Developing a GIS approach to routes

All of the above GIS techniques were designed to address spatial questions that are relevant to the study of routes and interaction along them. ‘Network analyses’

involve the explicit definition of relationships between nodes (sites), so they must be firmly linked into particular temporal contexts in which both nodes (sites) and linkages (routes) are already known; only then can the ‘connectivity’ and accumulation of movement be simulated. In contrast, ‘cost-surface analyses’

start from a generic geographic perspective (although the landscape model is normally static and so taking account of environment change is different), which is independent of time-period. Subsequently, the accumulation of potential movement can be simulated by performing calculations based on the geographical location(s) of archaeological phenomena/sites. Both sets of analysis may allow the development of complex visualizations of interaction networks, but neither of them offers a straightforward way to ‘reconstruct’ past route networks that look like the traditional ‘road-map’. ‘Least-cost-path’ analyses, which are often seen as the ultimate aim of cost-surface procedures in archaeological application, do indeed create linear ‘road-like’ maps, but are very problematic as a predictive tool. It is clear from various studies that a path generated by ‘least-cost-path’ is fairly arbitrary (Gietl, Doneus and Fera 2008) – it represents one hypothetical path amongst many. This suggests that, at least at the small-scale, ‘least-cost-path’

should normally only be used for comparison with pre-known paths rather than for reconstructing or predicting unidentified ones.

3.4.1 ‘Cost-of-passage’, ‘cost-surface’ and ‘cost-corridor’ approaches Instead, less-restrictive forms of cost-surface techniques may be the most productive way to access the potential of the landscape in order to create a model of ‘natural routes’. Thus ‘cost-corridor’ may be seen as a more flexible and realistic version of ‘least-cost-path’ that gives less arbitrary results because it ‘brackets’ an area rather than suggesting a single path – this certainly fits better with the definition of a route as corridor (see Section 3.2); similarly resistivity models based on the flow of electrons across a field of variable conductance may offer more realistic cumulative maps (for an application from within animal ecology, see Shah and McRae 2008). But there are various potential objections to the application of

‘cost-surface’ to human behaviour. One of the most serious is the suggestion that

the underlying model of movement is defective: i.e. that the idea of cost and its assumption of economic rationality produces unrealistic results and cannot make useful predictions. As the anthropological literature on ritual can attest, human agents are interested in many different factors beyond resource-maximization, and paths wind around and past cultural and natural features. Even better walking models (which take anisotropic costs into account or can create more realistic

‘spiral’ paths up steep hills) remain simplifications of real human motivations (and leave out the other modes of travel and transportation).

There are a number of ways to pre-empt this kind of critique. First (1) I would argue that it is disingenuous to suggest that travel ‘cost’ did not play a major role in the production of the ancient world and the remains we find: it may be more comfortable for an anthropologically-informed archaeologist to use the word

‘friction’ in place of ‘cost’ to disassociate the procedure from monetary economics – though as the Old Assyrian karum records attest, monetary and time costs of routes were themselves often of great interest to ancient peoples. Actually, the more serious problem with cost/friction models is to do with the restrictions of the algorithms, which require the user to condense many different factors into a single quantitative value. Real decision-making by individuals (ancient and modern) is much more adaptive, allowing different factors to come to the fore at different times (as we note from the routes taken by armies – for examples, see Section 2.4.2). To some extent this can be overcome by using the more advanced

‘anisotropic models’ – although the introduction of greater complexity increases the difficulty of creating and comparing models.

Second then, (2) scale should also be considered. Most applications of cost-surface techniques have made use of environmental data, normally topography or resource distribution, to study movement or paths on the local scale. The resultant analyses and interpretations often seem abstract and removed from the everyday life of ancient people. Less attention, on the other hand, has been paid to the use of cost-surface techniques on a macro- or supra-regional scale. By analogy to Braudel’s argument that different sets of evidence are appropriate sources to study historical processes for particular temporal scales (Braudel 1996), we may likewise suggest that the larger the geographic and temporal scale of analysis, the more likely that environmental factors (which are the ones most easily modelled with GIS) will determine the structure of human movement. Such GIS approaches are actually far better suited to study macro-scale of inter-site, inter-region and inter-cultural movements than they are to the procurement of resources within a local field because at the macro-scale, localized deviations from straight-line paths (everyday movements along fields, between villages etc. – the pathways described in Section 3.2.1) will become slowly smoothed out into more general aspects of the landscape (slope, climate). The condensed single variable of ‘cost’ is thereby less problematic because, at the large-scale, adaptive decisions are subsumed by large-scale processes. Conversely, by providing a baseline model of the landscape, the particular cultural dynamics and route systems of the period can perhaps be brought more starkly into relief.

Finally, (3) we need to dislocate GIS techniques from their traditional association with positivist and hypothesis-testing epistemologies. In a positivist-processualist approach characteristic of much GIS analysis, the analysis of ancient route systems would rely on combining the above factors into a single model or set of rules, which could be programmed into the GIS. The model would then be ‘tested’

against the evidence to measure its ‘goodness of fit’. There are a number of reasons why such an approach is unhelpful, some of which will be briefly summarized below, and not only because of objections derived from phenomenology that mathematical cost-surface analysis itself might be irrelevant to human societies because of its emphasis on non-human values (cf. Ingold 2000). Instead we should treat cost-surface and other techniques as tools for visualizing archaeological problems – designed to make historical problems easier to think through. Explicit models allow us to simultaneously foreground our already active but frequently unstated assumptions but apply them rigorously to the data. What this means in practice is that the ‘models’ created are not abstract ones in the Popper-ian or mathematical sense, but rather visual ones akin to conventionalized drawings of objects or the scale models of ancient architecture or objects. While ultimately based on empirical foundations, these are qualitatively orientated documents that are designed to facilitate understanding or interpretation of the objects in question (in this case landscapes) rather than be manipulated statistically. In my view, rich visualization of interaction across landscapes is far more important to archaeological narratives than the reconstruction of any ‘true’ road, which is in any case an impossible task especially when the concept of route is so slippery in the first place. This book thus adopts a non-positivist, or rather ‘soft’ approach to the use of the computer modelling.

The following sections will explore how various factors that contribute to the cost of movement across the landscape can be incorporated into a model of

‘cost-of-passage’.

3.4.2 Theory: natural and anthropogenic cost factors

i. Topography – slope

Topography has been the basis for most previous cost-surface analyses, and it is obvious that the shape of the landscape is very important to the ability of armies, traders and migrants to travel. This can be easily gleaned from the weight of importance given to mountain passes in descriptions of military campaigns, real or imagined, from Alexander the Great to the British Navy at the turn of the 20th century³⁰. Additionally the topography itself and the environmental conditions such as soil and climate that are formed as a result, may structure the pattern of settlement, in turn structuring movement.

Where bulky, heavy or large volumes of goods are involved, mountainous areas and steep valleys will be avoided at all costs in favour of river valleys and water transport. On the other hand, on certain scales, goods may have high enough value across mountain ranges to be worth transporting in this way. Recent work by Irene Good in the Pamirs of Tajikistan (Good 2006) suggests that mountain valleys may often act as conduits of exchange, an idea which has also been put forward for the Alps. In any case, the slope limits the intensity of movement by reducing the number and ‘size’ of exchanges possible. If looking at particular kinds of exchange, it might be helpful to categorize slopes into qualitative values (‘too steep’, ‘steep’,

‘not-steep’), as has been done for some previous analyses. For example, for tracing or predicting Roman roads, whose routes tend to reflect major highways (and thus

30 As shown in the military ‘Handbook’ to Asia Minor and Mesopotamia researched and published by the British Naval Intelligence Dept. – see Section 2.3.2.

bulk transport routes), some slopes are simply too steep to make sense, and these areas may be discarded in the predictions. On the other hand, if a more general pattern of movement is under analysis, small-scale exchanges between villages and small communities are also important to cultural interaction: thus a fuller degree of nuance can be achieved by using a continuous slope value.

Slope is an obvious measure by which a cost-surface can be obtained, since steep slopes significantly increase the cost of travel. As mentioned earlier, however, its relationship to cost is not straightforward. On a small scale, angle/percentage of the slope cannot be used as a direct value in cost. Instead a model of cost/slope must be applied taking into account relative difficulty of upward and downward slopes.

This may differ according to transportation technology, but most archaeological models currently depend on laboratory measurements of energy expenditure or time of walking. Whether this problem also applies to the macro-scale has not yet been theoretically or experimentally investigated. It is possible that at this macro-scale, the differences between transportation types are blurred, and if two-way equal exchange is dominant between all regions, a fairly straightforward equation might be used. Macro-economic theories, especially world-systems, emphasize the inequality of exchange between regions, however. This could militate against such a straightforward equation if inequality is taken to be normal.

ii. Topography – elevation, inequality and directionality

Another related aspect of the topography is relative elevation. Even if the value of exchanges were to be taken as equal, the distribution of resources and their relative bulk or transportability is not. For example, the sources of metals and precious stones may often tend to be found in higher altitude areas, precisely because they are created by the tectonic activity that also creates mountain ranges (such as the Taurus-Zagros, or Pamirs-Himalayas). Transferring bulk goods downhill is significantly easier than transporting them uphill, as is especially the case where water may involved, such as along the river Euphrates³¹ – though this is very difficult to model practically. Such directionality in exchange is also important where sea-routes are considered: the rotational currents and winds of the Eastern Mediterranean have long been an important consideration for the study of ancient trade, but it is difficult to see how we could build a comparative model which would include directional aspects of both sea and land together.

iii. Seasonality

Seasonality of travel is a frequently ignored aspect of ancient routes. Archaeologists who tend to travel to their study regions only during favourable summer conditions (normally using motorized transport) rarely see the same landscape under snow, and it is easy to forget the extent to which modern transport and air conditioning technologies considerably lessen the discomfort even in the summer. In the past, regions under considerable amounts of snow may have been relatively or completely impassable for long periods in the year. If data are available, it is relatively easy

31 One example of such ‘directionality’ in transportation along a river comes from the late 19th century AD: “The Tigris is much utilized by large rafts from Diarbekr [sic] downwards to Baghdad. These consist of about 150 inflated sheepskins tied underneath a light wood framework, and are utilized to carry merchandise or travellers. These rafts cannot ascend the river, but are broken up on arrival in Baghdad and the skins taken back by caravan.”(Maunsell 1894, 6).

to impose a modelling ‘handicap’ on a region on the basis of the number of days of snow cover (or conversely, favour regions which have relatively little), by increasing the cost-calculations appropriately. The only problem here is the extent to which modern data on snow cover can be used to represent past environments:

ideally, one would want to take into account periods of climatological cooling and warming – but these data are not easily available over a large geographical area in sufficient detail.

iv. Climate – humidity/aridity and extreme temperatures

Little research has been done into the extent to which general climate, but particularly humidity/aridity in combination with extreme temperatures, can affect travel ability. Anecdotal evidence from regions such as Cilicia shows that high humidity and heat can significantly increase discomfort for travel, though to some extent it may be a matter of acclimatization³². Similarly the extreme heat of Taklamakan summers encouraged the tradition of winter caravans across the desert (Potts 2004a, 147). For those travellers engaged in particularly long distance exchange, movement between climatic zones might perhaps represent a cost factor, but it is difficult to place a measure on this. As with other factors, the problem remains a lack of experimental data upon which suggested ratios of ‘cost’

might incorporate the factor into an equation.

v. Rivers

A related aspect of seasonality is rivers, and their ability to be forded or navigated depending on seasonal variables: this is particularly the case for glacial and rain-fed rivers (though with contrary results) whose levels vary depending on the season.

In general navigable rivers facilitate parallel trade, but hinder perpendicular trade.

That is to say that it is often easy to travel along the river, on waterborne transport along larger, stable, rivers, or alongside on towpaths, but finding appropriate crossing points may often be difficult, even for smaller streams, and in some cases fraught with danger (flash floods, drowning, and goods/provisions/clothing becoming wet/ruined)³³. The same is true for crossing irrigation streams: huge detours must sometimes be taken simply to be able to cross safely with vehicles or animals. Rivers also have directionality, related to the slope of the topography, in that downstream is often (though not always) less costly than upstream.

Rivers have been sometimes been included in cost-models by assigning the route of the river a markedly lower cost value than the surrounding land. The problem is how to include the flow direction of the river: whilst theoretically possible to include in an anisotropic model, it would be rather complex. Only certain parts of certain rivers are navigable by boat. As a basic means of modelling the difficulty of crossing the river., an additional parallel ‘buffer’ of ‘high cost’

can be added. Where rivers are navigable or fordable, it is usually from particular points, i.e. ports, fords and bridges, and not from the entire length of the river.

32 An easy way to experience this shift in humidity or temperatures is to travel from the Konya plain across the Taurus into the Göksu valley, or from the Lake Van region into Batman province.

33 E. Newby’s A Short Walk in the Hindu Kush illustrates the problem of crossing rivers with pack animals in his description of his travels in Nuristan, Afghanistan (Newby 1974, 234-236). Other nineteenth and early twentieth century travel accounts offer similar though less colourful descriptions of the time taken to locate suitable fording or bridging points.

vi. Water sources

Access to clean drinking water sources is a particularly complex problem. The main difficulty with trying to include this factor in route models is in identifying the density of water sources across a landscape: first these have rarely been systematically documented even in modern topographic maps (and maps that might show this are generally harder to obtain); and second there is a strong chance that springs may have changed positions through time depending on water levels, or irrigation/water management schemes. In the hills of the Caucasus and Anatolia, it is rare to be completely without nearby water sources, but, depending on routes taken, the time between sources can prove problematic. By contrast, in the deserts of Turkmenistan, a much better specific local knowledge is required to identify springs or drinkable water – although fountain monuments of the Islamic era have recently been identified along a path between Merv and the Oxus (Wordsworth 2010). Being able to predict availability of clean drinking water has a profound effect on the ability of armies, traders and migrants to move effectively, even if other means of purification (the addition of alcohol, boiling teas) might overcome some of the dangers. Lee Ullman has argued that Hittite monuments are located precisely to take account of the travelling Hittite army’s need to camp and replenish near water sources, both bodily and spiritually (Ullmann 2011).

Modern springs in Turkey are usually made highly visible by surrounding white-washed brick, stone or concrete constructions, and a water trough for animals – often next to mosques. The common factor in these examples is that of water monuments – which are not identifiable in all periods.

Modelling the effect of water sources on travel/exchange routes would be possible if one could establish a comprehensive database of their distributions (converted to a density raster) or else by using proxy indicators such as rainfall or maps of groundwater distribution. The increased ‘cost’ would be calculated as a factor of the low density of water sources, since greater effort might be spent finding water in ‘low density’ regions.

vii. Sea

In coastal regions, sea transport can facilitate bulk movement of goods in particular. As mentioned before, sea-transport often has directionality, in that the sea currents and prevailing winds tend to promote movements across water in certain directions. In the case of the Eastern Mediterranean currents and winds favoured an anti-clockwise rotation for sail-based transport, for example. Areas that receive frequent storms and high seas will also resist movement across them.

The complicated factor is how one converts this qualitative knowledge into a model which can be directly compared to movement across land. Directionality

The complicated factor is how one converts this qualitative knowledge into a model which can be directly compared to movement across land. Directionality