

35 Asari Kefala

6 Chamaizi Liopetro/ Liopetra 36 Chandras Voila Kastri 25 Chrysopigi Korakia 63 Dasonari/ Ellenika

27 Kalamafki Kypia

39 Kategari Pigadi 60 Kato Chorio Profitis Elias

28 Kavousi Azoria 29 Kavousi Kastro 30 Kavousi Vronda 40 Kephala Vasilikis 12 Koutsoulopetres: Kastro 64 Koutsounari Karfi 31 Krya Agios Georgios

24 Lastros Vigla

52 Lithines Andromyloi Anginares

41 Mega Chalavro

(19)

19

43 Monastiraki Katalimata 14 Myrsini Kastello/ Ellinika 45 Oreino Epano Ellinika

46 Oreino Kastri

47 Oreino Kato Ellinika (Petrokopia) 48 Pefki Kastellopoulo 49 Pefki Stauromenos 65 Phobolies 32 Praisos 22 Sfakia Kastri 53 Stauvrochori Skalia 23 Tourloti Kastri 66 Vainia Skouros (StoSkouro) 67 Vainia Stavromenos

54 Agios Ioannis Katalimata/ Kastrolakkos 55 Agios Ioannis Plagia

56 Agios Ioannis Psychro 34 Agios Stefanos Kastello 57 Anatoli Elliniki Korifi 58 Anatoli Mesokastella 59 Anatoli Sochores 10 Andrianos (Arkoudas) Fortetsa

35 Asari Kefala

6 Chamaizi Liopetro/ Liopetra 36 Chandras Voila Kastri 37 Christos Skistra 25 Chrysopigi Korakia 63 Dasonari/ Ellenika 3 Dreros 7 Elounda Oxa 26 Istron Vrokastro 4 Itanos 38 Kalamafka Kastello 27 Kalamafki Kypia 39 Kategari Pigadi 60 Kato Chorio Profitis Elias

28 Kavousi Azoria 29 Kavousi Kastro 30 Kavousi Vronda 40 Kephala Vasilikis 12 Koutsoulopetres: Kastro 64 Koutsounari Karfi 21 Kritsa Kastello

(20)

20

31 Krya Agios Georgios

24 Lastros Vigla

13 Lato

52 Lithines Andromyloi Anginares

41 Mega Chalavro

1 Milatos Kastellos 42 Monastiraki Chalasmeno 43 Monastiraki Katalimata 14 Myrsini Kastello/ Ellinika 61 Mythoi Kastello

62 Mythoi Zonari

5 Neapoli (Vrachasi) Kastri

44 Oleros

47 Oreino Epano Ellinika

46 Oreino Kastri

47 Oreino Kato Ellinika (Petrokopia)

15 Palaikastro

16 Palaikastro Kastri

17 Palaikastro Plakalona Kalamafka 48 Pefki Kastellopoulo 49 Pefki Stauromenos 65 Phobolies 32 Praisos 50 Prina Stauromenos 22 Sfakia Kastri 53 Stauvrochori Skalia 18 Tapes Epano Kastello 19 Tapes Kato Kastello 23 Tourloti Kastri 66 Vainia Skouros (Sto Skouro) 67 Vainia Stavromenos

2 Vrachasi Anavlochos 8 Vryses Drasi Xeli 9 Vryses Profitis Elias 33 Zakros Ellinika

51 Zakros Gorge Kato Kastello (Kastelas) 20 Zenia Kastrokefala

3. Methodology

Methodology is a central part of this research. As stated in Chapter 1, the effect of scale is also evaluated in regards to the computation process. For this research, the term computation process describes the computation of the variables and the analysis

(21)

21

process. The purpose of this chapter is to define the methodological details of the SLMs. Chapter 3.1 is dedicated to providing more details on the SLMs as well as the presentation of the SLMs of this research. The next two chapters have the same pattern as they present the technical details of the variables used in the analysis. Chapter 3.2 is dedicated to the description of the variable used to represent the first order effects. Those variables are divided in three groups, environmental, distance and visibility, based on the type of data and tools used. Each group is discussed in a different subchapter. Additionally, in the environmental variable subchapter there is also a brief description of the distribution of the landscape based on each variable in each research area. Chapter 3.3 is focused on the presentation of the second order effect variable. The last chapter of methodoloy is dedicated to the description of the methods used for the analysis of the factors in relation to each other.

3.1. Settlement Location Models

SLM application has the potential of addressing the relations between settlements and specific landscape features quantitatively. They can be used in multiple ways. Usually SLMs are applied as a tool to test the validity of analogies on a case study, to examine the spatial relations of between the catchment area and the settlement or create the basis of predictive models for archaeological research or cultural services (Mason 1972; Hodder 1972; Wheatley and Gillings 2002, 165- 138). However, the efficiency of the SLMs and their results have been the centre of controversy due to environmental determinism (Gaffney and van Leusen 1995).

SLM can model how settlement systems are spatially ordered and reflect a human priority order in the broader environment within a specific period that might relate to everyday subsistence-based issues (Spencer and Becan 2018, 72). To create a meaningful result, landscape features and contemporary human relations should be included in the model. The first order effects are the variables connected with the underline properties of the local environment and constant exogenous influences (O’ Sullivan and Unwin 2003, 79). The second order effects represent the variables that measure the local interaction between settlements and describe the influence of neighbouring points (Palmisano 2014, 349). Previous researches on Crete have used

(22)

22

this type of models to understand the settlement patterns and social relations of the Bronze Age period (Spencer and Bevan 2018; Bevan and Wilson 2013; Déderix 2017; Fernandes et al. 2012; Paliou and Bevan 2016). These previous researches are the map for the selection of the first order effects and as a guiding tool to test the second order effects. The settlements of the dataset were located based on legacy data. The coordinates were not collected by GPS sensors, nor an in-person survey was conducted. However, the application of Google Earth was used for the location of the settlements based on the text, and the extraction of the coordinates (Déodat and Lecoq2015) (Appendix).

The SLMs were created as evenly as possible, regarding the distribution of valleys and mountains. SLM1 covers half of the Mirabello bay and the Isthmus of Ierapetra and Kavousi. Archaeological projects have covered the area in survey campaigns and excavations (Figure 2). The geographical range of the SLM2 is stretched further E covering the Siteia bay and the W side of the Siteiaka Mountains. It is approximately double the size of SLM1. The area of between Siteiaka Mountains and Thrypiti has been surveyed and the Siteia bay has been excavated with focus on the Minoan phases of occupation (Figure 3). The size of SLM3 is approximately triple of SLM1’s. In relation to SLM2, it extends to the W and E. The extension to the W covers the other half of the Isthmus and the eastern side of Dikti mountain range. The area around the northern passage has been investigated by excavation and survey projects. The E extension covers the whole of the Siteiaka Mountains and the peninsula of Cavo Sidero. Nowadays, this general area is not intensely populated, and the towns are located on small plateaus. Archaeological projects have covered the area Cavo Sidero and the NE side of the Siteiaka Mountains have been extensively covered by survey projects. The settlements are described as N and S in relation to the mountain ranges and as W and E in relation to the Isthmus and the valley of Siteia (Figure 4).

(23)

23

Figure 2 Settlements within research area of SLM1

(24)

24

Figure 4 Settlements within research area of SLM3

3.2. Variables used for the first order effects

The first order effects consists of twelve variables divided into three groups, based on the type of analysis (Table 6). The first group is the environmental ones; those factors are calculated from the Digital Elevation Model (DEM).1_{The second group are}

variables that describe the distance of the settlements from different features. Lastly, the third group includes variables that describe visibility relations.

(25)

25

Variables Type of Data ArcGIS tool2

Elevation Settlement Number _(m) Elevation Catchment Number _(m)

Slope Settlement _{(Usable/ Non-usable)}Binary Slope (Spatial Analyst)

Slope Catchment _{(Usable/ Non-usable)}Binary Slope (Spatial Analyst)

Aspect Settlement Text Aspect (Spatial Analyst)

Aspect Catchment Text Aspect (Spatial Analyst)

Distance the coastline Number _{( km)} Near (Analysis)

Distance from travel

corridors Number ( km) Cost distance (tool), Near (Analysis)

Local Visibility

(0-1 km) Percentage Viewshed(Spatial Analyst) Regional Visibility

(1-6 km) Percentage Viewshed(Spatial Analyst) Visibility of the travel

corridors Binary (0/1) Viewshed(Spatial Analyst)and Cost Distance (Distance toolbox)

Visibility Towards Site Ratio _(0-1) Viewshed(Spatial Analyst)

Table 6 First order effects, type of data and the tool and the name of the tool and toolbox used to render them.

3.2.1. Environmental variables

As environmental variables, I define elevation, slope and aspect, those were calculated based on the attributes of the DEM3_{(USGS, NGA and NASA 2015). To}

make the analysis of these variables more meaningful, human behaviour was taken into account. Human activities are not limited only to the settlement. The surrounding area of the settlements could be used for rural, domestic or later urban activities. In this analysis the surrounding area is referred to as catchment area. The environmental variables reflect those human activities by including and calculating the settlement and catchment area separately. the values were not only calculated from the point features that were created initially. The values presented have been summarized statistically. Taking into account the estimation of the size of the settlements and the resolution of the DEM, the environmental variables of the settlements represent the

2_{SoftwarepackageArcGIS® 10.5.1.}

3_{Each pixel contains the average elevation value of 25 sq m in the real landscape, meaning that it has a}

(26)

26

values within an area of 50 sq. m.4_{Accordingly, the size of the catchment is}

calculated to 500 sq. m.

Each environmental variable required a different computation process and analysis. The elevation values were measured directly from the DEM. The slope variable classifies the land in a binary format based on the degree of slope. The values 0 to 12 degrees classifies the slope are ‘usable’ for long term habitation and agricultural purposes and anything over that as ‘non usable’. This range has been used for models throughout the Mediterranean Basin (Koh 2015; Tartaron et al. 2003). The results do not take into account anthropogenic alteration of the terrain, such as the construction of terraces for domestic or agricultural use. Aspect represents the orientation of the land and contains values (-1) for flat area and a range from 0 to 360 degrees representing the azimuth. In this analysis, the results of the tool were translated to cardinal and intercardinal directions.

In Chapter 2.3, the similarity of the topographic features was discussed. The examination of the distribution of the environmental variables among the models was calculated and the similarities were calculated and accounted for. The landscape is divided into percentages to classify the environmental factors within their general landscape. In SLM1, the 0 to 200 m zone elevation consists of 36% of the landscape, then 200 to 400 m is 23%, and the percentages are reduced by 4% till the 1,460 m. The usable slope is 38% of the landscape (Figure 5). The orientation of the landscape is equally distributed. For SLM2, the elevation values of the general landscape consist of 31% of 0 to 200 m, 24% 200 to 400 m and 22% in the next 200 m. In regards to the slope, 47% of the landscape consists of usable slope. Lastly, the orientation is evenly distributed (Figure 6). Lastly, the distribution of environmental variables in SLM3 is as follows. Half of the area is from 0 to 400 m from the sea level; the other half contains values till 2300 m. Moreover, usable slope cover 46% of the landscape. Finally, the orientation of the landscape is equally distributed (Figure 7). In general, SLM1 is more mountainous and SLM2 and SLM3 have similar distributions

4_{The tool used was ‘Zonal Statistics as Table’ (Spatial Analyst/ Zonal Toolbox) and the statistic type}

(27)

27

Figure 5 Names of landscape features and distribution of elevation, slope and aspect variables (%) in the research area of SLM1

(28)

28

3.2.2. Distance Variables

The distance variables are two, coastline and travel corridors, and represent the meassured distance in a straight line from each settlement to the selected feature in km. The first distance variable is representing the distance between the settlement and the coastline. In the maps this variables is visualised as lines with 1 km starting from the coast. Researchers agree that the coastline of Crete is the same from Bronze Age until today (Spencer and Bevan 2018, 76). For this reason, the coastline of the DEM was used for the calculations.

The second distance variable is representing the distance from the travel corridors. The travel corridors in this research are computer generated paths that follow the most traversable part of the landscape. The movement represented is on foot. The traversability of the landscape is calculated based on a cost surface. The factors of energy and time spent were not included (van Leusen 1999, 217). For this research, the coast surface was created by calculating the slope layer for the whole island:

(29)

29

Slope (Persent)/105

The current coastline is the guideline for the creation of the points. For this analysis, the same points were used as starting points and destinations.6 Based on their location the points were divided in opposing pairs (N and S, E and W). Then, the algorithm calculated the most traversable paths with four different directionalities (Whitley and Hicks 2003, 80- 83). In the end, the separate paths were merged in one layer. The visualisation of this variable is a continous layer was created that displays high and densities of paths based on a predefined colorramp.7_{The paths were}

calculated independently of the models so there was no edge effect.

3.2.3. Visibility Variables

The visibility variables represent four visibility relations based on the directionality of the sight and the area of analysis (Local and Regional visibility, ratio of visibility and visibility over the travel corridors). Those variables aim to allow a better understanding of the physical landscape surrounding the settlements and their position within it, based on visibility relations (van Leusen 1999). These variables are computated by the same tool. By changing the settings, it was possible to create different layers and different analyses (Llobera 2001; 2005). Before presenting the details of the different variables, it is essential to explain briefly the tool that was used. The viewshed analysis tool requires at least one point feature, an observer, and a DEM. The observer is set as the starting point of the analysis and a new binary layer is created with pixels that have a value 0, when the pixel is not visible from the observer, or 1, when visible.Whether the pixel is visible or not is calculated based on the obstructions on the line of sight of the observer created by the elevation values of the DEM. Additional factors that could obstruct the eye sight, such as vegetation or built constructions were not taken into account as this information was not available for all the cases. By manipulating the number and location of observers as well as the size of the area of analysis, it was possible to create different variables.

5_{Calculated with Map Algebra}

6_{The points were created on the borders of the terrain in an equal distance of 3 km. Each origin point}

was used to calculate the cost distance and the cost backlink and based on the layers the least cost paths to the destination points were calculated.

(30)

30

More specifically, the local and regional visibility variables share the same number of observers but cover different areas. 8_{Each analysis created a binary layer}

that classified the defined landscape to visible (1) and non-visible (0) for all the observers. The values of these variables are represented by the percentage of visible over non-visible landscape in the form of a percentage. 9 Local visibility variable represents the percentage of visible landscape within a radius of 0 to 1 km from the settlement. On the other hand, the regional visibility variable represents this percentage for the pixels within a 0 to 6 km radius from the settlements. The differentiation is meant to take into consideration the eyesight (Ogburn 2006, 411- 412). Those distances are used in previous researches in the area (Spencer and Bevan 2018, 75).

The visibility of the travel corridors represents the overlap of the regional visibility layer with the travel corridors in a binary form. Depending on the result the settlements were assigned a number, 0 when the travel corridor is non-visible, and 1 when the travel corridor is visible.

Visibility Towards is representing the topographic prominence of the settlements (Llobera 2001). This analysis uses the same tool as the previous ones; however, the number and the location of the observers are different.10_{The observers}

are located within a 0 to 6 km radius from the settlement and create a layer that provides results of the broad landscape.11_{This layer contains values from 0 (not}

visible) to n (maximum amount of times a pixel is visible). This variable represents the topographic prominence of the settlements within their regional landscape in the form of a ratio. Taking into consideration that the presence of a settlement can be recognised by landscape features in its close proximity, the values within the catchment area represent the prominence of the settlement.

8_{A set number of observers (5) were randomly distributed within a 100 m radius from the settlement;}

the standard height of the observer was set to 1.6m for all the computations.

9_{The initial analysis creates a layer with a range of values from 5(which would mean that a pixel is}

visible from every observer) and 0 (not visible from observers).

10_{A set number (100) of random observers were rendered within a 6 km radius from each settlement.}

The values were combined into a layer that represents the range of visibility containing values that could range between n (maximum number of observers that can see one pixel) and 0.

(31)

31

Due to the relief of the landscape, some locations have in general poorer visibility than others. To make the results comparable a ratio was created using the following formula:

n of general layer/ n of catchment area

The n of general layer represents the maxinum value of the whole layer, and n for catchement area is the maxinum number value within the catchment area. Both factors belong in the same layer. The difference is in the extent of the areas. This process creates a comparable ratio that ranges from 0 to 1 and can be used to describe the visibility of the settlement in regards to its surrounding landscape. The ratio is classified in five groups. I created the range as a tool to describe the ratio from ‘very low’ to ‘exceptionally high’ (Table 7). This ratio is more of an indication of the visibility of the settlements. Factors such as the elevation of the house, the architecture, material used, the colours and the difference between day and night could have been used to make the settlement stand out or conceal its existence.

Ratio Interpretation

0-0.2 Very low visibility

0.2- 0.4 Low visibility

0.4-0.6 Moderate visibility

0.6-0.8 High visibility

0.8-1 Exceptional visibility

Table 7 Classification of visibility towards

3.3. Variable used for the second order effects

Second order effects represent the local interactions between the settlements. Those relations can be calculated by the distance between the sites or by creating networks between the settlements to examine the potential social relations between them. The distance between the settlements can be a factor in the selection of the location. Distribution of point patterns is analyzed with different statistical analysis. The points can represent either settlement, sites, artefacts or any type of archaeological record. However, the results of the second order effects regarding clustering, settlement

(32)

32

hierarchy and spatial distribution can be a direct result of uneven fieldwork and site survival, among other reasons (Hodder and Orton, 1976, p85). Due to the uneven data and the data collecting process, the results will be biased by clustering on the sites that have been the centre of local intense surveys. Taking into account the diversity of the data, the different survey methods, and chronological range, clustering will not be included as a second order effect for this research. For this analysis, the second order effects are only examined with networks that are created based on the geographical relations between sites.

The networks are depicted in the form of a one-mode graph network (Brughman 2010; Peeples 2019, 453- 454). They represent hypothetical networks and indicate the probability of traversing between all the contemporary sites within the predefined area and topographical relief. The effect of topographical relief on human movement in the landscape is calculated by a cost surface. The cost surface is created with the same formula as the compuation of the travel corridors, see Chapter 3.2.2. However, these networks are created with a different tool.12_{The social networks have}

no directionality, so there are no starting points or destinations. The area that was used for a cost surface was not confiend by the research area of each model. The sea routes and connection were not included in the calculations.

The settlements are desrcibed based on their position in the networks (Table 8). The positions is described based on the number of paths that cross within the catchment area of the settlements (500 sq. m). In addition to the description of the position of individual settlements, the social networks are displayed with a density map, created with the same tool as the travel corridor one. The maps were an invaluable tool in visualising the change in the location of densities over different periods and models. The second order effects are examined separately on the settlements for each period and model.

(33)

33

Value Position Example

1 Edge

2 Passage

3 Edge and two diverging paths

4 Edge and three diverging paths

5 Passage and a diverging paths

6 Passage and two diverging path

Table 8 Explanation of the description of the position of the settlements in the social networks. The settlements are depicted with a black trianglem the catchement ares is the grey circle and the lines are the networks between the

(34)

34

3.4. Analysis of the variables

The previous processes merely describe the factors and the type of data. To complete the SLM and make archaeological interpretations, the variables have to be combined and analyzed. Due to the small size of the dataset, the options for analysis were limited as most statistical tests have a minimum number of observations. The analysis of the models is based on the description of the relations between the variables with the use of maps and graphs.

Two types of graphs are used to display different relations. The relations between the same variables are depicted by scatter plots. This method is used to describe the relation between catchment and settlement area for the environmental variables. The same plot is also used to display the relation between local and regional visibility. The second type of graph that is used is the boxplot. Boxplots display the concentration of the observations. Half of the values are concentrated in the form of a box with a line for medium values, the extended lines depict the ¼ of additional values, and lastly, values that are not within the majority of observations are depicted by circles. Boxplots are used in two ways. Firstly, they describe chronological trends of each environmental variable for each model. Secondly, boxplots were used to depict the trends of all the variables within each SLM per chronological period. Since the variables are represented by different type of data, their values had to be converted in a ratio of 0 to 1. This made the comparison between the variables possible.

Another method that is used is the chi-square test. It is a statistical hypothesis test.13_{First of all, a null hypothesis has to be formulatd. For this research, the}

hypothesis is: The variable has no effect on the selection of the settlement location. To complete the test, the number of actual observations (N) has to be compared with the expected ones. The number of expected observations is created with the following formula:

13_{The chi-square test was calculated in the computer environment of Microsoft Excel 2007, with the}

formula:

(35)

35

N* area covered by the classification of variable/ the defined area

The area was calculated with the number of pixels, classified as usable and non-usable slope. If the expected number of observations is below 5, the test is not meaningful. If the value is over 5, it is. The range of expected observations is compared with the range of actual ones. Depending on the resulting value it is possible to say whether the hypothesis is true or it is true or false, respectively. The predefined values depending on the degree of freedom (Df):

Df= (number of rows- 1)* (number of columns- 1)

Due to the numbers of the dataset, this analysis was only possible in the case of slope analysis that is binary. So the predefined value was 0.05. If the values from the range are above this number, then the null hypothesis is true, and the variable did not affect the selection. If the number is lower, then the null hypothesis is false, so the variable had an effect on the selection of the location of the settlement.

4. Results

The results of the models are presented per chronological period. Each chronology is a subchapter and follows the same structure. Firstly, there is a presentation of the number and distribution of settlements over the models and the results of the variables that represent the first order effects. The environmnetal variable are presented indivivually and the distance and visibility variables are presented as groups. At the end of each subchapter, the results of the social networks of the models are presented. Furthermore, the results of each SLMs are presented. Each SLM is a subchapter. Within the SLM subchapters, the results of each environmental are presented separately with the use of graphs for all the period.

(36)

36

4.1. Late Minoan IIIC

In SLM1, there are twenty LMIIIC settlements distributed equally regarding N and S of the island, and two settlements are located at the W of the Isthmus. SLM 2 includes thirty-one settlements and the same distribution of observations as SLM1. The LMIIIC period of SLM3, includes fifty-six observations that cover the whole of the research area. Twenty-six settlements are located to the N. The greater concentration of settlements is in the area between the valleys, on the mountain areas of Orno and Thrypiti. The second bigger group is to the W of Isthmus. Lastly, there are four settlements to the E of Siteiaka Mountains.

The analysis of the elevation shows little difference between the models. In SLM1, the majority of settlements are grouped between the 200 and 400 m. The rest of the observations are scattered in two groups. The first group is between 500 and 600 m and the second from 700 to 900 m. Lastly, there is a settlement at over 1000 m (Figure 8). The same trends are visible on the scatter plot of SLM2. The only exception is that the groups between 500 and 900 m are denser, in comparison to SLM1. In SLM3, the groups that were visible in the previous models are now merged in one group from 100 to 900 m. The edges of the group are a settlement at 42m elevation and the settlement over 1000 m from SLM1. The densest group is between 400 and the 600 m. In all models, the leading trend is that the settlement are higher than the catchment area.

The settlements in SLM1 in their majority are on non-usable slopes, and four of the observations, either in their settlement or in catchment area have a usable slope (Figure 9). The same trend is also visible in SLM2, although there is an increase in the number of observations with usable slope as six settlements are in usable land, and three catchment areas have usable slopes, only one observation has both classified as usable. In the SLM3, fourteen settlements are located in usable slope, and four catchment areas are usable, only one observation has both.

The orientation of the landscape is equally distributed in the landscape, as seen from the presentation of the environmental factors in Chapter 3.2.1. In SLM1, there is a group with orientation from E to S, three observations have a W orientation, and the

(37)

37

rest NE and NW (Figure 10). In SLM2, there is the same trend. However, settlements have a more variable orientation with orientations from NE to NW. The only orientation that is not favoured is N. The variation of the settlement orientation is not mirrored to the orientation of the catchment area, with the majority facing S. There are two exceptions that are facing N. In SLM3, SW to SE are the most popular orientation with thirty-eight settlements, the rest are facing in majority E and W. The SW to SE is the preferred orientation for the catchment areas with forty-seven observations.

(38)

38

Figure 9 Distribution of LMIIIC settlements by slope. There is no meaningful difference visible in the graphs.

(39)

39

The results of distance variables show that in SLM1 most of the sites are located in general zone 3to 5 km from the sea. Only one site is less than 1 km away from the sea. The settlements that are located less than 1 km from the coastal are defined as coastal. As for the distance from the travel corridors, most of the sites are within a range of 0 to 3 km (thirteen settlements), the rest are within 3 to 5 km, and one over 7 km away (Figure 11). In SLM2, the distances from the coastline are the same. The majority of them are located in the zone of 2 to 5 km, and there are five settlements located over 7 km away. In regards to the distance from the travel corridors, the settlements are divided into two groups. In the first group there are fifteen settlements positioned from 0 to 2 km. The second group is the same size and covers the distance from 2 to 4 km; one settlement is over 6 km away. Those patterns are the same in SLM3. The distance from the coast ranges from 0 to 9.5 km, the most populated groups are within the 3 to 5 km with ten observations. The rest of the settlements are evenly divided into the other zones. In relation to the travel corridors, the majority is within 1 to 3.5 km.

Figure 11 Distribution of LMIIIC settlements in 1 km zones from the coastline and the density of travel corridors within the study area.

(40)

40

In regards to the visibility variables, the local visibility values in SLM1 is mainly between 16% and 50% (Figure 12). For the regional visibility, most of the percentages are within a range of 2% to 37%, with one exception of a settlement with 74%. The majority of settlements have visibility over the travel corridors. Lastly, the settlements are classified as moderately visible based on their ratio. The results of local visibility variable in SLM2 are mainly between 11% and 57%, but the two have higher than 70%. In regards to regional visibility, most of the sites have percentages within a range of 2% to 37%, with the exception of one settlement with 74%. Twenty-six settlements have visibility over the travel corridors, the majority of them are within 2 km distance. Finally, the majority of settlements in SLM2 are in moderately visible locations. There are some settlements that are exceptionally visible. The exceptionally visible observations do not have a high percentage of local or regional visibility. On the other hand, the settlements that have the higher local and regional visibility are moderately visible. To complete the presentation of the results of the visibility variables per model, the values of SLM3 local visibility range from 11% to 75%. Most of the settlements have values between 30%- 40%. Regional visibility is in general lower than the local and ranges from 2% to 37%. There are two cases with regional visibility 47% and 74%. In those cases, regional visibility is higher than the local. Ten settlements have no visibility over the travel corridors. Most of the sites are moderate visible. Two settlements are slightly visible, and a group of eight settlements are exceptionally visible.

Figure 12 Scatter plots of the LMIIIC results that show the relation between local and regional visibility in the three different models. There are no meaningful differences visible in the graphs.

(41)

41

The probabilistic networks share some similarities in regards to the density points. In SLM1 nine settlements are edges of the network, four are passages, two nodes with proximity to a passage, one edge close to two diverging paths and four passages with one diverging path. The main uninterrupted passages cross from N to S through the Isthmus. Isthmus is the meeting points for the sites of the W and the group of sites in the S. The higher densities are located in the Kavousi area and the SE side of the Isthmus (Figure 13A). In SLM2, Kavousi is again an area of high network density as it is the meeting point of the paths from the settlements from N, E and S. Within this area, there are three settlements, two of which are passages close to diverging paths. Another area of density is to the E of the Isthmus where the paths of five settlements join in one. There is no settlement on the exact point, but there are two in proximity (Figure 13B). Lastly, the SLM3 has four locations of high-density. The higher densities are at the W side of the Isthmus and Kavousi. In this area, five settlements are passages with proximity to diverging paths. Another area of high-density of paths is located to the N of the Dikti Mountains, and three settlements close to passages and diverging paths. Lastly, there is a density in the Siteia valley. Another density is located to the SE of Dikti Mountains (Figure 13C).

(42)

42

(43)

43

4.2. Subminoan

The SM dataset is low in numbers in all SLMS. In SLM1 there are five settlements most of which are in the S of the island. In SLM2 there are fifteen settlements which are also located in the S. As opposed to the similar distribution of the previous models, in SLM3 there are thirty-two settlements. Based on their orientation, they are divided in three groups. The biggest group is located on the NE of the Dikti Mountains, the second is to S of the Thrypiti Mountains, and the smaller group is located to the SE of Dikti.

In regards to the elevation, the SLM1 settlements are located over 300 m elevation (Figure 14). In SLM2, the SM settlement group is between 250 to 600 m. The SLM3 has a greater variation. Settlements are located from 180 to 805 m, with the more prominent grouping in the areas of 400 m and 600 m. As with the previous period, the settlement is located in higher ground than its catchment area.

The observations of slope and aspect variable are homogenous in the three models, for this reason the variables are described together. Regarding the slope variable, all the SLM1 settlements are in non-usable slopes (Figure 15). In SLM2, three settlements are at a usable slope, and one of them has a usable catchment. Lastly, in SLM3, ten settlements and two catchment areas have usable slope. One observation has usable slope in the settlement and catchment area. The aspect variable is the same in all the models. All the settlements and catchment areas are facing S (Figure 16).

The models had similar results in the distance variables as well (Figure 17). In SLM1 and 2 the majority of settlements are located within a range of 2 to 5 km and the rest are over 7 km away. For SLM3 the settlements indicate a preference in the zone of 2 to 5 km from the sea with the exception of two coastal settlements. As for SLM1 and SLM2 the distance of 2 to 5 km from the travel corridors is also preferred. In SLM3, Fourteen settlements are located less than 2 km distance from the travel corridors. From the rest, the majority of the other sites are located in the 3 km zone and three over 6 km away.

(44)

44

Figure 14 Distribution of SM settlements by elevation. Only the SLM3 graph can be used to describe a more specific pattern.

(45)

45

Figure 16 Distribution of SM settlements by aspect. There are no meaningful differences visible in the graphs.

Figure 17 Distribution of SM settlements in 1 km zones from the coastline and the density of travel corridors within the study area.

Tipping the scales G U

U

NIVERSITY OF

G

RONINGEN

M

T

Tipping the scales

A multi-scalar approach on settlement location modelling

Contents

Acknowledgements

1. Introduction

2. The contextualisation of the SLM

3. Methodology

4. Results