«Terraced fields and Mediterranean landscape structure: an analytical case study from Antikythera, Greece Andrew Bevan and James Conolly Postprint of ...»
Terraced fields and Mediterranean landscape structure: an analytical case
study from Antikythera, Greece
Andrew Bevan and James Conolly
Postprint of a paper in Ecological Modelling 222: 1303–1314.
Terraces are important capital investments in a range of agricultural landscapes
worldwide, typically enduring well beyond any single farming cycle and over many
human generations. This paper begins by emphasising that, while human population
growth may often be loosely linked to terrace construction efforts, the association is by no means a straightforward one. We then argue that the choice of which parts of the landscape to terrace is driven by a range of cultural and environmental priorities that are most usefully explored by a combination of global, local and auto-correlative modelling, as well as via simulation-based methods. The results demonstrate that surficial geology, terrain slope, pre-existing terraces and pre-existing patterns of human habitation are all important structuring features. We also consider terraces as method for soil conservation, question the uncritical use of meso-scale erosion models and argue that patterns of catastrophic soil loss are often overstated in Mediterranean contexts. However, erosion modelling can, if deployed cautiously and comparatively, nonetheless be used to explore ways in which terraces do indeed manage localised soil movement in agriculturally- favoured parts of the landscape, with our results suggesting that a substantial proportion of the erosion in this regions is indeed ameliorated by such measures.
1 Introduction Terraces are highly flexible forms of agricultural construction used by farmers in tropical to semi-arid environments to convert hillslopes into stepped linear units of relatively flat ground that are suitable for cultivation. Although they share the common function of increasing available arable land, they vary considerably in shape, complexity and degree of labour investment, as well as in the environmental setting in which they are deployed.
Their ubiquity, variability, and importance for agricultural productivity have motivated a range of studies designed to understand them better both as cultural and ecological features (e.g., Donkin 1979; Evans 1990; Lansing and Kremer 1993; Dunning and Beach 1994; Price and Nixon 2005, Rodríqez 2006, among many others). Terracing processes are often thought to involve increased exploitation of marginal landscapes by expanding local populations and/or by new colonists (e.g., Coote and Whitelam 1987: 129; Donkin 1979: 33, 133; Kunen 2001), although it is far from clear whether this is the motivating factor in all situations in which terraces are found.
Terraced agricultural fields are a defining feature of Mediterranean landscapes and their antiquity and socioecology has been the subject of considerable interest (e.g. in the Aegean, Whitelaw and French 1999; Frederick and Krahtopoulou 2000; Grove and Rackham 2001; Price and Nixon 2005). However, the socioecological contexts under which such built features emerge and persist remains poorly understood, despite their obvious relevance to modern concerns about sustainable land use, erosion control, water management and food production. This paper seeks to address this problem by exploiting an exhaustive mapping of terraces across the Greek island of Antikythera. There are three objectives. First, we aim briefly to address the demographic context in which terraces emerge and demonstrate, at least for the more recent history of the island’s occupation, that there is no immediate link between population growth and terrace expansion.
Second, we develop a spatial model to explore the aggregate human preferences behind the landscapes chosen for terraced agriculture in order to formalise and develop further our understanding of their locational characteristics. Finally, we examine the role that terraces play in soil conservation via two kinds of meso-scale erosion model. In so doing, we also revisit some of the key methodological challenges associated with quantitative assessment of built structures at the landscape scale.
2 Research Context In the Mediterranean, up until the recent and more haphazard use of bulldozers, terrace construction has typically involved the use of stacked and/or interlocked dry stone risers that act as the walls or retainers to support beds of level soil (‘treads’) of varying depth for a wide assortment of cereal crops, vegetables, vines and fruit trees. In addition, although explicit control of soil movement has not always been the primary concern of the farmers who construct dry-stone wall terraces, nevertheless such features do reduce the effects of hillslope erosion. For these reasons, terraces and other dry-stone wall features, such as enclosures and houses, are in fact an important form of niche construction (Odling-Smee et al. 2003) that provide significant adaptive benefits for the constructors as well as for later inhabitants. Such investment therefore has both anticipated and unanticipated uses beyond the current farming cycle and/or over many human generations, making them a well-regarded form of landesque capital (Blaikie and Brookfield 1987: 9-10; Widgren 2007).
Figure 1. (a) A map of Antikythera (as a hillshade surface) and its extant systems of agricultural fields, (b) a set of hillslope and cross-channel terraces on the north side of the island, (c) an aerial photograph of terrace systems in the centre of the island (see fig.
1a inset box), taken in February 1944 just before the forced removal of the island’s inhabitants to Crete (courtesy of the Aerial Reconnaissance Archive).
Our focus below is on the development of appropriate modelling methods to understand the environmental niches in which terraced agriculture is most common, and to consider what subsequent landscape impact terraces might have had, particularly with regard to erosion. We draw on recent landscape research on Antikythera―a small (20.8 sq.km, maximum altitude 379 m) and relatively remote Greek island, lying midway between Kythera and western Crete in the southwest Aegean (figure 1a). Antikythera’s terraced fields are not necessarily special or unusual by comparison to others in the Mediterranean, but as an analytical context, the island offers four important advantages.
First, it is possible to acquire a close to complete feature-by-feature sample of the relict terraces across Antikythera during the last phase of the island’s occupation. Second, we can also consider earlier possible terrace use in the context of a relatively comprehensive understanding of prior settlement patterns, due to the rare, if not unique, case of having conducted an intensive archaeological survey of the island’s entire extent (for the method, see Bevan et al. 2008; Bevan and Conolly 2009). Third, the island has experienced a dynamic history of human exploitation, cycling between phases of relatively substantial settlement for its size (in recent times, 10–45 people per km2, with what appears often to be comparable levels in the past) and periods of near complete abandonment. This episodic settlement history is analytically more convenient that the complex and continuous conditions that typically prevail in other regions and offers a useful set of simplifying conditions under which to consider terrace construction and particular phases of the island’s past.
Terraces are notoriously difficult to date, but slow progress has been made in assessing their antiquity in the Mediterranean and some systems of this kind clearly extend back to as early as the Bronze Age (e.g. Gibson 2001; Krahtopoulou and Frederick 2008). Whilst some terrace systems were undoubtedly constructed in antiquity (see also Price and Nixon 2005), there is little direct archaeological evidence to support (or refute) the relationship between historic population growth and terrace construction, largely because of the difficulty of directly dating terrace systems. Developing a better understanding of the social and ecological context of terraces, including their relationship to cycles of past population growth, is therefore an area requiring concerted effort especially if we are to make sensible decisions about the appropriateness of such strategies in the present-day.
Human and animal labour, in particular, is clearly a key variable in the formation of terraces, as their construction, often in agriculturally marginal and less accessible portions of the landscape, requires substantial and on-going investments in time and energy.
Figure 2. Demography and 19th century agricultural expansion: (a) population from ca.
1770 to present (the grey-shaded area is the time period covered by figure 2b), (b) proportion of the island under cultivation or one-year fallow (1826-1847, total island extent is 20.8 sq.km). Sources: UK National Archives, Colonial Office 136/1332-40, National Statistical Service of Greece.
On Antikythera, a range of evidence (circumstantial, stratigraphic, and chronometric), suggest pre-modern terracing at several different stages of the island’s history. However, of particular interest for our purposes here is the most recent phase of terracing on the island, beginning in the late 18th and lasting up until the mid 20th century AD. Census, taxation and agronomic records from the period 1826–47 are particularly useful in indicating a phase of rapid agricultural expansion during which the amount of declared land under crop rose from 7% to 43% of total island area and more marginal areas became increasingly exploited (figure 2b), a large proportion of which was demonstrably invested with enclosure walls and terraces. A key point to note is that agricultural expansion, at least at the sub-decadal and decadal scale, does not match shifts in human population on the island over this period (figure 2a). For example, a large spike of immigrants in 1824–5 was the result of refugees from the political uprisings in Crete to the south, but these do not lead to any immediate or delayed incorporation of new land into cultivation. Rather it is only in the 1830s and 40s, during a period of slight population decline, that we see the latter agricultural expansion occur. This pattern suggests that population pressure is not, in this case at least, the sole determining factor behind terrace building and associated agricultural expansion in the period, but confirms the impression also left by the historical sources that direct political stimulus and more favourable overall economic conditions were very important. In particular, the enclosure of and physical investment in new tracts of land was part of a deliberate strategy within the Ionian islands at this time, promoted both through formal agronomic guidance and tax exemptions for newly-cleared and recently-walled land (Jameson 1836: 274; Leontsinis 1987: 217-41).
With these demographic issues and their consequences in mind, our documentation of Antikythera’s terraces has involved on-the-ground mapping of terrace formations, satellite and air-photo mapping, and geoarchaeological investigations (e.g. standing sections, soil samples, radiocarbon and optically-stimulated luminescence dates). The detail from this program cannot be fully reported on here, and we thus concentrate on modelling of aggregate spatial patterns. In total, over 12,000 relict enclosure walls and terraces are clearly visible on the island and have been individually mapped as a vector line dataset (figure 1a), with the suggestion of more ephemeral traces in several further locations. Relict terraced fields are particularly visible over about 26% of Antikythera’s area, with at least 10–15% of island land far less obviously terraced and enclosed, but still occasionally used for agriculture.
3. Location Modelling Although commentators regularly note the association of terraces and agricultural fields with specific types of landscapes, it is surprising how few clear and formalised attempts have been made to assess the locational logics behind such systems. Here we consider methods for inferring spatially structured relationships between the location of humanbuilt features in the landscape, such as terraced agricultural fields, and a variety of other landscape-scale variables.
3.1. Methodology One obvious approach to modelling the presence or absence of terraced fields is multivariate logistic regression. For example, we can take as our starting point a division of the island into 10x10 m cells of either terraced agricultural fields (n=51,290) or unterraced, non-agricultural regions (n=149,793). This analytical scale was adopted for two reasons: a) it matches a 10m elevation model we have constructed for the island from 2-4m contours and spotheights, and b) it is roughly the size of the smallest discrete terrace units observable in the landscape (ca.10m long). The Antikythera field systems do not exhibit any sharp separation of ‘flat’ versus ‘terraced’ fields, with even the least inclined field enclosures exhibiting a typical combination of slope change and soil accumulation behind a wall that we generally consider to be terraces. We therefore initially consider all of these areas as ‘terraced’ (whose distribution is therefore not solely on hillslopes, see figure 3a, before addressing the more extreme terraced hillslopes as a separate case thereafter.