Somogyi Múzeumok Közleményei 18. (Kaposvár, 2008)

FEKETE CSANÁD: Predictive archaeological modelling in Somogy county

Used softwares I used fGIS (forestry GIS) to digitize and edit vector data and represent raster result maps. It is a free, compact shape file editing software and has the advantage of its small size and portability.

7 I used GRASS 6.0 (Geographic Resources Analysis Support System) to transform the different raster data (military surveys, topographical maps) into projection and create the releif map and the terrain model. This software is also free, open source, its updated version can be downloaded from the Internet.

8 Used data and maps I used the VI N/23 and VIII/24 sections from the First Military Survey and the XXVI/59, 60 and the XXVII/59, 60 sections from the Second Military Survey. As a soil map I used the map on the scale of 250 metres to the centimetre of Kreybig's map. This map is the only one of its kind, which is large scale and covers the whole territory of Hungary to date. The relevant section, transformed into the EOV projection, was given by RISSAC

9 GIS Laboratory. As topographical map I used the following sections from the EOV projection: 23-121, 23-122, 33-332, 33334, 33-343 and 33-344. Steps of defining the inhabitable areas Parameterizing of EOV projection could not directly be carried out in GRASS because the geodesic datum and the ellipsoid of EOV cannot be found on the parameterizing list. I used the Swiss Oblic Mercator projection (its ellipsoid: grs 67), which is the most similar to EOV. Its rotating and shoving paramètres compared to WGS84 was given according to the paramètres of HD72 geodesian datum (dX: +57, dY: -70, dZ: -9).

10 The very first step of GIS work was to transform the different maps into the same (EOV) projection. After digitizing the EOV sections I georeferred them with the help of the coordinate-pairs of the four sections' corners. The fitting of the EOV sections is about 1 metre. The military survey's maps were fitted into projection with the help of common joining points that can be found on both the EOV and the old military maps. The fitting of the First Military Survey was not as accurate as in case of the Second Military Survey, because the first one has no projection. The first one was transformed with about 100 metres error, while the later one was transformed with less than 50 metres error. The software was originally developed for the Wisconsin State Forestry handling natural resource data. 8 Originally developed by the U.S. Army Construction Engineering Research Laboratories as a tool for land management and environmental planning by the military. 9 Research Institute For Soil Science And Agricultural Chemistry Of The Hungarian Academy Of Sciences 10 The projection paramètres was determined by Gábor Tímár. (TÍMÁR 2003, http://sas2.elte.hu/tg/hd72.htm) As for the vector data, I started digitizing the contour lines of the topographic map. Where the model needed higher accuracy I digitized the contours by two metres. In some cases - because of the altitude data was not given or could not be deduced -there was not enough contour data so I digitized every single lines whose height data was known. In those cases, where this accuracy was not needed, the contour lines were digitized by four-five metres. The altitude value of some contour lines was not integer, in these cases their z data were rounded, (e.g. 132,5 -> 133) The physical paramètres of the sample area was defined that the deduced raster elements could cover the sites' polygons in the largest possible percentage. At first an aspect map of the sample area (Figure 1.) was created with the help of the terrain model (Figure 2.). The northern direction was determined between 315° and 45°, the eastern between 45° and 135 °, the southern between 135° and 225° and the western between 225° and 315°. Most of the sites of the sample area (23 sites entirely, 31 partly) can be found on the terrain with southern-eastern aspect, while only 6 can be discovered on norhern-western aspect. More than 75% of the sites can be recognized and identified with their parts falling to southern-eastern aspect. The next step was to measure the distance between the sites and the watercourses. At first I specified the distance from living waters where can be hardly found any sites out of. The buffer zone of the streams, where all sites can be identified, is approximately 600 metres. This distance would set too large limits to the indicated area, so I chose such a buffer zone, where not all of the sites were within but most of them. The distance of 300400 metres seems to suit the model's requirements. 44 out of 60 sites can be found within the 300 metres buffer zone with their entire area. 11 sites are partly located within this buffer and there are only 5 sites whose entire area are out of this range. 92% of the sites can be identified within this range if those sites are also considered to be recognizeable that are only with their parts within the 300-metre-buffer. Without these sites the probability of the sites being identified is still 75%. At second I determined the distance from living waters where no sites can be found within. This zone is approximately 60-80 metres from the watercourses, which is equal to the humid, reedy, marshy region that, in fact, is the border of the flood area. Within this zone sites could hardly be located. I chose the 70-metre-buffer that almost entirely free from sites. Only one site can be found within this distance with its entire area and there are merely 3 sites that are located within this zone with their area bigger than their half. Most of the sites (more than 90%) are out of the 70-metre-buffer.

11 (Figure 3.) Afterwards I created a map of the slope categories that was generated from the terrain model. (Figure 4.) Six different cathegories were distinguished: 0-3%, Only a few sites are located within 70 metres with very little parts. They do not influence the formation of the model.