Somogyi Múzeumok Közleményei 18. (Kaposvár, 2008)

FEKETE CSANÁD: Predictive archaeological modelling in Somogy county

then less than half of the sites of the control area can be observed on the same aspect. The later result is more adverse than in the case of the sample area, but the majority of the sites can be identified on the control area, as well. Furthermore, the number of those sites whose territory does not touch southern-eastern aspect surface are limited on both areas: 10% and 12%. It is also noteworthy that on the vast area with relatively small difference of altitude and decisively western aspect, situated between the Fajszi-patak and the Bontakeszi-árok, there are no sites predicted by the model. The 300-metre-bufferzone measured from the watercourses is almost statutory. 93% of the sites are located within 300 metres on the control area. There are only 5 sites (Somogyfajsz 5, 22, 23 and Pusztakovácsi 49, 50) that are out of this buffer. Their common feature is, as in the case of the sample area, that they possess very sporadic finds. It indicates that they were dwelled for a short period of time. In the case of the site of Somogyfajsz 5 is worth to mention that huge number of metallurgical dross was found, that indicates the presence of iron metallurgy. This site is situated on the eastern bank of an artificial drain, which was not included into the survey of the water network. This fact was proved by a fieldwalking observation: this site can be found next to an oval, flowless cavity. I mentioned the importance of watersources more times in this study. This fact is also confirmed by the measuring data. Determining the distance from living waters where no sites can be found within was not as unambiguous as in the case of the sample area. The sites are closer to the streams in the case of Somogyfajsz and Pusztakovácsi.

21 The common feature of the sites that are within the 70-metre-buffer (Pusztakovácsi 6, 16, 22, 28 and Somogyfajsz 11) is that all of them are small and have only sporadic finds. Nevertheless, these sites can be identified during a contingent field-survey, so the inaccuracy of this model element is negligible. As far as the slope of the terrain is concerned there is a definite rule, just as in the case of the sample area: the areas above 12% slope are practically uninhabited, only a few sites are partly affected. 79% of the sites of the control area are located on the slope less than 12% with their whole territory. After analysing the soil data of the control area the same conclusion can be drawn as in the case of the sample area: the 70 metres distance equals to the uninhabitid zone, marked by the alluvial deposit of the floodplain. As for the chemical feature of the soil, the neutral reaction type with the most prosperous capability does not exist on the control area. 34 sites can be found on slightly acid soil, while an important number of sites can be observed on intensely acid soil. As far as the physical attributes are concerned, most of the sites are located on the soil type with strong water driving and highly water content ability. A significant part of them can be found on sandy soil with very strong water driving and sparse water content ability. After analysing the different physical paramètres I found that the result is slightly worse on the control area than it was in the case of the sample area (Table 4.), particularly in the case of the aspect examination. Much more sites can be observed on the surface with northern-western aspect in contrast to the expectations" of the model. As far as the physical and chemical features of the soil are concerned I could not use a generaly true rule despite of the tendency that appeared in both main features' respect (Table 3.). After surveying each elements the ideal inhabitable zone made probable by the model and the real geographical position of the sites can be compared. (Figure 8.), (Table 4.) 10 out 71 sites are entirely, while 7 are partly situated on the predicted area. Whereas 38 sites are located on the predicted area with less than the half of their territory and there are 16 sites that do not touch it in the least. The number of the identifiable sites are 43 (61%). This is a much worse result statistically than it was on the sample area. However there are some remarkable sameness between the ideal inhabitable zone and the archaeological sites. There are some well-modelled sites on the territory bordered by the Fajszi-patak and its different sidebranches, eastwards from the centre of Pusztakovácsi. Their extension and direction accord with the field data. This area is the most successfull one in the modelling's respect because the predicted patches are located site-like, modelling perfectly not only the sites but the empty areas among them. This feature cannot be observed on the most western and eastern (Kürtöspuszta) parts of the area, where quite large territories were predicted but they did not possess any sites. As an experiment I examined the relationship between the archaeological sites and the ideal inhabitable zone in different ages (Table 5.). In the case of the prehistoric and Roman Age sites of the sample area the polygons of the sites fit better to the predicted zone than the Middle Age sites do. This observation is, of course, true in the case of those Middle Age sites that possess prehistoric finds. Being farther away from the ideal inhabitable zone is peculiar to those sites that have pure Middle Age archaeological finds. In the peripheral, eastren part of the sample area there are 7 sites in a group that hardly touch the predicted inhabitable zone.

22 In the case of the control area there are worse scales in each age than they can be experienced on the sample area (74, 66, 58%). The poor fitting of the Middle Age sites is remarkable. Acknowledgements Hereby I would like to thank Dr. József Szabó for handing on the soil map and my college, Péter Szőcs for his helping ideas and advice. 21 Only 6% of the sites of the sample area are within the 70-metrebuffer, while in the case of the control area this rate is 22%. In the case of the Migration Age sites it does not worth to examine the statistical distribition because of the limited number of sites.