ARTÍCULOS

Probabilistic survey and prehistoric patterns of land and resource use in Mendoza Province, Argentina

 

Raven Garvey

Raven Garvey. University of Michigan, 4013 Ruthven Museum, 1109 Geddes Avenue, Ann Arbor, MI 48109-1079. E-mail: garveyr@umich.edu

Received 24 January 2014.
Accepted 16 April 2014

---

ABSTRACT

This paper describes the theory, methods and findings associated with a recent regional-scale, probabilistic surface survey designed to examine prehistoric hunter-gatherers' landscape and resource use as environmental conditions fluctuated throughout the Holocene in Mendoza, Argentina. Survey identified 67 previously undocumented sites in six environmental zones across the region. Correlations between key site attributes suggest that the mountains, foothills and plains were all used extensively, though perhaps at differing intensities and in different ways as a function of environmental and demographic factors. Probabilistic surface survey produces broadly comparable samples that can be combined with data from stratified sites for a better understanding of regional settlement and subsistence systems, and to address larger ecological and evolutionary questions.

Keywords: Regional archaeology; Surface record; Human ecology.

RESUMEN

Análisis probabilístico y patrones prehistóricos de uso de la tierra y de los recursos de la provincia de Mendoza, Argentina. Este trabajo describe la teoría, los métodos y resultados asociados a una prospección probabilística de superficie a escala regional en Mendoza, Argentina.

Dicha prospección fue diseñada para examinar el uso del paisaje y los recursos de los cazadores-recolectores prehistóricos dadas las condiciones ambientales fluctuantes durante el Holoceno, y permitió identificar 67 sitios no documentados anteriormente en seis zonas ambientales de la región. Las correlaciones entre atributos clave sugieren que las montañas, colinas y planicies fueron utilizadas extensivamente, aunque quizás de distintas formas e intensidades debido a cambios del medio ambiente y la demografía. La prospección probabilística de superficie produce muestras ampliamente comparables que pueden combinarse con los datos de sitios estratificados para entender mejor los sistemas regionales y también para abordar preguntas más amplias de ecología y evolución cultural.

Palabras clave: Arqueología regional; Registro de superficie; Ecología humana.

---

 

INTRODUCTION

This paper describes the theory, methods and findings associated with a recent regional-scale, probabilistic surface survey in Mendoza Province, Argentina (Figure 1; Garvey 2012). The project was designed to examine prehistoric hunter-gatherers' use of the landscape in relation to resource availability as environmental conditions fluctuated throughout the Holocene, and it combined survey with geochemical sourcing and obsidian hydration data to track trans- Holocene settlement and subsistence patterns in southern Mendoza. Probabilistic survey methods are rooted in probability theory and employ random sampling strategies in the collection of archaeological data to minimize biases, particularly the tendency to focus archaeological attention where previous research or a particular theory suggests sites should be found. Sampling, of course, is a means of generating "representative and reliable data within the bounds of [the researcher's] restricted time and monetary resources" by deliberately reducing the whole of available data to a manageable portion for detailed study (Binford 1964: 427). A sample is then assessed statistically to make generalizations about the whole. Random or probabilistic survey, then, proceeds by dividing the study "universe" into units of equal size and selecting a random subset of them for analysis; to be a random sample, each unit must have an equal chance of being selected for survey. Binford's (1964) description of a variety of probabilistic methods remains an invaluable resource for archaeologists designing random regional surveys.
In the present study, systematic random sampling provided a means to address prehistoric patterns of landscape use and ecological relationships through time (e.g., hunter-gatherers' strategic positioning relative to key resources) in Mendoza. This technique has been used similarly and to great advantage in a number of places including the North American Great Basin -a region usefully compared to Mendoza given ecological similarities between them- where a tradition of random sampling began in the 1960s and 70s (e.g., Thomas 1971, 1973, 1975; Davis 1975; Bettinger 1977). Prior to this time, much archaeological work in the Great Basin had been focused on cave sites, which represent of only a small portion of prehistoric groups' settlement and subsistence systems (Ebert 1992: 58). The so called "New Archaeologists" saw random sampling as way to counteract such biases, to better understand regional systems, and to produce broadly comparable samples that could be used to address larger ecological and evolutionary questions. For example, to understand whether one valley was used more intensively than another and to then make inferences about underlying behaviors, these researchers acknowledged that each valley must be sampled at same level of intensity. Otherwise, we cannot know whether observed site densities owe to differing prehistoric use of the valleys or to unequal sampling intensity. With this new sampling design came recognition of the importance of the surface record. Indeed, a clear picture of prehistoric life in places like the Great Basin requires attention to the surface record since surface scatters "may be the only remnants of some prehistoric task activities" (Thomas 1973: 167, 1974). This is particularly true in arid regions, including parts of southern Mendoza, where geomorphologic and pedologic systems leave archaeological sites exposed on the surface (i.e., soil formation is weak) or erode overlying sediments through time.
In the Great Basin, the combination of random sampling and surface survey afforded a level of scientific rigor and comparability not previously achievable. These methods were a key part of the present assessment of trans-Holocene settlement and subsistence patterns in Mendoza.

METHODS

The survey described here builds on a significant body of work aimed at understanding patterns of regional land and resource use throughout the Holocene in southern Mendoza Province (Gil and Neme 2002, 2010; Durán et al. 2004; Gil 2005; Gil et al. 2005, 2011; Neme et al. 2005, 2006; Zárate et al. 2005; Neme 2007; Neme and Gil 2008; Morales et al. 2009). As in similar environments, including the North American Great Basin (e.g., O'Connell and Hayward 1972), hunter-gatherers in northern Patagonia almost certainly exploited a wide range of resources, many of which were available at different times and in an assortment of biotic communities. However, we know relatively little about resource scheduling in and use of certain environment types in the region, particularly the plains east of the Andes (Figure 1). The regional surface survey described here was designed to test a wide range of environmental zones in order to generate the representative sample necessary to assess settlement and subsistence decisions, particularly in lesser-known areas and in the context of resource scheduling.

Figure 1. Project area map.

The survey universe consisted of 120 quadrats (1 km²) within a 140 × 20 km corridor extending from the Andes to the eastern plains (Figure 2). The main survey corridor was centered on the Atuel River, which, like other major river systems in Mendoza, has probably always served as a focal point of human activity. The corridor begins in the Andes at approximately 2500 m, and runs to El Nihuil Dam (1300 m), southwest of the city of San Rafael. Previous surveys indicate limited and fairly recent (~2000 14C BP) use of landscapes above 2500 m, perhaps owing to the long, severe cold season and low total biomass characteristic of the Andes at this latitude (Neme 2007), or to attrition of the record at higher elevations due to greater erosion, temperature variation, and other factors that affect preservation. The eastern terminus of the survey corridor, El Nihuil Dam has created an artificial lake, east of which modern development has affected a significant portion of the landscape. The main survey corridor crosscuts a variety of environment types, including lower elevations of the Andes, foothills, and the eastern plain, each of which is described in the following section.

Figure 2. (Top) Project area map indicating probabilistic survey transects and units. The main survey corridor follows the Atuel River from the mountains to the plains. Survey transects are oriented perpendicular to the river at a 20 km interval. Each transect is subdivided into twenty 1 km² quadrats. Quadrats selected at random are indicated in white. (Bottom) Detail of project area map indicating additional survey locations (white dots). Given the scale of the image, survey areas within 1 km of each other are indicated by a single point.

Along the main corridor, survey transects were established every 20 km, centered on the Atuel River and oriented perpendicular to its channel; each of the six 1 × 20 km survey transects was further divided into twenty 1 km² quadrats (Figure 2). While the distribution of survey transects was designed to capture resource variability owing to elevation and distance from the region's primary water source, the only justifiable sampling stratification prior to survey was to distinguish riverine (those adjacent to the Atuel) from non-riverine quadrats. A coin toss determined which of the two riverine quadrats in each transect would be surveyed, which ensured a sample of river use at various elevations while avoiding preferential survey of a particular river margin. Three additional quadrats within each transect were selected using a table of random numbers. The resulting distribution of survey quadrats sampled a wide range of riverine and non-riverine environmental zones at various elevations. Between January 2008 and July 2009, four short survey campaigns targeted twenty-four quadrats selected in the manner described above. Survey teams were unable to access five of the originally selected quadrats on account of impassible terrain, lack of infrastructure, private ownership, or extreme conditions. Five alternative units in adjacent or comparable locations replaced these and four additional units increased the sample to 28 successfully surveyed quadrats (of 33 attempted; Figure 2). Teams of 2 to 6 people inspected the ground surface in each location with a 50-meter interval between surveyors. Each team maintained a pace of approximately 3 km/hr (depending on terrain), keeping parallel to each other and the appropriate quadrat boundary, and following a boustrophedon configuration (Figure 3) to survey the quadrat systematically and efficiently.

Figure 3. Boustrophedon survey pattern.

Random survey revealed six usefully distinguished environment types: montane, foothills, riparian, alluvial fan, plains and lacustrine. The number of potential survey quadrats within each environment type was roughly proportional to its prevalence within the survey corridor, though riverine units were "over-" sampled given the initial decision to sample at least one per transect, and plains units were under-sampled relative to both the number of potential survey units on the plains and the sheer size of the plains environment (Table 1). The random distribution of probabilistic survey units did not include any of the region's lacustrine areas. To assess resource availability and prehistoric activities in lacustrine environments, three additional 1 × 1 km quadrats were sampled in the area between Transect 6 and El Nihuil Dam (Figure 2). In addition to the probabilistic sampling described above, nonrandom survey of 24 locations outside the main survey corridor was implemented after completion of the probabilistic survey in order to generate a larger and more balanced sample of environmental zones, targeting locations in the foothills and near upland lakes. Non-random survey included five sub-regions: the valleys of Arroyos La Manga and Blanco (major north-south trending tributaries of the Atuel River), the upland plain around Laguna Blanca (south of the Atuel River), the upper Atuel Valley in the vicinity of the river's confluence with Arroyo de la Manga, and the Loma de la Mina formation between the Andean foothills and eastern Plain (Figure 2). Virtual survey, using open-access satellite imagery (Google Earth 2010), identified sixteen previously untested areas in the greater Atuel drainage that, in conjunction with the probabilistic sample, provided reasonable coverage of the environment types in the survey universe (Figure 2). During ground truth of these, the crew opportunistically surveyed a small number of adjacent areas as well. Within each location, survey proceeded in a systematic way: crewmembers maintained 10 m to 15 m spacing between them and covered the entirety of natural landforms (e.g., benches, ridges, saddles, vega margins).

Table 1. Distribution of potential survey units and surveyed units across environment types. "# possible" is the total number of units within each environment type. "% universe" is the proportion of all potential survey units within each environment type.

ECOLOGICAL CONTEXT: PHYSICAL FEATURES AND AVAILABLE RESOURCES

The six relevant environment types identified during survey -montane, foothills, alluvial fan, plains, riverine, and lacustrine- are described here to give ecological context to the cultural material located during survey and to inform reconstruction of settlement and subsistence patterns.

Montane
In the Mendozan Andes (and, to a lesser extent, the adjacent foothills), resource zones change quickly with elevation. The highest elevations are glaciated and the mountains receive approximately 1000 mm of precipitation annually, primarily as winter snow, and for much of the year the entire ground surface is snow-covered. Between 3800 m and 2300 m, vegetation is of the Andean phytogeographic province, which is dominated by cushion plants (Böcher 1972; Neme 2007); steep terrain, runoff, poor soil development, high winds and solar radiation result in a flora consisting primarily of low, slow-growing plants adapted to withstand extreme conditions. Birds (large raptors, migratory waterfowl), guanaco (Lama guanicoe), puma (Puma concolor), vizcacha (Lagidium viscacia), and tuco-tuco (Ctenomys spp.) are among the few animal species found at these elevations. Thus, temperature, length of growing season and primary productivity all contribute to a relatively low, patchy and highly seasonal edible biomass, and human use of the Mendozan Andes is currently restricted to a small portion of the year, as it appears also to have been prehistorically (Neme 2007). This montane environment is occasionally interrupted by fertile vegas, marshes that develop in flat valleys bottoms. Here, rushes (Juncaceae) and grasses dominate the comparatively lush and varied flora that, while not edible to humans, attracts a variety of animals. Plants are sparse on adjacent hills and can include a variety of edible plants such as algarrobo (Prosopis flexuosa), molle (Schinus polyganus), solupe (Ephedra chilensis), and two species of cacti (Roig 1972; Böcher et al. 1972; Neme 2007). Much of survey Transect 1 is located in the montane environment, as are the non-randomly surveyed upper stretches of the Arroyo la Manga and Arroyo Blanco valleys.

Foothills
In the foothills, between roughly 2300 and 1500 m, conditions are generally milder and flora and fauna more diverse than in the mountains despite the lower average precipitation (Neme 2007). This environmental zone receives an average of 350 mm of precipitation annually, which falls relatively uniformly throughout the year, and there are numerous arroyos and springs that provide water year-round. Vegetation is of the Patagonian phytogeographic province (Böcher 1972) and edible plants include algarrobo, molle, solupe, and epazote (Chenopodium ambrosioides). Fauna include guanaco, ñandú and choique (Rhea Americana and R. pennata, respectively), pichi ( Chaetophractus vellerosus), culpeo ( Lycalopex culpaeus) and puma. Portions of survey Transects 1 and 2 are located in the foothills, as are the lower reaches of the Arroyo la Manga and Arroyo Blanco valleys and the Loma de la Mina formation (Figure 1).

Alluvial Fans
The Atuel and Salado Rivers produce an expansive alluvial fan at their confluence just out of the Andes. The alluvial fan biome characterizes an area approximately 22 × 17 km and descends gently from 1565 m near the village of El Sosneado to ~1400 m on the eastern plains. The fan itself is characterized by a substrate of coarse sub-angular to rounded gravels, very sparse vegetation (e.g., Schinus spp., Larrea spp.) and few permanent water features. East of El Sosneado, much of the Atuel's flow seeps into the gravelly soils at the western edge of the alluvial fan, to emerge as hundreds of dendritic arroyos, dry channels and gravel bars towards the transition from alluvial fan to plains. These arroyos eventually merge to re-form the main fork of the Atuel River but the position of the main channel has likely fluctuated for millennia. Portions of survey transects 2 and 3 are located on alluvial fans and many of the survey units were inaccessible during both summer and late winter attempts; the arroyos were swollen with melt water over a meter deep and moving too swiftly to cross either by vehicle or on foot.

Plains
The vast plains (1400 to 1000 m) are characterized by the Monte phytogeographic province (1800-500 m; Böcher 1972; Neme 2007), which consists of grasses and legumes (Poaceae, Fabaceae). The area receives little rainfall (350-250 mm per year), but the dry expanse is broken up by numerous, scattered water features including the Atuel River and its tributaries, shallow lakes and marshes, and salares (salt pans) that fill with rainwater in wet years. More generally, the terrain is characterized by modern and ancient dry stream channels, meanders and oxbows, sand dunes, salares and playa lakes. Towards the west, the surface is undulating and vegetation includes woody shrubs; as one moves east, the surface is flatter and sandier with sparse, low grasses and excellent visibility. Salares also become more prevalent towards the east and the whole area may have been submerged periodically in the past. Seed-bearing grasses and waterfowl (both resident and migratory) are common in riparian and lacustrine zones. Edible plant resources decrease with increased distance from riparian and lacustrine zones but guanaco, rhea, pichi and algarrobo are available both within and outside water-based biomes. This area is active pasture for goats. Survey transects 3-6 are located on the plains.

Riverine
Riverine habitat changes dramatically as the Atuel River descends from the Andes to the plains. At the westernmost edge of the probabilistic survey transect (2500 m), the river runs in braided channels through a fairly broad, flat valley. The substrate in this area is gravelly and vegetation is sparse except in backwaters that form along the margins of the main channel. East of the alluvial fans, described above, a rich riverine biome develops and includes stands of dense shrubs that attract a variety of animals. The river itself, however, is deeply incised in parts of the plains. Aerial photographs and archaeological survey data suggest that the river's channel has meandered across a wide area over the millennia.

Lacustrine
In the mountains and foothills, lacustrine zones around lakes including Lago el Sosneado (2000 m) and Laguna Blanca (1900 m) attract local and migrant waterfowl, ñandú (below 1500 m) and choique, guanaco and puma. Lacustrine zones on the plains are dominated by xerophilous and halophilous vegetation including Atriplex spp. and Prosopis spp., which may have been important foods prehistorically. Around Laguna Llancanelo, one of the region's major lakes, migratory aquatic birds, resident passerines, terrestrial fauna and fish form a rich lacustrine ecosystem within the arid plain. Mara, ñandú and choique are also common. No survey transects are located in lacustrine zones, but additional sampling targeted them sufficiently to rectify this bias.

RESULTS

The methods described above led to successful identification of 67 sites, small artifact scatters, and isolated artifacts (collectively "archaeological locations") in previously untested areas. Here, "site" refers to an accumulation of artifacts in an area greater than 10 m in diameter, regardless of artifact density in that area; "artifact scatters" to accumulations of artifacts less than 10 m in diameter; and "isolates" to artifacts found singly. These definitions are somewhat arbitrary, but it should be noted that distinctions between them were clear; sites are significantly larger and scatters generally smaller than 10 m in diameter. Probabilistic surface survey identified 40 archaeological locations at an average density of 1.42/km². Twenty-seven additional locations were identified by non-random survey at an average density of ~2.7/km² (Tables 2 and 3). These densities indicate a rich surface record across the study area, particularly considering the relatively wide survey interval necessitated by time and resources (50 m).

Table 2. "Encounter rates" of archaeological locations (sites, scatters, isolates) for probabilistic and non-probabilistic survey.

Table 3. Sites located during survey listed by environment type. Dimensions are in meters; area = m²; rich = artifact richness (# distinct artifact classes); date = obsidian hydration age estimate; n/m² = estimated number of artifacts per square meter; deb = lithic debitage; haft = haftable projectile point; form = formal tool; exped = expedient tool; mill = milling stone (mano/ metate); pot = pottery; org = organic; feat = feature; AVG = average per environment type for each quantitative variable and percentage of sites that contain particular artifact types (e.g., debitage, formal tool) for each.

Correlations between key attributes of archaeological locations and environmental features informed interpretations of landscape use and mobility through time. These attributes and features include: site/location size, date of occupation, artifact density, artifact class richness, and dominant lithic material types, environmental zone, elevation, and distance to water. With the exception of the riverine units and those on the alluvial fan, site size is not strongly correlated with environment type. The single site found in the riverine environment is large (~5000 m²) while sites on the alluvial fans are always small (isolates and small artifact scatters). The largest recorded site (~315,000 m²) is in the foothills, but both the widest range of site sizes and the highest proportion of sites larger than 1000 m² are in the mountains (N = 9). Site size is negatively correlated with distance to permanent water features. Sites within one kilometer of water range widely in size, from isolates to the largest site mentioned above (315,000 m²). Conversely, all sites located farther than one kilometer from water have a maximum dimension of 100 m or less, with two exceptions including a possible secondary lithic source (~500 × 60 m) that is 16 kilometers from any known permanent water source. Subsurface testing was not part of the original research program and assigning ages to surface sites requires an alternative to radiocarbon dating. Several obsidian flakes and tools were found during survey but, because obsidian hydration dating (OHD) is still in the earliest stages of development in Mendoza (Garvey 2012; Garvey et al. 2015), the age estimates derived by this method were grouped preliminarily into early Holocene (older than 8000 14C years BP), middle Holocene (8000 - 4000 14C years BP), and late Holocene (younger than 4000 14C years BP). While only 12 of the sites could be dated by OHD (or any means), the data suggest that sites in the mountains were occupied during the middle and late Holocene, sites in the foothills were occupied during the early and middle Holocene, and sites on the plains were occupied during all periods (Figure 4; Table 2). Parsed by elevation, the data show that sites above 2000 m are more common during the middle and late Holocene (Figure 5).

Figure 4. Distribution of dates among environmental zones.

Figure 5. Distribution of dates by altitude.

Here, artifact class richness is a simple count of discrete, descriptive artifact categories (e.g., biface; unimarginal, unifacial tool) present at each site. Artifact class designation follows Andrefsky's (1998) broad, macroscopic, morphological (rather than functional) typology, and observed classes are provided in Figure 6. Other classification schemes or microscopic usewear analyses might indicate greater variety among the artifacts than is presented here; this analysis presents a conservative estimate artifact class richness. Estimates of richness provide a means of distinguishing sites of limited use from ones where more activities were performed. "Limited use" or "limited activity" sites may have been the short-encampments of small groups, or the logistical encampments of groups whose residential bases were located elsewhere. Here, "logistical" refers to specific-resource or generally limited procurement and is not meant to imply "complex" hunting and gathering or a "collector-like" subsistence strategy (sensu Binford 1980; Kelly 2007), though certainly these were possible. Smaller, less-diverse sites (i.e., those with fewer artifact classes) are interpreted as limited activity areas or logistical camps while those with more artifact types are interpreted as longer-term base camps where a wider variety of activities were performed. As might be expected (Jones et al. 1983), artifact class richness is positively correlated with site size (R2 = 0.52; Figure 7). When all isolates and lithic scatters smaller than 5 m in diameter are removed from the sample, there is a nearly bimodal distribution: sites with a maximum dimension less than 2000 m have richness measures between 1 and 4 (mode of 1) while those with a maximum dimension greater than 2000 m have richness measures between 4 and 7 (mode of 4). These may represent different site types, namely logistical camps and base camps, respectively.

Figure 6. Diagram showing artifact classes (adapted from Andrefsky 1998:74).

Figure 7. Correlation between site size and artifact class richness among sites found during survey.

As with site size, and granting the autocorrelation described above, artifact class richness is highest within one kilometer of water. With one exception (a site with six artifact classes), sites farther than one kilometer from water contain just one or two artifact classes (mode of 1). Artifact class richness appears to be negatively correlated with date of occupation: younger sites contain more artifact classes on average than older ones. This pattern may owe to either site formation or prehistoric human behaviors. Earlier material records have simply been exposed to destructive agents longer than younger ones; increased time since discard / deposition equates to increased opportunity for artifacts to be moved, scavenged or buried. Conversely (or in addition), the observed pattern could reflect decreased residential mobility during the late Holocene. Results of within-site sampling performed during survey suggest that basalt is frequently the dominant lithic material at sites in the mountains (86% of sites) and foothills (42%) and on alluvial fans (50%). On the plains, however, obsidian and cryptocrystalline silicates (CCS; e.g., chert) are each the dominant lithic material at approximately 40% of sites. There may be a secondary source of CCS in the vicinity of the Embalse del Nihuil (on the plains), but XRF analysis indicates that the obsidian comes almost exclusively from the Andean sources of Las Cargas and Laguna del Maule (see Giesso et al. 2011; Garvey 2012 for source descriptions).

REGIONAL PATTERNS OF LAND AND RESOURCE USE

In the context of previous research (Gil and Neme 2002, 2010; Durán et al. 2004; Gil 2005; Gil et al. 2005, 2011; Neme et al. 2005, 2006; Zárate et al. 2005; Neme 2007; Neme and Gil 2008; Morales et al. 2009), these summary statistics permit interpretations of regional patterns of land and resource use through the Holocene. Namely, prehistoric hunter-gatherers' use of different environmental zones appears to have changed in response to fluctuations in resource availability caused by climatic and demographic factors (Garvey 2012). Sites in the montane region (between 3000 and 2000 m; N = 14) appear to have been used as residential base camps during the middle and late Holocene, as indicated by their average size and the fact that they are characterized by both more artifact classes (richness) and higher artifact densities on average than sites in other environment types. Paloenvironmental data indicate that the middle Holocene was a time of hotter, drier conditions in a number of regions worldwide (Nuñez and Grosjean 1994; O'Brien et al. 1995; Grimm et al. 2001), during which plant and animal communities may have migrated upslope (Chen et al. 2011). This, in combination with less severe winters may have made this habitat more attractive to humans during the middle Holocene. After 4000 BP, when climatic conditions ameliorated, increased human population densities across the region (Gil et al. 2005) may have required continued use of montane sites even if other environmental zones became more attractive. More detailed study of these montane sites -and the montane region more generally since neither random nor nonrandom survey went above 3000 m- is necessary to understand whether they represent seasonal or longerterm occupations during the middle and late Holocene.
During both periods, the overall pattern of mobility and subsistence may have been logistical, given the presence of small, limited-activity sites across the region that also date to the middle and late Holocene. Based on both previous research and results of this survey, the foothills appear to have been a preferred habitat throughout the Holocene, perhaps because elevations between 2000 and 1500 provided easy access to both upland and lowland resource including guanaco, rhea, armadillo, algarrobo and a wide variety of small animals, birds and edible plants. However, of the 26 foothills locations identified during survey, the majority (73%) are small artifact scatters or isolates (richness = 1). Given that the stratified site record (e.g., Gil et al. 2005) documents a large number of longduration or serially reoccupied sites of all periods in the foothills, these scatters and isolates may represent logistical locations associated with base camps in the foothills or other environmental zones (mountains and plains). Interestingly, the relatively large size, moderate to high artifact densities, and predominance of basalt at middle Holocene-aged sites in the foothills suggests low residential mobility during that time (Garvey 2015). Survey data indicate that the plains were also used throughout the Holocene (Figure 4). While the majority of archaeological locations found on the plains (11 of 17) are small artifact scatters or isolates (richness = 1), a moderately sized (10,000 m²), multi-use site (richness = 7) indicates that there were residential occupations on the plains at least seasonally. Despite the relative aridity of the plains, resources may have been diverse and abundant, particularly during the summer when migratory waterfowl flocked to the shallow lakes and marshes, which would also have been rich in fish and edible seed-bearing grasses. Indeed, archaeological data from comparable regions including the Great Basin of North America indicate that such lowland habitats may have been preferred prehistorically (e.g., Kelly 1999, 2001; Zeanah 2004). Still, on average, plains sites found during survey are less diverse than sites above 1500 m and many of them were likely short-term or limited-use.
The pattern of lithic raw material use at upland (montane and foothills) versus lowland (plains) sites invites further interpretations of land use and mobility. If the predominance of CCS and obsidian at plains sites -versus that of basalt above 1500 m- is an accurate reflection of prehistoric behavior, it suggests plains activities more often required precision in tool manufacturing and/or sharp cutting edges. The small grain size that makes CCS and obsidian both easy to work and very sharp might have been desirable for the production of small projectiles used to hunt birds and small animals, for fileting fish or harvesting grasses. It is also interesting to note that, among the sites found during survey, the obsidian was most commonly from sources in the Andes, suggesting high mobility or trade. It remains puzzling that prehistoric people seemed to prefer Andean obsidian sources to ones on the plains despite similarities in their quality and abundance, particularly in light of the fact that only the plains sources would have been available year-round (Giesso et al. 2011). Regarding prehistoric use of the plains, it bears noting both that this environmental zone remains proportionally under-surveyed and that it is a highly dynamic landscape. The plains are composed largely of loose, friable eolian and alluvial sediments; winds across the plains lift loose soils, creating palimpsests and burying sites beneath dunes. During survey of the plains, sites were most commonly found in blowouts suggesting that a failure to observe sites in particular survey quadrats is insufficient evidence that they do not exist there. Likewise, the Atuel floodplain is several kilometers wide and the river channel has meandered across it for millennia, surely eroding some plains sites and deeply burying others. This may also explain the low densities of sites found in alluvial fan, riverine and lacustrine survey units. The single riverine site located during survey may represent a base camp given its size (500 × 10 m) and moderate richness (four artifact classes). Without other riverine sites for comparison, we can't know whether this site is representative, but it is likely that the river and other water features were foci of human activity throughout prehistory in this semi-arid region; during all time periods and in all environment types, site size and artifact diversity are negatively correlated with distance to water. Still, results of the random survey make clear that human activity was not restricted to water margins and these less obvious site locations are key to understanding prehistoric patterns of land and resource use.

CONCLUSIONS

Data from 67 archaeological contexts (sites, scatters and isolated finds) identified during probabilistic and limited non-random surface survey, combined with data from the region's stratified sites, clarify trans- Holocene patterns of landscape and resource in Mendoza, Argentina. Specifically, these data suggest that the montane, foothills and plains regions were all important resource zones prehistorically but that each may have been occupied more or less intensively and, perhaps, used in different ways as environmental and demographic conditions fluctuated throughout the Holocene. That is, all three environments contain both base camps and smaller, limited-activity (logistical) sites. However, these data are presently insufficient to clearly identify a particular seasonal, annual or suprannual cycle of mobility. Nonetheless, limited evidence suggests stays of longer duration during middle Holocene droughts, increasing sedentism and larger populations during the late Holocene and, perhaps, activities on the plains that required a substantially different tool kit than that of other resource zones. Riverine and lacustrine environments were almost certainly more important prehistorically than indicated by the surface record, which may have been disproportionately disturbed by the flooding and meandering of the river itself. The plains remain under-sampled.
As noted by Great Basin archaeologists (e.g., Thomas 1973), probabilistic surface survey reduces sampling biases inherent to other techniques and produces broadly comparable samples that can be used to better understand regional settlement and subsistence systems. That is, if we restrict our archaeological attention to places known to have been important prehistorically, we may ultimately do little more than confirm what we already know about both the prehistory of a particular area and human adaptability more broadly; looking in unlikely places may yield unexpected results. The combination of random sampling and surface survey is particularly important in arid regions where site formation processes lead to an extensive surface record that offers a more complete picture of prehistoric behavior and produces the sample necessary to identify trans- Holocene settlement and subsistence patterns, from which we can then interpret larger ecological and evolutionary questions.

Acknowledgements

The author extends her sincere thanks to Adolfo Gil and Gustavo Neme, who provided invaluable advice and logistical support, and to Robert Bettinger for advice and feedback over the course of this project's development. Three anonymous reviewers provided insightful comments. Fieldwork was funded through support from the National Science Foundation (award #0914578), the Fulbright Foreign Scholarship Board, and the following University of California - Davis divisions: Department of Anthropology, Office of Graduate Studies, Consortium for Women and Research, Hemispheric Institute for the Americas and Institute of Government Affairs.

REFERENCES

1. Andrefsky, W. 1998 Lithics: Macroscopic Approaches to Analysis. Cambridge University Press, Cambridge.         [ Links ]

2. Bettinger, R. L. 1977 Predicting the Archaeological Potential of the Inyo- Mono Region of Eastern California. In Conservation Archaeology, edited by Michael B. Schiffer and George J. Gumerman, pp. 217-226. Academic Press, New York.         [ Links ]

3. Binford, L. R. 1964 A consideration of Archaeological Research Design. American Antiquity 29: 425-441.         [ Links ]

4. Binford, L. R. 1980 Willow smoke and dogs' tails: Hunter-gatherer settlement systems and archaeological site formation. American Antiquity 45: 4-20.         [ Links ]

5. Böcher, T., J. Hjerting and K. Rahn 1972 Botanical Studies in the Atuel Valley Area, Mendoza Province, Argentina. Parts I, II and III. Dansk Botansk Arkiv 22 (3). Copenhagen.         [ Links ]

6. Chen, I., J. Hill, R. Ohlemüller, D. Roy and C. Thomas 2011 Rapid Range Shifts of Species Associated with High Levels of Climate Warming. Science 333 (6045): 1024-1026.         [ Links ]

7. Davis, E. L. 1975 The "Exposed Archaeology" of China Lake, California. American Antiquity 39: 43-53.         [ Links ]

8. Durán, V., M. Giesso, M. Glascock, G. Neme, A. Gil and L. Sanhueza 2004 Estudio de fuentes de aprovisionamiento y redes de distribución de obsidiana durante el Holoceno Tardío en el sur de Mendoza. Argentina. Estudios Atacameños 28: 25-43.         [ Links ]

9. Ebert, J. I. 1992 Distributional Archaeology. University of Utah Press, Salt Lake City.         [ Links ]

10. Garvey, R. 2012 Human behavioral responses to middle Holocene climate changes in northern Argentine Patagonia. Doctoral Dissertation, University of California, Davis.         [ Links ]

11. Garvey, R. 2015 A Model of Lithic Raw Material Procurement. In Lithic Technological Systems and Evolutionary Theory, edited by N. Goodale and W. Andrefsky, Jr., pp. 156- 171. Cambridge University Press, Cambridge.         [ Links ]

12. Garvey, R., T. Carpenter, A. Gil, G. Neme, and R. Bettinger 2015 Archaeological Age Estimation Based on Obsidian Hydration Data for Two Southern Andean Sources. In review.         [ Links ]

13. Giesso, M., V. Durán, G. Neme, M. Glascock, V. Cortegoso, A. Gil and L. Sanhuesa 2011 A Study of Obsidian Source Usage in the Central Andes of Argentina and Chile. Archaeometry 53 (1): 1-21.         [ Links ]

14. Gil, A. 2005 Arqueología de la Payunia (Mendoza, Argentina): El poblamiento humano en los márgenes de la agricultura. BAR International Series S1477. Archaeopress, Oxford.         [ Links ]

15. Gil, A. and G. Neme (editors) 2002 Entre Montañas y Desiertos: Arqueología del sur de Mendoza. Sociedad Argentina de Antropología, Buenos Aires.         [ Links ]

16. Gil, A. and G. Neme 2010 Registro arqueológico en la cuenca media del Atuel: viejos y nuevos problemas; viejos y nuevos datos. In Condiciones paleoambientales y ocupaciones humanas durante la transición Pleistoceno-Holoceno y Holoceno de Mendoza, edited by M. Zárate, A. Gil y G. Neme, pp. 239-276. Sociedad Argentina de Antropología, Buenos Aires.         [ Links ]

17. Gil, A., G. Neme, A. Ugan and R. Tycot 2011 Oxygen Isotopes and Human Residential Mobility in Central Western Argentina. International Journal of Osteoarchaeology. doi:10.1002/oa.1304.         [ Links ]

18. Gil, A., M. Zárate, and G. Neme 2005 Mid-Holocene Paleoenvironments and the archaeological record of southern Mendoza, Argentina. Quaternary International 132: 81-94.         [ Links ]

19. Grimm, E., S. Lozano-García, H. Behling and V. Markgraf 2001 Holocene Vegetation and Climate Variability in the Americas. In Interhemispheric Climate Linkages, edited by V. Markgraf, pp. 325-370. Academic Press, San Diego.         [ Links ]

20. Kelly, R. L. 1999 Theoretical and archaeological insights into foraging strategies among the prehistoric inhabitants of the Stillwater Marsh wetlands. In Prehistoric Lifeways in the great Basin Wetlands, edited by B. E. Hemphill and C. S. Larsen, pp. 117-150. University of Utah Press, Salt Lake City.         [ Links ]

21. Kelly, R. L. 2001 Prehistory of the Carson Desert and Stillwater Mountains: Environment, mobility and subsistence in a Great Basin wetland. Anthropological Records Number 123. University of Utah Press, Salt Lake City.         [ Links ]

22. Kelly, R. L. 2007 The Foraging Spectrum: Diversity in Hunter- Gatherer Lifeways. Eliot Werner Publications, Clinton Corners, New York.         [ Links ]

23. Morales, M., R. Barberena, J. Belardi, L. Borrero, V. Cortegoso, V. Durán, A. Guerci, R. Goñi, A. Gil, G. Neme, H. Yacobaccio and M. Zárate 2009 Reviewing human-environment interactions in arid regions of southern South America during the past 3000 years. Palaeogeography, Palaeoclimatology, Palaeoecology 281 (3-4): 283-295.         [ Links ]

24. Neme, G. 2007 Cazadores-recolectores de altura en los Andes meridionales: el Alto Valle del río Atuel. BAR International Series 1591. Archaeopress, Oxford.         [ Links ]

25. Neme, G., V. Durán, V. Cortegoso, S. Diéguez, M. Giardina, C. de Francesco, C. Llano, A. Guerci and A. Gil 2006 A paleoecological approach to the archaeology of southern Mendoza. In Histories of Maize Multidisciplinary Approaches to the Prehistory, Biogeography, Domestication, and Evolution of Maize, edited by J. Staller, R. Tykot and B. Benz, pp. 199-214. Academic Press, New York.         [ Links ]

26. Neme, G. and A. Gil 2008 Biogeografía Humana en los Andes Meridionales: Tendencias arqueológicas en el Sur de Mendoza. Chungara 40: 5-18.         [ Links ]

27. Neme, G., A. Gil and V. Durán 2005 Late Holocene in Southern Mendoza (Northwestern Patagonia): Radiocarbon Pattern and Human Occupation. Before Farming 2005/2 (5): 1-18.         [ Links ]

28. Núñez, L. and M. Grosjean 1994 Cambios ambientales Pleistoceno-Holoceno: ocupación humana y uso de recursos en la Puna de Atacama (Norte de Chile). Estudios Atacameños 11: 11-24.         [ Links ]

29. O'Brien, S. R., P. A. Mayewski, L. Meeker, D. Meese, M. Twickler and S. Whitlow 1995 Complexity of Holocene Climate as Reconstructed from a Greenland Ice Core. Science 270 (5244):1962-1964        [ Links ]

30. O'Connell, J. and P. S. Hayward 1972 Altithermal and Medithermal Human Adaptations in Surprise Valley, Northeast California. In Great Basin Cultural Ecology: A Symposium, edited by Desert Research Institute, pp. 25-41. University of Nevada, Reno.         [ Links ]

31. Roig, V. 1972 Espozo General del Poblamiento Animal de la Provincia de Mendoza. En Geología, geomorfología, climatología, fitogeografía y zoogeografía de la provincia de Mendoza, pp. 81-88. Suplemento del volumen XIII de la Sociedad Argentina de Botánica, Mendoza.         [ Links ]

32. Thomas, D. H. 1971 Prehistoric Subsistence-Settlement Patterns of the Reese River Valley, Central Nevada. Ph.D. Dissertation, University of California, Davis.         [ Links ]

33. Thomas, D. H. 1973. An empirical test of Steward's model of Great Basin settlement patterns. American Antiquity 38: 155-176.         [ Links ]

34. Thomas, D. H. 1975 Nonsite sampling in archaeology: Up the creek without a site? In Sampling in Archaeology, edited by J. W. Mueller, pp. 61-81. University of Arizona Press, Tucson.         [ Links ]

35. Zárate, M., G. Neme and A. Gil (editors) 2005 Mid Holocene Paleoenvironments and Human Occupation in Southern South America. Quaternary International 132.         [ Links ]

36. Zeanah, D. W. 2004 Sexual division of labor and central place foraging: a model for the Carson desert of western Nevada. Journal of Anthropological Archaeology 23: 1-32.         [ Links ]