Applying Geoarchaeological Methods on an Iron Age site: Part two of a two-part study discussing Archaeological Prospection for Ytings,

Jessica Coleman

Master thesis, 30 hp Archaeology Masters Programme / Masters Thesis Spring term 2017

Abstract:

Geoarchaeology has had a long history within archaeology around the world, but not so much so in Gotlandic archaeology. This study is aimed at drawing attention to this by using magnetic susceptibility (MS) and phosphate (P) analysis of an Iron Age site at Ytings, Gotland. This is where a small silver was discovered in 1888, and in 2009 a geophysical survey was done, via metal detector, and concluded with the theory of there being a workshop in the southern field and a farmstead in the north (ArkeoDok, 2011). The first part of this study discussed predictive modeling and whether or not the information available at the time would be enough to generate a reliable model (Coleman, 2016). The first study concluded with not being able to do so since the only discrete data available was from the metal detecting survey, which when used alone is not the most reliable instrument for archaeological prospection (Coleman, 2016). This led to this current study, which is the second part of a two-part study of Ytings. This study is aimed at using geoarchaeological methods for archaeological prospection to illustrate the benefits and need for these types of studies on Gotland, by comparing the MS and P results with the 2015 excavation report.

2

1. Introduction………………………………………………………..………………..6 • Defining Geoarchaeology and archaeological prospection and their role in archaeology. • Continuation of previous discussion about predictive modeling on Gotland. 2.Purpose and Research Questions…………………………………………….……..7 • To provide context to the 2015 excavation and how it compares to the metal detecting survey from 2009. • Theoretical prospective • Present a hypothesis as to where possible archaeological sites may be located. o How to interpret the finds from metal detecting surveys? o What explanations could account for the concentration of metal finds in the South from the 2009 metal detecting survey? o Are there any areas that exhibit archaeological potential and what areas can be delimitated from the search? 3. Advantages of Archaeological Prospection…….….…....….….….……...…..…...8 • Archaeological excavation and budget • Example Case studies: o “Unveiling the prehistoric landscape at Stonehenge through multi-receiver EMI” o “Soil chemical analysis as an interpretive tool for ancient human activities in Piedras Negras, Guatemala” o “A review of geophysical archaeological prospection in Sweden” 4. Review of Predictive Modeling and Background of Gotland……..…………… 14 • Predictive Modeling within archaeology • Case Studies: o Gustaf Svedjemo: “Predictive Model for Iron Age Settlements on Gotland, 200 – 600 AD.” o Gustaf Svedjemo and Erland Jungert: “Examples of Historical Maps in GIS and Databases” o Dan Carlsson: “The Ancient Cultivation of Arable Land” • Archaeological sites and Features of Ytings • Gotland’s Geology and Landscape • Iron Age Gotland • What Next: Phosphate analysis and Magnetic Susceptibility

5. Archaeological Prospection and Excavation at Ytings………………………… 32 • Sampling Strategy and Collection • 2015 Excavation Report

3

6. Results…………………………………………………………………………… 38 • Lab work • Magnetic Susceptibility • Phosphate Analysis • Munsell Color 7. Discussion……………………………………………………...……………..…… 55 • Magnetic Susceptibility and Phosphate results • Results and Discussion 8. Conclusion……………………………..…….……...………………………….…. 59 9. Bibliography…………………………………………………………………….… 60

4

Table of Figures: FIGURE 1: FEATURES AND FINDS REGISTERED WITH RIKSANTIKVARIEÄMBETET 22 FIGURE 2: A CLOSER LOOK AT THE REGISTERED SITES FROM THE STUDY AREA. 22 FIGURE 3: METAL DETECTING MAP FROM 2009. 23 FIGURE 4: BEDROCK OF GOTLAND. 25 FIGURE 5: THE DIP IN THE STRATUM OF THE BEDROCK. 25 FIGURE 6: A GIS MAP OF THE SOIL IN THE AREA. 26 FIGURE 7: THIS IS A GIS MAP OF PRESENT DAY WETLANDS 27 FIGURE 8: YTINGS REFERENCE SAMPLING AREAS 31 FIGURE 9: SAMPLING AREAS. 33 FIGURE 10: 2015 EXCAVATION PLAN. 36 FIGURE 11: UPDATED METAL DETECTING MAP. 37 FIGURE 12: GRAPH: SUM OF MS DATA BY AREA. 41 FIGURE 13: GRAPH: SUM OF MSLF AND MSQ BY AREA. 42 FIGURE 14: GRAPH: SUM OF MSLF AND MSQ IN ’GRID’ BY ROW. 42 FIGURE 15: GIS MAP OF MSQ. 44 FIGURE 16: GIS MAP OF MSLF DATA WITHIN THE 'GRID'. 45 FIGURE 17: MSLF IN RELATION TO THE 2015 EXCAVATION 46 FIGURE 18: GRAPH: SUM OF P DATA BY AREA. 48 FIGURE 19: GRAPH: SUM OF CITP AND PQ DATA BY AREA. 48 FIGURE 20: GRAPH: SUM OF CITP AND PQ IN ’GRID’ BY ROW. 49 FIGURE 21: GIS MAO OF PQ DATA BY AREA 50 FIGURE 22: GIS MAP OF CITP DATA IN GRID 51 FIGURE 23: GIS MAP OF CITP IN RELATION TO THE 2015 EXCAVATION 52 FIGURE 24: GRAPH: MUNSELL HUE BY AREA 54 FIGURE 25: GIS MAP WITH THE HIGHEST CITP AND MSLF RESULTS 56 FIGURE 26: GIS MAP 2015 EXCAVATION MAP WITH CITP AND MSLF DATA . 57

5

1. Introduction

Archaeology has come a long way from its humble beginnings into a comprehensive science that utilizes multiple concepts and methods from other sciences in order to better understand past societies. This continued growth of archaeology and it's mergence with other disciplines enables archaeologists to capture a clearer image of the past than was available before. One example of this amalgamation is geoarchaeology. Geoarchaeology is defined as a multi-proxy approach that implements the methods and concepts of earth science into archaeological research (Butzer, 1982; Rapp and Hill, 1998; Ghilardi and Desruelles, 2009). Geoarchaeological methods and concepts have been applied with an increasing frequency since the 1970's, with efforts to distinguish itself as a separate discipline that uses geoscience methods for when evaluating archaeological records (Rapp and Hill, 1998). This created the problem of geoarchaeology not being clearly defined since it encompasses other disciplines such as: sedimentology, pedology, stratigraphy, geophysics, remote sensing, etc., mentioning other popular disciplines that overlap and sometimes considered to be synonymous with geoarchaeology such as; geochemistry, environmental archaeology, archaeopedology, etc. these disciplines overlap with other fields of science (i.e. geology, pedology, chemistry) (Ghilardi and Desruelles, 2009).

This study reviews a geophysical method measuring magnetic susceptibility and a geochemical method of phosphate analysis on soil samples to use for archaeological prospection. Archaeological prospection is a predictive method that can use multiple types of data collected to hopefully find and/or provide context to archaeological sites (Neubauer, 2004). Two concerns archaeologists have today is the growing threat of destruction that urban expansion presents and the accelerated growth of erosion (natural or anthropogenic) have upon archaeological sites (Neubauer, 2004). These threats to archaeological sites creates the demand for quick, accurate, and non- destructive survey methods that can identify, therefore protect and preserve to be able to study those areas of historical interest. This is where archaeological prospection and geoarchaeology come in.

The first part of this study started as an attempt to raise awareness about the potential that geoarchaeology has with hopes to reverse its slow progression in Swedish archaeology, especially when compared to other areas in the world (Coleman, 2016).

6

This idea led to predictive modeling as a means for archaeological prospection. The first part of this study covered predictive modeling in detail, analyzing various methods and data needed for the modeling process (Coleman, 2016). The conclusion of that study is what created the need for a second part or continuation, because at the time there was not enough appropriate data which then affected the ability to create a model for the Iron Age site at Yitngs, Gotland. The primary data source was from a metal detecting survey done in 2009, this led to the hypothesis that there might be a farmstead (boplats) in the north field and a workshop in the south (ArkeoDok, 2011). Hoping to confirm this hypothesis an excavation was done in 2015, yielding no evidence to support this theory (Arendus, 2015). The first study discussed that the use of a metal detector and how alone, it collects a biased set of finds that may have led the archaeologists to misinterpret the finds and that the data was not enough to ‘predict’ or assume where archaeological sites could be located (Coleman, 2016).

The significance of these two studies is that there are very few geoarchaeological studies on Gotland, hopefully making this a stepping stone for further research. This study discusses magnetic susceptibility and phosphate analysis as potential methods to collect information of (about?) an area and their accuracy for archaeological prospection. Then compare those results to the metal detecting survey and the resulting excavation report that has since occurred at Ytings.

2. Purpose and research questions:

Other than addressing the need and benefits of exercising geoarchaeological/geophysical methods in Sweden, and advocating for other studies such as this. The goal of this study is to further illustrate how the magnetic susceptibility and phosphate analysis from the soil samples collected at Ytings provides useful contextual information and apply this to the 2015 archaeological excavation. Lastly to demonstrate how singular element analysis is not a recommended approach for prospecting archaeological sites.

This study is approached from a Processual Archaeological perspective; in that I believe that my conclusion should not be based upon my work and interpretations as a

7

researcher, but rather based upon the framework and structure of my conclusion and is open for further testing or investigations (Renfrew and Bahn, 2010).

Two of my research questions were inspired by Carlssons 2015 archaeological report. In his report he posed the question; “how should the results from the 2009 metal detecting report be interpreted?” considering that the 2009 metal detecting report led to the hypothesis of a workshop in the south and a farmstead in the north, which will be discussed further on the excavation was not able to confirm this to be the case (Arendus, 2015). The question that I want to answer here take this one step further and asks: How should the metal detecting report from 2009 and the 2015 excavation be interpreted? Another question to ask is: What explanations are there for the metal concentrations found by the 2009 metal detecting survey? Lastly are there any areas that exhibit any signs containing any archaeological sites, or any areas that can be delimitated from the search?

3. Advantages of Archaeological Prospection

One component of archaeological projects that many people fail to comprehend is the sheer amount of time and energy, not to mention how much money that is needed. Not to negate other projects and experiments that also cost their fair amount of money and time, but archaeology has the added stress-factor of only having one chance to collect all of the information possible. Research is the first step in preparing for an archaeological project, this helps decide whether or not an excavation is worth doing or not. This is dependent upon not only the accuracy of the individual's research, but also any and all previous documentations and information that would concern the site (Neubauer, 2004); including geological data, time period, maps, letters/journals, etc. To understand the development of past societies collecting information such as; historical records, maps, cultural and economic activities, etc., including all, if any previous archaeological work that has been done in the area and/or anywhere similar are all useful sources (Neubauer, 2004). This stage tends to be very time consuming regardless what work has been done; you are not only relying on other researchers studies of relatable archaeological sites, but there is the precarious nature of

8

the archaeological material itself that must be taken into account. This is referencing the biased nature of the archaeological material itself, making it difficult to attain the necessary information needed to continue (Verhagen, 2007a). Having reliable information is key when trying to build a case to be granted permission to conduct an excavation. Once this stage has been completed and if an excavation is authorized, the next step would be to structure a budget and find the funding needed to excavate. How this is done in Swedish contract archaeology is, after the search trenches are completed and a full excavation is approved, a cost estimate is then made up (Reuterdahl, 2013). Cost estimates include expenses such as; travel for fieldwork, lodging, soil sampling (and the analysis), artifact cleaning, lab supplies, field supplies, work permits, etc. ("Sample Project Budget Acquisition & Development Application"). If the cost is below 890,000 Swedish Crowns (≈104,000 Euro) the County Board can decide who is to conduct the excavation, if it costs more, the money has to be obtained by some other means (Reuterdahl, 2013). This instance is when geoarchaeological methods and archaeological prospection could save time and money for archaeological excavations. Choosing which method to use for prospection will depend heavily upon the location (i.e. city, open field, forest), the geology (i.e. sandy, rocky, glacier, water) what is being searched (i.e. settlement, battlefield, workshop), and of course how much is in the budget and the time frame for the project. Using the incorrect method or instruments where they are least effective will be a waste of money, but using the correct method in the appropriate setting the results have the potential of producing useful information which can yield answers about an area.

There are several ways to approach archaeological prospection that are constantly being developed and tested to better understand their strengths and weaknesses. Some of the popular geophysical prospection instruments and methods are electrical resistivity, ground penetrating radar (GPR), and magnetometry (Batayneh, 2011; De Smedt et al., 2014). Geochemistry is another popular approach in searching for anthropogenic activities; studies have shown that certain chemical signatures that can indicate anthropogenic presence. Human activities past or present have an altering effect upon the levels of Macro and Micro plant nutrients, Macro nutrients are the chemical elements that are used when measuring the geochemistry of the soil for archaeological prospection. Chemical elements that have been studied with success in being able to detect the presence of anthropogenic influence are elements such as:

9

Nitrogen (N), potassium (k), magnesium (Mg), calcium (Ca), sulfur (S), sodium (Na), and most popular of all is phosphorus (P) (Holliday and Garnter, 2006; Rapp and Hill, 1998; Wilson et al., 2007).

Case Studies:

An important aspect that has been reiterated by several studies and researches is how focusing upon only one element or physical property can have an adverse effect on the conclusion, and is not advised (Walkington, 2010). Having a singular focus without considering the big picture can lead to unclear or biased results, taking into account that there are natural occurrences which can result in similar readings. This is why a minimum of two complementing methods that shed light upon two different aspects of geological and anthropogenic events, because each property can shed light upon a new aspect of ancient life. Applying just two techniques can provide an improved insight into the soil's history, and geoarchaeological analysis can give additional information that can prove to be useful in other areas beyond what was intended. Examples of these will be summarized in the following case studies.

“Unveiling the prehistoric landscape at Stonehenge though multi-receiver EMI” De Smedt et al. 2014

Stonehenge is a UNESCO World Heritage Site (WHS) and has earned the privilege of being one of the most studied archaeological landscapes in the world (De Smedt et al., 2014). Other than the iconic Henge there are several prehistoric monuments in the landscape that contains at least 12,000 years of human occupation, much of which is still locked away in the landscape (De Smedt et al. 2014). Most archaeological investigations have centered around the monuments, but current research perspectives have broadened to emphasize the geography and archaeology of the surrounding area by using a non-invasive approach to “tackle current gaps in knowledge concerning the archaeological landscape” (De Smedt et al., 2014, pp. 16; Gaffney et al., 2012). The Stonehenge Hidden Landscape Project (SHLP) is a series of research projects that focus upon the archaeological landscape of Stonehenge that have

10

covered themes such as; Light Detection and Ranging (LiDAR), geophysical surveying, pedalogical variations, geological surveys characterizing the general stratigraphy of the Salisbury Plain, and soil micromorphological analysis (De Smedt et al., 2014, pp. 16- 17; Gaffney et al., 2012). This particular case discusses a large-scale geophysical survey using a multi-receiver electromagnetic induction (EMI) to reconstruct the palaeotopography of the area, and hopefully pinpoint both palaeoenvironmental and archaeological sampling locations (De Smedt et al., 2014). This study also addresses the potential problems that recent land-use and events that have taken place may have and further examine how using the multi-layered EMI data set can discriminate between recent topsoil disturbances and the underlying archaeology (De Smedt et al., 2014). Large areas upon the landscape have been disturbed from military activities mostly dating to the early 20th century, more recently during the 1970's and 80's additional magnetic material was left behind, which can affect geophysical surveys creating 'noise' in the data sets that may appear as false positives for other geophysical instruments (De Smedt et al., 2014). Other popular geophysical survey techniques (i.e. electrical resistivity, GPR, magnetometry) focus upon only either electrical conductivity or magnetic susceptibility, unlike the multi-receiver EMI used here (De Smedt et al., 2014). EMI is able to map soil variations by recording the apparent electrical conductivity and apparent magnetic susceptibility simultaneously, and the multi-receiver EMI soil sensors offers the potential to discriminate the changes in electrical conductivity and magnetic susceptibility in three dimensions (De Smedt et al. 2014). The results from this survey concluded with being able to discriminate between different types of natural and anthropogenic soil variation, this exposed certain anomalies that may possibly offer new archaeological information that was not previously recorded (De Smedt et al., 2014). Being able to identify modern soil disturbances along with remnants of flattened earth works, demonstrates the benefits of using multi-receiver EMI as a means for distinguishing archaeological features in this ‘chalkland’ environment (De Smedt et al., 2014).

“Soil Chemical Analysis Applied as an Interpretive Tool for Ancient Human Activities in Piedras Negras, Guatemala” - Parnell et al. 2002

11

The objectives of this study was to compare different methods of analyzing P and trace element analysis in a horizontally excavated household group; also to compare those chemical signatures with archaeological features and artifact distribution in the excavation in the Mayan region of Guatemala (Parnell et al., 2002, pp. 379-380). There was a trend in the Mayan culture where it seems there was a gradual abandonment of these settlements, meaning that the discovery of an artifact “may have very little to do with its original area of use because of discard or cleaning prior to abandonment” (Parnell et al., 2002, pp 379). Not using the traditional approach of using artifact discoveries and knowing how they are distributed in an area as the foundation for interpreting these household groups is the advised, relying exclusively upon artifact distribution as the basis of interpretation would be misleading (Parnell et al., 2002). Abandonment activities do not affect the chemical signatures left behind in the soil, which means that the chemical signatures then reflect the use of space during the time of occupation (Parnell et al., 2002). The hypothesis of Parnell, Terry, and Nelson (2002, pp. 380) is that the chemical signatures would be able to locate areas that may have high concentrations of artifacts, and indicate the ancient activity areas within the excavation. The P and trace element analysis results were compared to the artifact distribution of the site; no attempt was made to use these methods to try and understand the relationship between specific artifacts with the P and trace element analysis, rather all artifacts were grouped together as a single entity from an excavated unit (Parnell et al., 2002). The results showed a strong relationship between areas with high chemical concentrations and where high concentrations of artifacts were, with the exception of only one deposit that did not correspond as well as the others (Parnell et al., 2002). All of the middens that were predicted by the P and trace element analysis were all found during the excavation (Parnell et al., 2002). The results of this study also noticed patterns where there were higher concentrations of P were areas associated with food production, this was later verified by the artifacts recovered; and where there were lower P signatures but high concentrations of metals concluded that these were areas associated with craft production (Parnell et al., 2002). This study is an example of how soil chemical analysis can be used as a predictive tool before excavations to find areas of artifact concentration, and also as an interpretive tool to be able to reconstruct the use of space and indicate what activities occurred (Parnell et al., 2002).

12

“A Review of the Use of Geophysical Archaeological Prospection in Sweden” This review is when I first learned of the lack of geophysical prospection methods in Scandinavian archaeology, notably so when compared to the rest of Europe (Viberg et al., 2011; Coleman, 2016). Interestingly enough in Swedish archaeology, geochemical prospection techniques such as P analysis have been applied at archaeological sites since the 1930’s (Viberg et al., 2011; Sjöberg, 1976). A point of interest: Sweden has been one of the leading developers in mining geophysics and prospection, but has not had much success in implementing this focus to archaeological prospection (Viberg et al., 2011). Viberg (2011) listed probable reasons for the underdevelopment of archaeological prospection, explanations such as; geological, pedological and geomorphological conditions, the character and expression of common archaeological sites and features, archaeological tradition in research and exploration archaeology, disappointing initial experiences with geophysical archaeological prospection, etc.. Geological and environmental conditions of the country I believe may be the main reason for this lack of archaeological prospection. Sweden is dominated by uneven surfaces and rough terrain of forested and mountainous areas; forests cover about 75% of the country and the last glacial period left its mark in forms of large boulders, outcrops of bedrock, and glacial tills each of which comes with a set of obstacles that some geophysical instruments are not able to overcome (Viberg et al., 2011). There may be a lack of using geophysical prospection for archaeological purposes, but geologically there is a long history of geophysical prospection to locate iron ores in Sweden (Viberg et al., 2011). It is believed that the first archaeological survey conducted by some kind of geophysical method was during the late 1800’s, but the first official documented archaeological geophysical survey done occurred in 1959 when a metal detector was used at Öland (Viberg et al., 2011; Hagberg, 1961). There are recent studies within Swedish archaeology of other methods and approaches being used, the issue is that these studies used imprecise survey methods, coarse sample spacing, and covering a small area resulting in substandard results that led archaeologists to doubt the usefulness of using geophysical prospection surveys for archaeology (Viberg et al., 2011). Despite this, the metal detector is still a widly popular instrument for archaeological prospection in Sweden, especially on Gotland (Viberg et al., 2011; Coleman, 2016; Pettersson, 2009). This can be explained by the simplicity of this instrument and the low cost of using it, which has led to several problems from amateurs that use metal detectors to loot archaeological sites (Viberg et

13

al., 2011; Coleman, 2016). Presently geophysical prospection for archaeology is a developing practice that is growing (Viberg et al., 2011), but there is still work to be done in order to develop new and better ways in archaeological prospecting in Sweden.

These case studies hopefully further illustrated the range of applications that archaeological prospection and with a sufficient background of the area (i.e. geology, history, and previous research) is crucial when applying prospection methods. These studies show how these non-invasive instruments can shed light upon previously unknown information about land use and spatial analysis of prehistoric societies on already excavated areas without threatening the integrity of the area. The following portion of this thesis will be a brief recap of the first part of this study, to act as background information to which the second part of this study is built upon.

4. Review of Predictive Modelling and Background of Gotland

At this stage of studying Ytings the focus was discussing the various aspects of predictive modeling; the types of data, methods, criticisms, and the types of models that can be created, and how they apply to Ytings. In addition to the topic of Predictive Modeling, this study also went into detail regarding geology of Gotland, nearby archaeological sites and surveys that were related to Ytings, and the summary of a few case studies that point out important variables that must be considered when investigating an Iron Age site on Gotland. For the purposes of this paper only a brief recap of Predictive Modeling will be reviewed since it is not the focus here.

Predictive Modelling:

The basic function of predictive modeling is to identify or delimit areas of archaeological significance by observing patterns within an area or by using theory to interpret the distribution of sites and/or features found (Svedjemo, 2003; Coleman, 2016). Since this is almost identical to the definition of archaeological prospection, for the purposes of this paper predictive modeling will be addressed as the statistical

14

application of using geographical information system (GIS) to identify patterns or generate predictions based upon whatever method is used; archaeological prospection will be referred to as the encapsulating process of data collection, processing the data, and observing the results to then make a hypothesis.

A distinguishing quality that predictive modeling offers is that it takes into account the archaeological remains and their relationship with the environment, rather than just be scattered randomly on the landscape; better understanding the relationship anthropogenic activities have with the landscape (Verhagen, 2007a; Coleman, 2016). There are several factors that can have an effect upon the quality of the model and its process, such as; the purpose of the model (i.e. Cultural Resource Management (CRM) or Archaeological Research), theoretical perspective, and what data is available for the model (Svedjemo, 2003; Svedjemo, 2014; Svedjemo and Jungert, 2007; Coleman, 2016); not all of these will be covered here.

Data variables for predictive modeling typically fall under two main categories, independent or dependent (Svedjemo, 2003). Dependent variables are described as “archaeological sites or features whose distribution is sought” and independent variables are the characteristics recorded at each land parcel (Svedjemo, 2003, pp.5). The most used dependent variable is known as; site presence or site absence, other types of dependent variables are; multiple site types, counts or site density, and site significance (Svedjemo, 2003). Site presence or site absence is a binary technique that uses dichotomous dependent variables to predict the presence or absence of a site (Svedjemo, 2003). Multiple site types manage several different site types to then provide a prediction of whether or not there is a site of a specific class or not (Svedjemo, 2003). Counts or site density calculates the frequency or density of sites within an area, this is useful for when surveying a large area (Svedjemo, 2003). Lastly site significance is an arbitrary ranking model that ranks sites based upon their level of importance or archaeological significance (Svedjemo, 2003).

Independent variables have four categories that can be recorded at each land parcel; environmental variables, positional characteristics, cultural and social factors, and radiometric characteristics (Svedjemo, 2003). Environmental variables such as; elevation, soil types, vegetation, etc. that are then measured at every unit to then calculate aspects such as, slope, relief, distance to water, etc (Svedjemo, 2003; Coleman, 2016). Positional characteristics are the autocorrelation of variables that

15

observe the clustering or dispersal of variables in the model (Svedjemo, 2003). Cultural and social factors are such factors that are still visible, such as; roads, ruins, graves, etc. that can be recorded in an area (Svedjemo, 2003; Coleman, 2016). Radiometric characteristics are what is measured by the reflectivity of radiation within a unit, or the data from remotely sensed instruments such as, satellites images or aerial photos (Svejdemo, 2003; Coleman, 2016).

One of the primary complaints critics have with predictive modeling is that models are environmentally deterministic (Lock, 2003; Verhagen, 2004; Svedjemo, 2003). This stems from models that are built from the present day landscape characteristics and excludes cultural and social factors, running the risk of the model being environmentally deterministic (Verhagen, 2004; Lock, 2003). Building a model without understanding what changes occurred in the landscape results in a biased model; but since the present condition of the landscape is what is visible and quantifiable, makes avoiding this tricky (Svedjemo, 2014; Verhagen, 2004; Verhagen, 2007a). This is when understanding the underlying landscape and its social aspects are vital to the modeling process (Svedjemo, 2003). In many cases the decision to exclude cultural and social factors out of the modeling process is because, sometimes these factors are too ambiguous or intangible to be included in the models with any degree of certainty (Verhagen, 2004; Lock, 2003). Remembering that archaeological material itself is biased, the importance of how (or what) data was collected and incorporating this information into models carries a level of risk that if not cautious will undermine the full potential of the model (Svedjemo, 2014; Svedjemo, 2003; Verhagen, 2007a). This was the case of Ytings, where the only method in locating archaeological sites was done by a metal detector.

Case Studies:

One aspect of the first study that was beneficial for understanding the state of archaeological prospection in Gotlandic archaeology were these following case studies. These studies acted as a foundation for this geoarchaeologcal study.

Gustaf Svedjemo: “Predictive Model for Iron Age Settlements on Gotland, 200-600 AD.

16

This paper was useful in that it gives a clear explanation of predictive modeling by breaking down and defining the types of variables, explaining some of the methods, criticisms and strengths of those methods, and how to apply this to Iron Age sites on Gotland. The goal of this study was to test the level of efficiency of predictive models for locating Iron Age sites that are no longer present on the landscape on Gotland (Svedjemo, 2003). This was the only study that found that used predictive modeling for locating archaeological sites on Gotland (Coleman, 2016). It is estimated by Dan Carlsson (1979) that around 1,800 Iron Age farmsteads are still visible on the landscape, and it is thought that around 40% (≈800 total number) of Iron Age farmsteads are still missing from the landscape (Svedjemo, 2003). This represents the demand for there to be a more effective way of surveying areas in order to locate these farmsteads before they are lost. The models for this study were created by using a Logistic Regression Analysis (LRA) method: this is a statistical, nonparametric, regression technique that has no underlying assumptions as to how the data is distributed, unlike other regression or correlation models that have an assumption on how the data is distributed (Svedjemo, 2003; Coleman, 2016). LRA is considered to be one of the best methods for archaeological prediction, but its biggest drawback is the complexity of this type of model and the success of the model does depend on the quality of the data that is used (Svedjemo, 2003). This technique is able to use dichotomous dependent variables such as site presence and site absence combined with independent variables such as environmental variables and positional characteristics within the same model to produce a probability map (Svedjemo, 2003; Coleman, 2016). To test the effectiveness of modelling Iron Age sites on Gotland three areas were selected; a Training area for developing the model, a Verification area to use to for authentication, and a Prediction area for where actual predictions would be made (Svedjemo, 2003). The Training area was an area that is known to contain several Iron Age settlements, the Verification area is also an area that is documented to have Iron Age settlements; the difference between these two areas is that the Training area has a higher degree of moraine soil and the Verification area is more forested (Svedjemo, 2003). The Prediction area does not have as many documented Iron Age sites and over half the area has undergone cultivation (Svedjemo, 2003).

The variables used within this study were; distance to moraine soils, distance to 18th century settlements, distance to 18th century meadows, the distance to

17

18th century fields, distance to limestone, and the location of the bogs and/or wetlands recorded on the 18th century maps (Svedjemo, 2003). The critical variable for this study is the ‘distance to moraine soils’ based on a previous study that showed up to 80% of all Iron Age farmsteads are located on moraine soils (Svedjemo, 2003; Coleman, 2016). This was first noticed by Arrihenius when he conducted a phosphate survey in the 1930’s that concluded that Iron Age houses are typically built on or near non-productive land; more often than not these houses are built upon bare rock (Svedjemo, 2003 pp. 14; Carlsson, 1979; Coleman, 2016). One variable that was excluded from this model that in most other models would use is the location of streams and other freshwater sources (Svedjemo, 2003). This was chosen to be excluded here because of the widespread drainage of nearly all bogs and other major wetlands by ditches in the late 19th and early 20th centuries that resulted in altering the natural water flow of these areas (Svedjemo, 2003) (Figure 7). Another element that was excluded when creating these models were DEM’s, which Svedjemo (2003) decided against using them due to the quality of the DEM’s at that time, concluding that they would not be usable for such a large-scale model (Svedjemo, 2003, pp. 22).

This study is a good beginning for predictive modeling on Gotland, drawing attention to the need for more studies such as this need to be done. This study did also illustrate the usefulness of cadastral maps for creating a predictive model (Svedjemo, 2003; Coleman, 2016), cadastral maps and their uses is further explained in the next case study.

Gustaf Svedjemo and Erland Jungert: “Examples of Historical Maps in GIS and Databases”

This study used Gotland as an example for illustrating cadastral maps usefullness within GIS (Svedjemo and Jungert, 2007). A cadastral map defined as ‘A legal map for recording ownership of property; the map describes both the boundaries and the ownership of properties’ (investorwords.com, 2017). Using historical maps is not an unusual concept, but Gotland's historical maps are considered to have a unique quality of also mapping out the cultural landscape dating as far back as 500 B.C. (Svedjemo and Jungert, 2007). The cultural landscape is thought to have begun during the Iron Age and still relatively maintained as such today (Carlsson, 1986; Coleman,

18

2016; Elison et al., 2010; Svedjemo and Jungert, 2007; Svedjemo, 2003). These maps can be useful for showing the cultural landscape as far back as two thousand years ago; based on Dan Carlsson’s theory that the organization and the structure of the Gotlandic farms that are visible today would have been founded during the Iron Age (Carlsson, 1978; Svedjemo and Jungert, 2007). Proper analysis of these maps together with other data it is possible to recreate landscapes from the medieval/ and maybe even further back, this is known as retrogressive analysis, whose practice is firmly rooted into Swedish and European historical geography studies (Svedjemo and Jungert, 2007, pp. 5). Important factors to take into account when studying these maps are land parcel names, their location, and the site quality class; these were graded according to fertility of the soils and recorded on the maps which tends to indicate older, healthier farmland (Svedjemo and Jungert, 2007; Carlsson, 1986). Old agricultural societies relied on distance; this affects time, cost, and effort of maintaining a farmstead, which is why farmsteads are typically placed near fields that had a higher economic value in order to turn out more profit (Svedjemo and Jungert, 2007). Another correlation that have been noticed on Gotland are the areas labeled as ‘Kämpegravar’ which are now understood as foundations of Iron Age houses, and the land use found in maps dating between 1693 and 1705, to which they call GM1700 (Gotlandic Map of 1700) maps (Svedjemo and Jungert, 2007).

What this study has done is present an approach that is well suited for when using historical maps in a model. This gives the ability to be able to capture all concepts within each map series with GIS and was able to produce positive results (Svedjemo and Jungert, 2007).

Dan Carlsson: “The ancient cultivation of Arable Land”

This thesis does not have much to do with predictive modelling, but this paper presents a theory which could be used for relative dating purposes in Gotland (Carlsson, 1978). Carlsson wrote this thesis to present his hypothesis concerning the desertion event that occurred during the 6th century Scandinavia, he claims that this was not due to depletion of nutrients in the soil (Carlsson, 1978). Instead countered with the idea that the continued use of the farmland actually increases the fertility of the soils, Carlsson responds with the theory that it is possible to relatively date farmsteads by

19

understanding the spatial distribution of arable land (Carlsson, 1986). After the previous studies of predictive modeling that demonstrate how the location of Iron Age settlements shown on the cadastral maps and the location of fertile soils on these maps this hypothesis does not seem so far-fetched (Svedjemo, 2003; Svedjemo and Jungert, 2007; Carlsson, 1978). This thesis is a valuable source for this study because of his analysis of spatial distribution of the arable fields; he tested this by carbon dating (14C) fields that were closer to the (known) farmsteads and compared how they date to fields that are further away (Carlsson, 1986). The results revealed that the fields located closest to the farmsteads were older than the fields further away (Carlsson, 1986). Further demonstrating how well preserved the cultural landscape is and how using cadastral maps and the present day locations of agricultural fields can be a viable factor when creating a predictive model or trying to decided where or what to conduct the archaeological prospection.

Archaeological sites and Features of Ytings

This summary of archaeological sites is what was known at the time of the first study, new data has since been collected and will be presented in a later portion of this paper. Ytings is located within Othem Parish just east of . This area, much like the rest of Gotland is surrounded by a number of registered sites and features (Figure 1). Only 2 km east of Ytings is Spillings, this is where possibly the most famous discovery of Gotland was made (Waugh, 2011). Within the immediate area there are four registered sites; Othem 265 has been categorized as a ‘boplatser och visten’, in English, a living area and/or place of social gathering, this category also includes Othem 264 and Othem 266; the site Othem 150;1 is labled as 'Silverskatt' or silver treasure, this is the site that drew Dan Carlsson to investigate Ytings (Figure 2)(Fmis.raä.se, 2016; ArkeoDok, 2011; Coleman, 2016). In 2009 a metal detecting survey was conducted to investigate a small silver hoard that was found in 1888 in connection with ‘trench digging in a field’ (‘dikesgrävning i en åker’) (ArkeoDok, 2011). The metal detecting survey had 111 spikes consisting of primarily bronze finds with a few that were made of lead, iron, and a couple silver finds from within the northern and southern fields (Figure 3) (ArkeoDok, 2011). The datable finds were primarily from the Viking Age, including a fragment of an Arabian silver coin (ArkeoDok, 2011). A few exceptions

20

that date outside the Viking age was a bronze sewing needle from the northern field that was dated to the Vendel time period which was just before the Viking Age, and one silver coin dating to the 1500’s recovered in the southern field (ArkeoDok, 2011). Notably, before and after the 2009 survey there was evidence of looting before and after the survey was done (ArkeoDok, 2011), meaning that there is a distinct possibility of missing material that could have been pertinent for interpreting the character of the area. The conclusion of this survey led Dan Carlsson and his team to hypothesize that there may be a farmstead (boplats) in the northern field and possibly a workshop in the southern field (ArkeoDok, 2011; Coleman, 2016). Since this metal detecting survey the only source of discrete data for Ytings and the same was the case for the other neighbouring sites registered with Riksantikvarieämbetet, became the main problem when it came to creating a reliable predictive model, because of the biased nature of how these areas were identified (Coleman, 2016). With metal detecting being the only source of inquiry for the area, this could easily lead to an incomplete and biased interpretation of the area. Not to imply that metal detecting is not a fruitful instrument for archaeological surveying, it is useful when surveying disturbed topsoil and as a preliminary technique of clearing iron for magnetometer surveys (Clark, 1990). Still, metal detecting is only focusing upon a narrow set of finds that can prove to be inconclusive, unless partnered with another geophysical technique to achieve a more holistic understanding of the area in question (Viberg, 2012).

21

Figure 1: An image of the features and finds registered with Riksantikvarieämbetet overlain a google map image. Enclosed is the study area. Source: Google, 2016; Riksantikvarieämbetet, 2016

Figure 2: A closer look at the registered sites from the study area.

22

Figure 3: Metal Detecting Map from 2009. Red circles are mints (coins/coin fragments) dating to Viking age, Red Squares are other finds from Viking Age. Yellow Diamonds are other finds that were not identified or dated. Source: ArkeoDok, 2011

23

Gotland’s Geology and Landscape

Geology is an essential need for geoarchaeology and archaeological prospection. The first part of this study covered the geology of all of Gotland which will not be repeated, this geologic portion will focus mainly upon the vital aspects that pertain to Ytings.

The bedrock of Sweden is divided into three formations; the primary formation here is the third formation that emerged during the Phanerozoic period (ca. 530 to 1.4 million years ago), this formation is the sedimentary bedrock of Gotland (Viberg, 2012; Coleman, 2016). Gotland's sedimentary bedrock primarily consists of limestone and marlstone, with a few select areas of sandstone (figure 4) (Elison et al., 2010). Geologically, Gotland has been greatly affected by the glacial events that occurred during the various ‘ice ages’, as evidenced by the isostatic rebound that increased the surface area of the island and the gentle dip of the stratum (a layer or bed of sedimentary rock) in the island’s bedrock (figure 5). The glacial retreat also left a thin soil cover on the island which is predominantly made up of glacial till (also known as moraine) consisting of gravel, sand, and clay (Thompson and Turk, 1994: Elison et al., 2010; Viberg, 2012). Moraine is the primary type of soil at Ytings, which as previously mentioned is understood as being a preferred soil type of Iron Age farmsteads, estimating that 80% of all Iron Age houses are situated on moraine soils (clay and silt) (figures 6) (Svedjemo, 2003, p.14; Carlsson, 1974). The cultural landscape is thought to have begun during the early Iron Age; this coincides with a change in climate causing the islanders to alter their husbandry and agricultural practices to acclimate to the new, cooler climate (Elison et al., 2010; Svedjemo, 2003; Coleman, 2016). After agriculture and husbandry became the dominant form of sustenance, over time this created a demand for more open land which triggered a trend of forest clearing and draining all but two wetlands on Gotland, to supply the need of more farmland and fodder to maintain upkeep (figure 7) (Elison et al., 2010; Svedjemo, 2003).

24

Figure 4: Bedrock of Gotland. Source: Elison et al., 2010

Figure 5: The dip in the stratum. The highest point is Visby in the northeast and Hammarshagehällar in the south. Source: Elison et al., 2010

25

Figure 6: A GIS map of the soil in the area. The yellow dot represents where study area. Source: Coleman, 2016

26

Figure 7: This is a GIS map showing present day wetlands. Source: Coleman, 2016

27

Iron Age Gotland

Focusing upon the Iron Age of Gotland, this is when Gotland was the pinnacle point of trade that connected the east to the west and north to the south (Waugh, 2011; Coleman, 2016). There is overwhelming amount of evidence that support this, the sheer number of silver that have been found so far, picture stones which are unique to Gotland, and a number of other artifacts, graves, journals, folklore, etc. (Coleman, 2016; Waugh, 2011; Pettersson, 2009; Elison et al., 2010). Gotland’s location in the middle of the Baltic Sea gave it an advantage of being a pit- stop between the east and west spreading wealth to various lands. Gotland is considered to have been the wealthiest provinces of Sweden during this time, with 2,700 finds recovered in all of Sweden more than 247,500 coins have been recovered, and Gotland claims two-thirds of this total of coin finds (Jonsson, 2009). The Spillings hoard being the most famous silver hoard in the world, containing over 14,000 coins, 67 kg of silver and 20 kg of bronze objects (Waugh, 2011; Östergren, 2009). The estimated time of burial is around 871 BC, based upon the youngest coin in the hoard which dates to 870/871 BC, the earliest coin identified dates to BC 539 (Östergren, 2009; Waugh, 2011).

The accidental discovery of Spillings Hoard occurred in 1999 when Jonas Ström took reporters out to the field where the landowner Björn Engström previously found Viking Age bronze artifacts and some Islamic coins, as a way of raising public awareness to the problem of looting ancient monuments on Gotland (Östergren, 2009). Ironically once the reporters had wrapped up and had left is when the first signal was heard from the metal detector (Östergren, 2009). They continued on with the metal detector, extensively surveying the area, eventually covering the entire area (Pettersson, 2009; Carlsson, 2009; Östergren, 2009). This is one of the few examples where a phosphate survey was done at an archaeological site on Gotland and the metal detecting results were teamed with the P analysis to positively locate a farmstead (Carlsson, 2009; Widerström, 2009).

28

What Next: Phosphate analysis and Magnetic Susceptibility

The conclusion of this first study ended with the statement that there was not enough discrete data and information to create a predictive model, but proposed a soil sampling survey of Ytings to acquire more information and to be able to predict or delineate areas of archaeological significance (Coleman, 2016). The metal detecting survey done in 2009 led to the hypothesis that there is a farmstead (boplats) in the northern field and a workshop (verkstadsområde) in the south (ArkeoDok, 2011). To confirm this hypothesis the first study concluded that phosphate analysis and magnetic susceptibility tests would be the most suitable methods (Coleman, 2016).

These two methods have been proved that when used together they have yielded positive results of ancient activity and/or prehistoric settlements (Linderholm, 2007; O'Connor and Evans, 2005). This will just be a short overview of these methods to further elaborate why these methods were chosen. A more in-depth description of these will be discussed later.

Magnetic susceptibility (MS) is the measurement of the induced magnetization of a sample to the degree that the substance can be magnetized; not to be confused with magnetic enhancement, which refers to the changes in the magnetic mineralogy of the upper soil layers to which then results in a higher susceptibility values of the surface horizons (Dalan, 2008, p. 3). This method is also used for geologic and environmental studies, and in archaeology MS has proven to be an effective asset (Linderholm, 2007). Magnetic enhancement can happen from a number of natural and anthropogenic events such as, firing or burning, and as a part of organic and/or inorganic pedogenic processes (Dalan, 2008, p. 3). Natural magnetization of the soil can occur along with non-natural events, this depends on the iron content of the parent material, the pedogenic process, and the lithology of the area (Aubry et al., 2001; Clark, 1990); which is why it is important to understand the landscape as it is and how it was formed in order to better understand the MS readings. MS within archaeology is useful when indicating areas of fire exposure or high levels of heat, i.e when identifying the location of the ‘workshop’ that is hypothesized in the southern field (O’Connor and Evans, 2005; Clark, 1990; ArkeoDoc, 2011; Coleman, 2016). MS readings are found where there were high levels of heat; the heat converts the weak magnetic hydroxides and the iron oxides in the soil to stronger magnetic forms which is what makes this

29

detectable (O’Connor and Evans, 2005; Wynn, 1990; Clark, 1990; Coleman, 2016). Certain archaeological features are known for being associated with high MS readings resulting from fire are kilns and slag deposits; other features that are not associated with fire that give MS readings are wells, pits, wells, or other backfilled features (Aubry et al., 2001). Gotland's sedimentary bedrock has a magnetic homogeneity which has a low natural magnetic susceptibility readings, making it ideal for magnetometry studies (Gustafsson and Viberg, 2012; Viberg, 2012; Coleman, 2016).

Phosphate analysis has long been a popular subject area within soil sciences and is one of the most studied (Holliday and Gartner, 2007). Phosphate (P) analysis is instrumental within Northern European archaeology, especially in Swedish archaeology where there is a long history of its use in identifying areas of ancient human activity (Holliday and Gartner, 2007: Sjöberg, 1976). Olaf Arrihenus is one of the pioneers in developing P analysis as a suitable method in identifying areas that may contain ancient human activity, especially when there is a lack of physical evidence on the landscape (Holliday and Gartner, 2007; O'Connor and Evans, 2005). Anthropogenic phosphorus is generally the result of human waste (urine and excrement), organic refuse (bone, meat, plants), burials, and the refuse of animals which tends leave a higher residue of phosphorous than humans (Sjöberg, 1976; Holliday and Gartner, 2007). The anthropogenic deposition of P are normally unaffected by many present day human activities and is less susceptible to leaching, oxidation, reduction, and plant uptake (Holliday and Gartner, 2007), making this an ideal investigatory method in identifying the proposed ‘farmstead’ that is thought to be located in the northern field.

The plan was to conduct MS and P on the samples to look for areas with high MS and lower P to indicate a workshop and a lower or concentrated MS readings and a high and dispersed P for indication of a farmstead. The following figure (Figure 8) shows the proposed sampling strategy, areas to use as a framework for the samples. Meaning that where livestock is kept and fertilizer has been applied would yield high non-anthropogenic P results, giving an outlier to use as a frame of reference for when looking for the farmstead. The same applies to the plowed fields for the MS data, ideally providing another outline to use to locate the workshop.

30

Figure 8: The areas to be sampled from. The solid blue area in the left is where the livestock was kept and where there should be a higher degree of P. The Red area to the left is where seasonal plowing has occurred ideally having higher MS readings. The Yellow section in the middle is where a grid was placed. . Source: Coleman, 2016

31

5. Archaeological Prospection and Excavation at Ytings

From this point on, the information presented is what came after the first part of this study of Ytings. This study could be considered unorthodox because the excavation occurred before the data from the field sampling was processed, and so the ‘prospection’ in this case occurs after the excavation. This is done to best present how metal detecting alone may not be the most suitable prospection method.

Sampling Strategy and Collection

I was on Gotland from June 06th to June 16th collecting soil samples at Ytings to take to Umeå University, where MS and P analysis would be conducted. With hopes of understanding the dynamics of the landscape and identify if and where a workshop and/or a farmstead in the area. A total of 163 samples were collected with a soil auger, detailed notes were kept about each soil column, noting the texture, how compact or loose the samples were, the color of the sample, total depth reached, and any other facet that could be recorded. Sampling was split up into 8 different areas: ‘Southeast’ and ‘Northeast’ (of the Viking road), ‘North field’ (where the proposed farmstead is thought to be located), ‘Middle’ (the dividing areas between the North and South field), ‘Meadow’ (the area west of the North field), ‘Sheep’ (where the sheep were kept during the time of sampling), ‘South field’ (outside where the ‘workshop’ is thought to be located), and lastly the ‘Grid’ (located over where the workshop is thought to be located and where the excavation trenches would be located) (Figure 9). The 'Grid' was measured out in the southern field covering where the metal detecting survey had the most spikes. The ‘Grid’ covered an area of 50x50 meters and the samples were taken systematically at every 10 meters and plotted by hand on a separate sheet (Renfrew and Bahn, 2007). A handheld GPS was used when collecting samples in an unsystematic method from outside and around the ‘Grid’ location (Renfrew and Bahn, 2007).

32

1 4 3

2 5 6 8

7

Figure 9: Sampling areas. (1) White is labeled ’Sheep’. (2) Green is ‘Meadow’. (3) Yellow is ‘North’. (4) Purple is ‘NE of road’. (5) Orange is ‘Middle’. (6) Red is ‘SE of road’. (7) Pink is ‘South, outside of grid’. (8) Blue is where the ‘Grid’ was placed. Source: Coleman, 2016

The initial impression of the pedology at Ytings was that the soil in many areas was very compact and the texture was described as being clayey, which is expected considering the moraine soils in the area further indicating the likelihood of there being a Viking Age farmstead in the vicinity (Svedjemo, 2003; Carlsson, 1986). When collecting the soil samples, several rocks (mostly Limestone) inhibited the soil probe making it difficult to reach an ideal depth, but the average total depth achieved by the soil probe was 21.5 cm. There was a homogeny in the soil samples (depending where I was) that I noticed; in many of the samples there was only a slight color variation between samples according to the Munsell Color Chart, which will be addressed shortly.

33

2015 Archaeological Excavation

The 2015 excavation was structured around the 2009 metal detecting results. The archaeological excavation at Ytings lasted for four weeks during the months of June and July in 2015, with students who were doing their archaeological field school and students from the local Folkhögskolan (Arendus, 2015). The focus area for the excavation was spread over RAÄ Othem 255 and Othem 256, documentation of the trenches were done with a total station, photography, and notes regarding the stratigraphy of the profile were kept, lastly another metal detector survey was done, the spikes were recorded with a handheld GPS (Arendus, 2015).

Three trenches were positioned in the sampling area labeled ‘Grid’; the purpose here was to ‘find evidence of any constructions or features that would indicate a ‘workshop’ or any kind of prehistoric occupations (Arendus, 2015). When those trenches were opened (1, 2, 3) it was apparent that there were no traces of occupation and no cultural layer was identified (Arendus, 2015). Two more trenches opened (5, 6) east of the first three trenches and another (7) located just south of trench 5 and 6 (Figure 10) (Arendus, 2015). Again, no evidence of any structures or features that indicated any kind of prehistoric activity, which seems that this area may not have been settled or contained a ‘workshop’ (Arendus, 2015). The earth was consistently sterile and no sign of any ‘out plowed’ cultural layers; however, interestingly enough there was a lack of animal bones recovered from the area as a whole, which is normally indicative that there was an ‘out plowed’ settlement (Arendus, 2015).

Having no luck in the southern field, a series of test pits were opened in the northern field with hopes of finding or delimiting areas of the farmstead for future excavations (Arendus, 2015). A total of 35 test pits were opened numbering 9 to 43, twenty of these were opened in the south-west corner of the ‘North’ field, and the other eleven were located in the ‘Meadow’ (Figure 10) (Arendus, 2015). The test pits were also unsuccessful in finding evidence suggesting that there were any constructions here or finding a cultural layer in the pit's stratigraphy (Arendus, 2015). These test pits may not have recovered anything in helping to find a location for a farmstead, but the upside is that these test pits delimitated a few areas for locating the prehistoric settlement; which can then be taken into account when looking at the results from the lab.

34

With the lack of material coming from the trenches and test pits, the decision was then made to do another metal detecting survey with hopes of finding something new that would be able to answer or at least help lead to a conclusion as to what did once occur here (Figure 11)(Arendus, 2015). The survey over the entire southern area uncovered a few new artifacts including; a belt buckle, a pair of nails, and a fragment of an Arabian coin all of these dated to the Viking Age (Aredus, 2015). The ‘North’ field contained a few prehistoric artifacts, but several of the finds were dated to be much later, finds such as; capsules, cartridges, parts for kerosene lamps, etc. none of which help in finding a Viking Age settlement (Arendus, 2015).

This excavation report concluded with more questions rather than any definitive answers. The first metal detecting survey was done in 2009 and led to the theory that Ytings contained a possible farmstead and workshop, however the lack of evidence from the trenches and test pits could not confirm this. Therefore, the question arises; ‘how should the material be interpreted’? (Arendus, 2015). It is thought that most likely 'Ytings farm' was located further north in the prehistoric period, where the one piece of evidence that favours the notion of possibly being an Iron Age farm are the 4 known ‘malstenar’ or grinding stones in English (Arendus, 2015).

The excavation then shifted towards the grave fields or "gravfält” located further north, registered as Othem 67:1 (Arendus, 2015). Seeing as the graves are not within the immediate study area, no more will be said regarding the graves.

35

Figure 10: A layout of the 2015 excavation plan. The trenches are located in the southern section and the test pits are the clustered numbers in the north. Source: Arendus, 2015

36

Figure 11: This is an updated metal detecting map. The red dots are the new finds from the 2015 excavation overlain upon the 2009 metal detecting spikes which are represented by the black dots. Source: Arendus, 2015

37

6. Results

Representing the information gathered from the lab is the most crucial part of any published study. These days good collaboration between prospecting and excavating archaeologists is more important than ever. Especially when trying to keep up with (and eventually get ahead of) the rate of destruction that is threatening archaeological sites around the world (Neubaer, 2004). Collaboration between the two is possible if the 'interface between them is defined and the data flow is possible in both directions'; essentially understanding all parties goals, methods, and establishing a standard to exchange the information that is readily available is key for successful collaboration (Neubaer, 2004). The most significant interface used to facilitate these needs between the prospector and excavation is the geographical information system (Neubaer, 2004). Geographical information system (GIS) is described by Neubaer (2004) as "a system of complex computer hardware and software, and geographical data, which enables the identification, distribution, manipulation, analysis and storage of all possible types of geographically referenced information". GIS has become a fundamental tool for archaeological interpretation of prospection data and for processing and analyzing excavation data along with any other archaeological significant information that are collected digitally within the same coordinate system (Neubaer, 2004).

Lab Work

163 samples were brought to Umeå Universities' Archaeological Laboratory to undergo Magnetic Susceptibility (MS) and Phosphate (P) analysis. Before testing began the samples were dried out in the designated drying room, this removes moisture to prevent possible contaminations in the results. After 24 hours the samples were then ground down to grain sizes small enough to be sieved through a 1,25mm sieve, analyzing and documenting the excess material before disposal. Grain size variation can affect the MS and P results, so grinding down the samples into smaller grain sizes avoids this (Holiday and Gartner, 2007; Rapp and Hill, 1998). Once the samples were ground down, soil color was then documented according to the Munsell Color chart to understand soil characteristics and its composition.

38

MS is the measurement of the magnetic properties in the soil; this is done by placing a given volume of a sample in an applied magnetic field (Linderholm, 2007). There are five categories of magnetic behavior found in the soil; Ferromagnetism, Ferrimagnetism, Canted antiferromagnetic, Paramagnetism, and Diamagetism (Dearing, 1999). The MS calculated from the environmental material is the sum of the ferrimagnetism, canted antiferromagnetics, paramagnetism, and the diamagnetism components that is measured in the Bartington MS2 single sample dual frequency sensor magnetometer (Dearing, 1999). 10 grams were weighed out from the samples and recorded by the Bartington MS2 at both low (MSlf) and high frequencies (MShf). The samples were first recorded at low frequency in designated plastic cups; keeping track of the orientation of each sample so that when the samples underwent another round at high frequency the orientation would be the same (or at least very close) to prevent small directional variations in susceptibility which could affect the readings (Dearing, 1999).

The phosphate analysis conducted followed the O. Arrhenius method (1934) which is known to be well suited for acid, Fe-rich environments (Linderholm, 2007). A citric acid extraction was used followed by spectroscopic determination on molybdenum blue complex formed in the acid solution to be able to extract the inorganic P (CitP) (Linderholm, 2007). Due to time restrains the pH of the soil samples were never taken, which is something to consider since pH has various effects in P solubility and precipitation (Holliday and Gartner, 2007).

Once the first round of tests was done the samples were then burned in porcelain crucibles at 550c for 4 hours, weighing the samples before and after ignition to know how much organic material was removed. Once burned the samples can give new readings that help understand the landscape and its use, this process gives us a Loss of Ignition (LOI) measurement. For P analysis, igniting the organic material is done to extract a fraction of the organic and inorganic P (CitPOI), this comes from using the citric acid complex which releases mostly Fe-bound P and insignificant amounts of P bound to organic substances (Linderholm, 2007). Igniting the soil also changes the solubility of Fe compounds, which may affect the P analysis (Linderholm, 2007)

39

Magnetic Susceptibility Results

As mentioned previously MS is measuring the ability of a substance to be magnetized, which within an archaeological context is typically associated with areas that were exposed to fire (Rapp and Hill, 1998; Linderholm, 2007). This would mean that the location of a ‘workshop’ would theoretically have strong MS readings. We know that the excavation was unsuccessful in finding anything resembling any type of construction or features that indicated intense fire exposure (Arendus, 2015). The most common Fe-oxides in magnetic susceptibility are magnetite, maghemite, hematite, and geotite (Linderholm, 2007); the Munsell Color Chart tells us that the area is full of geotite and maghemite, meaning that areas that exhibit high MS readings may have natural causes rather be a result of anthropogenic influence. Other than the prospection uses the MS has; MS also gives information about the mineralogy and geochemistry of environmental materials (Dearing, 1999). There are several ways to present this information, sometimes making illustrating and understanding the data a challenge. How the data is presented was decided based upon what graphs and maps best showed the data and its distribution in the area. The following MS readings are labeled as MSlf (MS at low frequency), MS550 (MS after ignition), and MSQ (quotient of MSlf and MS550).

The following graph (Figure 12) shows the Sum of all MS data by area. This graph is an example of how important and difficult presenting the data can be; here the MS550lf data overshadows the MSlf and MSQ data. This is expected since the MS550lf is after ignition, magnetizing the samples at a higher degree than the MSlf and MSQ would show.

40

Figure 12: Sum of MS data by Area. Notice that MS550lf is overshadowing the other data.. Here you want to notice the total readings of all three and how they vary from each other.

The following two figures exclude the MS550lf data in order to better illustrate the other MS data and its distribution. The areas of most interest in figure 13 are 'North field', 'South, outside of grid', and 'NE of road'. Looking back to figure 12 the 'grid' has the highest sums of both MSlf and MSQ, coupling this with the data shown in figure 13 'South, outside of grid' are exciting to see considering these areas are where the hypothesized workshop is located. The problem with graphs is that we have no idea how the MS is distributed in these areas, we are only seeing the sum. The 'North field' results have the highest sum of MSlf overall, this is interesting since the 'North field' is where the hypothesized farmstead is located, this area is also where during the 2009 metal detecting survey had the least amount of spikes.

41

Figure 13: Sum of MSlf and MSQ by area. This is a better illustration that shows the distribution of two types of MS data. Notice that again the ‘North Field’ still shows the highest readings, however ‘SE of road’ and ‘South, outside of grid’ both also show strong readings that are worth paying attention to. The next figure (figure 14) shows the MS distribution of the 'grid'. Each row was lettered A - F orienting East to West, the sampling from the 'Grid' was taken systematically in a 50x50 meter at every 10 meters. The overall results from the 'Grid' are disheartening low to be indicative of a workshop. However, the stronger readings radiate westward which is not where the trenches were located; they were placed in the east portion of the 'Grid'.

Figure 14: Sum of MSlf and MSQ in ’Grid’ by Row. Here breaks down the ‘Grid’ layout going from East (A) to West (F). Here the sum of the MSlf data is quite low to be indicative of a workshop within this 50x50 meter grid.

42

Based upon the graphs alone one could deduce that it is not likely that a workshop is located in the ‘Grid’, however the readings of MSlf and MSQ do shine a light upon a few other areas that may warrant more investigation, such as; ‘North Field’, ‘SE of road’, and ‘South, outside of grid’ (Figure 14). One issue that graphs have is the level of disconnect that GIS maps do not have, again hammering in the importance of having a common interface that can be clearly understood by all parties. The following GIS images all better show how the MS spreads over the area.

The following figure (figure 15) shows the MSQ data and the sampling areas. MSQ is the quotient of MSlf and MS550lf; a few areas with high MSQ readings seen in figure 16 do not match up very well with what the graphs were showing. For example when looking at figure 13 the sum of MSQ in the grid was the highest sum of all the MSQ, second was the 'North field'. Looking at the GIS map of MSQ (figure 15) the 'Grid' doesn’t show the strong readings that the graph showed, but looking at the 'North field' with its one very dark spot could be the reason that the sum from the graph showed that 'North field' was the second strongest.

43

Figure 15: GIS map of MSQ. Three areas worthy of mentioning is the dark areas in the 'NE of road', the one dark spot in 'North field', and in the 'South, outside of grid'.

The next figure (figure 16) is a close up of the MSlf data from the 'Grid'. Unlike the previous MSQ map this following map does correspond with the MS graph from the 'Grid' (figure 14). The graph showed that the strongest readings started in row C and continued through D, E, and F, which is shown in the map. Interestingly row B seems to contain the strongest readings in the GIS map but this did not appear in the graph.

44

Figure 16: GIS map of MSlf data within the 'Grid', with row A starting from the NE corner.

The following GIS map is the MSlf data in relation to the 2015 excavation. This map seems to also correspond with the graph (figure 13) that showed the sum of the 'North field' being quite high. It is noticeable that the MS spreads over a large area in the 'North field' and spreads over into 'NE of road'. Another area worth mentioning is

45

the feature that appears in the 'South, outside of grid', this feature could be there due to the inverse distance weighted (IDW) interpolation method used here, but similar features appeared in the MSQ map (figure 15) and in the graph (figure 13), which is interesting.

Figure 17: MSlf in relation to the 2015 Excavation

46

Phosphate Analysis Results

Phosphorus (P) according to many researchers is considered the best indicator for prehistoric agricultural (or pre-agricultural) settlements (Holliday and Gartner, 2007). There are 4 primary applications of P analysis in archaeology; first and most popular application is determining soil P levels as a pre-excavation prospection method to locate or delimit possible locations for sites, researchers also use P analysis to locate kitchens, middens, and aid in the orientation of excavations; P analysis is also used as a tool in landscape archaeology to examine past agricultural societies; lastly, delineation of features and activity areas through P analysis during and after excavations (Parnell et al., 2002 pp. 3). Here soil P analysis is used as a combination of the first and last application, using soil P as a pre-excavation prospection technique but is applied after the excavation concluded as a delineation of activity areas. Phosphorus in its most common form is Phosphate; it is relatively stable in soils and is less susceptible to leeching, oxidation, or plant uptake like many other elements are (Holliday and Gartner, 2007). There are several ways of extracting P out of the soil; the extraction method is what defines the analyzed form of P, not the chemical composition (Linderholm, 2007). This is due to how many forms of P can exist in the soil and the various methods in which to extract P (Holliday and Gartner, 2007). The main focus here are the inorganic P and organic P, these can occur in various forms but can remain in the soil for an extended period of time (Holliday and Gartner, 2007). The P presented here are separated into three groups; CitP (inorganic P), CitPOI (fraction of inorganic and organic P), and PQ (the quotient of CitP and CitPOI).

The following two graphs show the sum of P data by area, similar to how the first two MS graph was structured. The first graph (figure 18) also had the same issue of representation that the MS graph had with the MS550lf overshadowing the other data, as does the sum of CitPOI does here, which is why it is followed by a graph that excludes the CitPOI data to better show the P distribution. Similar to the first MS graph (figure 12) areas with significant readings are 'Grid', 'North field', and 'South, outside of grid'. Again the limitations that graphs have are the lack of visual distribution of the data.

47

Figure 18: Sum of P data by Area. This distribution of P data shows that the ‘Grid’ is significantly higher than the others, however the areas The distribution of P was expected to be strong in the 'Meadow' and 'Sheep' areas, however in the following graph (figure 19) the strongest readings are from the 'North field'. This is what you want to look for when searching for a farmstead. 'NE of road'. 'SE of road', and 'South, outside of grid' interestingly enough have quite strong readings, which is noteworthy.

Figure 19: Sum of CitP and PQ data by Area.

48

The following graph (figure 20) shows the P data in the 'Grid' by Row. This graph also shows a similar trend of high readings coming from row C westward like the MS data from the 'Grid'. There seems to be low readings of PQ overall, but with the relatively high readings of CitP in the 'Grid'.

Figure 20: Sum of CitP and PQ in ’Grid’ by Row. Here there seems to be low readings of PQ overall, but with the relatively high readings of CitP in the area starting from row C westward is interesting.

The previous graphs show the expected readings coming from the 'Meadow' and 'Sheep' areas as being relatively high; this is most likely left behind from the animal waste. The strong readings from the 'North Field' is hopeful when looking for evidence of the suspected 'farmstead', however the almost identical readings from the 'NE of road', 'SE of road', and 'South, outside of grid' are interesting, and unexpected. The 'Grid' data was interesting, considering they follow suit with the MS data from the 'Grid'; with highest readings in the rows C and D, but the second two are E and F rows again the data is stronger westward than east. The following images are the GIS maps showing the P data, giving a much clearer understanding of its distribution.

The following GIS map is of the PQ data, here it shows the expected strong readings in the 'Sheep' and 'Meadow' areas since these are unplowed fields and contain sheep (i.e. animal waste), and high PQ can also be an indicator for significant quantities of fertilizer usage (Figure 21). The 'North Field' seems to be quite low compared to the

49

graphs, also taking into account that this is where the supposed 'farmstead' is located. The 'NE of road', 'SE of road', and 'South, outside of grid' all corroborate with the graphs, but this map shows us how this data is distributed in those areas which is something the graphs are not able to do.

Figure 21: PQ data by area

The following figure (Figure 22) is the close up of the CitP data in the ‘Grid’. There is an interesting feature that appears in the northwest area, as was similarly indicated by the previous graph (Figure 20). It is also important to point out that this feature is outside where the 2015 excavation trenches were placed.

50

Figure 22: CitP data in Grid, with row A starting in the NE corner.

Another visualization is the CitP data overlain the 2015 excavation map (Figure 23). This figure shows that there are strong readings coming from the 'North Field', which was not present in the PQ data (Figure 21). What is also interesting is in the following CitP figure; it is noticeable that many of the test pits from the ‘North Field’ were located where there are higher P readings. The test pits however were not able to yield any positive results indicating that there is a farmstead in this area.

51

Figure 23: CitP in relation to the 2015 excavation

52

Munsel Color Chart

The Munsel color chart describes three aspects of the soil; hue, value, and chroma (Rapp and Hill, 1998). Hue is defined as the quality of color described by words, Value is the quality of lightness or darkness of the color that is measured, and Chroma is the degree of color saturation from a gray color that is of the same value (Rapp and Hill, 1998, pp. 36). The samples from Ytings were analysed with the Munsell color chart once the samples were dry, which can alter the Value of the sample (Rapp and Hill, 1998). The Value of the sample has been known to register at two units higher in Value than when wet and can change the Chroma of the sample, but the Hue generally stay the same (Rapp and Hill, 1998). The information from the Munsell Color Chart presented here is mostly derived from the Hue rather than the Value and Chroma, because the samples were dried when recorded. The Hue can still provide a general picture of the parent material and how the soil formed, which some indication of its chemical make-up. Most of the samples were from the 10YR and the 2.5YR hue with a few from the 5YR hue, which are indicative of specific oxides that are occurring in the soils. The 10YR samples could be indicating the presence of geothite, alpha FeO(OH), which is one of the most common iron oxides in soils, but when it occurs alone is when it falls under the hue classification of 10YR and the 7.5YR (Rapp and Hill, 1998). The

2.5YR hue could be indicating the presence of maghemite, gamma Fe2O3, which ferromagnetic, however this could also be indicating the oxidation of magnetite (Fe3O4) or other iron oxides that were transformed into maghemite by fire that is located in organic matter with limited exposure to oxygen (Rapp and Hill, 1998, pp.37). The few samples that were classified under the 5YR hue could be indicating the presence of hematite, alpha Fe2O3, which are formed at fairly low temperatures (Rapp and Hill, 1998). Hematite coexisting with goethite is not uncommon, however in specific conditions such as lower water availability, high temperature, a near-neutral pH, bedrock with high iron content, and a rapid turnover of biomass all tend to favor hematite over goethite in the soils (Rapp and Hill, 1998). This is important to consider since most of the samples were documented as the Hues that indicate goethite and maghemite, which as mentioned previously are some of the most common Fe-oxides associated with magnetic susceptibility (Rapp and Hill, 1998; Linderholm, 2007).

53

Figure 24: Graph of Munsell Hue by Area

54

7. Discussion

With all of the data from Ytings having been presented (2009 metal detecting survey, MS, P, 2015 excavation and test pits, and 2015 metal detecting survey) and their specific implications already discussed. Here is where all of the data is interpreted as a whole in an attempt to further understand or explain the history of Ytings.

The Magnetic Susceptibility data we know shows the induced magnetization of a sample, used here as a technique for identifying a potential workshop and/or anthropogenic burning activities. Phosphate analysis we know can measure anthropogenic waste, here it was used to locate a possible farmstead. Focusing upon only one of these elements creates a degree of variability that cannot be understood to its fullest extent until partnered with a complementing elemental analysis or methodology. The following figure (Figure 25) shows that the strongest CitP and MSlf readings are in the 'North Field’ that is what we want to see when looking for a farmstead. There is however an interesting feature of P that appears in the South field from where the 'Grid' was placed.

55

Figure 25: Areas with the highest CitP and MSlf results

The issue with the previous figure (Figure 25) is when taking into consideration the disappointing results from the 2015 excavation, which concluded with no reliable evidence indicating the presence of a workshop or a farmstead in the 'North Field' or in the South field. Attempting to better understand these readings I queried the results at a higher degree and placed those results upon the 2015 excavation map, this yielded some interesting results (Figure 26). This shows that the highest MSlf readings appear in the 'North' and that the highest CitP readings appear in the western section of the 'Grid'.

56

Figure 26: 2015 Excavation Map with CitP and MSlf data at a higher degree

The purpose of the previous figure (Figure 26) seems to show the strongest readings of P and MS are not where the excavations took place. This exemplifies one of the core concepts of this paper which are to caution future archaeologists of the pitfalls and uncertainties that single-element analysis can have for prospecting archaeological sites. First and foremost I do not want to in any way discredit any of Carlsson's findings and conclusions from his surveys and excavation at Ytings, his work has been instrumental in understanding and interpreting my data. It was because of his work that I had to explore my data more thoroughly (i.e. exampled in the previous figure), in order to possibly indicate areas of anthropological significance.

57

Carlsson presented a question within his 2015 excavation report asking "How should the 2009 metal detecting survey results be interpreted?" since that survey seemed to evidence a 'workshop' in the south field and a 'farmstead' in the north field (ArkeoDoc, 2011; Arendus, 2015). This study takes Carlsson’s question one-step further asking: How should the finds from the metal detecting surveys (2009 and 2015) be interpreted? My answer to this is that they should be interpreted at face value; these are metal finds recovered in a plowed field. These finds have been fractured and/or broken down into fragments, possibly redeposited meaning that where the finds were recovered may not have been the artifacts original location, and leaving non-metallic finds behind creating the possibility of misinterpreting the results, giving only one piece of the puzzle that makes up archaeological prospection; which is what seems to have occurred here at Ytings.

This leads to the next research question which "what explanations are there for the metal concentrations found in the 2009 metal detecting survey?” This is tricky to answer due to my lack of understanding the physics involved in artifact redistribution within a plowed field; there are too many factors to take into consideration to be able to answer definitively. One possible explanation comes from the pattern of plowing that occurs here. The 'North Field' and the south field are plowed separately from each other, leading to the assumption that all of the metallic finds recovered from the south field did originate from somewhere in the southern field, and those spikes in the ‘North Field’ originated in the north, little chance of blending. Looking back to the MSQ map of the whole area (Figure 15) there is a feature that appears further south from where the excavation occurred, there is a chance that the artifacts may have originated further south and have been relocated and distributed further northward assuming that the pattern of plowing goes south to north.

The last research question could be considered the primary objective of this study asks; "Are there any areas that exhibit any signs indicating possible archaeological sites, and/or what areas can be delimitated from the search?"

This question in part has already been answered. The areas that may exhibit signs archaeological significance or areas worth investigating I would say are; in the ‘North Field’ east of where the test pits were opened, within the ‘Grid’ area northwest of where the trenches were opened, and further south in the ‘South, outside of Grid’ area shown in the MSQ map (Figure 16). Areas that I feel comfortable eliminating from

58

the search would be the ‘Sheep’ area, ‘Meadow’, and the ‘Middle’ areas. This leaves two areas I have not mentioned yet which is ‘NE of Road’ and ‘SE of Road’. These areas have exhibited certain features that could be explained both anthropogenically and naturally. Meaning that having only MS and P data is not enough to understand what is occurring here. With the data that has been collected I feel confident that these areas do not take priority in the search for a farmstead or a workshop.

8. Conclusion

This study was conducted not only to help me come to a better understanding of how archaeological prospection works by applying this to a place that has not had much exposure to geoarchaeological methods, it was also done with hopes of drawing more attention for more studies such as this to be conducted. Hopefully with the information collected and presented in this study further investigation can be put into Ytings and its historical background. Archaeology has entered a new age where there are more resources available than there really has been before. With programs such as GIS and others like it there is a common medium to which several data sets and types can be entered and viewed at one place. To be in this new age there are certain challenges that are present, however by developing new practices and applications in Archaeology and testing/perfecting these, archaeologists have a remarkable chance of collecting more information about past societies than we could ever before.

59

9. Bibliography

Arendus (2015). Ytings 1:9, Ytings 1:43. Arkeologisk undersökning, Diarienumer; 431-21-15. Gotland: Arendus.

ArkeoDoc, (2011). Rapport över efterundersökning med metalldetecktor vid fyndplats för halsringar av silver. Visby, Sweden: Länsstyrelsen i Gotlands Län, pp.1-13.

Arrhenius, O. (1938). Den gotländska åkerjordens fosfathalt (Sveriges geologiska undersökning. Serie C, Avhandlingar och uppsatser, 412). Stockholm: SGU, pp.1-15.

Aubry, L., Benech, C., Marmet, E. and Hesse, A. (2001). Recent achievements and trends of research for geophysical prospection of archaeological sites. Journal of Radioanalytical and Nuclear Chemistry, 247(3), pp.621-628.

Barham, A. and MacPhail, R. (1995). Archaeological sediments and soils. London: Institute of Achaeology. University College.

Batayneh, A. (2011). Archaeogeophysics–archaeological prospection – A mini review. Journal of King Saud University - Science, 23(1), pp.83-89.Barham, A. and MacPhail, R. (1995). Archaeological sediments and soils. London: Institute of Achaeology. University College

Butzer, K. (1982). Archaeology as human ecology. 1st ed. Cambridge University Press.

Carlsson, D. (1986). The Ancient Cultivation of Arable Land. Nordic Late Quaternary Biology and Ecology, [online] 24, pp.89-91. Available at: http://www.academia.edu/3821363/Ancient_Cultivation_of_Arable_Land

Carlsson, D. (1979). Kulturlandskapets utveckling på Gotland. 1st ed. Visby: Press.

Carlsson, D. (2009). Owner Missing? The Hoard, the Farm, and the Community. In: A. Pettersson, ed., The Spillings Hoard, 1st ed. Visby: The , pp. 65-104.

Clark, A. (1990). Seeing beneath the soil. London: B.T. Batsford.

Coleman, J. (2016). Discussing Predictive Modeling on Gotland. First Year Master Thesis. Umeå University.

Dalan, R. (2008). A review of the role of magnetic susceptibility in archaeogeophysical studies in the USA: recent developments and prospects. Archaeological Prospection, 15(1), pp.1-31.

De Smedt, P., Van Meirvenne, M., Saey, T., Baldwin, E., Gaffney, C. and Gaffney, V. (2014). Unveiling the prehistoric landscape at Stonehenge through multi-receiver EMI. Journal of Archaeological Science, 50, pp.16-23.

Dearing, J. (1999). Environmental Magnetic Susceptibility Using the Bartington MS2 System. [online] Available at: http://gmw.com/magnetic_properties/pdf/Om0409%20J_Dearing_Handbook_iss7.pdf [Accessed 24 Oct. 2016].

Eliason, S., Bassett, M. and Willman, S. (2010). Geotourism highlights of Gotland. Tallinn: NGO GEOGuide Baltoscandia.

6 0

Fmis.raa.se, (2016). Riksantikvarieäbetet - Fornsök. [online] Available at: http://www.fmis.raa.se/cocoon/fornsok/search.html

Gaffney, C. (2008). Dectecting Trends in the Prediction of the Buried Past: A Review of Geophysical Techniques in Archaeology. Archaeometry, 50(2), pp.313-336.

Ghilardi, M. and Desruelles, S. (2009). Geoarchaeology: where human, social and earth sciences meet with technology. S.A.P.I.E.N.S, [online] 2(2.2). Available at: http://sapiens.revues.org/422[Accessed 9 Apr. 2016].

Gustafsson, N. and Viberg, A. (2012). Tracing High-temperature Crafts: Magnetometry on the Island of Gotland, Sweden. Archaeological Prospection, 19(3), pp.201-208.

Holliday, V. and Gartner, W. (2007). Methods of soil P analysis in archaeology. Journal of Archaeological Science, 34(2), pp.301-333.

InvestorWords.com. (2017). What is Cadastral Map? definition and meaning. [online] Available at: http://www.investorwords.com/14150/cadastral_map.html [Accessed 31 May 2017].

Jonsson, K., 2009. Eastern contacts based on the coin finds. Situne Dei, pp.57–67. Available at: http://urn.kb.se/resolve?urn=urn:nbn:se:su:diva-36986.

Linderholm, J. (2007). Soil chemical surveying: A path to a deeper understanding of prehistoric sites and societies in Sweden. Geoarchaeology, 22(4), pp.417-438.

Lock, G. (2003). Using computers in archaeology. London: Routledge.

Neubauer, W. (2004). GIS in archaeology—the interface between prospection and excavation.Archaeological Prospection, 11(3), pp.159-166.

O'Connor, T. and Evans, J. (2005). Environmental archaeology: principles and methods. 2nd ed. Stroud: Sutton. Oonk, S., Slomp, C. and Huisman, D. (2009). Geochemistry as an aid in archaeological prospection and site interpretation: current issues and research directions. Archaeological Prospection, 16(1), pp.35-51.

Parnell, J., Terry, R. and Nelson, Z. (2002). Soil Chemical Analysis Applied as an Interpretive Tool for Ancient Human Activities in Piedras Negras, Guatemala. Journal of Archaeological Science, 29(4), pp.379-404.

Pettersson, A. (2009). The Spillings hoard. 1st ed. Gotland: The Gotland Museum.

Quine, T. (1995). Soil analysis and archaeological site formation studies. In: A. Barham and R. Macphail, ed., Archaeological Sediments and Soils: Analysis, Interpretation and Management, 1st ed. London: Institute of Archaeology. University College.

Rapp, G. and Hill, C. (1998). Geoarchaeology. 1st ed. New Haven, Conn.: Yale University Press.

Renfrew, C. and Bahn, P. (2010). Archaeology essentials. 2nd ed. London: Thames & Hudson.

Reuterdahl, Magnus. "ABC Of Swedish Planning Archaeology And An Archaeological Paradox | Day Of Archaeology". Dayofarchaeology.com. N.p., 2016. Web. 6 July 2016.

61

Rypkema, H., Lee, W., Galaty, M. and Haws, J. (2007). Rapid, in-stride soil phosphate measurement in archaeological survey: a new method tested in Loudoun County, Virginia. Journal of Archaeological Science, 34(11), pp.1859-1867

Sjöberg, A. (1976). Phosphate Analysis of Anthropic Soils. Journal of Field Archaeology, 3(4), pp.447- 454.

Svedjemo, G. (2003). Predictive Modeling for Iron Age Settlements on Gotland. [online] pp.1-26. Available at: http://www.divaportal.org/smash/get/diva2:415702/FULLTEXT01.pdf

Svedjemo, G. (2014). Landscape Dynamics - Spatial analysis of villages and farms on Gotland AD 200- 1700. Occasional Papers in Archaeology, (58), pp.1-255.

Svedjemo, G. and Jungert, E. (2007). Example Applications of Historical Maps in GIS and Databases. In: 11th Scandinavian Research Conference on Geographical Information Science. [online] Linköping: Linköping University, pp.1-16. Available at: http://uu.diva- portal.org/smash/record.jsf?dswid=2160&faces- redirect=true&language=en&searchType=SIMPLE&query=&af=%5B%5D&aq=%5B%5B%5D%5D&a q2=%5B%5B%5D%5D&aqe=%5B%5D&pid=diva2%3A775183&noOfRows=50&sortOrder=author_so rt_asc&onlyFullText=false&sf=all&jfwid=2160

Thompson, G. and Turk, J. (1994). Essentials of modern geology. Fort Worth: Saunders College Pub.

Verhagen, P. (2007)a. Testing Archaeological Predictive Models: A Rough Guide. In: Computer Applications and Quantitative Methods in Archaeology. pp.285-291.

Verhagen, P. (2007)b. Case studies in archaeological predictive modelling. Leiden: Leiden University Press.

Viberg, A., Trinks, I. and Lidén, K. (2011). A review of the use of geophysical archaeological prospection in Sweden. Archaeological Prospection, 18(1), pp.43-56.

Viberg, A. (2012). Remnant echoes of the past. Archaeological geophyisical prospection in Sweden. Doctorial Thesis. Stockholm University.

Walkington, H. (2010). Soil science applications in archaeological contexts: A review of key challenges. Earth-Science Reviews, 103(3-4), pp.122-134.

Waugh, D. (2011). The Spillings Hoard in the Gotlands Museum. The Silk Road, [online] 9, pp.165-169. Available at: http://www.silkroadfoundation.org/newsletter/vol9/srjournal_v9.pdf

Widerström, P. (2009). Spillings Farm – The Home of a Rich Viking in the Northeast of Gotland. In: A. Pettersson, ed., The Spillings Hoard, 1st ed. Visby: The Gotland Museum, pp.41-62.

Östergren, M. (2009). Spillings. In: A. Pettersson, ed., The Spillings Hoard, 1st ed. Gotland: The Gotland Museum, pp.11-39.

62