Examining Land-Use through GIS-Based Kernel Density Estimation: A Re-Evaluation of Legacy Data from the Berbati-Limnes Survey

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=yjfa20

ISSN: 0093-4690 (Print) 2042-4582 (Online) Journal homepage: https://www.tandfonline.com/loi/yjfa20

Examining Land-Use through GIS-Based Kernel

Density Estimation: A Re-Evaluation of Legacy Data from the Berbati-Limnes Survey

Anton Bonnier, Martin Finné & Erika Weiberg

To cite this article: Anton Bonnier, Martin Finné & Erika Weiberg (2019) Examining Land-Use through GIS-Based Kernel Density Estimation: A Re-Evaluation of Legacy Data from the Berbati- Limnes Survey, Journal of Field Archaeology, 44:2, 70-83, DOI: 10.1080/00934690.2019.1570481 To link to this article: https://doi.org/10.1080/00934690.2019.1570481

© 2019 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group

Published online: 22 Feb 2019.

Submit your article to this journal

Article views: 278

View Crossmark data

Examining Land-Use through GIS-Based Kernel Density Estimation: A Re-Evaluation of Legacy Data from the Berbati-Limnes Survey

Anton Bonnier , Martin Finné , and Erika Weiberg Uppsala University, Uppsala, Sweden

ABSTRACT

The use of archaeological survey data for evaluation of landscape dynamics has commonly been concerned with the distribution of settlements and changes in number of recorded sites over time.

Here we present a new quantitative approach to survey-based legacy data, which allows further assessments of the spatial con figuration of possible land-use areas. Utilizing data from an intensive archaeological survey in the Berbati-Limnes area, Greece, we demonstrate how GIS-based kernel density estimations (KDE) can be used to produce cluster-based density surfaces that may be linked to past land-use strategies. By relating density surfaces to elevation and slope, it is also possible to quantify shifts in the use of speci fic environments on a regional scale, allowing us to model and visualize land-use dynamics over time. In this respect, the approach provides more multifaceted information to be drawn from archaeological legacy data, providing an extended platform for research on human-environment interactions.

KEYWORDS landscape archaeology;

legacy data; archaeological GIS; kernel density estimation; archaeological survey; ancient land-use;

Berbati-Limnes survey

Introduction

Digital approaches o ffer new ways through which archaeolo- gical legacy data may be approached and re-evaluated to examine past landscape dynamics. In the current study, we demonstrate how GIS-based kernel density estimation (KDE), can be used as a suitable method to quantitatively assess shifts in land-use patterns by using distributions of sites from previously published archaeological datasets.

More speci fically, we test how density surfaces produced through KDE can be intersected with topographic data pro- vided by high resolution Digital Elevation Models (DEMs), and how these can be used to examine and visualize shifts in the use of speci fic environments over time.

Previous discussions on past landscape dynamics have often involved diachronic comparisons of site numbers, including more qualitative discussions of distribution patterns (for Greece, see for example Alcock [1993], Bintli ff [ 1997], Stewart [2013], and Weiberg and colleagues [2016]). Such approaches have been useful in illustrating settlement fluctuations linked to socio-economic developments, but are at the same time limited in the type of quantitative information they convey (Whitelaw 2000; Witcher 2008). Through the method outlined here, we demonstrate how KDE can be used to quantitatively assess land- scape dynamics, moving beyond diachronic comparison of site numbers. Given the simplicity of GIS-based KDE, as well as the subsequent intersecting applications demonstrated in the cur- rent study, it forms a suitable and easily applied method through which we can obtain new information that allows robust com- parisons of land-use patterns between periods.

In the current study, our method of GIS-based KDE will be exempli fied through analyses of data derived from the Ber- bati-Limnes Archaeological Survey (Wells and Runnels 1996), conducted by the Swedish Institute at Athens in the

late 1980s and early 1990s in the Berbati valley, which is situ- ated in the northern Argolid, in the Northeastern Pelopon- nese, Greece ( FIGURE 1 – 2). We will argue that KDE can provide additional perspectives on patterns of land-use in the Berbati valley and the neighboring Limnes area from pre- history to the 4th century A.D. , adding to the interpretations of site dynamics and distributions presented in the original publication of the survey results.

Mediterranean Survey, Legacy Data, and GIS

The method outlined in the current study has been speci fi- cally developed for evaluations of legacy data in the form of published site distributions recorded through archaeological field surveys. In Greece (as well as in other parts of the Med- iterranean region), a wide range of data has been made avail- able over the past decades through survey projects utilizing di fferent methods and employing varying degrees of spatial coverage (for a geographic overview of Peloponnesian survey projects, see Figure 1). Intensive surveys engaged in the sys- tematic recording of artifact scatters in plow zones provided a signi ficant advancement in regional research agendas from the 1970s and onwards (for projects forming part of the new wave of intensive survey in Greece in the 1970s and 1980s, see Bintli ff and Snodgrass [ 1985, 1988a], Wright and colleagues [1990], Cherry and colleagues [1991], Jameson and colleagues [1994], and Bintli ff and colleagues [ 2007]).

Intensive investigation methods provided ways to examine the distribution and location of “sites” as well as “off-site archaeology ” (i.e., scatters of surface artifacts between places de fined as sites [Bintliff and Snodgrass 1988b; Alcock et al.

1994; Bintli ff and Howard 1999; Bintli ff 2000; Pettegrew 2001; Bintli ff et al. 2002]). The published data have, however,

© 2019 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

CONTACT Anton Bonnier anton.bonnier@antiken.uu.se Department of Archaeology and Ancient History, Uppsala University, Box 626, 751 26 Uppsala, Sweden

JOURNAL OF FIELD ARCHAEOLOGY 2019, VOL. 44, NO. 2, 70 –83

https://doi.org/10.1080/00934690.2019.1570481

usually been concerned with the distribution of sites or localities carrying specific archaeological signatures (the semantic qualities attached to the word “site” itself have been much debated in the archaeological survey literature,

see for example Gallant [1986], Wells [1996a: 16–18], and Jameson and colleagues [1994: 221]). More recent survey pro- jects have also employed sophisticated spatial models for evaluating the distribution of both artifact scatters and sites.

Figure 1. Areas previously examined through systematic archaeological surveys in the Peloponnese (adapted from Pullen [2011] and Weiberg and colleagues [2016])

and the location of the Berbati-Limnes survey area. The map has been created using the ASTER GDEM, a product of METI and NASA.

At Antikythera in Southern Greece, for example, the results of intensive artifact level survey have been tested through spatial analyses, which demonstrate that archaeological locations (or sites) are not randomly distributed on the island but highlight the signi ficance of access to resources such as spring water, flat land, and lighter soils (Bevan and Conolly 2013: 106 –109). The spatial analysis employed at Antikythera further demonstrated how the patterning of archaeological artifacts in the landscape re flects multiple and complex aspects of habitation and land- use in di fferent periods (Bevan and Conolly 2013: 112 –157).

Legacy data derived from site-based surveys will not always provide such a high-resolution picture of artifact distributions as that available from Antikythera, but they can nevertheless be re-evaluated through spatial analyses that provide additional perspectives on land-use. More complex interpretations and quantitative evaluations of site distributions are highlighted by Whitelaw (2000), for example. In a paper focusing on survey data from the island of Keos, Whitelaw demonstrates how site data can be correlated with speci fic locations, incorporating di fferent topographic parameters (Whitelaw 2000: 234 –237).

The Keos study highlights the importance of quanti fication for comparisons of dynamics between time periods in a speci fic regional setting. A substantial di fference between Whitelaw’s method and our approach here is the use of GIS, providing pos- sibilities for more exact quanti fication of spatial configurations and the use of hypothetical land-use surfaces for di fferent periods. The quanti fications produced through GIS-based KDE further allow for diachronic analysis of changes in land- use systems utilizing identical digital approaches for the chrono- logically structured datasets.

GIS-based research within Greek archaeology has been growing at a steady pace in recent years, including studies

of legacy data and other approaches that move beyond field-based data recording as part of survey and excavations (Constantinidis 2001; Farinetti 2011; Donati 2016; Jazwa and Jazwa 2017; Argyriou et al. 2017). The use of GIS, of course, needs to be approached critically, since digital methods and mapping run the risk of masking uncertainties in the data and consequently in the archaeological interpret- ation (Sharon et al. 2004; Witcher 2008). It is therefore crucial to de fine the methods, assumptions, and theoretical frame- works informing such research.

In the current study, we follow a similar approach to that carried out by Argyriou and colleagues (2017) for di fferent areas of Bronze Age Crete. However, while their study was speci fically concerned with the relationship between sites and landforms, utilizing the topographic position index (TPI) of sites recovered through archaeological survey, we have focused on kernel density surfaces and examined how these can be used to understand shifts in land-use patterns through time. In part, this follows the recent calls by Gupta and Devillers (2017) for more time-sensitive visualization of archaeological data in GIS, although we do not primarily employ map-based visualizations of dynamics in the current study. Diachronic developments are instead illustrated through a series of graphs highlighting spatial dynamics.

Kernel Density Estimation (KDE) and Land-Use Patterns

The premise of KDE is to produce smoother visualization of break values for quantitative groups, building on the principle of a heat map distribution between core areas (kernels) and surrounding neighborhoods. KDE has increasingly been

Figure 2. The Berbati-Limnes area and ground covered through the intensive pedestrian survey. The map has been created using a 5 m DEM, acquired from the Greek National Cadaster and Mapping Agency Ktimatologio (Copyright © 2012 NATIONAL CADASTRE & MAPPING AGENCY S.A.).

72 A. BONNIER ET AL.

used within archaeology to examine the spatial distribution and frequency of both archaeological sites and artifacts in di fferent (global) contexts (Baxter et al. 1997; Wheatley and Gillings 2002: 186 –187; Conolly and Lake 2006: 175 –177;

McMahon 2007; Herzog and Yépez 2013; Lindholm et al.

2013; Sayer and Wienhold 2013).

GIS-based KDE is carried out using a radius input (some- times de fined as bandwidth) through which the various den- sity levels are calculated. The radius can be either manually or automatically de fined. For the present study, we have used a manual 2.5 km input for all time frames, which is based on previous discussions of an idealized catchment zone of agri- cultural communities, representing an approximate one- hour walking distance for farmers commuting to their fields (Bintli ff 2012: 271).

The first use of a defined catchment radius was presented by Vita-Finzi and Higgs, who argued for a catchment com- posed of 5 km for agriculturalists and a 10 km radius for hun- ter-gatherers in the Near East (Vita-Finzi et al. 1970). This type of catchment analysis was later reworked by Flannery (1976) for Mesoamerican villages, for which a radius of 2.5 km was proposed for mature agricultural landscapes.

The 2.5 km radius has subsequently been used to represent a general idealized catchment for Greek agricultural settle- ments, particularly by Bintli ff and others working with settle- ment archaeology and survey data in the central Greek region of Boiotia (Bintli ff 1999, 2012: 271; Farinetti 2011: 42 –43).

The 2.5 km radius should not, however, be understood as the de fined catchment area of the individual sites included in the KDE. Speci fic catchment areas should be defined using more re fined site-based cost-distance application within GIS, as is highlighted by recent research (Farinetti 2011;

Becker et al. 2017). The radius instead provides a spatial fra- mework speci fically used for the purpose of GIS-based KDE, establishing an outer limit to the density estimation and using the input as a measure through which we can model and compare density patterns over time.

Density levels and “Extent of Possible Land-use” (EPLU) Based on the KDE heat maps produced through the period speci fic site inputs, we have extracted hypothetical land-use surfaces, each de fined as the “Extent of Possible Land-Use”

(EPLU). The EPLUs represent a possible land-use surface based on the clustering of archaeological sites, measured according to the de fined radius input, without any regard to the interpreted function or status of the sites. For the cur- rent study, two versions of ESRI ArcMap (10.3 and 10.5) have been used to perform all stages of the analysis. For the first step, we used the Kernel Density tool available in the Spatial Analyst tool box. The KDE rasters were then re-classi fied into a three-level distribution —consisting of a maximum extent, medium extent, and high-density areas —using natural breaks division (so-called Jenks), available in the ArcMap software, which divides data into groups inherent in the numerical data, based on the arrangement of values into groups follow- ing deviations from group means. The data are divided into uneven classes and separated according to signi ficant changes (i.e., the natural breaks) in the data (Jenks and Caspall 1971).

Within this three-tiered division, the “maximum extent” cor- responds to the full area of the kernel, the “medium extent”

forms the second tier of the kernel, and the “high-density area ” corresponds to the identified hotspots within the heat

maps. For further analysis and data extraction, each density surface was transformed into a polygon using the Raster to Polygon tool.

In the current study, no weighting was performed in the KDE, either by site size or by function, since our aim here was to examine how we can explore land-use dynamics based on site density and the presence of varying site clusters rather than densities of human populations. The heat maps produced through KDE should therefore not be understood as representing relative densities of people living in the land- scape but rather the density of human activity (recoverable through the imprint of archaeological sites) available in the de fined geographical context. With size-weighted distri- butions, the area of the maximum extent would remain unchanged (compared to the result of the current analysis), while areas of the medium extent and, in particular, high- density areas would be altered due to the increased weight of some of the points within the kernel. The unweighted heat maps nevertheless demonstrate that larger sites are gen- erally located within the high-density areas, highlighting a non-random distribution of site clusters.

Case-Study: The Berbati-Limnes Survey

The Berbati-Limnes Archaeological Survey was carried out by the Swedish Institute at Athens between 1988 –1990 (Wells et al. 1990; Wells and Runnels 1996), in the Berbati Valley and in the neighboring upland valley to the east, surrounding the modern village of Limnes, in the northern part of the Argolid ( FIGURE 2). The survey was motivated by a wish to add a regional perspective to the results from earlier exca- vations at the site of the Mastos Hill initiated by Swedish archaeologists in the 1930s (Sä flund 1965; Åkerström 1968, 1987). A further rationale behind the project was to docu- ment archaeological surface evidence threatened by rapid advances in mechanized agriculture (Wells 1996a).

The survey itself was performed on a field-by-field basis, where field walking was carried out at a 10 or 15 m interval between field walkers. The defined survey universe consisted of approximately 61 km

, out of which approximately 25 km

were investigated intensively (Wells 1996a: 16). Very little o ff- site archaeology was recorded and the catalog of find spots for each period provides the primary evidence for human activity in the area in antiquity (Wells 1996a: 18). The original Ber- bati-Limnes survey was further complemented with an inten- sive artifact level investigation of the Mastos Hill in 1999, which added detailed information on the extent and chronol- ogy of habitation on this particular location (Lindblom and Wells 2011). The site of Mastos has therefore been included in the KDE analysis carried out in the current study, even though the Mastos Hill was not surveyed during the course of the fieldwork carried out in the late 1980s.

Findspots and chronological resolution

As a prerequisite for the GIS analyses, we have structured the

digitized site distributions according to a number of de fined

time frames corresponding to the reported relative archaeolo-

gical periods. The data collection and analysis in the current

study have formed part of a broader initiative of synthesizing

Peloponnesian site data dating from the Neolithic until the

Middle Roman period and we have used the same chrono-

logical breaking points for the current study. We have

therefore performed KDE analyses for 19 “time frames” (i.e., period-de fined distribution of sites used for the spatial analy- sis) that are based on the reported chronology for the findspots, and which span a period from the Early Neolithic until the Middle Roman (ca. 6800 B.C.– 300 A.D. ) ( FIGURE 3).

Within this broad chronological focus, there are also periods where the lack of recorded archaeological remains suggests hiatuses in activity in the Berbati valley. The survey could not identify any secure findspots for the period between EH III and LH II periods (ca. 2200–1420 ^B.C. ), even though material of LH IIB date was recorded at low quantities in some tracts (Schallin 1996: 169). The gap is filled by the later intensive survey on the Mastos Hill (not included in the original survey), during which abundant pot- tery dating to these periods was recorded (Lindblom and Wells 2011). These results suggest that Mastos was the only active site for this long period of time. A combined single point analysis was therefore performed for these periods, based on the Mastos Hill. The EPLU does corre- spond to the de fined radius input and no real emphasis should therefore be placed on the results for this frame. In this sense, a cost-distance based catchment analysis of the Mastos Hill would be more informative, but the KDE/

EPLU approach still produces data that allow their incorpor- ation in the long-term model of land-use dynamics of the region.

In addition, very little secure evidence can be identi fied for the LH IIIC to Middle Geometric periods (ca. 1200–750 ^B.C. ) in the Berbati valley, suggesting a hiatus at least of permanent habitation and land-use (Wells 1996b). These periods have

therefore not been incorporated into the current analysis and are given null values in the quanti fication of the spatial extent derived from the KDE.

Patterns of nucleation and dispersal

The heat maps produced as part of the KDE are exempli fied in Figure 4, showing the results from the analysis of some selected time frames, which help to visualize patterns of dis- persal and nucleation of land-use systems. The assessment of spatial shifts based on the heat-maps will, however, remain primarily qualitative in terms of the resulting analysis. A strength of GIS-based density estimation is that we can also extract spatial values from each EPLU (we have consistently used hectares as the preferred spatial measure), allowing us to track quantitatively the spatial changes occurring in the EPLUs over time, as is highlighted in Figure 5.

Previous assessments of nucleation and dispersal rates in other areas of the Greek mainland have largely been based on changes in the number of sites recorded in survey areas, in addition to more qualitative discussions on site distri- butions according to physical topography and environmental factors (Alcock 1993; Bintli ff 1997). An increase in site num- bers has usually been equated with a dispersed pattern of settlement and a reduction in site numbers has been associ- ated with a more nucleated pattern (Jameson et al. 1994:

252 –257).

By comparing the extent of density levels with the shifts in site numbers, we can state that the dynamics of the EPLUs over time are generally matched by the overall trend in site

Figure 3. Time frames used for the GIS based KDE and chronological information. The time frame labeled LH II also represents the values for the Early Helladic III, Middle Helladic I and II and III, as well as Late Helladic I, since the only active site during this time occurs at the Mastos Hill. For the prehistoric periods (Neolithic to Late Bronze Age) we have used the absolute dates presented by Manning (2010), while for the historical periods (Late Geometric to Early and Middle Roman) we have primarily used the absolute dates provided by the final report of the Berbati-Limnes survey (Ekroth 1996; Penttinen 1996; Forsell 1996). EBA = Early Bronze Age, LBA = Late Bronze Age.

74 A. BONNIER ET AL.

quantities, but we can also observe some interesting variabil- ity between periods. In time frames for which site numbers have been recorded in similar quantities, we can instead

note variable trends of expansion and contraction, as well as nucleation and dispersal as highlighted by the KDE. Site quantities are, for example, relatively similar for the EA,

Figure 4. Examples of site clusters and resulting EPLUs (all of the individual time frame density maps produced as part of the KDE are not shown here). Four pre-

historic time frames (A –D) and four historical time frames (E–H) are illustrated. The map has been created using a 5 m DEM, acquired from the Greek National

Cadaster and Mapping Agency Ktimatologio (Copyright © 2012 NATIONAL CADASTRE & MAPPING AGENCY S. A.).

Figure 5. Quanti fication of the Berbati-Limnes EPLUs (in hectares) based on the KDE compared with the site quantities for each time frame. A) Long term variability in extent according to the full duration of all time frames. B) Dynamics in land-use extent between the Late Bronze Age and the Roman period.

Figure 6. Quanti fication of the Berbati-Limnes EPLUs (in hectares) based on the KDE according to absolute chronology.

76 A. BONNIER ET AL.

LA, HL, and ER–MR time frames (between 9 and 11 sites per time frame), but the EPLUs, particularly in the medium and maximum extent, vary distinctively owing to the fluctuating distribution of sites in the valley and in the neighboring Limnes uplands ( FIGURE 5).

A period in which we can observe a reduction in site num- bers can still maintain a dispersed distribution of sites, as in the case of the HL time frame, while a period with a similar number of sites (for example the EA and LA time frames in terms of the Berbati Valley) can provide a much more nucleated distribution. In terms of the Late Bronze Age time frames, we can see that although the EPLU for LH IIIA2–LH IIIB1–LH III B2 decrease with the site numbers, the decrease in the EPLU is less dramatic. While the site num- bers for LH IIIB2 are less than half of those from the LH IIIA2 time frame, much of the EPLU is retained, highlighting that the dispersal of the remaining sites remained high.

The current method thus provides a more nuanced picture of the rate of dispersal and nucleation, as the resulting quantification is based on the spatial configuration of site dis- tribution rather than the number of sites. It should be noted, however, that the values presented here are ordered according

to periods de fined by relative chronology and not in time slices of equal length ( FIGURE 5). The division according to relative archaeological periods means that changes in land- use patterns may appear to be more rapid and dynamic when structured as part of relative archaeological periods.

The long extent of the Neolithic and Early Bronze Age (between 6800 B.C. and 2200 B.C. ), in flates the results com- pared with later periods ( FIGURE 6), and in the case of the Neolithic to Early Bronze Age time frames, sites were prob- ably not in continuous use. We have therefore focused our subsequent analysis to the period between 1450 B.C. and

A.D. 300, which mostly consists of time frames of roughly similar length ( FIGURES 3, 6).

Intensi fication and extensification

The rate of dispersal and nucleation was further explored via a land-use density index, which was determined by dividing the di fferent density surfaces of the EPLUs by the number of sites within each time frame ( FIGURE 7). Low values will consequently suggest more nucleated land-use, while a higher value is indicative of a more dispersed land-use. Overall, the broad trends in the Berbati-Limnes data suggests increasing land-use density in most of the historical time frames, owing to a generally more nucleated EPLUs compared to the late Bronze Age frames ( FIGURE 7).

Both the extent of EPLUs and the land-use density index can be used to suggest long-term dynamics in terms of spatial intensi fication and extensification, which have previously been discussed for agricultural dynamics in Mediterranean landscapes, particularly for prehistoric periods (Halstead 1999; van der Veen 2005; Currie et al. 2015; Bogaard 2017;

Styring et al. 2017). In this context, spatial intensi fication may be understood as a higher degree of investment of labor (per hectare) into agricultural land-use, resulting in a smaller area that is more heavily worked. Extensi fication should instead be understood as a process of expansion into the landscape utilizing less labor-intensive agriculture, result- ing in a lower input/output per hectare, but possibly provid- ing a higher overall output (van der Veen 2005: 158).

These results from the KDE suggest a scenario favoring, or at the very least enabling, extensive agricultural strategies in the prehistoric time frames. Conversely, given the lower values on the density index for the historical time frames, this suggests a scenario that would usher in an increasing use of more labor-intensive agricultural strategies. The alternative is that during the historical period, people utilized land farther away from the settlement to a greater degree. The e ffect of the potentially lower preservation rate of prehistoric ceramic should not be overlooked and may have a ffected the quantity of sites recorded during the survey (Bintli ff et al.

2003; Davis 2004), but the KDE results provide both new and multifaceted data to expand such contextualization.

EPLU and topography

After the initial calculations of the EPLUs, the second step in the KDE method employed here is an analysis of how EPLUs are spatially distributed in relation to physical topography, using data extracted from a digital elevation model (DEM).

The topographic data used for the current analysis has been derived from a 5 m resolution DEM, acquired from the Greek National Cadaster and Mapping Agency Ktimatologio.

Figure 7. Land-use density index for each time frame, based on the correlation

between EPLU and site numbers a ffecting the density distribution. A) Density

index for the maximum extent. B) Density index for the medium extent. C) Den-

sity index for the high-density areas.

All surfaces de fined by the EPLU in the different time frames were vectorized as shape files (.shp), which were subsequently correlated with topographical data using the Intersect tool available in ArcMap. Although other types of data could easily be used for the intersect analysis, we will here focus on the correlation between EPLU surfaces and rasters with de fined elevation and slope values (extracted as polygons for the intersect analysis).

ELEVATION

For the analysis of the distribution of EPLUs in relation to elevation, we created a raster with a de fined 50 masl equidis- tance that could be intersected with the extracted density sur- faces for the various time frames. The dynamics of the EPLUs regarding elevation point to a general trend of increased land- use at higher elevation in periods of substantial land-use expansion in the maximum and medium extent, as is visible for the LH IIIA2 (1390 –1330 ^B.C. ) to LH IIIB1 (1330 –1250

B.C. ) frames, and in the LCL –EHL (350–250 ^B.C. ) frame ( FIGURE 8).

As expected, the highest degree of variability can be found in the maximum extent of the EPLUs, while the high-density areas reveal a much more static picture in terms of elevation being used. This re flects a long-term focus of high-density site clusters within the Berbati Valley rather than the surveyed uplands to the east in the Limnes area. Moving away from the maximum extent, land at higher elevation is generally reduced in all time frames, but in the medium extent, the trend of expansion into land at higher elevation is still visible for periods with visible expansion of land-use areas.

It is important to stress here that land at high elevation is not the same as a high gradient landscape. High elevation

areas may consist of flat ridges, plateaus, or upland valleys situated at a higher elevation. The comparison between EPLUs and elevation therefore provides little direct infor- mation on the use of marginal areas with thinner soil cover.

A separate analysis has therefore been performed to correlate the EPLUs with the degree of slopes within them.

SLOPE VALUES

For the investigation of correlation between EPLUs and slope, we created a slope raster based on the DEM data, which was subsequently reclassi fied. For a first analysis we employed the de finitions previously employed by Farinetti (2011: 17) for landform classi fications in Boiotia, Central Greece. This classi fication system is constructed around five different slope categories ranging from flat or nearly flat ground to very steep slopes ( ^FIGURE 9). For a second analysis, we utilized the slope classi fication provided by Whitelaw (2000: 234), in which the division of categories has been based on the impact of slopes on plowing and the potential use of terracing ( FIGURE 10). The classi fication used by Whitelaw makes a clear distinction between land that can be suitably plowed without terracing (Slope class 1), land that may have been cultivated without terracing with an inherent risk of soil erosion (Slope class 2), and land requiring terracing for cultivation (Slope class 3). The slope classi fication presented by Whitelaw, while less detailed than the one based on Farinetti ’s ( 2011) landform classi fication, is more directly connected to agricultural strategies and land-use.

To a certain extent, the high-gradient land in the EPLUs will simply re flect the physical topography of the area. The Berbati Valley is bounded by steep slopes formed by the

Figure 8. Distribution of land according to elevation in the Berbati-Limnes data. A) Trends observed for the Prehistoric time frames. B) Trends observed for the Late Geometric, Early and Late Archaic time frames. C) Trends observed for the Classical to Roman period time frames.

78 A. BONNIER ET AL.

Psili Rachi range to the north and the Euboia (Prophitis Elias) and Rachi Kalogirou peaks to the south. Steep slopes are also significantly present in the uplands surrounding Limnes. An effect of this topographic situation is that the proportion of high-gradient land in the maximum extent is still high in periods with a more nucleated distribution ( ^FIGURES 9, 10).

However, this pattern is not maintained at all density levels.

In the medium extent and high-density areas, the amount of steep land is significantly reduced in the more nucleated frames, demonstrating that high-gradient areas primarily form part of the outer edges of EPLU in the maximum extent.

In the context of the Berbati-Limnes data, therefore, the med- ium extent provides the most accurate picture of dynamics since it reduces the amount of marginal land at the edges of the EPLUs. In the high-density areas, low-gradient land always forms the greatest part of the EPLU, but an increase in the amount of steeper ground is still visible for the dis- persed distributions, for example in the LCL–EHL time frame. For purposes of comparison between time frames, the medium extent and high-density areas provide the most meaningful observations.

Previous discussions of site distribution have often been focused on the location of individual sites and their geo- graphical positions according to factors such as slope, elevation, soil types, and geology (Jameson et al. 1994: 257–

258; Mee and Forbes 1997; Stewart 2013). By using estimated density surfaces in the form of the EPLUs for this type of analysis, we are not merely evaluating the topographical cir- cumstances of individual site locations, but also those of site clusters to extrapolate the proportion of topographic classes within the EPLU. The topographically-defined EPLUs thus provide data through which we can observe more large-scale changes in the use of more marginal soils situated on sloping ground. It is important to stress that the presence of land within an EPLU will not always signify actual usage. However, the available land is a measure of the types of land that could be used and in the medium extent and high-density areas more likely were used. In extension, the type of land-use and the total pressure of land-use has bearing on the resources needed to use/farm that land and vulnerabilities such usages may have inferred on societies.

The topographically defined EPLUs allow us to re-evaluate and expand upon the conclusion reached in final report of the Berbati-Limnes survey. One example can be seen in terms of the distribution pattern of sites dating to the Late Bronze Age.

Many of the recorded LH IIIA2–LH IIIB1 findspots were located on slopes (FS43, FS44, FS12), including findspots situ- ated at a low elevation (FS428) but in a high-gradient environment (Schallin 1996). The favoring of sloping and/

or elevated terrain (e,g., mounds or knolls) for settlements

Figure 9. Slope variation in the Berbati-Limnes data using an adapted version of Farinetti ’s ( 2011: 17) slope classi fication. A, C, and E) Line graph of hectare values

projected according to an absolute time scale based on the duration of the time frames in the maximum extent (A), medium extent (C), and high-density areas (E). B,

D, and F) Proportion (%) of the di fferent slope values in each time frame (defined according to the relative archaeological period) in the maximum extent (B), medium

extent (D), and high-density areas (F). The proportion of each slope category is established using the hectare value of the speci fic slope class in relation to the extent

of each EPLU for the di fferent density levels. The proportions given in B are therefore dependent on the values provided in A, the proportions given in D are depen-

dent on the values provided in C, and the proportions given in F are dependent on the values provided in E.

brings to the fore the sometimes reappearing interpretation that sloping ground was sought to keep the plains available for cultivation (see, for example, Forsén [1996: 117], who argues for the presence of this type of settlement strategy in the Berbati Valley during the Early Helladic period).

Although the location of a settlement certainly can have econ- omic motivations, as well as both functional (defense, con- trol) and cognitive ones (Weiberg 2011), the suggestion that the choice of settlement location was driven by a wish to maximize farmland seems to be negated by the correlation of Roman period sites with flat land. Investment in agricul- ture including cash-cropping seems to have occurred in di fferent parts of the Greek mainland during the Roman period (Rizakis 2013), and we can therefore assume that the Berbati Valley would have been part of similar land-use sys- tems. If site locations were primarily determined by a wish to maximize available farmland, we would expect to find the estate situated on sloping ground at the edge of the plain.

The KDE approach further demonstrates that site location alone will only provide one piece of the puzzle. The potential increases in use of high-gradient land in expansive time frames suggest an increasing use of more marginal soils, in a pattern that goes beyond the locations of the sites them- selves. Furthermore, the greater inclusion of high-gradient land also in the medium extent EPLUs increases the likeli- hood that sloping ground was used for farming. The higher

proportion of sloping ground in the medium extent also increases the likelihood that agricultural terracing was in use in the Berbati-Limnes area during the latter part of the Bronze Age and in the Late Classical and Hellenistic periods.

Early Modern and more recent terrace structures form a recognizable part of the slopes surrounding the valley and in the uplands bordering on the Limnes Plateau (Wells 1996a), but firm evidence for ancient agricultural terracing is largely absent and widespread use of terracing in ancient Greek agriculture has been questioned within previous research (Foxhall 1996, 2007). The reinterpretation of the Berbati-Limnes data presented here nevertheless gives some weight to the use of terracing, since the expansion of land- use into high-gradient land during boom periods in the Ber- bati-Limnes area would otherwise have been di fficult to main- tain (Whitelaw 2000: 234). The results of the KDE thus suggest that that terraces would have been needed in the Ber- bati-Limnes region in certain periods, such as LH IIIA2 and LH IIIB1, as recent studies have proven them to be in the rela- tively nearby area of Kalamianos to the northeast (Kvapil 2012).

Conclusion

GIS-based KDE provides a useful tool for the quanti fication of possible land-use dynamics and can facilitate a more

Figure 10. Slope variation in the Berbati-Limnes data using Whitelaw ’s ( 2000: 234) tripartite slope classi fication. A, C, and E) Line graph of hectare values projected according to an absolute time scale based on the duration of the time frames in the maximum extent (A), medium extent (C), and high-density areas (E). B, D, and F) Proportion (%) of the di fferent slope values in each time frame (defined according to the relative archaeological period) in the maximum extent (B), medium extent (D), and high-density areas (F). The proportion of each slope category is established using the hectare value of the speci fic slope class in relation to the extent of each EPLU for the di fferent density levels. The proportions given in B are therefore dependent on the values provided in A, the proportions given in D are dependent on the values provided in C, and the proportions given in F are dependent on the values provided in E.

80 A. BONNIER ET AL.

nuanced and multifaceted discussion of boom-and-bust pat- terns occurring over time. The KDE approach allows us to visualize changes in the spatial extent and shifts in the topo- graphic context of possible land-use over time, both through the GIS-generated heat maps but also through graphs plotting quantitative changes in the spatial extent of possible land-use.

The method employed here has been developed with an expli- cit comparative intent. In the present study, comparisons are performed on a temporal scale, indicating di fferences between the prehistorical and historical periods. Through KDE, we can thus start to extract patterns of variability in land-use con figurations between periods, which will provide new input to discussions on changing agricultural strategies and settlement dynamics. For example, this approach provides new quantitative data to complement earlier assessments of nucleation and dispersal patterns.

The analysis of the EPLUs in relation to topography further demonstrates how new and complementary types of land-use data can be extracted through the GIS-based KDE.

By incorporating the slope and elevation data derived from the high-resolution DEMs, we can track and quantify changes in the topographical context of land-use in di fferent periods.

In general, the integration of slope values into the di fferent levels of the EPLU provides the most interesting aspect of the potential shifts in land-use patterns, since we get a picture of dynamics relating to the use of sloping ground and conse- quently the use of more marginal soils and terraces in di ffer- ent periods.

In a second stage, this perspective should be extended by comparisons on a geographical scale, between di fferent regions and survey datasets that have been produced through di fferent intensive and extensive field methodologies. By uti- lizing density surfaces (here represented as the EPLUs) related to the clustering of archaeological sites (not only sites de fined as settlements) we have sought to minimize some of the problems associated with comparing site quan- tities and integrating digitized legacy data. Ideally, we would further like to compare results of the KDE utilizing site inputs with o ff-site distributions recorded through inten- sive field walking. In the case of the Berbati-Limnes data, such information is not available in the final report limiting poss- ible comparison with the result of the KDE.

Other intensive surveys in Greece where the distribution of o ff-site scatters has been more thoroughly published suggest that there is usually a de fined presence of material surrounding both settlements and smaller rural sites, usually in the form of halos that tend to drop o ff at varying distances away from the recorded sites (Bintli ff et al. 2007: 23 –26).

However, these distributions do not always provide the full geographical range of land-use associated with recorded sites, but may rather re flect the presence of more intensive land-use areas such as in fields and gardens (Bintliff et al.

2007: 23 –26; Winther-Jacobsen 2010: 271 –272; Forbes 2013) for the complexities of interpreting o ff-site distri- butions as part of ancient agricultural processes). KDE ana- lyses of recorded site distributions therefore o ffer complementary perspectives of possible land-use patterns and allow for possible evaluation of EPLUs in regards to the picture presented by o ff-site scatters.

In summary, the density surfaces and the type of quanti- tative data produced through the GIS-based KDE provide information related not merely to habitation patterns but also to aspects of land-use dynamics and ultimately the

potential human pressure on the landscape in di fferent periods. Such quanti fications have considerable bearing on our understanding of the resilience of communities within these regions. Recent research has demonstrated the impor- tance of incorporating high-resolution paleoenvironmental records into comparative e fforts involving survey data to facilitate a better understanding of how changes in climate and environment may have impacted societies, and vice versa (Weiberg et al. 2016). Such integrated approaches demand quanti fication of possible land-use dynamics, going beyond changes in site and settlement numbers. GIS-based density tools such as KDE o ffer ways in which such new quantitative data can be created also on the basis of legacy data. The present study shows how GIS-based analysis of the spatial con figuration of possible land-use can provide the basis for more multifaceted quanti fications and open new ways to explore the linkages between environmental conditions, agricultural strategies and socio-economic transformations.

Acknowledgments

The methods of GIS-based KDE described in the current study was developed and carried out within the framework of the Domesticated Landscapes of the Peloponnese (DoLP) Project hosted at the Department at Archaeology and Ancient History, Uppsala University and generously funded by the Swedish Research Council (Grant #421-2014-1181). The project has an interdisciplinary scope and the aim is to investigate long-term human-environment interactions in the Peloponnese, from the Neolithic until the 4th century

. The initial plotting of georefer- enced site data was originally carried out by Emanuel Savini within the framework of the 1999 Mastos Survey directed by Berit Wells. We are also grateful towards the Greek National Cadastre and Mapping Agency, Ktimatologio, for providing us with high resolution DEM data of the Peloponnese (Copyright © 2012 NATIONAL CADASTRE

& MAPPING AGENCY S. A.). We are further grateful for the comments provided by the three anonymous reviewers, who provided good input for improving the original manuscript. Any mistakes or idiosyncrasies remain our own.

Disclosure Statement

No potential con flict of interest was reported by the authors.

Notes on Contributors

Anton Bonnier (Ph.D. 2010, Stockholm University) is a researcher at the Department of Archaeology and Ancient History, Uppsala University.

He is a classical archaeologist and ancient historian focusing on land- scape archaeology, GIS, and digital applications in archaeology, as well as the economic and environmental history of Classical, Hellenistic, and Roman Greece. He is heading the GIS and database development section of the Domesticated Landscapes of the Peloponnese (DoLP) project.

Martin Finné (Ph.D. 2014, Stockholm University) is a researcher at the Department of Archaeology and Ancient History, Uppsala University.

His main research focus is on Holocene climate in the eastern Mediter- ranean and the paleoclimatology and socio-environmental dynamics of southern mainland Greece. He is heading the paleoclimate section of the Domesticated Landscapes of the Peloponnese (DoLP) project.

Erika Weiberg (Ph.D. 2007, Uppsala University) is a researcher at the Department of Archaeology and Ancient History, Uppsala University.

She is an Aegean prehistorian with a research focus on landscape and

settlement archaeology of mainland Greece, including studies of socio-

politically transformative periods and human –environment–climate

dynamics, among other things. She is PI of the Domesticated Landscapes

of the Peloponnese (DoLP) project, which seeks to integrate archaeology

and history with the paleoenvironmental records from the region and

take as such an active part in the development of new methods for fruit- ful interdisciplinary collaboration.

ORCID

Anton Bonnier http://orcid.org/0000-0002-6386-5293 Martin Finné http://orcid.org/0000-0001-7433-268X Erika Weiberg http://orcid.org/0000-0001-6583-387X

References

Åkerström, Å. 1968. “A Mycenaean Potter’s Factory at Berbati near Mycenae. ” In Atti e Memorie Del 1. Congresso Internazionale Di Micenologia, Roma 27 Settembre-3 Ottobre 1967, 48 –53. Incunabula Graeca 25. Rome: Ed. dell ’Ateneo.

Åkerström, Å. 1987. Berbati. 2: The Pictorial Pottery. Skrifter Utgivna Av Svenska Institutet i Athen 4

36, 2. Stockholm: Åström.

Alcock, S. E. 1993. Graecia Capta: The Landscapes of Roman Greece.

Cambridge: Cambridge University Press.

Alcock, S. E., J. F. Cherry, and J. L. Davis. 1994. “Intensive Survey, Agricultural Practice and the Classical Landscape of Greece. ” In Classical Greece. Ancient Histories and Modern Archaeologies, edited by I. Morris, 137 –170. New Directions in Archaeology. Cambridge, UK: Cambridge University Press.

Argyriou, A. V., R. M. Teeuw, and A. Sarris. 2017. “GIS-Based Landform Classi fication of Bronze Age Archaeological Sites on Crete Island.”

PLOS ONE 12 (2): e0170727.

Baxter, M. J., C. C. Beardah, and R. V. S. Wright. 1997. “Some Archaeological Applications of Kernel Density Estimates. ” Journal of Archaeological Science 24 (4): 347 –354.

Becker, D., M. de Andrés-Herrero, C. Willmes, G.-C. Weniger, and G.

Bareth. 2017. “Investigating the Influence of Different DEMs on GIS-Based Cost Distance Modeling for Site Catchment Analysis of Prehistoric Sites in Andalusia.” ISPRS International Journal of Geo- Information 6 (12): 36.

Bevan, A., and J. Conolly. 2013. Mediterranean Islands, Fragile Communities and Persistent Landscapes: Antikythera in Long-Term Perspective. Cambridge: Cambridge University Press.

Bintliff, J. 1997. “Regional Survey, Demography, and the Rise of Complex Societies in the Ancient Aegean: Core-Periphery, Neo- Malthusian, and Other Interpretive Models. ” Journal of Field Archaeology 24 (1): 1–38.

Bintli ff, J. 1999. “Settlement and Territory.” In The Routledge Companion Encyclopedia of Archaeology, edited by G. Barker, 505 – 545. London: Routledge.

Bintli ff, J. 2000. “The Concepts of ‘Site’ and ‘Off Site’ Archaeology in Surface Artefact Survey. ” In Non-Destructive Techniques Applied to Landscape Archaeology, edited by M. Pasquinucci and F. Trement, 200 –215. Oxford: Oxbow Books.

Bintli ff, J. 2012. The Complete Archaeology of Greece: From Hunter- Gatherers to the 20th Century A.D. Chichester, UK: John Wiley &

Sons.

Bintli ff, J., E. Farinetti, P. Howard, K. Sarri, and K. Sbonias. 2002.

“Classical Farms, Hidden Prehistoric Landscapes and Greek Rural Survey: A Response and an Update. ” Journal of Mediterranean Archaeology 15 (2): 259 –265.

Bintli ff, J., and P. Howard. 1999. “Studying Needles in Haystacks. Surface Survey and the Rural Landscape of Central Greece in Roman Times. ” Pharos 6: 51–91.

Bintli ff, J., P. Howard, and A. M. Snodgrass. 2007. Testing the Hinterland: The Work of the Boeotia Survey (1989-1991) in the Southern Approaches to the City of Thespiai. McDonald Institute Monographs. Cambridge, UK: McDonald Institute of Archeological Research.

Bintliff, J., and A. Snodgrass. 1988a. “Mediterranean Survey and the City. ” Antiquity 62 (234): 57–71.

Bintliff, J., and A. Snodgrass. 1988b. “Off-Site Pottery Distributions: A Regional and Interregional Perspective. ” Current Anthropology 29 (3): 506 –513.

Bintliff, J., and A. M. Snodgrass. 1985. “The Cambridge/Bradford Boeotian Expedition: The First Four Years. ” Journal of Field Archaeology 12 (2): 123 –161.

Bogaard, A. 2017. “The Archaeology of Food Surplus.” World Archaeology 49 (1): 1 –7.

Cherry, J. F., J. L. Davis, and E. Mantzourani. 1991. Landscape Archaeology as Long-Term History: Northern Keos in the Cycladic Islands from Earliest Settlement Until Modern Times. Monumenta Archaeologica 16. Los Angeles: UCLA Institute of Archaeology.

Conolly, J., and M. Lake. 2006. Geographical Information Systems in Archaeology. Cambridge Manuals in Archaeology. Cambridge, UK:

Cambridge University Press.

Constantinidis, D. 2001. “Introspective Sitescaping in GIS.” In Computing Archaeology for Understanding the Past: CAA 2000:

Computer Applications and Quantitative Methods in Archaeology:

Proceedings of the 28th Conference, Ljubljana, April 2000, edited by Z. Stan čič and T. Veljanovski, 165–172. BAR International Series 931. Oxford, UK: Archaeopress.

Currie, T. E., A. Bogaard, R. Cesaretti, N. Edwards, P. Francois, P.

Holden, D. Hoyer, A. Korotavev, J. Manning, J. C. Moreno Garcia, O. K. Oyebamiji, C. Petrie, P. Turchin, H. Whitehouse, and A.

Williams. 2015. “Agricultural Productivity in Past Societies: Toward an Empirically Informed Model for Testing Cultural Evolutionary Hypotheses. ” Cliodynamics: The Journal of Quantitative History and Cultural Evolution 6 (1): 24 –56.

Davis, J. L. 2004. “Are the Landscapes of Greek Prehistory Hidden? A Comparative Approach.” In Side-by-Side Survey: Comparative Regional Studies in the Mediterranean World, edited by S. E. Alcock and J. F. Cherry, 19 –32. Oxford: Oxbow Books.

Donati, J. C. 2016. “The Identification of Land Divisions in the Ancient Mediterranean through Integrated Remote Sensing and GIS Applications. ” LAC Proceedings 2014. https://doi.org/10.5463/lac.2014.17.

Ekroth, G. 1996. “The Late Geometric and Archaic Periods.” In The Berbati-Limnes Archaeological Survey, 1988-1990, edited by B.

Wells and C. N. Runnels, 179 –227. Skrifter Utgivna Av Svenska Institutet i Athen 4° 44. Stockholm: Svenska Institutet i Athen.

Farinetti, E. 2011. Boeotian Landscapes: A GIS-Based Study for the Reconstruction and Interpretation of the Archaeological Datasets of Ancient Boeotia. BAR International Series 2195. Oxford:

Archaeopress.

Flannery, K. V., ed. 1976. The Early Mesoamerican Village. Studies in Archeology. New York: Academic Press.

Forbes, H. 2013. “Off-Site Scatters and the Manuring Hypothesis in Greek Survey Archaeology: An Ethnographic Approach. ” Hesperia 82 (4): 551 –594.

Forsell, R. 1996. “The Roman Period.” In The Berbati-Limnes Archaeological Survey, 1988-1990, edited by B. Wells and C. N.

Runnels, 285 –343. Skrifter Utgivna Av Svenska Institutet I Athen 4° 44. Stockholm: Svenska institutet i Athen.

Forsén, J. 1996. “The Early Helladic Period.” In The Berbati-Limnes Archaeological Survey, 1988-1990, edited by B. Wells and C. N.

Runnels, 75 –120. Skrifter Utgivna Av Svenska Institutet i Athen 4° 44. Stockholm: Svenska institutet i Athen.

Foxhall, L. 1996. “Feeling the Earth Move: Cultivation Techniques on Steep Slopes in Classical Antiquity. ” In Human Landscapes in Classical Antiquity: Environment and Culture, edited by G. Shipley and J. Salmon, 44 –67. Leicester-Nottingham Studies in Ancient Society 6. London: Routledge.

Foxhall, L. 2007. Olive Cultivation in Ancient Greece: Seeking the Ancient Economy. Oxford: Oxford University Press.

Gallant, T. W. 1986. “‘Background Noise’ and Site Definition: A Contribution to Survey Methodology. ” Journal of Field Archaeology 13 (4): 403 –418.

Gupta, N., and R. Devillers. 2017. “Geographic Visualization in Archaeology. ” Journal of Archaeological Method and Theory 24 (3):

852 –885.

Halstead, P. 1999. “Surplus and Share-Croppers: The Grain Production Strategies of Mycenaean Palaces. ” In Meletemata: Studies in Aegean Archaeology Presented to Malcolm H. Wiener as He Enters His 65th Year, edited by P. P. Betancourt and M. H. Wiener, 319 –326.

Aegaeum 20. Liège: Université de Liège.

Herzog, I., and A. Yépez. 2013. “Least-Cost Kernel Density Estimation and Interpolation-Based Density Analysis Applied to Survey Data. ” In CAA2010: Fusion of Cultures: Proceedings of the 38th Annual Conference on Computer Applications and Quantitative Methods in Archaeology, Granada, Spain, April 2010, edited by F. Contreras, M.

Farjas, and F. J. Melero, 367 –374. BAR International Series 2494.

Oxford: Archaeopress.

82 A. BONNIER ET AL.

Jameson, M. H., C. N. Runnels, T. H. Van Andel, and M. H. Munn. 1994.

A Greek Countryside: The Southern Argolid from Prehistory to the Present Day. Stanford, CA: Stanford University Press.

Jazwa, C. S., and K. A. Jazwa. 2017. “Settlement Ecology in Bronze Age Messenia. ” Journal of Anthropological Archaeology 45: 157–169.

Jenks, G. F., and F. C. Caspall. 1971. “Error on Choroplethic Maps:

De finition, Measurement, Reduction.” Annals of the Association of American Geographers 61 (2): 217 –244.

Kvapil, L. A. 2012. “The Agricultural Terraces of Korphos-Kalamianos a Case Study of the Dynamic Relationship between Land Use and Socio-Political Organization in Prehistoric Greece. ” Ph.D. diss., University of Cincinnati.

Lindblom, M., and B. Wells. 2011. Mastos in the Berbati Valley: An Intensive Archaeological Survey. Skrifter Utgivna Av Svenska Institutet i Athen 4

54. Stockholm: Svenska Institutet i Athen.

Lindholm, K.-J., E. Sandström, and A.-K. Ekman. 2013. “The Archaeology of the Commons. ” Journal of Archaeology and Ancient History 10: 3 –49.

Manning, S. W. 2010. “Chronology and Terminology.” In The Oxford Handbook of the Bronze Age Aegean (ca. 3000-1000 BC), edited by E. H. Cline, 11 –28. Oxford: Oxford University Press.

McMahon, T. C. 2007. “Discerning Prehistoric Landscapes in Colorado and the Mesa Verde Region Using a Kernel Density Estimate (KDE) Method. ” In Digital Discovery. Exploring New Frontiers in Human Heritage, CAA 2006, edited by J. T. Clark and E. M. Hagemeister, 151 –166.

Budapest: Archaeolingua. https://doi.org/10.15496/publikation-3499.

Mee, C., and H. A. Forbes, eds. 1997. A Rough and Rocky Place: The Landscape and Settlement History of the Methana Peninsula, Greece: Results of the Methana Survey Project, Sponsored by the British School at Athens and the University of Liverpool. Liverpool Monographs in Archaeology and Oriental Studies. Liverpool:

Liverpool University Press.

Penttinen, A. 1996. “The Classical and Hellenistic Periods.” In The Berbati-Limnes Archaeological Survey, 1988-1990, edited by B.

Wells and C. N. Runnels, 229 –283. Skrifter Utgivna Av Svenska Institutet i Athen 4° 44. Stockholm: Svenska Institutet i Athen.

Pettegrew, D. K. 2001. “Chasing the Classical Farmstead: Assessing the Formation and Signature of Rural Settlement in Greek Landscape Archaeology. ” Journal of Mediterranean Archaeology 14 (2): 189–209.

Rizakis, A. D. 2013. “Rural Structures and Agrarian Strategies in Greece under the Roman Empire. ” In Villae Rusticae. Family and Market- Oriented Farms in Greece under Roman Rule. Proceedings of an International Congress Held at Patrai, 23-24 April 2010, edited by A. D. Rizakis and I. P. Touratsoglou, 20 –51. Meletēmata / Ethniko Idryma Ereun ōn, Instituto Istorikōn Ereunōn, Tomeas Ellēnikēs Kai R ōmaïkēs Archaiotētos 68. Athens: De Boccard.

Sä flund, G. 1965. Excavations at Berbati 1936-1937. Stockholm Studies in Classical Archaeology 4. Stockholm: Almqvist & Wiksell.

Sayer, D., and M. Wienhold. 2013. “A GIS-Investigation of Four Early Anglo- Saxon Cemeteries: Ripley ’s K-Function Analysis of Spatial Groupings Amongst Graves. ” Social Science Computer Review 31 (1): 71–89.

Schallin, A.-L. 1996. “The Late Helladic Period.” In The Berbati-Limnes Archaeological Survey, 1988-1990, edited by B. Wells and C. N.

Runnels, 123 –175. Skrifter Utgivna Av Svenska Institutet i Athen 4° 44. Stockholm: Svenska Institutet i Athen.

Sharon, I., Y. Dagan, and G. Tzionit. 2004. “The [Awful?] Truth about GIS and Archaeology. ” In Archaeological Field Survey in Cyprus: Past History, Future Potentials: Proceedings of a Conference Held by the Archaeological Research Unit of the University of

Cyprus, 1-2 December 2000, edited by M. Iak ōvou, 151–162.

British School at Athens Studies 11. London: The British School at Athens.

Stewart, D. R. 2013. Reading the Landscapes of the Rural Peloponnese:

Landscape Change and Regional Variation in an Early “Provincial”

Setting. BAR International Series 2504. Oxford: Archaeopress.

Styring, A. K., M. Charles, F. Fantone, M. M. Hald, A. McMahon, R. H.

Meadow, G. K. Nicholls, A. K. Patel, M. C. Pitre, A. Smith, A.

Sołtysiak, G. Stein, J. A. Weber, H. Weiss and A. Bogaard. 2017.

“Isotope Evidence for Agricultural Extensification Reveals How the World’s First Cities Were Fed.” Nature Plants 3 (6): 17076.

van der Veen, M. 2005. “Gardens and Fields: The Intensity and Scale of Food Production. ” World Archaeology 37 (2): 157–163.

Vita-Finzi, C., E. S. Higgs, D. Sturdy, J. Harriss, A. J. Legge, and H.

Tippett. 1970. “Prehistoric Economy in the Mount Carmel Area of Palestine: Site Catchment Analysis. ” Proceedings of the Prehistoric Society 36: 1 –37.

Weiberg, E. 2011. “Topography and Settlement: Perception of the Bounded Space. ” In Helike IV: The Early Helladic Peloponnesos:

Abstracts Volume of the Fourth International Conference on Ancient Helike and Aigialeia, Nikolaiika, Diakopton, 1-3 September 2007, edi- ted by D. Katsonopoulou, 47 –64. Athens: Hetaireia Philōn Archaias Helik ēs.

Weiberg, E., I. Unkel, K. Kouli, K. Holmgren, P. Avramidis, A. Bonnier, F. Dibble, M. Finné, A. Izdebski, C. Katrantsiotis, S. R. Stocker, M.

Andwinge, K. Baika, M. Boyd, and C. Heymann. 2016. “The Socio- Environmental History of the Peloponnese during the Holocene:

Towards an Integrated Understanding of the Past. ” Quaternary Science Reviews 136: 40–65.

Wells, B. 1996a. “Introduction.” In The Berbati-Limnes Archaeological Survey, 1988-1990, edited by B. Wells and C. N. Runnels, 9 –22.

Skrifter Utgivna Av Svenska Institutet i Athen 4° 44. Stockholm:

Svenska Institutet i Athen.

Wells, B. 1996b. “The Early Iron Age.” In The Berbati-Limnes Archaeological Survey, 1988-1990, edited by B. Wells and C. N.

Runnels, 177. Skrifter Utgivna Av Svenska Institutet i Athen 4° 44.

Stockholm: Svenska Institutet i Athen.

Wells, B., and C. N. Runnels, eds. 1996. The Berbati-Limnes Archaeological Survey, 1988-1990. Skrifter Utgivna Av Svenska Institutet i Athen 4

44. Stockholm: Svenska institutet i Athen.

Wells, B., C. Runnels, and E. Zangger. 1990. “The Berbati-Limnes Archaeological Survey. The 1988 Season. ” Opuscula Atheniensia 18:

207 –238.

Wheatley, D., and M. Gillings. 2002. Spatial Technology and Archaeology: The Archaeological Applications of GIS. New York:

Taylor & Francis.

Whitelaw, T. 2000. “Reconstructing a Classical Landscape with Figures:

Some Interpretive Explorations in Northwest Keos. ” In Extracting Meaning from Ploughsoil Assemblages, edited by R. Francovich, H.

Patterson, and G. Barker, 227 –243. Oxford: Oxbow Books.

Winther-Jacobsen, K. 2010. “The Classical Farmstead Revisited. Activity Di fferentiation Based on a Ceramic Use-Typology.” The Annual of the British School at Athens 105: 269 –290.

Witcher, R. 2008. “(Re)Surveying Mediterranean Rural Landscapes: GIS and Legacy Survey Data. ” Internet Archaeology 24. https://doi.org/10.

11141/ia.24.2.

Wright, J. C., J. F. Cherry, J. L. Davis, E. Mantzourani, S. B. Sutton, and

R. F. Sutton, Jr. 1990. “The Nemea Valley Archaeological Project a

Preliminary Report. ” Hesperia 59 (4): 579–659.