Methodology: Reconstruction of the OIS5a/c ecosystems of the Deba and Urola valleys.


1. the base map


The topographic data for the analysis was acquired as a digitised version of the Servicio de Información Territorial’s 1:5000 series of maps of the province of Guipuzcoa, downloaded in .dxf format from the website of the Diputacion Foral de Guipuzko[1]. As initially downloaded the data comprised more than thirty layers, including information on roads, railways and amenities such as swimming pools and playing fields. The number of layers was therefore reduced to 5:

1.     25 metre interval contour lines

2.     natural bodies of water

3.     rivers

4.     streams

5.     stream/river beds


This data was then cleaned and georeferenced[2] using AutoCAD Map 2000 and imported into GRASS 5.0.2. as a vector file[3] and the lines reconstructed[4] before labelling[5]. Once contour labelling was complete, contour heights were checked against the original paper maps and found to be highly accurate.


The downloaded data did not, however, include the bathymetric elevation data necessary to model rising and falling Pleistocene sea levels. Bathymetric contour lines were thus digitised by hand from georeferenced .tiff files[6].


Once the bathymetric and terrestrial contour maps were joined[7] and transformed to a raster format[8], a raster mask of ‘null’ values was layered over the resulting map[9] to maintain its original boundaries[10].  One issue of concern was the difference between the contour intervals in the terrestrial (at regular 25m intervals) and bathymetric data (at uneven intervals: -5, -10, -20, -50, -100 and -200m). As distances between the bathymetric contour lines increase, so the spatial interpolation module produces a ‘stepped’ effect in the raster map, whereby the interpolated values do not result in a smooth surface (a common problem, see e.g. Wheatley & Gillings 2002 for discussion of potential solutions). However, the affected regions were only really a factor during periods of maximum sea regression, as for example at the Last Glacial Maximum, and so impact on only a small subset of the analysis, and proved insignificant at the scale of the analysis, as described below.


2. preparation of the OIS5a/c timeslices


The first part of the process of ‘placing’ animals into the landscape involves its classification in terms relevant to the animal species’ distributions. A number of systems of classification have been presented in the literature, including Sturdy et al.’s comparable work in Epirus, (1997), Butzer and Clark’s work in Palaeolithic Cantabria (Butzer 1981, 1986; Clark 1983), and in the Holocene, van Hove’s work in Calabria, Southern Italy (2003) and Hammond’s ‘classes of land-surface form’ for the Holocene USA (1964) as well as other systems designed for non-archaeological uses, such as agricultural potential and land-use maps produced by many agricultural agencies around the world such as the Soil Survey of England and Wales’ Land Use Capability Classification (Bibby and Mackney, 1977).


The two major axes, corresponding to the major kinds of data required, remain reasonably constant:

1.     topographic: data on elevation, with its associated temperature and vegetational variation, as well as slope and drainage systems and what Sturdy et al. (1997) call general ‘ruggedness’ of terrain, which essentially describes the ease of access of parts of the landscape by different animal species.

2.     edaphic: ‘the underlying soil and subsoil characteristics which make a piece of ground more or less attractive to animal species in terms of their nutritional needs’ (ibid.: 593)


Topographic factors are of course covered by the downloaded data; edaphic factors, however, are more difficult to examine. Sturdy et al. (1997) use a combination of geological and chemical analyses of the major soil types of the area, but although it was originally hoped that a geological element would be incorporated into this reconstruction, geological data for the region proved extremely difficult to source as the 1:25,000 maps produced by the Instituto Geologico y Minero de Espana (IGME) were unavailable for digitising. IGME’s 1:50,000 Mapa Geologico de Espana sheet 5/12 (Bermeo/Bilbao) and that published by Galan (1988) served to supplement the topographically-based reconstructions as necessary. Chemical analysis is beyond the scope of this project, and no relevant work has been done in the region to date. Modern soil distribution maps are of course available, but modern soils have been subject to various processes of change throughout the course of prehistory and particularly in the modern era with the adoption of intensive farming practices.


However, the development of particular soils is in any case a highly context-specific process dependent on a multiplicity of factors including the duration of development, climatic, chemical and physical characteristics of the immediate environment and particularly the parent material, the local bedrock (Buol, et al. 1973, 108-9; Wild, 1993, 49). Thus there are no one-to-one linkages between rock and soil types, and this study draws on geological and edaphic data only in a very general way to supplement the data from palynology and ecology used for reconstruction of the ‘timeslices’.


Development of these ‘timeslices’ maps required ‘translation’ of the palaeoenvironmental data into essentially topographically-based categories that would make sense in terms of a GIS model: such descriptive terms as ‘sheltered valley bottoms’ therefore needed to be broken down in terms of variables handled by the computer model, i.e. altitude, slope, aspect, distance to water. Such variables describe the environment in terms which are hugely significant for the movement of embodied entities through it.  Specific details for individual timeslices (and how they relate to the habitats preferred by various animal species.


Changing Pleistocene sea levels are perhaps the most obvious altitudinal consideration. Although it is important to consider the potential roles of marine and littoral species, this analysis focuses mainly on the terrestrial animal species with which hominid populations interacted and thus areas of land below the sea level estimated for each timeslice are simply assigned to the category of ‘sea’ and treated as functionally impassable (by being assigned a high ‘movement’ cost). At the opposite extreme, areas at altitudes above the snowline can also be considered out of bounds (Bailey 1983)


Other altitudinal effects are of course not so binary but nevertheless significant in terms of the experience of the landscape to people moving through it: altitude is associated with differing climatic and geological and thus vegetational regimes and potential views. Perhaps the most obvious example of this is the treeline; in Vasco-Cantabria today beech forests grow at altitudes of around 1600m, and deciduous oaks to 1100m (Butzer 1986, 212). However, such altitudinal associations are of course hugely affected by climatic regime and would have varied considerably over the course of the Middle and Upper Palaeolithic. Butzer (ibid.: 204) used Hammond’s synthetic terrain classes to reconstruct the vegetation patterns of a hypothetical full-glacial Cantabria with four major altitudinal categories (which were in practice cross-cut by slope categories): <100m, 100-300m, 300-1,000m and >1,000m. These categories were designed to be directly relevant to the distribution of large herbivore species, and are thus a useful guide for the relevant (‘cold’) timeslices.


Similarly, Clark’s altitudinal and slope categorisation system for early Holocene/Boreal Vasco-Cantabrian Spain (1983, table 9.2.) divides the territory into three categories: <100m (further subdivided by other criteria); 100-200m; and all territory above 200m or 100m if it is of a ‘steep’ gradient (‘Alpine’). This system was used as a guide for the modelling of ‘warmer’ timeslices during which altitudinal effects would have been less pronounced; the lowered snowlines of colder phases would have compressed ecotones, with higher elevations far more subject to exposure and rigorous climatic regimes (Bailey 1983)


The gradient of a slope is also highly significant in terms of the way a landscape is perceived and experienced. It has a significant effect on the vegetation able to grow (for example, steeper slopes in the coldest palaeoenvironmental phases are likely to have been highly unstable, with minimal soil development and thus largely bare; e.g. Butzer 1986), and is therefore indirectly as well as directly (in terms of access) related to the animal species that might be associated with parts of the landscape.


For the purposes of this analysis I have projected that slope gradients would not have changed significantly over the course of the Pleistocene, and have used the same map layer for all timeslices. Of course this is certainly overly simplistic – slope gradients, particularly those of river valleys, will have changed almost constantly and sometimes dramatically throughout the various palaeoenvironmental phases, with changing moisture regimes and drainage patterns in particular altering the landscape through slope movement processes, erosion and the deposition of colluvium and alluvium. However, such change would be virtually impossible to model within the scope of this analysis, and in any case it seems likely that the overall structure of the landscape has not changed significantly. Provided the influence of slope gradient is not taken as an exact reconstruction of the palaeoenvironmental phase in question, but rather a guide to large-scale landscape patterns, this will not be a significant problem for the analysis.


Although the model can provide a more or less precise measurement of slope in terms of either degree or percentage, for the reasons discussed above, using these raw figures would provide a spurious accuracy to the analysis. In any case, human movement around the landscape is based less on calculation of exact slope gradients than on their mental categorisation of them as perhaps ‘steep’ or ‘gentle’. Exactly how slope gradient may be divided in terms of human judgement is a matter of some debate. Clark, for example, considers any gradient of greater than 40 degrees as ‘steep’ (1983: table 9.2.), while Hammond (1964) prefers to term any slope of less than 8% or 5 degrees ‘gentle’ (see also Butzer 1986). Van Hove (2003) defines slopes of gradients 0-10% as ‘low’, 10-90% as ‘high’, and slopes of 100% ‘cliffs’. Munier et al. suggest categories of <5, 5-10, 10-20 and >20 degrees (2001, table 2), and Vogt et al. (2003, table 1) used 5 categories of 0-2%, 3-13%, 14-20%, 21-55% and >56%, while the Scottish Avalanche Information Service’s Avalanche Hazard Scale defines ‘steep’ slopes as being greater than 30% (c25 degrees)[11].


Consideration can also be given, of course, to the energy costs and changing experience of bodily movement over slopes of varying gradients; energy costs of moving uphill appear to increase monotonically with slope much as would be expected, but the energy costs associated with walking downhill decrease until -10% (c5 degrees), and then begin to increase again. When running downhill, the decrease lasts until the slope reaches -20% (c11 degrees; Susta et al. 2000). Working from these examples, I have settled on a figure of 0-10%/0-5 degrees as a ‘gentle’ slope, 11-30%/6-c25 degrees as ‘moderate’ and anything greater than 30%/c25 degrees as ‘steep’.


Another potentially significant topographic factor is aspect. More ‘sheltered’ south-facing locations were more likely to provide suitable conditions for acting as refugia for vegetational species which could not survive on more exposed north, coast-facing slopes, and the exposure or shelter of an area to weather or sunlight, for example, would have been a factor in the perception and experience of the landscape.


An aspect map was automatically generated by GRASS at the same time as the slope map[12], and was further reclassified into three categories: Exposed or north-facing[13], ‘sheltered’ or South/East/West facing[14] and ‘flat’ (land with a gradient of <5 degrees), given a ‘null’ value as it cannot sensibly be said to have an ‘aspect’ (Munier, et al. 2001). The resulting map could of course be used for all timeslices, albeit with the same caveats as discussed in the section above.



The topic of changing sea-levels was considered above; however, proximity to open coastline also has a significant effect on vegetational patterns. Although Butzer defines the ‘coastal plain’ of full glacial Vasco-Cantabria solely in terms of altitude and slope (1986, 204), Clark prefers to define a category of ‘open coastline’ <100m from the shoreline, as well as creating an ‘estuarine’ category defined as the ‘First 500m of river flood-plain’ (1983, 100). After some experimentation with the modern data, I found that a category of land within 4km of the coastline appeared to correspond well with the modern coastal plain; such a category was created using the GRASS module r.buffer[15] separately for each timeslice.


Proximity to rivers and streams is also a factor, particularly in ‘warmer’, more humid palaeoenvironments. Van Hove (2003) used a category of ‘river channel’, defined as being within 50m of a river and at an altitude of between 15m and 1000m. The maximum width of the modern flood plain around both modern rivers is c200m; in this analysis, therefore, the GRASS module r.buffer was used to create a landscape category of ‘watercourse’ <200m from each river. This category was used unaltered for all timeslices, subject, of course, to the caveats discussed above.


Sea level


In Cantabria, sea level changes would have had an important effect on the extent of the presently narrow coastal plain. Full glacial sea levels would have been between 100m and 130m lower than today, although the narrowness of the continental shelf just off the northern coast of Spain meant that even in full glacial conditions only an extra 4-12km of coastal plain was exposed (Straus 1992).




At the opposite altitudinal extreme, snowlines probably marked the upper limit of human and animal activity:


Although browsing is possible under all but the heaviest snowfalls, feeding would not be easy above the snowlines, even if there was not much snow. Vegetation would have been ice-encrusted and there would have been relatively little browseable scrub (Turner & Sanchez Goni, 1997; see also Gilbert and Beckinsale 1941, cited Bailey 1983, 150).


In the present climatic conditions there is no permanent snowline in the region (Bailey 1983, 150), although further west in the higher Picos de Europa there are some year-round snowfields at heights of c. 2400m-2600m, with patchy snow still lying at heights of above 2200m as late as June. Current permanent snowlines in the French Pyrenees are around 2800m, and (theoretical) permanent snowlines in the Sierra de Aralar c. 10km to the southeast of the head of the Urola valley are calculated at 2400m (Kopp 1965, 14). Sturdy et al. cite a general modern permanent snowline of 2400m, with a pleniglacial snowline depression of around 700m (Sturdy, et al. 1997, 591).


Estimates of the permanent snowline during cold phases of the Pleistocene range between 1,650m and 1,025m above sea level see references in (Straus 1992, 21). At the Last Glacial Maximum (LGM), glaciers existed in the Sierra de Aralar, with terminal moraines found at 825m above (current) sea level, 25km from the coast (ibid.) and perhaps even lower, down to 460m on the northeast slopes (Kopp, 1965: 14). Glacier tongues of the Atlantic catchments extended as low as 500m in the west and 900m in the North, while Regional Climatic Snowlines (RCS) appear to have been at 1100-1700m along the Atlantic-Duero watershed, with the higher mountains snowbound year-round (Butzer 1986, 206).


In the Sierra de Aralar permanent snowlines are calculated at 1050m for the LGM, a depression of 1360m (Kopp 1965, 14). In the Picos de Europa, Butzer (1973), placed the permanent Pleniglacial snowline around the level of 1400 – 1500m above sea level, and mentioned that some evidence of earlier (Middle Pleistocene) glaciations suggested a level of around 1450m. Winter snowfall in the area today means that much of the terrain at altitudes above c1000m is impassable (Gilbert and Beckinsale 1941, cited Bailey 1983), but the estimation of lower limits for winter snowfall in the past is problematic because of uncertainties about relative precipitation and snowfall (see Bailey 1983, 151): clearly the 1350m descent in the permanent snowline indicated by Kopp (1965, 14) for the Pleniglacial cannot simply be applied to the line of snowfall; the ameliorating proximity of the sea would have maintained the coastal plain as a relatively favourable winter zone (snow rarely falls on the coastal plain today; (Altuna 1972, 17). Although 4.5 - 5m of snow was probably common in lowland Vasco-Cantabria from early December to late April in pleniglacial phases, the impression is one of a ‘moderately thick snow cover that would only occasionally pose a problem for grazing animals or hunting forays’ (Butzer 1986, 216). Sturdy et al., working in Epirus, Greece, give suggested seasonal snowlines detailed in the table below, and as the suggested permanent modern snowline of 2400m they give is very close to suggested snowlines for northern Spain, I have used these as the bases for my own estimations.





Early summer

High summer











Suggested Palaeolithic snowlines in Epirus, Greece (after Sturdy, et al. 1997) Table 30.2.)




In terms of vegetational regime, variations between warmer and colder paleoenvironmental phases tend to show up as alternations between open and more forested conditions (van Andel, 1998, 491; Kukla, 2002: 9). Conditions in the north were harsh during the coldest Pleniglacial phases, with areas of almost barren polar desert and open tundra communities.


However, reconstruction of palaeoenvironments is not simply a case of making latitudinal shifts: Guipuzcoa (and Vasco-Cantabrian Spain more generally), represent low-latitude tundra-steppe, at the southern, more productive end of the biome, and reconstructions demonstrate that ‘although categorized as equivalent to modern biomes, the simulated paleovegetation is structurally different from the equivalent modern biome’ (Huntley, et al. 2003, 209). Colder phases here would have been characterised by a more steppic flora than the tundras of further north, with heath and grasslands communities highly conducive to cold-adapted large ungulates. Much of the coastal plain and river valley floors would have been dominated by grasses. During milder glacial phases, although steeper, north-facing slopes were probably denuded of vegetational cover and frequently geomorphically unstable (Straus 1992, 51), considerable localised patches of hardier tree species such as pine and birch survived on sheltered south-facing slopes and along valley bottoms, with relatively rapid tree expansion during warmer phases.


Warmer phases, in general, were characterised by mixed (although probably mainly coniferous) forest with some thermophile deciduous trees, arriving in familiar succession with birch and pine through the elm, oak, hazel and hornbeam forests, gradually shifting back to pine, spruce and other cold-tolerant species when colder conditions returned. Tree cover, however, was probably always rather open to judge from avian faunas (Adams & Faure 1998) and arboreal pollen (AP) values (Mellars 1996, 27), although reforestation might have been quite significant during longer or warmer phases – though still not attaining the dense forests recorded in the area at the onset of the Holocene (d'Errico & Goni 2003, 777). Open pinewoods and parkland probably dominated during ‘warm’ phases, then, with isolated oak, hazel and birch groves in more sheltered spots, and the coastal plain and river valleys providing rich grass pastures.



Although the generalised ‘warm’ and ‘cold’ phase regimes discussed in section 7.4. form the basis for the reconstruction of palaeoenvironments in this analysis, these large-scale patterns were subject to considerable variation during the different palaeoenvironmental phases of the Vasco-Cantabrian Pleistocene














Typical Mousterian

Range from 288,000+34,000-26,000 B.P. to 231,000+92,000-49,000 B.P. - see table A5.8. and section A5.6. for discussion.


Levels from the Early Glacial/OIS 5a or c/St.German I or II


In many ways, as Mellars has commented,


the most striking feature of the early glacial period is not the severity of the colder periods but the relative warmth of the intervening “interstadial” periods represented by stages 5c (St. Germain I) and 5a (St. Germain II) of the oxygen-isotope records (Mellars 1996, 21).


The total volume of global ice sheets during these phases – although still much greater than that during fully interglacial periods – was only around half of that of the cold stages of 5b and 5d.


Both OIS 5c (correlated with the pollen Interstadial St. Germain I) and OIS 5a (correlated with St. Germain II) appear to have been fairly marked and significant warming phases, marked by the shrinking of the Scandinavian ice sheet during OIS 5c (van Andel & Tzedakis 1998,  490), and the retreat of the North Atlantic polar front[16] (Mellars, 1996). July temperatures are estimated to have ranged from c12 degrees C (approximately 6 degrees below present values) in southern Scandinavia, to 18-20 degrees C (only 1-2 degrees C below present) in the southern parts of France and along the Atlantic coast (Zagwijn, 1990). These estimates are supported by pollen data from Les Echets and La Grande Pile; at the former peak warm conditions again appear to have been around 1-2 degrees C cooler than present, and at the latter the pollen demonstrates conditions almost identical to those in the same region today (Mellars 1996, 23).

Sea level

Lambeck et al. argue that global sea levels were around 20-30m lower than at present during these two phases (2002). However, European research suggests a figure at the lower end of this estimate, with van Andel and Tzedakis suggesting a 20m drop (1998, 491) and Mellars estimating around 10-20m (1996, 23). It seems likely that OIS5c levels were slightly higher than those of 5a, as the ice volume curve shows a gradual global increase in 5a beyond that of the previous Interstadial (van Andel & Tzedakis 1998, 491), but this has yet to be quantified satisfactorily and in this analysis I have used the figure of -20m for both phases.


With temperatures only a little lower than those of today, permanent snowlines during OIS5a and 5c were certainly well above the highest peaks above the study region; the ‘summer’ timeslice, therefore, does not include a snowline, and for ‘winter’ I have used Bailey’s altitude of 1000m as a functional upper limit for human and animal activity (1983, 150).


While the climatic oscillations of OIS5, as noted above, appear in the pollen records as alternations between expanding open vegetation during the colder phases and returning forest conditions in the warmer periods (van Andel & Tzedakis 1998), the exact ways in which the interstadials of 5a and 5c were reflected in the local records varied significantly across Europe (Mellars 1996, 22). South central Europe and France were apparently rather rapidly covered by a mixed forest of birch and pine, with some elm, oak, hazel and hornbeam during the warmer periods (ibid.; van Andel & Tzedakis 1998), although fossil bird faunas suggest that the Interstadial tree cover was always rather open (Adams & Faure 1998).


These data were combined into a broad-scale reconstruction of the ecosystem of OIS 5a/c, and also the habitats with which various animal species were associated


‘placing’ animals in the landscape


Having established some of the relevant parameters of the palaeoenvironments, this section will consider how different animal species might have behaved within it. A number of avenues are relevant here, including:


·        Preferred feeding regimes – e.g., whether animal species prefer to graze in more open conditions or browse in woodlands.

·        Accessibility of and preferences for different types of terrain – the landscape offers e.g. ibex and horse very different kinds of affordances in terms of elevation and slope.

·        Seasonal variation in behaviour; aggregation, dispersal and patterns of movement, reproductive changes and variation in condition over the course of the year.


Of course species’ behaviour and habitat preferences today are not necessarily an accurate guide to their behaviour in the past, especially where past environments have no precise modern analogue (Sturdy, et al. 1997, 587-8). Some dimensions of behaviour are certainly more predictable than others, and thus provide a more secure basis for extrapolation back onto past palaeoenvironments, particularly those regarding feeding behaviours. Animal species’ behaviour and habitats relevant to their ‘placing’ in the ecosystem; reviews of these and other species’ behavioural ecology given by, for example, Kurtén 1968; Jochim 1976; Winterhalder & Smith 1981; Clark 1983; Boyle 1990; Mithen 1990; MacDonald & Barrett 1993; West 1997, and allow the construction of a comprehensive picture of the characteristics of the major species represented at faunal sites in the region.


Generating pathways


These habitat preferences were used to associate animals with the topographic/vegetational categories represented in the timeslices illustrated.

The next step was to identify potential paths of movement between the areas of species’ preferred habitats and the particular sites at which their remains were recovered.


Such pathways are generated automatically in GIS by calculating the cumulative ‘cost’ of moving between two points. Calculation of such costs requires the specification of a cost surface, in which each individual ‘cell’ of information is associated with a number representing the ‘cost’ of traversing it, differing from previous methods of geographical analysis such as catchment analysis (van Leusen 1999, 216), which assumed that geographical space was ‘flat’ and homogeneous. However, definitions of ‘cost’ inevitably vary widely (this is as true of GIS computer systems as it is of Processual Archaeology), as do the parameters and algorithms used to calculate the cost of movement through a landscape (see ibid.: 216-7 for review).


Algorithms can be both isotropic (the same in all directions) or anisotropic (where the cost of movement may differ with direction – e.g. swimming upstream rather than down); the cost of travel obviously combines components of both: ‘the former exemplified by costs relating to the type of terrain (soil, vegetation, wetness), the latter by costs relating to slope and streams’ (ibid. 217). However, there are certain advantages to using isotropic calculations of cost in this analysis. When traversing particularly rather rugged terrain, for example, descent is often as tiring – if not more so – than ascent (see e.g. Susta et al. 2000; Llobera 2000, fig 2; Wheatley & Gillings 2002, 156 fig 7.4.). In addition, while the faunal remains recovered from sites clearly travelled there from the species’ preferred habitats, it is less certain that hunters travelled to these hunting grounds from that particular cave site; in the absence of evidence regarding the direction of travel, it seems more prudent to use isotropic methods.


However, there are a considerable number of ways to calculate even isotropic cost surfaces. Most studies have taken degree of slope as the most significant factor for calculating the cost of movement, although there are now several examples of more complex calculations based on physiological measurements of actual energy expenditure, for example that of Gorenflo & Gale (1990; cited van Leusen 1999), who specify the effect of slope on travelling speed by foot as:

V = 6e –3.5 | s + 0.05 |

Where V = walking speed in km/h, s = slope of terrain (calculated as vertical change divided by horizontal change, and e = the base for natural logarithms).


A much simpler alternative is provided by Diez (cited van Leusen 1999, 217), who recommends:

Effort = (percent slope) / 10


This has several advantages, particularly its very simplicity; more complex calculations tend to produce costs related to actual physiological expenditure and derived from modern human observations: even if we could assume that Palaeolithic ‘modern’ humans had identical metabolic systems to our own (despite evidence that they may well have been considerably more ‘robust’ than ourselves; e.g. Klein 1999), we certainly cannot assume this of Neanderthal populations. In addition, most people do not generally base their daily activities on precise calculations of the likely expenditure of energy and this does not seem a sound basis for exploring the ways in which their movements reflect their interactions with other aspects of the ecosystem. And lastly, given the coarse temporal scale of the study and the inevitably high level number of unknown variables and assumptions involved, any attempt to calculate actual physical costs of the movement of individuals or populations in the Palaeolithic would give a spurious accuracy to the results that would simply not be supported by the data itself – more complex calculations of the cost of movement can always be used in any subsequent, more detailed analyses.


However, as Bell and Lock have pointed out, the cost of climbing a slope is by no means directly proportional to the degree of slope:


thus surmounting a 45 degree slope is not simply 45 times as difficult as moving on the level, a 0 degree slope. Taking this to its logical conclusion would suggest that climbing a vertical slope of 90 degrees is ninety times as difficult as walking on the level, a 0 degree slope, an absurd simplification which would not stand up to scrutiny (2000, 88).


Instead, they suggest that by taking the tangent of the slope angle and then dividing the result by the base cost of traversing entirely flat ground (1 degree, to avoid a division by 0), the relative cost of movement across a landscape, rather than any absolute cost, can be established. Using this equation, the relative cost of climbing a 60 degree slope is not 60 but 100 ‘units’.


Diez’s equation was therefore modified slightly, and the equation used to generate the cost surface from the original ‘slope’ layer was:


Effort = tan slope / tan 1


A cost surface was thus generated from the slope basemap[17] in which the above equation was applied to each cell to give a value in (deliberately) vaguely termed units of ‘effort’ (see e.g. Wheatley & Gillings 2002, 152) representing the cost of traversing that particular ‘cell’ of landscape.


However, other factors than slope of course play a role in the ‘cost’ (however defined) of movement through a landscape, including barriers, transportation routes and the effects of differential terrain types (flat grassland, for example, presents a very different experience in terms of bodily movement to the dense undergrowth of mature woodland):


The vast majority of archaeological applications have so far accepted the simplification that energy expended or time taken to move around in a landscape is a function of slope … this is a worrying oversimplification (Wheatley & Gillings 2002, 155),


and further modification of this equation was necessary to account for this. We have no real way of knowing if there were any cultural ‘no-go’ zones in the area during the Palaeolithic, or even any territorial boundaries that would have influenced movement: really the only significant factors here are the potential relative costs of crossing rivers (as river transport is unlikely to be an issue in the Palaeolithic) and of moving through different forms of vegetation.


The cost of traversing rivers may be considerable and was addressed by the addition of an extra variable t into the equation. A high cost (200, arbitrarily chosen to be greater than the highest base cost derived solely from slope) was added to cells in the timeslice maps categorised as  ‘sea’ or ‘snow’ (altitudes below sea level and above the permanent snowline for the timeslice in question), and for cells categorised ‘river’, t = 50 (also chosen arbitrarily relative to the ‘cost’ of sea/snow).


Of further concern is the effects of changing (whether seasonally or climatically) vegetation on the cost of movement – my own visits to the region have highlighted the fact that walking in summer when undergrowth is thick consumes much more time and energy than that in autumn and winter when it has died back[18].


Much of the seasonal variation should be negated by the generation of ‘summer’ and ‘winter’ variants of the reclassified timeslice maps, and as the pathways will be generated within specific timeslices, climatic variation should also not impact negatively on the analysis. However, assessments of the differing relative ‘friction’ of different types of vegetation on movement have been suggested by, for example, Glass et al. (1999) and are integrated into this analysis by the addition of a further variable v. The ‘base’ cost of traversing terrain – represented by the tangent of the slope gradient - is multiplied by this value: where vegetation is moderately difficult to move through (e.g. grassland with stands of trees), v = 1.5 – hence it is considered half as ‘costly’ again to traverse than other terrain of a similar gradient. Where vegetation is more difficult to travel through (open woodland and parkland environments), v = 2 (the cost of movement is doubled). And in areas of denser vegetation (e.g. dense woodland), v = 2.5.[19]. Travelling across open grassland, steppe and bare rock, it is assumed, incurred no significant extra cost above the ‘base’ terrain cost derived from the slope gradient.


The final equation used to derive the ‘costs’ of movement in each timeslice, therefore, was:

Effort = ((tan s/tan 1) v) + t


Where s = slope, v = vegetation and t = terrain.


The map layer with the base costs derived from slope tangents was thus amalgamated with the timeslice vegetation map[20] and values re-calculated to reflect the addition of the terrain value t and vegetation value v[21] as described above.


The result was a raster map in which each cell was associated with a cumulative ‘cost’ of traversing across it from the specified starting point of a particular cave site[22]: i.e. the landscape in the immediate vicinity of the specified cave site has a low cumulative cost because it requires less effort to reach from the starting point of the cave than points at a distance. This timeslice-specific cost surface was then used to derive a least-cost pathway (defined by a sequence of cells of lowest cumulative ‘cost’) of potential movement between these two points using the GRASS module r.drain. This module is designed to model the run-off patterns of water, and traces a path from the higher cost areas of a user-defined starting point (within an area of habitat associated with a particular species) to the ‘low’ cost of the cave site from which the cost surface was generated (and where that species’ bones are represented).


In order to provide some form of temporal framework for consideration of the resulting paths, the area that could have been traversed in a single day was calculated. Rather than consider merely distance in such a calculation, a timed (2 hour) walk over varied terrain was undertaken[23] and the ‘cost’ of the walk then computed in GRASS to take the terrain and vegetation factors considered above into account. From this it was calculated that an average hour’s walk represents 1673.5 units of ‘effort’. In a (generous) 15 hour (mid)summer day[24], then, areas lying beyond 12551 units of ‘effort’[25] from the site under consideration could probably not have been reached as part of a return day’s travel. For a (mid)winter day (estimated at 8.5 hours of daylight; Butzer 1986), the corresponding figure was 7112 units[26]. The module r.reclass was used to generate a ‘limit’ to the day’s activity.


Of course, these are rather coarse estimates of the area that could have been walked in a day by the Pleistocene inhabitants of the region. Both Pleistocene populations were rather robust and probably highly adapted to moving fast over difficult terrain (Trinkaus 1995) and therefore modern analogues are unlikely to underestimate potential distances. In addition, the area reflects an uninterrupted walk between two known points, which may not represent a good analogue for hunting and/or gathering parties’ movements, which are likely to have been more meandering (at least on the outward portion) as people searched for game, vegetable foods and/or other resources such as flint and paused at various places to pursue/stalk, kill and butcher the animals they successfully killed, check traps or snares, pick/dig up vegetable foods and/or extract raw materials. Nor did parties necessarily return to the same site they left from, perhaps travelling instead to the nearest cave site available. The limits to days’ walks provided in the following analysis, therefore, are given only as a guide and serve solely to put the potential paths generated into general temporal context.





Adams, J. & Fauré, H. 1998. Europe during the last 150,000 years. vol. 2002. Quaternary Environments Network.


Altuna, J.1972. Fauna de Mamíferos de los Yacimientos Prehistóricos de Guipúzcoa. Munibe 24: pp. 1-464.


Altuna, J. & Merino, J. M. 1984. El Yacimiento prehistorico de la Cueva de Ekain (Deba, Guipuzcoa). San Sebastian: Sociedad de Estudios Vascos con la colaboración del Socieded de Ciencias Aranzadi.


Bailey, G. 1983. Economic Change in Late Pleistocene Cantabria. In G. Bailey (ed), Hunter-Gatherer Economy in Prehistory, pp. 150-65. New Directions in Archaeology. Cambridge: Cambridge University Press.


Bell, T. & Lock, G. 2000. Topographic and cultural influences on walking the Ridgeway in later prehistoric times. In G. Lock (ed), Beyond the Map: archaeology and spatial technologies, pp. 85-99. NATO Science Series A: Life Sciences. vol. 321. Oxford: IOS Press.


Bibby, J. S. & Mackney, D. 1977. Land Use Capability Definition. Technical Monograph No. 1: 1-12. Harpenden: Soil Survery of England and Wales.


Boyle, K. V. 1990. Upper Palaeolithic Faunas from South-West France: a zoogeographical perspective. B.A.R. International Series 557. Oxford: British Archaeological Reports.


Buol, S. W., Hole, F. D. and McCracken, R. J. 1973. Soil Genesis and Classification. Ames: Iowa State University Press.


Butzer, K. W. 1981. Cave Sediments, Upper Pleistocene Stratigraphy and Mousterian Facies in Cantabrian Spain. Journal of Archaeological Science 8: pp. 133-83.


- 1986. Paleolithic Adaptations and Settlement in Cantabrian Spain. Advances in World Archaeology 5: pp. 201-52.


Clark, G. A. 1983. Boreal Phase settlement/subsistence models for Cantabrian Spain. In G. Bailey (ed), Hunter-Gatherer Economy in Prehistory: a European perspective, pp. 96-110. Cambridge: Cambridge University Press.


d'Errico, F. & Goñi, M. F. S. 2003. Neanderthal extinction and the millennial scale climatic variability of OIS 3. Quaternary Science Reviews 22: pp. 769-88.


Galan, C. 1988. Zonas karsticas de Guipúzcoa: los grandes sistemas subterraneos. Munibe (Ciencias Naturales) 40: pp. 73-89.


Glass, C., Steele, J. & Wheatley, D. 1999. Modelling Human Range Expansion Across a Heterogeneous Cost Surface. In L. Dingwall, S. Exon, V. Gaffney, S. Laflin and M. v. Leusen (eds), Archaeology in the Age of the Internet: CAA 97, Computer Applications and Quantitative Methods in Archaeology, Proceedings of the 25th Anniversay Conference, University of Birmingham, April 1997, pp. 67-72. B.A.R. International Series. vol. 750. Oxford: British Archaeological Reports.


Hammond, E. H. 1964. Analysis of properties in land form geography, an application to broad-scale land form mapping. Annals, Association of American Geographers 54: pp. 11-9.


van Hove, D. 2003. Imagining Calabria: a GIS approach to Neolithic landscapes, Unpublished PhD Thesis, University of Southampton.


Huntley, B., Alfano, M. J., Allen, J. R. M., Pollard, D., Tzedakis, P. C., Beaulieu, J.-L. d., Gruger, E. & Watts, B. 2003. European Vegetation during Marine Oxygen Isotope Stage-3. Quaternary Research 59: pp. 195-212.


Jochim, M. A. 1976. Hunter-Gatherer Subsistence and Settlement: a predictive model. Studies in Archaeology. New York: Academic Press.


Klein, R. G. 1999. The Human Career: human biological and cultural origins. Chicago: University of Chicago Press.


Kopp, K.-O. 1965. Límite de la nieve perpetua y clima de la época glaciar Wurmiense en la Sierra de Aralar (Guipúzcoa, Navarra). Munibe 17: pp. 3-20.


Kukla, G.J., Bender, M.L., de Beaulieu, J.-L., Bond, G., Broecker, W.S., Cleveringa, P., Gavin, J. E., Herbert, T. D., Imbrie, J., Jouzel, J., Keigwin, L. D., Knudsen, K.-L., McManus, J. F., Merkt, J., Muhs, D. R., Muller, H., Poore, R. Z., Porter, S. C., Seret, G., Shackleton, N. J., Turner, C., Tzedakis, P. C. & Winograd, I. J., 2002. Last Interglacial Climates. Quaternary Research 58: pp. 2-13.


Kurtén, B. 1968. Pleistocene Mammals of Europe. London: Weidenfeld and Nicolson.


Lambeck, K., Esat, T. M. and Potter, E.-K. 2002. Links Between Climate and Sea Levels for the Past Three Million Years. Nature 419(12 September): pp. 199-206.


van Leusen, M. 1999. Viewshed and Cost Surface Analysis Using GIS (Cartographic Modelling in a Cell-Based GIS II). In J. A. Barcelo, I. Briz and A. Vila (eds), New Techniques For Old Times: computer applications and quantitative methods in archaeology. Proceedings of the 26th Conference of the CAA, Barcelona, March 1998, pp. 215-23. International Series. vol. 757. Oxford: British Archaeological Reports.


Llobera, M. 2000. Understanding movement: a pilot towards the sociology of movement. In G. Lock (ed), Beyond the Map: archaeology and spatial technologies, pp. 65-84. NATO Science Series A: Life Sciences. vol. 321. Oxford: IOS Press.


MacDonald, D. & Barrett, P. 1993. Collins Field Guide: Mammals of Britain and Europe. London: HarperCollins.


Mellars, P. 1996. The Neanderthal Legacy. Princeton: Princeton University Press.


Mithen, S. 1990. Thoughtful Foragers: a study of prehistoric decision making. New Studies in Archaeology. Cambridge: Cambridge University Press.


Muñier, B., Nygaard, B., Ejrnæs, R. & Bruun, H. G. 2001. A biotope landscape model for prediction of semi-natural vegetation in Denmark. Ecological Modelling 139: pp. 221-33.


Straus, L. G. 1992. Iberia Before the Iberians: The Stone Age prehistory of Cantabrian Spain. Albuquerque: University of New Mexico Press.


Sturdy, D., Webley, D. & Bailey, G. 1997. The Palaeolithic Geography of Epirus. In G. Bailey (ed), Klithi: Palaeolithic Settlement and Quaternary Landscapes in Northwest Greece, pp. 587-614. McDonald Institute Monographs. vol. Volume 2: Klithi in its local and regional setting. Cambridge: McDonald Institute for Archaeological Research.


Susta, D., Minetti, A. E., Moia, C. & Ferretti, G. 2000. Energy Costs of Walking and Running on Extreme Uphill and Downhill Slopes. Journal of Physiology 523: pp. 227.


Trinkaus, E. 1995. Neanderthal Mortality Patterns. Journal of Archaeological Science 22: pp. 121-42.


van Andel, T. & Tzedakis, P. C. 1998. Priority and opportunity: reconstructing the European Middle Palaeolithic climate and landscape. In J. Bayley (ed), Science in Archaeology: an agenda for the future, pp. 37-45. London: English Heritage.


Vogt, J. V., Colombo, R. & Bertolo, F. 2003. Deriving drainage networks and catchment boundaries: a new methodology combining digital elevation data and environmental characteristics. Geomorphology 53: pp. 281-98.


West, D. 1997. Hunting Strategies in Central Europe During the Last Glacial Maximum. BAR International Series 672. Oxford: British Archaeological Reports.


Wheatley, D. & Gillings, M. 2002. Spatial Technology and Archaeology: the archaeological applications of GIS. London: Taylor and Francis.


Wild, A. 1993. Soils and the Environment: an introduction. Cambridge: Cambridge University Press.


Winterhalder, B. & Smith, E. A. (eds) 1981. Hunter-Gatherer Foraging Strategies: Ethnographic and archaeological analyses. Chicago: University of Chicago Press.


Zagwijn, W. H. 1990. Vegetation and climate during warmer intervals in the late Pleistocene of Western and central Europe. Quaternary International 3/4: 57-67.




19 http://www. B5m.gipuzkoa/liz5000

[2] UTM (zone 30) projection (ellipsoid: international, datum: ED50)

[3] using the module

[4] using the module

[5] using the module v.digit

[6] from the Instituto Geologico y Minero de España’s (IGME) 1:25,000 Mapa Topográfico Nacional de España sheets 63-I (Ondarroa) and 63-II (Eibar)

[7] using the GRASS module v.patch

[8] using the module, which interpolates values between vector contour lines using the ‘regularised spline with tension’ algorithm (RST). Resolution 20x20 (maintained from source data).

[9] Using the module r.mask

[10] The interpolation process of transformation from vector to raster does not ‘stop’ at the edges of the map and thus produces an effect whereby raster data extends into adjacent areas of ‘null’ values.


[12] using the command

[13] N, NE and NW: 45-135° - aspects are calculated in GRASS in degrees counter-clockwise from East

[14] S, SE and SW (203-248°); West-facing (136-224º) and East-facing (316-44º).

[15] This creates a new raster map layer showing buffer zones around any non-NULL category cells in an existing map layer. As the input map needs to be composed solely of cells with values of 1 and 0, the coastline of each timeslice was extracted as a separate thinned raster file before r.buffer was performed.

[16] Several hundred miles from latitude 50°N to above 60°N

[17] Using the GRASS module r.mapcalc

[18] Interestingly, this also has significant effects on the visibility of sites.

[19] In the study region, such vegetation occurred only on river valley floodplains, which are in and of themselves rather difficult to traverse, at least in temperate climates (Chambers and Hosfield pers comm.)

[20] Using the GRASS module r.cross

[21] Using the GRASS module r.reclass

[22] using r.cost, the ‘Knight’s move’ option

[23] from Ekain to the summit of Erlo; see also Altuna & Merino 1984.


[25] 1673.5 x 15 = 25102.5/2 = 12551.25

[26] 1673.5 x 8.5 = 14224.75/2 = 7112.38