4.5. The topography of burial cairns in Houtskär, Korpo and Nagu
4.5.1. Introduction
In regional archaeological studies, environmental variables – topography, hydrology, geology, vegetation and so forth – in relation to settlement, and burial or other sites are often examined as possible independent variables, factors explaining the locations of sites. This is based on the presupposition that the sites were chosen in relation to their natural environment, and that the sites were assumed to be chosen freely, without any social or territorial limitations, without any preceding history. In settlement prehistory, this presupposition may hold in special conditions: in taking possession of wilderness to practice a new subsistence strategy, during climatic or topographic changes (e.g., deglaciation or shore displacement), or in other similar situations. However, we cannot overlook in general the likelihood that the locations of new dwelling sites may be influenced by the locations of older existing dwelling sites. Because of the cultural meanings of burial sites, it was not unimportant, either, where the graves of the older generations were located.
The varying locations of sites and the significance of the natural environment in this variation is a problem which thus presupposes sufficient knowledge of the ages of the sites. On the other hand, monitoring of the distributions of environmental variables in the entire natural environment of archaeological sites is also necessary. We have to find out where the sites are located as well as where they cannot be found (at least as current customary research techniques are used). Probably the total network of the sites can very rarely be recovered and thus, in practice, it is a question of acquiring a representative sample of sites and non-sites. In the following, we will compare known sites with randomly selected places, the latter of which reflect the distributions of environmental variables. The scrutiny concerns morphometrical variables which here are called terrain variables whose values can be derived from the altitudes of the locations.
The comparison is based on previous case studies. For instance, on the East Coast of the USA, Douglas Kellogg (1987) compared distances from shell-middens and randomly selected sites with mollusc habitats and water sources. Kenneth Kvamme (1992) analyzed the locations of agave cultivations using a digital elevation model. Herbert Maschner and Jeffrey Stein (1995) performed a logistic regression in modelling site location, using randomly selected places as a reference set.
Using a randomly selected reference set, we can obtain morphometric characteristics differentiating burial sites from randomly selected places. What constitutes those possible differences? Are there any morphometric differences not directly originating from shore displacement or not detectable by mere visual examination of maps? Were there any signs suggesting that the cairns were built near the shore zone of the building period? If so, the altitude of the burial place could be an indication of the age of the grave. Furthermore, are the grave sites located in relation to each other differently than those chosen randomly?
Because the reference set was generated by randomizing positions, some differences between the groups can be regarded as expected. Since the oldest cairns were erected more than 3000 years ago, land uplift has had time to lift the oldest burial places 15–20 meters upwards. As a logical consequence, burial places are likely to be situated on the average higher than random locations. Another effect of land uplift is that early burial sites are likely to be stochastically on higher elevations than more recent burial sites.
4.5.1.1. The field work
A relatively great number of cairns were discovered in the target area of the sampling study. In the area, the inner and outer archipelago zones are represented. The area contains a large part of the parish of Houtskär, the main island of Korpo and its SW archipelago, the western part of Storlandet, Nagu, and the southern isles of Nagu, extending in the south as far out as Nötö (fig. 24). The area (929 km) covers 15 % of the total research area.
64 cairns have been discovered in the target area, in other words 17 % of all the cairns in Åboland. 12 of the graves form a group forming the Iron Age cemetery of Furunabb, Houtskär. Here the cairns – excluding the largest one – remind greatly of each other, and are situated at a distance of a few meters from each other. This clustering is an obvious cause of spatial autocorrelation and interdependence of cases in sampling (e.g. Liedes & Manninen 1975: 13–17). One possibility to avoid this effect would be to calculate the mean value for each terrain variable and treat the cemetery as one grave. However, I think it is more consistent with the data to reduce the weight of the cemetery in the data by excluding some graves which are very close to each other. Thus, four randomly selected graves in Furunabb were excluded. The number of remaining burial places was 60.
Figure 24. The target area, the known burial cairns (filled dots) and the reference set (cross-hatched dots). © The National Board of Survey (MAR/103/98).
The reference set was chosen using stratified simple random sampling (Haggett et al. 1977: 270–273) so that 15 of the locations were located in the outer zone and the rest in the inner zone. The stratification ensures the outer archipelago zone to be represented in the sample roughly in proportion to the cairns discovered previously. Otherwise the probability of any site in the outer zone to be chosen into the sample would have been rather small because the land areas in the outer archipelago are small. Any random pair of coordinates was accepted, provided that, theoretically, a cairn could have been erected in the place determined by the coordinates. The place had to be located at least 5 meters above sea level, and sites located in water, bog, cultivated land or on steep slopes were excluded.
To maintain a reasonable precision of the sample, the size of the sampling unit was kept relatively small (Plog 1976; Plog et al. 1978; Read 1975; Schiffer et al. 1978). A radius of 400 m was drawn around each point. The areas falling within the radius were examined for the possible occurrence of burial cairns. No previously unknown graves were found, but one previously known cairn, on Bussö in Korpo (349) happened to fall inside one of the areas of the reference set.
4.5.2. The digital elevation model
The comparison of the burial places with the reference set is based on terrain variables derived from a digital elevation model (DEM) developed by the Finnish National Board of Survey. The DEM is a file of altitude points, interpolated from the height contours and shore lines of the Finnish basic map and recalculated to a network of rectangular cells representing altitudes. The resolution of the model is 25 meters. Altitudes are given at a numerical precision of one decimeter.
On average, the standard error of the altitudes of the DEM is 1.76 m. Of the 90 map sheets acquired for this study, 30 sheets are reported to give a better accuracy than the average value of the country. While examining the material, it is indeed evident that in many cases the altitudes given by the DEM differ from those locally measured. These differences cannot, however, be explained merely by the standard errors of the DEM. Firstly, the altitudes of the DEM do not necessarily denote the same pieces of land surface as the known altitudes. The DEM altitudes represent the average values of the cells while the known altitudes of burial places represent points, either levelled on-site or estimated from the basic map. Secondly, the location of each burial place was determined on the map. Thus, an inaccuracy of graphical origin is involved in the location of each burial place, roughly amounting to the 25 meter side of the cells of the DEM. The correct cell of the DEM does not always coincide with the known location of the burial place. Especially on precipitous slopes the inaccuracy may cause distortion in the relation between the burial place and the cell. However, a detailed look at the data gives the overall picture that the inaccuracies are of minor practical significance. To eliminate the inaccuracies discussed above it is evidently recommendable to use GPS positioning in the future (Ollikainen 1997; Poutanen 1998: 202–204). This applies particularly to emphasizing the sensitive changes in visibility when walking in the terrain (see Tilley 1994).
It is in any case worth while to state a few words about the possible systematic differences between corresponding altitudes obtained from different sources. Table 7 gives statistics on the differences between on-site measured altitudes and corresponding altitudes derived from the DEM, using different measuring methods. Individual height levels may differ distinctly from each other, but the differences are distributed both above and below zero. The differences between altitudes derived from the basic map and from the DEM are relative small on average. The altitudes measured using levelling instruments or aneroids indicate more systematic differences. This may be due to the fact that a saddle point was often chosen when a cairn was erected.
Table 7. Sampling area of Houtskär, Korpo and Nagu. Differences between on-site measurements of altitudes of burial sites and altitudes derived from corresponding cells in the DEM.
| Determination of altitude | n | Smallest difference | Greatest difference | Mean of differences | Standard deviation of differences |
|---|---|---|---|---|---|
| Examined from basic map | 43 | -3.8 | 6.5 | 0.40 | 2.24 |
| Spirit levelling | 15 | -2.5 | 5.1 | 1.63 | 2.22 |
| Barometric levelling | 2 | 0.3 | 3.2 | 1.75 | 2.05 |
| Total | 60 | -3.8 | 6.5 | 0.75 | 2.27 |
In practice, the deviations of altitudes affect most conspicuously the cairns located at low heights above sea level. For instance, the altitude given by the DEM for the burial place on Hummelskär, Korpo, is only 4.9 m, whereas the value obtained by using a spirit level was 8.4 meters. The grave of Kistkär, Österskär, Korpo, one of the outermost burial sites in the archipelago, is according to the DEM situated in a cell with a given altitude of 5.7 meters, whereas the altitude obtained from the on-site measurement was 7.5 meters. When shore zone datings are determined for low-altitude Iron Age graves, an error of this magnitude is significant. For this reason, the altitudes of burial places near the shore were determined using a spirit level or an aneroid. In higher terrain, some distance away from the shore, the DEM errors do not have the same significance for shore zone datings, since there the cairns were obviously not built near the shore. Furthermore, the point errors are smoothed when large areas are covered by the model. In the following discussion, I will use DEM altitudes in analyses based on GIS techniques, and on-site measurements in the determinations of shore zone datings.
4.5.3. The distributions of the terrain variables
4.5.3.1. The variables
In the comparison between burial places and random locations, ten morphometric variables are used. Variables such as altitude, height difference in relation to summits, relative altitude, slope, and sum of distances to the nearest neighbours are calculated by comparing the burial site with its topographic surroundings. Other variables are derived from the area around the burial place. The area is defined by a circular window opened in the DEM, with a radius of 500 meters, and with the burial place as the centre (1245 cells). The latter variables are mean altitude, median altitude, standard deviation of altitude, quartile range of altitude and relative proportion of sea area.
To determine the values of the terrain variables, the DEM has to be overlaid with the points representing burial cairns and random locations, the distances between the points have to be calculated and each cell has to be compared with its surroundings using different neighbour functions. These tasks were performed mainly using the Idrisi and BMDP softwares (Eastman 1997; Dixon 1992). The distances were computed using the NEIG module by Keith W. Kintigh (Kintigh 1987; Kintigh 1992: 41–46).
The values were determined as follows:
The altitude is the value of the cell in the DEM.
The height difference in relation to summits equals to the highest altitude within 500 meters subtracted by the altitude of the cell.
The relative altitude is obtained by extracting, using the cumulative frequency distribution, the relative proportion of the cells with an altitude lower than or equal to the cell in the middle of the circular window.
The slope can be determined from the DEM by comparing the altitude of each cell with the altitudes in the nearest adjacent cells on the north, south, west, and east sides. The slope indicates the most precipitous slope near the observation point. The slope may obtain a high numerical value, if there is a precipice in the neighbourhood, even if the cell in the center of the circle is on an even ground. The slope cannot directly be interpreted as a slope of the basal surface under the cairn, although that is sometimes the case. The accuracy of the slope depends on the resolution of the DEM and the local topography (Chang & Tsai 1991). Usually the calculated slope is smaller than or equal to the actual slope (Guzzetti & Reichenbach 1994: 59).
The sum of distances to the nearest neighbours indicates how close the five closest neighbouring burial sites or, respectively, the random places are located. It is defined as the sum of the distances, each of them weighted with the inverse of the square root of the distance
where dj denotes the distances to the j:th closest neighbour. Due to the graphical inaccuracy involved in estimating the coordinates of the burial site from the map, apparent zero distances may occur. They are eliminated by adding one meter to each distance.
For the five nearest neighbours, 0 < nd ≤ 5. The closer to each other the neighbours are located, the greater is the value of the sum of distances. The further away the j:th neighbour is from the observation point, in relation to which the distance is measured, the less the neighbour affects the value of the sum. The square roots of the divisors are taken, since otherwise neighbours situated a little further away would not have any significant effect on the value of the variable.
The mean altitude is given by the mean of the non-zero altitudes of the cells inside the window. The sea surface is thus excluded, since the result of including the sea level in the calculation would reflect rather the amount of sea around the burial place than the actual mean altitude of the terrain.
The median altitude is given by the median of the non-zero altitudes of the cells inside the window.
The standard deviation of altitude is given by the standard deviation of the non-zero altitudes of the cells inside the window.
The quartile range of altitude is given by the quartile range of the non-zero altitudes of the cells inside the window.
The relative proportion of sea area indicates the percentage of cells inside the window with altitudes equalling zero.
4.5.3.2. The differences of the distributions
The statistics for the distributions of the terrain variables are listed in tables 8 and 9. The locations of the distributions are compared using the Mann-Whitney U-test, completed with Kolmogorov-Smirnov tests for the equality of the location, variation, and shape of distributions. In the latter, the test statistic D is equal to the greatest difference between the two cumulative functions to be compared (Conover 1980).
The altitude. The burial sites are located on average 7 meters higher than the places of the reference set. The altitudes of the burial sites show a greater variation, too. The differences are statistically significant. At least partly the difference is a logical consequence of the age of the earliest graves: when they were built, the land surface was at least 15 meters higher at present. Contrary to Iron Age burial sites or randomly selected places, the earliest graves cannot be located at low altitudes. However, even a preliminary look at the topography indicates that the differences of altitudes are also partly due to the choice of location of the burial places, the gravebuilder’s inclination to choose high places in the terrain. The cairns were, thus, often built on land which had risen from the sea long before the building work. If we applied the shore displacement chronology in an uncritical way in order to date the burial places according to their altitude, we would, of course, end up with too high age estimates. This is evident especially for the classical hiidenkiuas graves that were often erected on tops of rocks. Reversely: on an average, the cairns are later than the age suggested by a superficial look at the topography.
Table 8. Comparison of the burial sites and the reference set (the means, the standard deviations, the greatest values, and the smallest value for all variables).
| Variable | Burial places, n = 60 | Reference set, n = 60 | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Mean | Sd | Max value | Min value | Mean | Sd | Max value | Min value | ||
| Altitude (m) | 20.335 | 8.298 | 38.8 | 4.9 | 13.657 | 6.326 | 32.3 | 6.1 | |
| Height difference in relation to summits (m) | 7.837 | 6.508 | 22.1 | 0.0 | 11.388 | 6.242 | 27.0 | 0.5 | |
| Relative altitude (%) | 77.497 | 24.051 | 100.0 | 27.5 | 70.737 | 20.166 | 99.1 | 16.9 | |
| Mean altitude (m) | 13.030 | 4.379 | 24.869 | 3.935 | 10.617 | 4.551 | 21.190 | 3.794 | |
| Median altitude (m) | 12.680 | 4.838 | 25.0 | 3.7 | 9.908 | 4.752 | 21.3 | 3.0 | |
| Sd of altitude (m) | 6.081 | 1.384 | 8.8790 | 2.321 | 5.312 | 1.817 | 10.882 | 2.404 | |
| Quartile range of altitude (m) | 4.485 | 1.350 | 8.050 | 1.900 | 3.903 | 1.638 | 8.975 | 1.700 | |
| Relative proportion of sea area (%) | 24.173 | 21.468 | 93.3 | 0.0 | 32.070 | 32.460 | 95.8 | 0.0 | |
| Slope (º) | 3.754 | 2.621 | 12.311 | 0.115 | 5.241 | 3.523 | 18.783 | 0.115 | |
| Sum of distances to nearest neighbours | 0.3004 | 0.3023 | 1.1409 | 0.0471 | 0.1001 | 0.0275 | 0.1724 | 0.0548 | |
Table 9. The significance levels α of the differences between the terrain variables of the burial cairns and the reference set. The two-tailed Mann-Whitney’s U statistic and the two-tailed Kolmogorov-Smirnov’s D statistic.
| Variable | Mann-Whitney’s U | a | Kolmogorov-Smirnov’s D | a |
|---|---|---|---|---|
| Altitude | 2666.5 | 0.000 | 0.383 | 0.000 |
| Height difference in relation to summits | 1259.5 | 0.005 | 0.300 | 0.009 |
| Relative altitude | 2284.5 | 0.011 | 0.267 | 0.028 |
| Mean altitude | 2354.0 | 0.004 | 0.283 | 0.016 |
| Median altitude | 2384.0 | 0.002 | 0.300 | 0.009 |
| Sd of altitude | 2343.5 | 0.004 | 0.350 | 0.001 |
| Quartile range of altitude | 2311.0 | 0.007 | 0.367 | 0.001 |
| Relative proportion of sea area | 1700.0 | 0.597 | 0.250 | 0.047 |
| Slope | 1309.0 | 0.010 | 0.267 | 0.028 |
| Sum of distances to nearest neighbours | 2906.0 | 0.000 | 0.600 | 0.000 |
Height difference in relation to summits. This difference is statistically significantly smaller for the burial sites – almost 4 meters – than for the reference set. The height difference distribution of the burial sites is clearly more weighted towards small altitude differences than the distribution of the reference set. Five of the burial sites are the highest hill tops within 500 meters and 14 burial places are no more than 1.2 meters lower than the highest summits. The cairns were typically placed in the summit areas of rocks but not, however, necessarily on the highest points.
Relative altitude. The relative altitude of the burial places is on average almost 7 % higher than in the reference set. The Mann-Whitney statistic is significant. The distribution for the burial places is weighted towards great numerical values: five cells contain the full value of 100 per cent, and in 19 cells the relative altitude is at least 95 %.
The differences for slope are significant as well, since on burial sites the land surface inclines on average 1.5 degrees less than in the reference set. The variation for the burial sites is also less than for the reference set. Burial sites can thus be characterized as horizontal surfaces not quite near slopes or steeps. The basal surface under the cairn seems to be rather regularly horizontal. The burial site of Storskogen, Korpo, has the highest value of slope, 12 degrees. The high value is caused by a precipice close by on the NW side of the grave. The basal surface under the cairn does not incline steeply.
The sum of distances to nearest neighbours has small numerical values and a positively skew distribution in both groups. This is due to the fact that in the archipelago dry land is split up into islands where the places belonging to the samples lie in clusters separated by stretches of sea. However, for the burial places, the sum of distances on average as well as its variation are significantly greater than in the reference set, which indicates that burial places tend to be located near each other. We may conclude that at the time when graves were built, the presence of existing graves on the site was culturally significant.
The mean altitude of the landscape around the burial places is significantly greater (2.4 meters greater) than the corresponding statistic of the reference set. The variation of the naighbouring topography is significantly greater for burial places. The differences characterize burial places on rocky highlands in contrast to lower moraines and fine sediment lands.
The relative proportions of sea area do not on average diverge significantly from each other. There is, however, a significant difference in the shapes of the distributions. The distributions are skewed towards zero because especially in many locations of the reference set there is no sea at all within the window. Twelve of the burial sites have a proportion of three percent or less while the corresponding number for the reference set is 23.
The summary may be: the cairns are situated on rock uplands varying greatly in altitude but on the average on high lying places, often in the summit areas, high compared with the adjacent area, on horizontal or little inclining surfaces, often near each other, and near the sea. The fairly high terrain typical of burial sites can already be observed in an overall survey of the maps; the position is also in relation to shore displacement. Without a systematic scrutiny, it is often difficult to discover that the burial sites are mostly plane terrain and that, nevertheless, they are not primarily among the highest spots in the neighbouring district, and that they are situated in places near the sea with, presumably, visibility to the sea.
Figure 25. The relative proportion (%) of the sea area of the window for the burial sites (upper) and the reference set (lower).
The visibility to the sea from the burial site seems to have belonged to the criteria of site choice. Land uplift has, however, weakened the visibility to the sea, and a closer investigation requires dating of the cairns and an analysis of changes of views caused by shore displacement. Visibility to the sea does not imply any certain topographic relation between the burial site and the shore zone at the time when the graves were built; actually there is a certain variety to be found concerning this relation. Some cairns situated on a low level suggest, however, that, at their time, they must have been constructed near the shore zone. The most distinct example of this is the cairn at Djupklevudden (364), Korpo, on a rock rising from the shore to the height of 6.4 m from sea level.