Landscapes of death: GIS modelling of a dated sequence of prehistoric cemeteries in Vastmanland, Sweden.

Landscapes of death: GIS modelling of a dated sequence of prehistoric cemeteries in Vastmanland, Sweden. Introduction The Malaren Valley in central Sweden (Figure 1) has had a longhistory of social analysis based on more than 10 000 prehistoric pre��his��tor��ic? also pre��his��tor��i��caladj.1. Of, relating to, or belonging to the era before recorded history.2. Of or relating to a language before it is first recorded in writing. cemeteries, the remains of which are often visible above ground asmounds or monuments (e.g. Ambrosiani 1964; Hyenstrand 1974). Thedescriptions and locations of these cemeteries are recorded in theSwedish monuments record, which has now been digitised and madeavailable online in both ESRI (Environmental Systems Research Institute, Inc., Redlands, CA, www.esri.com) The world's leading developer of geographic information systems (GIS) software, including programs that plot ZIP codes and addresses, demographic information and detailed, color-coded data. Shapefile and MapInfo TAB format The MapInfo TAB format is a popular geospatial vector data format for geographic information systems software. It is developed and regulated by MapInfo as a proprietary format. by theSwedish National Heritage Agency (Riksantikvarieambetet), through FMIS FMIS Financial Management Information System (Military Sealift Command)FMIS Facilities Management Information SystemFMIS Force Management Information SystemFMIS Force Modernization Information System ,the information system for archaeological sites and monuments (Blomqvist& Gentay-Lindholm 2002; Haskiya 2002). To make use of this information to map social change, it isnecessary to establish a chronology chronology,n the arrangement of events in a time sequence, usually from the beginning to the end of an event. for the burial grounds. An ambitiousattempt in the late 1970s employed descriptions of the sites and theirlocation in the landscape to suggest an estimated date of use(Gustavsson & Liden 1980). Since then there have been new surveysand a large number of rescue excavations. This suggests that it might betime for a new study, making use of the technology available today. This article presents a renewed attempt at relating the location ofcemeteries to their date, choosing the district of Vastmanland as a casestudy. In this study 1034 burial grounds and 3649 features registered assingle graves were analysed using GIS (1) (Geographic Information System) An information system that deals with spatial information. Often called "mapping software," it links attributes and characteristics of an area to its geographic location. and compared with thechronological chron��o��log��i��cal? also chron��o��log��icadj.1. Arranged in order of time of occurrence.2. Relating to or in accordance with chronology. information from excavated burial grounds in the samearea. Variables were collected from both the monuments record, such astype and shape of the graves, as well as from the landscape, such astopographical features and soil type at each site. A statistical profilewas thereby created and used to classify the burial grounds recorded bysurvey in relation to those that had been excavated. [FIGURE 1 OMITTED] Although variables from the physical landscape were used for theanalysis, the model was not seen as environmentally deterministic 1. (probability) deterministic - Describes a system whose time evolution can be predicted exactly.Contrast probabilistic.2. (algorithm) deterministic - Describes an algorithm in which the correct next step depends only on the current state. .Rather, it assumed a cognitive relation between the contemporary meaningof a burial and its position in the landscape. The result of this oughtto be that burial location in the landscape varies significantly withtime. Among the advantages of a statistical analysis is the possibilityof identifying quantitative trends in burial location by period, and ofincluding new information from future surveys and excavations, and soupdate and improve the model. The database FMIS has been available online since 2003, and today all ofSweden's provinces have had their records digitised and included inthe database--at present over 1.7 million items in total (www.raa.se).However, this number may increase substantially in light of futureexcavations as, when a burial ground Burial GroundAceldamapotter’s field; burial place for strangers. [N. T.: Matthew 27:6–10, Acts 1:18–19]Alloway graveyardwhere Tam O’Shanter saw witches dancing among opened coffins. [Br. Lit. is excavated, it is common for onlyabout 50 per cent of the graves identified during the excavation to havebeen known from previous surveys. As such, like any other archaeologicalrecord The archaeological record is a term used in archaeology to denote all archaeological evidence, including the physical remains of past human activities which archaeologists seek out and record in an attempt to analyze and reconstruct the past. , the FMIS database carries ambiguities and uncertaintiesassociated with surface features that are often vague, damaged ordisintegrating. Hence, the use of this record is not straightforward,and the interpretations of the social landscape and settlementstructures must take note of these caveats (Sawyer 1983; Bennett 1987). The database for the province of Vastmanland contains 4683 sitesamounting to 18 205 individual graves (Figure 2). The superficialappearance of each grave is described by up to three variables (broadly:overall type, construction and shape) (Table 1; Haskiya 2002). Acombination of these variables results in 59 recurrent types of burialmonument observed in the dataset. The variables for each site were collected and aggregated from thedatabase using the FME FME Formal Methods EuropeFME Faculty of Mechanical Engineering (Brno University of Technology, Czech Republic)FME Feature Manipulation EngineFME Facultat de Matem��tiques I Estad��stica Workbench application (Lowenborg 2007). Of the 59different types, some are very rare whilst the majority (67.3 per cent)belong to a single category: round, earth-filled stone-settings. In order to compensate for the poor distribution of values over thedifferent variables, a factor analysis was performed (Shennan 1988:271-80). This replaced the 59 types of burial monument with 35variables, with factor scores for variables with an eigenvalue eigenvalueIn mathematical analysis, one of a set of discrete values of a parameter, k, in an equation of the form Lx = kx. Such characteristic equations are particularly useful in solving differential equations, integral equations, and systems of of 1 orhigher, together describing 70 per cent of the variance in the material. [FIGURE 2 OMITTED] The information from surveys, as recorded in the FMIS database, isthus not sufficient for chronological interpretations by itself. Anotheraspect that has often been put forward as a complementary chronologicalindicator is how and where the different burial grounds are located inthe landscape. For this study, a range of landscape variables weredefined and attributed to each cemetery. The landscape variables, whichincluded soil type, elevation, slope, density of graves and the local('fuzzy') landscape form are listed in Table 2. The digital elevation model A digital map of the elevation of an area on the earth. The data are either collected by a private party or purchased from an organization such as the U.S. Geological Survey (USGS) that has already undertaken the exploration of the area. used for the topographical analysis wasa grid of 20m resolution calculated from digital contours Contours may mean: Contour lines on a map indicating elevation The Contours, a Motown musical group notable for the hit single "Do You Love Me" See also: plain at 5mintervals, provided by Lantmateriet, the Swedish Mapping Agency. Allvariables were calculated in ArcGIS, except the fuzzy fuzz��y?adj. fuzz��i��er, fuzz��i��est1. Covered with fuzz.2. Of or resembling fuzz.3. Not clear; indistinct: a fuzzy recollection of past events.4. topographicalfeatures (channel, peak, planar A technique developed by Fairchild Instruments that creates transistor sublayers by forcing chemicals under pressure into exposed areas. Planar superseded the mesa process and was a major step toward creating the chip. and ridge), which were calculated inLandSerf GIS (www.landserf.org) and exported to ArcGIS before analysis.In LandSerf the fuzzy features were calculated at different resolutions(15 x 15 pixels, 31 x 31 pixels and 65 x 65 pixels) in order toestablish the scale for analysis. Analysis The fuzzy features layers were combined by statistical principalcomponents factor analysis and compared with the different chronologicalperiods of excavated burial grounds. All the landscape variables usedwere stored as continuous surfaces. Together, these variables give richinformation about the landscape context at each site, while still usingbasic data that would not have changed significantly since the period ofstudy. Which variables to include in the discriminant dis��crim��i��nant?n.An expression used to distinguish or separate other expressions in a quantity or equation. analysis weredetermined in an experimental fashion, by including the differentvariables to see what effect they had on the outcome--since it is thepredictive power The predictive power of a scientific theory refers to its ability to generate testable predictions. Theories with strong predictive power are highly valued, because the predictions can often encourage the falsification of the theory. of the combinations of different variables that wassought (Norusis 1985: 93). A few variables were tested but excluded fromthe analysis since they did not seem to improve the results, such as thepresence of kerbs, rocks on the soil map, Late Iron Age place names andsites with rock art. In order to describe the chronology of the sites, a model ofclassification was used that separated the information into twovariables: the time of establishment of the site and whether the sitewas in use for more than one period or not. This will, of course, beproblematic in some cases, since the main concentration of graves maybelong to a later period than the site's initial use. Those graveswill thus belong to a different period than the one to which the burialground is primarily assigned. On the other hand, since this analysisputs much more emphasis on the location in the landscape, it makes senseto look at the time when the site was first chosen for burials. Themodel does not account for cases where there is discontinuity dis��con��ti��nu��i��ty?n. pl. dis��con��ti��nu��i��ties1. Lack of continuity, logical sequence, or cohesion.2. A break or gap.3. Geology A surface at which seismic wave velocities change. and laterre-use of a site. The periods followed the traditional Swedish systemused for excavated sites (Table 3). Excavated sites were only included if the exact location was known,if the monuments had been recorded during survey and if there was a firmindication of date. On these criteria, only 51 of the 1034 burialgrounds registered in FMIS in the county of Vastmanland were consideredeligible for analysis. The excavated burials were grouped intochronological periods of different lengths, so that there weresufficient examples in each for statistical analysis. The first periodthus spans a rather long period of time since there are few early sitesexcavated. Using the statistical package SPSS A statistical package from SPSS, Inc., Chicago (www.spss.com) that runs on PCs, most mainframes and minis and is used extensively in marketing research. It provides over 50 statistical processes, including regression analysis, correlation and analysis of variance. 15.0 and discriminate analysis Discriminate analysisA statistical process that links the probability of default to a specified set of financial ratios. ,the excavated burial grounds were clustered by their parameters. Thesecombinations of variables were used to predict the class membership ofthe unexcavated cemeteries (Norusis 1985: 75). The output of theanalysis are two predicted group memberships, the highest likely and thesecond highest likely, together with a value describing how well eachcase fits the discriminate function created from the sample. The resultsof the classification are summarised in Table 4. Results The upper half of Table 4 re-allocates the excavated sites toperiod on the basis of the discriminant analysis. As can be seen, thereis a good fit with recorded dates (numbers along the diagonal). Thenumber of sites attributed statistically to each period also roughlymatches those originally given by excavation (cf. Table 3). The modelseems to have had some difficulties in distinguishing between Periods 2and 3. Period 1 contains fewer burials than the original and the onlyexample of a re-classification over more than one period (to Period 4).This might reflect a real difficulty in distinguishing between someearly burial grounds and Viking Age Viking Age is the term denoting the years from about 800 to 1066 in Scandinavian History[1][2][3]. The vikings explored Europe by its oceans and rivers through trade and warfare. sites. However, success rate, at90.2 per cent, is fairly high considering that there were four differentclasses to choose from. This reflects how well the discriminate functionis potentially able to correctly date sites using the variablescollected. A good way to give an estimate as to what extent the model mightalso be relevant for the rest of the material is addressed by theleave-one-out method of cross-validation (Table 4; Norusis 1985: 87-8;Huberty et al. 1987: 324). This also gives a fairly strong positiveoutcome of 70.6 per cent, although again, there seems to be somedifficulty in recognising Period 3. An even more robust method ofvalidating the classification would be to use the leave-half-out method,where the regression model is built on 50 per cent of the sample andtested on the other 50 per cent of the sample. This would, however,require a large set of sample data, and thus was not attempted for thisanalysis where the sample is rather small to begin with. The same method was used to calculate the variable of continuity ofuse, with two classes; no continuity and continuity to the next period.The outcome seems rather robust, with 79.2 per cent of the originalgrouped cases correctly classified and 77.1 per cent in thecross-validated cases correctly classified. However, using this set ofvariables for predicting continuity of use of burial grounds isdifficult. Thus these values should only be considered as roughestimates and act as a reminder that the predicted chronologicalvariable only gives a predicted starting point Noun 1. starting point - earliest limiting pointterminus a quocommencement, get-go, offset, outset, showtime, starting time, beginning, start, kickoff, first - the time at which something is supposed to begin; "they got an early start"; "she knew from the for each site. [FIGURE 3 OMITTED] As can be seen in Figure 3, the landscape near Lake Malaren wassettled for the whole prehistoric period. Although the landscape settingmight differ for different periods, the spatial changes in location ofsettlements and burial grounds are often rather small and would bedifficult to analyse by means of traditional distribution maps. It issuggested that the best way to understand the settlement dynamics overtime would be to use quantitative methods of landscape analysis in orderto account for differences in location for the material as a whole(Lowenborg, forthcoming). Building on the results of the analysis of the cemeteries, achronological prediction was computed in the same manner for all thesingle graves (3649) in the study area. Although several of these mightactually represent burial grounds with more graves present than thosevisible, no continuity was predicted for the single graves. Since themajority of single graves are the same type of grave, as discussedabove, the possibility of using the landscape in the analysis tocomplement the archaeological record, as in the method suggested in thisstudy, could probably be beneficial. The single graves were correctlyclassified to 73.8 per cent, or 70.2 per cent for the cross-validatedcases. Evaluation The method used in this analysis, the discriminate function ofSPSS, could perhaps be criticised due to the type of linear correlationsbetween variables that is at the base of the regression analysis In statistics, a mathematical method of modeling the relationships among three or more variables. It is used to predict the value of one variable given the values of the others. For example, a model might estimate sales based on age and gender. . Thiscan be seen as a poor model of the dynamic and diverse reality of thearchaeological material at hand (Shennan 1988). In order to try a moreflexible statistical method of classification, the same analysis wasattempted with different kinds of cluster analysis Cluster analysisA statistical technique that identifies clusters of stocks whose returns are highly correlated within each cluster and relatively uncorrelated across clusters. Cluster analysis has identified groupings such as growth, cyclical, stable, and energy stocks. methods, such ashierarchical and k-means cluster analysis. The result of this wasdisappointing and did not match the good results of the discriminateanalysis, and was thus rejected. One reason for this might be thatcluster analysis constructs classes of cases based solely on thesimilarity of features they present. These features might representother aspects than those attributes that were considered good indicatorsof the chronological periods by the discriminant analysis, and thus morerelevant for this analysis. It could well be that other methods mightprove valuable, such as Artificial Neural Networks Analysis, that hasbeen suggested as a good tool for handling the kind of correlations thatthe archaeological material represents, especially for classifying toolsand artefacts (cf. Barcelo 2009). From Table 4 it can be seen that the cases that were not correctlyclassified in the first instance were usually only one period out. Thissuggests that the model is able to distinguish some aspects in thematerial that would be relevant for the question of predictingchronology for the burial grounds. The output also provides a value ofhow well each case matches the profile of the period to which it hasbeen assigned. This could be seen as an indication of how reliable theresults are for each site, and is given by the posterior probability The posterior probability of a random event or an uncertain proposition is the conditional probability that is assigned when the relevant evidence is taken into account. formula: P(G = g|D = d) at a value from 0 to 1. This can be read as: probability of groupmembership being the group membership predicted, given the discriminantscore, in line with Bayesian probability Bayesian probability is an interpretation of the probability calculus which holds that the concept of probability can be defined as the degree to which a person (or community) believes that a proposition is true. theory (Norusis 1985: 82-3).This information seems to be of great value for how to use the resultsof the analysis for further research, especially for the single graves.If only the cases that have a posterior probability value of 0.6 orhigher were considered, representing 58 per cent of the material, then81.5 per cent of the cross-validated cases would be correctlyclassified. This could be seen as a way of accepting that some of thesingle graves would be difficult to classify in this way. By excludingthe chronologically most ambiguous part of the material, the remainingsites could be used for further analysis with greater confidence. In a previous study, Klas-Goran Selinge performed a detailed surveyof the burial grounds of one parish, Rytterne, in southern Vastmanland(Selinge 2002). By visiting all of the burial grounds he estimated theperiod to which they belonged in order to do further settlement studiesof the region. In that study, only three chronological classes wereused, and if the result of the present study is compared to the work ofSelinge, the classifications of the 22 sites match to 86 per cent. InJuly 2008, I visited all of the burial grounds, paying particularattention to the sites where there was a difference in the results.While I would agree with the interpretations of Selinge in some cases, Iwould say that these sites belong to the category of burial grounds thatwould be very difficult to define in terms of chronology. This was alsoreflected in low posterior probability values for these sites. Theopportunity to assess the appearance of the sites in person certainlygives better grounds for the interpretation, since qualitative aspectsof the construction of graves and the appearance of the landscape can beincluded. For very large assemblages, however, it might be necessary torely on statistical analysis, such as the present study, and it seemsthat the results are sufficiently accurate. Although some sites will bemore difficult to classify, this could, to some extent, be accounted forin the posterior probability value. Conclusions In this paper it is suggested that a quantitative method using GISto model the landscape and statistics to predict the chronology ofburial grounds in Vastmanland could contribute to the interpretation andimprove the results. If the variables calculated using GIS wereexcluded, only 49 per cent of the cross-validated cases were correctlyclassified. The predicted chronology must, however, be considered as anestimate and be used with caution. The predictions will not be correctfor each individual site, but it is suggested that the model can beoperable operable/op��er��a��ble/ (op��er-ah-b'l) subject to being operated upon with a reasonable degree of safety; appropriate for surgical removal. op��er��a��bleadj. for more generalised Adj. 1. generalised - not biologically differentiated or adapted to a specific function or environment; "the hedgehog is a primitive and generalized mammal"generalizedbiological science, biology - the science that studies living organisms studies at a landscape level. if the studyarea was expanded to include more excavated sites it would probably bepossible to increase both accuracy and precision (i.e. increase thenumber of periods used). GIS has proved to be a powerful tool for interrogating the physicallandscape in archaeological research, and as such can be of great valuefor understanding prehistoric social processes. To some extent, evenissues of landscape perception can be hinted, in variables that givedetailed description of soil and topography topography(təpŏg`rəfē), description or representation of the features and configuration of land surfaces. Topographic maps use symbols and coloring, with particular attention given to the shape and elevations of terrain. . There will always be otheraspects of the mental landscape that are not easy to measure, such asrelations to previous burials, traditions, sacred places Sacred PlacesAlphsacred river in Xanadu. [Br. Poetry: Coleridge “Kubla Kahn”]Delphishrine sacred to Apollo and site of temple and oracle. , taboos,centres of population and much more. Where the variability of thelandscape is low and presents few possibilities to express preferencesfor certain conditions, the GIS approach might also be limited. Theremight also be considerable differences in opportunities to chooselandscape setting even at a local scale. This can be seen inVastmanland, where Early Iron Age burials are often located on eskers orelevated moraine moraine(mərān`), a formation composed of unsorted and unbedded rock and soil debris called till, which was deposited by a glacier. The till that falls on the sides of a valley glacier from the bounding cliffs makes up lateral moraines, . Another challenge in modelling the landscape using GISis therefore to calculate and test variables that can contribute to theanalysis in the particular geographical and cultural context at hand.With the increasing number of case studies using GIS, the modernlandscape archaeologist is presented with an increasingly elaboratetoolbox to choose from, in order to seek a rich understanding oflandscape and environment. Acknowledgements Many thanks to Dag Sorbom at the Department of Information Science,Uppsala University Uppsala University (Swedish Uppsala universitet) is a public university in Uppsala, Sweden, 64 kilometres (40 miles) north-northwest of Stockholm.[1] Founded in 1477, it claims to be the oldest university in Scandinavia, outdating the University of Copenhagen , for friendly advice and support. Received: 4 November; Accepted: 30 January 2009; Revised: 16 March2009 References AMBROSIANI, B. 1964. Fornlamningar och bebyggelse: studier iAttundalands och Sodertorns forhistoria. Uppsala: Almqvist &Wiksells. BARCELO, J.A. 2009. Computational intelligence Computational intelligence (CI) is a successor of artificial intelligence. As an alternative to GOFAI it rather relies on heuristic algorithms such as in Fuzzy systems, Neural networks and Evolutionary computation. in archaeology. NewYork New York, state, United StatesNew York,Middle Atlantic state of the United States. It is bordered by Vermont, Massachusetts, Connecticut, and the Atlantic Ocean (E), New Jersey and Pennsylvania (S), Lakes Erie and Ontario and the Canadian province of : Hershey. BENNETT, A. 1987. Malaromradets jarnaldersgravfalt, in T. Andrae,M. Hasselmo & K. Lamm (ed.) 7000 ar pa 20 ar Arkeologiskaundersokningar i Mellansverige: 143-64. Stockholm:Riksantikvarieambetet. BLOMQVIST, M. & C. GENTAY-LINDHOLM. 2002. Developing aninformation system for archaeological sites andmonuments--administration and maintenance model, in G. Burenhult &J. Arvidsson (ed.) Archaeological informatics Same as information technology and information systems. The term is more widely used in Europe. : pushing the envelope.Proceedings of the 29th conference Computer Applications andQuantitative Methods in Archaeology, Gotland, April 2001: 449-52.Oxford: Archaeopress. GUSTAVSSON, J. & H.A. LIDEN. 1980. Malardalens bebyggelse. Omett kartprojekt vid Statens Historiska Museum. Fornvannen 75: 73-80. HASKIYA, D. 2002. Developing an information system forarchaeological sites and monuments--data model and construction, in G.Burenhult & J. Arvidsson (ed.) Archaeological informatics: pushingthe envelope. Proceedings of the 29th conference Computer Applicationsand Quantitative Methods in Archaeology, Gotland, April 2001: 49-52.Oxford: Archaeopress. HUBERTY, C.J., J.M. Wisenbaker & J.C. Smith. 1987. Assessingpredictive accuracy in discriminant analysis. Multivariate BehavioralResearch 22: 307-29. HYENSTRAND, A. 1974. Centralbygd--randbygd. Strukturella,ekonomiska och administrativa huvudlinjer i mellansvensk yngrejarnalder. Stockholm: Almqvist & Wiksell. LOWENBORG, D. 2007. Flexibility instead of standards? How to makedigital databases on cultural heritage useable to large audiences--aresearcher's perspective, in S. Hermon & F. Niccolucci (ed.)Communicating cultural heritage in the 21st century. The Chiron Projectand its research opportunities. Budapest: Archaeolingua. --Forthcoming. Digital perceptions of the landscape. NORUSIS, M. 1985. SPSS-X advanced statistics guide. New York:McGraw-Hill. SAWYER, P. 1983. Settlement and power among the Svear in the Vendelperiod, in J.P. Lamm & H.-A. Nordstrom (ed.) Vendel period studies:transactions of the Boat-grave symposium in Stockholm, 2-3 February1981:117-22. Stockholm: Museum of National Antiquities Antiquities, nearly always used in the plural in this sense, is a term for objects from Antiquity, especially the civilizations of the Mediterranean: the Classical antiquity of Greece and Rome, Ancient Egypt and the other Ancient Near Eastern cultures. . SELINGE, K-G. 2002. Rytterne--fran jarnalder till medeltid. Envastmanlandsk fornlamningsbild, in O. Ferm, A. Paulsson & K. Strom(ed.) Nya anteckningar om Rytterns socken. Medeltidsstudier tillagnadeGoran Dahlback: 29-46. Vasteras: Westeras Media Produktion. SHENNAN, S. 1988. Quantifying archaeology. Edinburgh: EdinburghUniversity Press Edinburgh University Press is a university publisher that is part of the University of Edinburgh in Edinburgh, Scotland. External linksEdinburgh University Press . Daniel Lowenborg, Department of Archeology and Ancient History,Uppsala University, Box 626, 751 26 Uppsala, SwedenTable 1. Values for the variables used to describe the graves inthe study area, translated by the author. VariableVariable one values Variable two values three valuesFlat surface grave Construction: Boulders Shape: Boat-shapedCairn Construction: Earth-filled Shape: IrregularChamber grave Construction: Erected Shape: Other stoneGrave marked by Construction: Other Shape: Oval stoneMound Construction: Stone-filled Shape: RectangularOther Construction: Un-filled Shape: RoundStone circle Type: Boulder Shape: SquareStone Gist Type: Erected stone Shape: Stone circleStone-setting Type: Fixed stone Shape: Triangular Type: Passage-graveTable 2. Variables used to describe the burial grounds forthe analysis.Variable DefinitionGrave types Factor scores describing the number of cases of the different types of graves from Table 1, at each site.Site composition Number of graves per 1 [m.sup.2] of the burial grounds.Fuzzy topographical Variables describing the topographical features situation at each site, in a value of 0 to 1 for each of the features 'peak', 'ridge', 'channel or 'planar'. Calculated in LandSerf GIS (see below).Soil map A soil map was reclassified in order to differentiate primarily between eskers (value = 10) and clay (value 1), with moraine in between (value = 6). This was analysed with a focal statistics function in order to express the dominating soils in the vicinity as a continuous raster surface.Curvature The topographical curvature was described by two values, curvature in plane and in profile.Slope Maximum change in elevation, in degrees, at each site.Aspect Compass direction (azimuth) of the slope, in degrees.Fire-cracked stone Concentration (density) of heaps of fire-cracked stone in the vicinity of each site.Elevation Unmodified elevation data, in metres above present sea level.Table 3. Definition of the chronological classes for theexcavated cemeteries.Analytical Period 1 Period 2 Period 3periodDate Up to AD 1 AD 1 to 550 AD 550 to 800Period name Bronze Age and Roman Iron Age and Vendel Period Pre-Roman Migration Period Iron AgeTotal no. 9 sites/17.6% 21 sites/41.2% 14 sites/27.5%of sites/%Analytical Period 4periodDate AD 800 to 1050Period name Viking AgeTotal no. 7 sites/13.7%of sites/%Table 4. Period attribution of all cemeteries (excavated cemeteriesare shaded grey). Predicted Group Membership Period 1 2 3 4Excavated Count Period 1 8# 0# 0# 1# Period 2 0# 19# 2# 0# Period 3 0# 2# 12# 0# Period 4 0# 0# 0# 7#Surveyed Unexc. 75 506 222 176Excavated % Period 1 88.9# .0# .0# 111.0# Period 2 .0# 90.5# 9.5# .0# Period 3 .0# 14.3# 85.7# .0# Period 4 .0# 0.0# .0# 100.0#Surveyed Unexc. 7.7 51.7 22.7 18.0Cross- Count Period 1 7 1 0 1validation Period 2 0 17 4 0 Period 3 0 4 7 3 Period 4 0 0 2 5 % Period 1 77.8 11.1 .0 11.1 Period 2 .0 81.0 19.0 .0 Period 3 .0 28.6 50.0 21.4 Period 4 .0 0.0 28.6 71.4 TotalExcavated Count 9# 21# 14# 7#Surveyed 979Excavated % 100.0# 100.0# 100.0# 100.0#Surveyed 100.0Cross- Count 9validation 21 14 7 % 100.0 100.0 100.0 100.0Note: Excavated cemeteries are shaded grey indicates #.