The first peoples of the Atacama Desert lived among the trees: A 11,600- to 11,200-year-old grove and congregation site

Significance

Hunter-gatherer mobility is widely debated in anthropology. Ethnographic data have led to hypotheses that hunter-gatherers in ancient times moved according to resource availability. This is a critical issue for hunter-gatherers from deserts, where water is usually the focus. Less studied is another immovable or hard to move resource: trees. Trees not only provide firewood and fruits, but also natural structures that offer protection from the elements and generate biotically rich environments. This study presents a mapped and dated record of trees associated with late Pleistocene/Early Holocene sites and explores different relationships between hunter-gatherers and trees, beyond that of using wood for fuel, but trees as part of the landscape.

Abstract

In deserts, water has been singled out as the most important factor for choosing where to settle, but trees were likely an important part of the landscape for hunter-gatherers beyond merely constituting an economic resource. Yet, this critical aspect has not been considered archaeologically. Here, we present the results of mapping and radiocarbon dating of a truly unique archaeological record. Over 150 preserved stumps around five Late Pleistocene/Early Holocene archaeological campsites (12,800 to 11,200 cal BP) show that trees were key features in the creation of everyday habitats for the first inhabitants of the Atacama Desert. At two of these sites, QM12 and QM35, the spatial and chronological correlation between trees and hearths reveals that people located their homes under the tree canopy. At residential site QM35, artifact distribution coincides with a grove dated to ~11,600 to 11,200 cal BP. A third residential area (QM32) occurred along the grove margins ~12,000 to 11,200 cal BP. Based on the distinct cultural material of these two camps, we propose that two different groups intermittently shared this rich wetland-grove environment. The tree taxa suggest a preference for the native Schinus molle, a tree scarcely present on the landscape today, over the endemic, nitrogen-fixing Strombocarpa tamarugo, both for toolmaking and firewood and even though the S. tamarugo was locally more abundant. Together with the spatial and chronological coincidence of campsites, hearths, and trees, we propose that people spared the most abundant and resilient species to create their homes, in turn promoting fertility oases amid the Atacama’s hyperaridity.

The first human groups to populate South America faced highly variable environmental conditions during the last glacial termination. By the end of the Pleistocene, they occupied almost every major continental environment. In turn, this triggered shifts in resource acquisition strategies (1–4). With the onset of Holocene warming, along with increased population pressure, groups began to move into less productive ecosystems (5, 6). One of these challenging places was the hyperarid core of the Atacama Desert, first settled between ~12,800 and 11,200 B.P. (7–10).

The Central Andean Pluvial Event (CAPE) transformed the Atacama’s hyperarid core compared to today. The Atacama has been hyperarid for ~14 Ma, receiving <5 mm/y in the basin between the Andes and the Coastal Range (7). However, increased rainfall in the high Andes during the Late Pleistocene and Early Holocene prompted runoff that penetrated the desert lowlands (<2,300 masl), and elevated the water table. This phenomenon occurred in two phases: CAPE I ( ~18,000 and ~14,100 cal BP, CAPE I), and CAPE II ( ~13,000 and ~9,500 cal BP, CAPE II) (11–13). During CAPE II increased biotic productivity facilitated the peopling of this desert, despite the overall lack of rainfall in the hyperarid core (9, 14–16). We found six archaeological localities around ancient water features (wetlands and rivers), particularly on the Maní and Guatacondo alluvial fans, dating to ~12,800 to 11,200 cal BP (Fig. 1). The broad divides between these productive areas were barren (~6 to 15 km), posing twin challenges for early settlers: how to use these patches of resources and how to move between them.

Fig. 1.

Map of the Atacama Desert with elevation profile, highlighting the PdT (yellow outline), the hyperarid core of the Atacama (broken black outline) and studied sites: Quebrada Maní 32, Quebrada Maní 35 and Quebrada Maní 40 (blue dot); Pampa Ramaditas 5, Pampa Ramaditas 7 (green dot), and Quebrada Maní 12 (red dot). For reference, Chipana 1 is also shown (yellow dot). (SRTM dataset, Mercator).

These archaeological sites reveal that people knew where to find water, good-quality local and nonlocal lithic raw materials, and wood fuel for their fires (9, 17, 18). Connections with people from the Andes (80 to 100 km away), the Pacific coast (60 to 80 km away), and the tropical lowlands east of the Andes (800 km away) were revealed by the presence of exotic materials, such as obsidian and marine shells (10, 17, 19). However, some of the sites suggest long-term use, indicating that this sector of the desert was recurrently visited, at least seasonally.

It remains unknown how people transformed this landscape into their home, that is, how they decided where to live and how to move across the desert. Water was certainly a factor, and in fact, all these archaeological areas are located next to either an ancient wetland or river terrace. We propose, however, that groves of trees, especially Strombocarpa tamarugo and Schinus molle, were key landscape features. In this environment trees grew as gallery forests or groves around wetlands (13, 16), constituting true oases amid extreme barrenness, aside from being one of the few visible landmarks in a mostly flat terrain. Therefore, the selection of groves to live in did not only consider firewood availability. People also looked for a biotically enriched harbor to rest, process food, shelter themselves, and structure social relationships.

Here, we present an exceptional record of over 150 subfossil tree stumps surrounding five of the six Paleoindigenous sites in the Pampa del Tamarugal (PdT) region of the hyperarid core of the Atacama Desert (Fig. 2A). By mapping all the visible surficial tree stumps around sites in Pampa Ramaditas (PR) and Quebrada Maní (QM), together with 85 radiocarbon dates on these stumps and 81 on archaeological materials, we present a study in the early peopling research that accounts for and reflects on the role of trees in the colonization of a new territory.

Fig. 2.

(A) Mapped tree stumps in the Guatacondo and Maní canyons alluvial fans and their association to archaeological sites. Small triangles represent one tree, medium-sized triangle is 10 trees, and large triangle corresponds to 30 trees. (B) Close-up to QM35 and surrounding sites (QM40, 32 & 34, 36, 37), displaying the distribution of trees (green triangles) in relation to lithic artifacts and bones on the surface (red dots). Artifact quantities for each site are shown in between parentheses. (C) Semiburned S. tamarugo/Neltuma sp. tree stump in PR8. (D) S. tamarugo and Neltuma sp. growing together and providing shade in a rest area near Colonia Pintados, Tarapacá (~170 cm humans for scale). This is a centenary tree and an example of how a habitation space under a tree could have looked like in the QM35 camp.

Results

Taxonomy, Preservation, and Chronology for Tree Stumps.

A total of 153 tree stumps preserved at the surface of the alluvial fans of the Guatacondo and Maní canyons were mapped and 85 were sampled, identified taxonomically, and radiocarbon dated (Figs. 1 and 2). These stumps have been preserved because of the extreme hyperaridity of the desert, i.e., lack of water prevents degradation and impedes fossilization. The trees likely became stumps due to a combination between wind erosion, termite activity, and “wood hunting” related to saltpeter mining (20, 21). The stumps are visible at the surface and covered by aeolian or fluvial sands. The roots usually go into a cemented Byz horizon impossible to excavate by hand, and an O horizon has not been preserved.

Nineteen individuals (24.7%) were identified as S. tamarugo or Neltuma sp., which are very similar anatomically. Given that Neltuma species were likely introduced into northern Chile several millennia later (22), it is possible that these 19 individuals correspond to S. tamarugo. An additional thirteen stumps (16.9%) were identified positively as S. tamarugo. Tamarugo is followed in abundance by S. molle, with 20 specimens (26%). Caesalpinia aphylla is represented by two securely identified individuals (2.6%), plus one stump that could be either C. aphylla or S. tamarugo (1.3%). Morella pavonis, a riparian tree that is not present today in the region, is represented by only one individual (1.3%). Twenty-one stumps (27.3%) could not be identified due to degradation issues, especially salts and termite activity (SI Appendix, Fig. S1).

We dated the bark or outermost portion of 68 of these stumps. To evaluate possible lifespan of these trees, we obtained additional dates from material sampled from the center of eight stumps. The resulting database was expanded by incorporating nine previously dated stumps (14). In total, we have obtained 85 radiocarbon dates for tree stumps in the PdT (SI Appendix, Table S1). The largest numbers of tree stumps are concentrated within the QM35 archaeological site (n = 25 stumps), and closely surrounding sites QM32, QM33, QM34, QM36, QM37, and QM40 (n = 47).

Nine stumps dated to 18,000 to 14,100 cal BP or CAPE I (N = 9). No tree stump dated to the period between 14,100 and 12,900 cal BP, providing further support for a dry phase, known locally as the PdT Desiccation Event or Ticaña Phase in the Altiplano (Fig. 3) (13, 16, 21). In contrast, tree stumps are much more abundant from 13,000 to 9,500 cal BP (N = 51) during CAPE II. Considering the unmodeled calibrated median dates, 40.3% (31 out of 77) of the tree stumps are contemporaneous with the Paleoindigenous occupation between ~12,800 and 11,200 cal BP (SI Appendix, Table S1). The rest of the trees (N = 46) correspond to either the late phase of CAPE II (11,200 to 9,500 cal BP) or later regional pulses of increased water availability. Eight trees also date to the Middle Holocene (~9,000 to 6,000 cal BP).

Fig. 3.

OxCal plot with all the CAPE I and CAPE II calibrated and unmodeled dates for tree stumps. CAPE I (18 to 14.1 cal) and CAPE II (12.8 to 9.5 cal) are shown in grey rectangles. White space in between CAPE I and II displays the PdT Desiccation Event (14.1 to 12.9 cal) as defined by Workman et al. (16).

We performed Bayesian modeling on 47 dates that match with CAPE II^*. The best-fit phase model delineates six phases (SI Appendix, Fig. S2), which hereafter is referred to as “the tree model.” The first three phases build up from isolated trees that grew from ~13,000 to 12,400 cal BP. Phase 4 incorporates 40 stumps mostly from the QM35-36-40 area that yielded similar medians. The obtained chronology suggests that this grove was active for ~1,110 to 850 y, between 11,620 to 11,345 and 10,650 to 10,490 cal BP. Phases 5 and 6 are also composed of isolated trees that grew between ~10,500 to 9,500 cal BP (SI Appendix, Table S2A).

Paired dates from eight bark (or outermost ring available) and tree stump centers show that specimens of both S. molle and S. tamarugo lived over ~100 y. One S. tamarugo may have lived for over ~600 y (SI Appendix, Table S3). This implies that the grove was probably alive for several centuries before the start of Phase 4, as our modeled phases are based on bark dates, which correspond to the death of the tree.

Chronology for the Paleoindigenous Occupation(s) of the PdT.

Summed probability densities (SPD) for 81 available archaeological dates on the six sites found and excavated at the PdT (QM12, QM32, QM35, PR5, PR7, and Chip1^†), show possible human activity between ~18,000 to 17,700 cal BP, ~15,200 to 14,800 cal BP, ~14,000 to 13,600 cal BP, ~13,000 to 12,700 cal BP, ~12,600 to 11,200 cal BP (SI Appendix, Fig. S3 and Table S4). All CAPE I dates come from charcoal (N = 12) and none from short-lived organic materials or artifacts (9). We consider these dates as evidence for the humans who lived at the PdT during CAPE II but were using old subfossil, mostly CAPE I-age wood for their fires. A previous study found evidence of different features between charcoal of the same taxa from CAPE I and CAPE II in the same prepared fireplace at site QM12, which were interpreted as subfossil wood utilization (18). Subfossil wood was highly degraded microscopically, showing deformed pores, tyloses, and insect galleries (18).

An SPD for 69 archaeological dates corresponding to CAPE II in the same six sites, presents no absolute gaps, but valleys and peaks could signal periods of less or more human activity in the area. The model that presents the best fit among dates is a four-phase model (SI Appendix, Fig. S4). According to this, human occupation in the region started sometime between 12,890 to 12,750 cal BP and ended between 11,250 to 11,200 cal BP. A first population event would have occurred between the initial date and 12,770 to 12,720 cal BP, represented only at site QM12, and lasting for 1 to 140 y. Phase 2 is represented by one date at QM12 between 12,620 and 12,480 cal y. However, it is not until Phase 3, starting at 12,490 to 12,010 cal BP, that most of archaeological campsites in the area aside from QM12 started to be occupied (i.e., QM32, PR5, and PR7). Phase 3 ends around 12,450 to 11,970 cal BP, thus spanning over 150 to 490 y. Phase 4 has a continuous signal from 11,980 to 11,830 to 11,250 to 11,200 cal BP, with representation at all sites. This phase lasted 670 to 780 y, overlapping for ~430 y with the grove of trees at the QM35 area, which was present between 11,680 to 11,410 and 10,730 to 10,580 cal BP.

The Archaeological Sites and a Late Pleistocene/Early Holocene Grove.

The QM35 is an archaeological site that covers a ~1.3 km² area on top of a coppice dune, alongside a deflated sector of a late Pleistocene paleowetland. The site is overlain by thousands of lithic artifacts that reveal almost complete reduction sequences (excluding nodules and cores) (10, 17, 23). Some areas appear superficially and stratigraphically denser (named with letters, such as QM35d), corresponding to both increased human activity and local taphonomic processes (e.g., wind deflation, salt heaving) (24). At QM35d, excavation revealed an unstructured hearth, and bone remains mainly from rodents and camelids, the latter including individuals of all ages, without selection of body parts that indicate a residential area. SPD for eight dates on bone collagen and wood suggest that QM35d reflects a single human occupation phase that lasted 450 to 750 y (SI Appendix, Fig. S5) beginning at 11,940 to 11,650 and ending between 11,250 and 11,180 cal BP (SI Appendix, Table S2B). Adjacent sites QM36, 37, and 40, are likely connected spatially and culturally to QM35, but have not been excavated yet. Wooden tools have not been recovered in this area, but lithic tools with a morphology related to woodworking (e.g., choppers, rabots, scrapers, side-scrapers, notches) are common on the surface and around tree stumps found specifically at QM40.

QM35 is the only site of the PdT where we were able to define a tree grove (i.e., a group of trees growing close together). Twenty-one stumps dated to 12,550 and 9,640 cal BP (Fig. 2B) are associated with QM35 over an area similar to the artifact distribution but more densely packed around the more intensely used portion of the site (QM35d). Sites QM36 and QM40 provided 10 additional stumps dated between 11,220 and 11,870 cal BP, which coincide with the chronology for QM35 occupation. This tree grove was in place between 11,620 to 11,340 and 10,730 to 10,580 cal BP (SI Appendix, Fig. S7), in full growth during most of the human occupation of the site but continued to thrive after humans abandoned the area between 11,200 and 10,600 cal BP (Fig. 4). The grove was composed of S. tamarugo/Neltuma sp. and S. molle equally (10 vs. 9 identified and dated individuals). The rest of the stumps (N = 13) could not be identified (SI Appendix, Fig. S6).

Fig. 4.

Calibrated range for each archaeological site, all CAPE II tree stumps, and tree stumps in the QM35-40 grove. Summed probabilities for all dates on each of these sites and assemblages are on top of the calibrated range (not to scale to probabilities density, just to show peaks and valleys). The green rectangle displays the period in which humans coexisted with the tree grove (~11,600 to 11,200 cal BP).

QM32, located one kilometer to the SW of QM35, probably represents residential occupations based on similarly varied lithic reduction sequences and the presence of several stages of cordage or textile manufacture, including local and foreign raw materials (10). The site also shows connections to the Pacific Ocean, the Andean highlands, and the tropical lands beyond (10). An SPD for 17 available radiocarbon dates show that the densest occupation could have started at ~11,900 cal BP, ending at 11,200 cal BP with a weak signal from 12,600 to 12,000 cal BP. A Bayesian 3-phase model best fitted the data (SI Appendix, Fig. S8 and Table S2B) with an initial phase defined by a single date (12,600 to 12,000 cal BP). The second phase lasted for 190 to 580 y, starting at 11,990 to 11,830 cal BP and ended at 11,810 to 11,420 cal BP. This phase is defined by 14 ¹⁴C dates, including dates on camelid hair cordage as well as on a piece of Ceiba sp. wood (11,840 cal BP) that was brought at least 600 km from the east (10, 25). The third phase (11,390-11,210 to 11,390-11,210 cal BP) is described by two dates on wood and cordage made of camelid hair.

Tree stumps from QM32 or from the nearby QM34 site (Fig. 2B) have been taxonomically identified as S. molle (N = 5), S. tamarugo or Neltuma sp. (N = 4) and M. pavonis (N = 1). Only one specimen could not be determined (N = 1; SI Appendix, Fig. S6). All these samples dated between 11,600 to 7,610 cal BP (N = 11), concentrating between 11,600 and 11,190 cal BP and thus partially coinciding with the main human occupation at the site (SI Appendix, Fig. S8). The main phase of human occupation at QM32 (~12,000 to 11,400 cal BP) chronologically overlaps with the QM35 tree grove (11,600 to 10,500 cal BP), and probably QM32 and QM34 were positioned on the fringes of this tree grove.

QM12 was the first location to be occupied by humans at the PdT and the Atacama Desert at large. The artifact assemblage—particularly the lithics which encompass a wide array of local and extralocal raw materials and full reduction sequences—suggests that the site corresponds to a residential camp, with cultural connections to the highlands and the coast (9, 10). According to our modeling of 19 dates on various materials, this site was occupied in three phases: Phase 1 between 12,880 to 12,750 cal BP and 12,620 to 12,480 cal BP (span between 148 and 355 y); Phase 2 between 12,460 to 12,000 and 12,460 to 11,950 cal BP (55 to 430 y), and Phase 3 between 11,940 to 11,760 and 11,830 to 11,690 cal BP (26 to 120 y) (SI Appendix, Table S2B). QM12 was also abandoned earlier than any other site, between (11,820 to 11,610 cal BP; SI Appendix, Fig. S9). The increase in dates toward the later periods of occupation is not visible in a simulation made in OxCal (SI Appendix, Fig. S10) and could signal an increase in the use of the site until its abandonment. This site is associated only with two S. tamarugo tree stumps dating to 12,900 to 12,780 cal BP and 12,960 to 12,760 cal BP (14), and thus are contemporaneous with the beginning of the human occupation. QM12 archaeological occupation span coincides chronologically with the beginning of the grove in QM35 (SI Appendix, Fig. S2), but the site is located 22 km NE of the grove.

Located 8 km north of the QM35, PR5 site rests on an unvegetated overbank deposit occupied in one phase between 12,490 to 12,100 cal BP and 12,460 to 12,000 cal BP (SI Appendix, Fig. S11 and Table S2B). We suspect that this phase is a short-lived event since available calibrated ages are statistically similar and this site lacks prepared fireplaces, exhibits a comparatively less diverse faunal record than residential sites, and finished lithic tools and plant fiber cordage are not found in association to manufacturing debris (23, 26), as seen at QM12, QM32, and QM35. The presence of a Punta Negra projectile point—a morphology linked to high-elevation basins from the Southern Atacama (27)—suggests that such an ephemeral site was related to the Salar de Punta Negra late Pleistocene occupations. Twelve stumps were mapped around PR5; six were identified as S. tamarugo/Neltuma sp., three as S. molle, and one as C. aphylla (SI Appendix, Fig. S6). These have a wide chronological range from 17,800 to 2,250 cal BP (SI Appendix, Table S1). None coincide with the reliable archaeological dates, especially those between 17,100 and 16,300 cal BP that likely represent relict stumps from CAPE I. These latest Pleistocene stumps are located 35 to 250 m from a fireplace excavated in PR5 in direct association with the diagnostic Punta Negra projectile point (SI Appendix, Fig. S12H). Charcoal from this fireplace was identified as S. tamarugo (N = 13) or Neltuma/Strombocarpa sp. (N = 7).

Site PR7 (7 km north of QM35) was also occupied in one phase with dates between 11,950 to 11,750 and 11,930 to 11,370 cal BP (SI Appendix, Fig. S13 and Table S2B). The phase model suggests two possibilities: a short-lived occupation span or a longer-term occupation (from a few years up to 498 y of occupation). The archaeological assemblage that includes a few young camelid bones and scarce expedient lithic tools morphologically suitable for processing animals and/or wood, suggesting a short-lived occupation (25). Charcoal sampled from a prepared fireplace evinces differences in managing firewood. Outside of the fireplace S. molle (N = 18) exceeds S. tamarugo (N = 3). Inside the prepared fireplace S. molle (N = 14) is also more abundant, but S. tamarugo (N = 5) and Neltuma/Strombocarpa sp. (N = 4) are slightly more frequent (SI Appendix, Fig. S14).

Before this human activity, an earlier phase between 12,470 to 12,000 cal BP is represented by only one date on charcoal from the same prepared fireplace. A third phase is defined by a bioapatite date on a rodent bone (11,610 and 11,240 cal BP). Apparently, this Caviomorpha individual entered the site after its abandonment by excavating a tunnel to levels below the human occupation and constructing a burrow with plant materials left by humans in this excavated fireplace and close by, not yet excavated fireplaces, in turn artificially producing phases 1 and 3 (25).

Only two tree stumps were mapped ~200 m south of the PR7 site, which date to 11,100 to 10,750 and 9,000 to 8,720 cal BP and have been identified as S. molle and S. tamarugo (SI Appendix, Table S1 and Fig. S6). The chronology for these stumps, however, does not coincide with the occupation of PR7. Occupation Phase 2, which we believe is when PR7 was occupied by humans, overlaps with the first half of phase 4 of the tree model, i.e., the tree grove at QM35 (SI Appendix, Fig. S2).

Discussion

Sites and Use of Wood and Trees.

During ~430 y between 11,600 to 11,200 cal BP, people inhabiting sites QM32, QM35, and PR7 had access to a S. tamarugo/S. molle grove located at site QM35. QM12 inhabitants may have had access to this grove during the last phase of their occupation. Considering that these trees can be long-lived (>100 y), it is possible that the grove and humans coincided for a longer span. In addition, most of these sites are contemporaneous: QM35 is contemporaneous with QM32, partially with PR7, and with the last phase of QM12. The range between ~11,900 and 11,200 cal BP, seems to be the period when people were inhabiting the southern sector of the PdT more intensively (Fig. 4).

Lithic and animal bone assemblages and our phase models suggest that QM12, QM35d, and QM32 were residential camps recurrently occupied for several to hundreds of years (9, 10, 17, 19). These three sites were located alongside or within areas of high productivity. Between ~12 and 11,500 cal BP these areas were wetlands. After this, they became vegetated floodplains (16). This process of wetland drying seemed to have occurred first at QM12. People abandoned this site and likely moved toward QM35 where a wetland and a grove were active. Isolated trees were present since the beginning of human occupation and even before ~12,000 cal BP, but the grove formed as the wetland shrank.

Human occupation in QM35d coincides, almost entirely, with the existence of a grove composed of S. tamarugo, and smaller numbers of S. molle. The abundance of S. molle, a facultative phreatophyte, at QM35, indicates that the water-table was close to the surface and that there was an active channel, a feature found and reported by Workman et al. (16). This grove was also large enough to reach sites QM36 and QM40 to the north. The fringes of the grove may have also covered site QM32 to the south.

In contrast, the sparsely vegetated floodplain of Pampa Ramaditas (PR) seems to have been occupied mainly as a hunting ground by people carrying a morphologically different set of projectile points, compared with those manufactured by people living in QM35 (ca. 7 to 8 km to the south). We do not know for sure who these people were, but the PR area is in between the residential sites of QM32 and QM35 and the Chipana workshop-quarry, source of a white silcrete that has been found in every Paleoindigenous site at the PdT (10, 17).

Ecological Traits of S. molle and S. tamarugo.

The investigated grove is composed of mainly two evergreen taxa: S. tamarugo and S. molle, both of which have ecological traits and uses that are important to understand the significance of the grove in terms of a residential place for hunter-gatherers. S. tamarugo is an endemic legume to the PdT that can reach 25 m in height, crown diameters of 20 to 30 m and ~145 m² canopy areas (28–30). Little is known on its longevity, but Rivera (31) reported 400-y-old trees. Our chronologies indicate that these trees lived between ~100 and 600 y during the Pleistocene–Holocene transition (25, 26). This obligate phreatophyte is well adapted to the extreme conditions (i.e., saline-alkaline soils, daily thermal oscillation, UV radiation) by deploying dual root systems, heliotropism, and stomatal regulation (28). Added to their nitrogen-fixing strategy, it increases biodiversity by modifying soil physicochemical properties, in turn creating fertile understories (32). Villagrán and Castro (33) indicate that S. tamarugo also provides forage, fuel, medicine, dyes, and raw material for construction and artifacts.

S. molle is native to tropical and subtropical zones of South America. This is a facultative phreatophyte tolerant to droughts and high temperatures, but in the Atacama it is only found in riparian environments. With hanging leaves and inflorescences, it can grow up to 15 m reaching 6 to 8 m in crown diameter, thus providing shade and coolness, good quality wood for crafts and fire as well as medicines (33, 34). Historical accounts refer to this tree as long lived (60 to 100 y). Nevertheless, to our best knowledge, there are no empirical data on its longevity.

Trees as Habitational Structures or Protection.

In two instances, one or more tree stumps can be directly linked spatially and temporally to an archaeological feature: i.e., fireplaces on sites QM12 and QM35 (Fig. 2). At QM12 a S. tamarugo tree was located ~15 m away to the northeast from the prepared fireplace. The first phase of occupation of this fireplace dates to ~12,900 to 12,500 cal BP, whereas the tree stump dates to ~12,900 to 12,700 cal BP. Considering a full-grown, healthy S. tamarugo canopy area the QM12 habitational area was partially under the shade of this tree in the earlier phase of occupation. The tree could have been a source of protection from the wind and the sun (Fig. 2D). It could have also served as a vantage point to the wetland located about 250 m to the east, where camelids may have gathered. Given that this terrace was elevated about 5 m above the active floodplain (9) and thus the water table was deeper, the terrace itself was probably not vegetated, except for tamarugos. At this site, we have also found several sharpened wooden stakes, vertically or diagonally hammered around the fireplace. These stakes might have been part of a hut, tent, or some kind of structure for processing animal skins (9, 24). All the stakes postdate the tree, however, with ranges between 12,000 and 11,600 cal BP. Consequently, if they were building a residential structure during this later phase of occupation, it was after the tree died.

At QM35, there are at least 12 trees in an area encompassing 300 × 150 m around the main human occupation, and 21 that have been dated for the whole site. We dated four of these trees surrounding the main area to medians between 11,380 to 10,940 cal BP, so some of them could have been providing protection during the second half of the occupation. The surficial artifact distribution roughly coincides with the mapped tree stumps as well (Fig. 2B). Therefore, more habitation areas might be found overlapping directly with the shade of a tree. This spatial organization draws a parallel with San hunter-gatherers. San residential camps are configured within tree groves, and each family made use of a shade tree to create their habitation area, by making a well-designated fire and sleeping area, and using unmodified tree branches as racks (35–38). This was also the case among the Aranda in the central Australian desert, who rarely built elaborate habitational structures, instead using trees to protect themselves while sleeping and holding meetings (39, 40).

Preferences of Material for Firewood.

Three fireplaces have been analyzed anthracologically for the studied sites. A prepared fireplace in QM12 was studied by Joly et al. (18), indicating that trees used as firewood were S. molle and to a lesser degree M. pavonis. QM12 is the only site that displays absolutely no use of S. tamarugo as firewood, even though the closest trees to the site were tamarugos. It seems that the people living there were avoiding Strombocarpa tamarugo as firewood. Previously, Joly and colleagues (18) concluded that S. molle and M. pavonis were complementary to each other in their utility. They did not discuss the question of why people were disregarding tamarugos. One possibility is that these people consciously chose not to use the tamarugos around the site because they were a fundamental element in their residential camp. Another, less likely option, if the QM12 occupants were originally from the highlands, as suggested by the presence of Las Cuevas projectile points (SI Appendix, Fig. S12I) and obsidian (9, 17, 19), could be that they preferred woods that were better known to them. M. pavonis and S. molle can grow near rivers at higher altitudes, while S. tamarugo is endemic and native to the PdT. This could be true for the earlier stages of occupation, but not very likely for when they were more established at the PdT.

S. tamarugo/Neltuma sp. have similar calorific values, and less moisture content as S. molle, although it can be less dense (18). S. tamarugo was used extensively at the PdT and Salar de Atacama in later times as firewood, for tools and construction. Based on the available information, the inhabitants of QM12 were collecting their firewood outside of the older, raised terrace where they set camp. Right beneath was a biotically productive floodplain, so people could have been getting their S. molle and M. pavonis wood from there.

Charcoal in the prepared fireplace at PR7 indicates a mix of S. molle and S. tamarugo/Neltuma sp., but S. molle is more abundant. Since this area seems to have been sparsely vegetated, if at all, it would make sense that these hunter-gatherers would have used whatever wood was available to them in their surroundings, including subfossil wood. However, they chose to use mainly S. molle, even though S. tamarugo was the more abundant species almost everywhere in the region. Additionally, this site does not contain any subfossil wood outliers, so the people were probably using trees or tree parts only recently dead as fuel. Because the two tree stumps found in the vicinity of the site do not correspond to the human occupation period, it is conceivable that these people carried wood into the site; perhaps from the QM35 grove, located 7 to 8 km from the PR7 camp.

The fireplace at PR5 consists of a shallow burned area with large pieces of charcoal mixed with ashes. This is the only site where S. tamarugo/Neltuma sp. predominates as fuel for fires. The absence of tree stumps contemporaneous with the human habitation of this camp suggests that these people were unable to bring recently dead wood to make fire, as was the case in PR7. Consequently, they had to depend on mainly subfossil wood for fuel. This assessment is consistent with the radiocarbon dating of the charcoal collected in this camp (SI Appendix, Table S4), all coinciding with the CAPE I period. The preference for S. tamarugo subfossil wood is explained by the fact that all the CAPE I tree stumps we found in the PR5 area are S. tamarugo/Neltuma sp. The QM35 grove (~7 km away), was not active during the PR5 occupation. Therefore, this behavioral pattern of selecting the wood that was closest at hand is consistent with an expedient occupation (41).

Trees as Raw Material for Tools.

Five pieces of unburned wood have been identified taxonomically from these sites. Two of them are wooden stakes from camp QM12, both S. molle. This is relevant because, apart from the use of S. molle as firewood, it reveals an additional preference as a raw material for artifact manufacture. Macrobotanical analysis of plant remains other than wood and charcoal revealed that leaves and seeds of S. tamarugo/Neltuma sp. are the most abundant unmodified plant remains from the camp (42). This could indicate that tamarugo was available in situ, whereas the other plants must have been brought in by humans or less likely by natural agents.

At PR7, located on a sparsely vegetated floodplain (SI Appendix, Tables S5 and S6) (16), people brought tree branches to the camp to make fire and tools. Some of the thin branches excavated at the site show cut marks made by lithic instruments. One of the pieces of wood brought to the camp is S. molle. Two other wood pieces found at QM32 and PR5, in a poor state of preservation, were probably intended for some special activity, since the wood came from the distant tropical setting beyond the Andes. The specimen found at QM32 was identified as Ceiba sp. and dated to ~11,900 cal BP. The cf. Erythrina L. or Ceiba sp. found at PR5 has not yet been dated (10, 43).

Our extensive collection of taxonomically identified tree stumps, along with other vegetation reconstructions for CAPE II, indicate that S. tamarugo dominated over S. molle. Thus, the utilization of woods other than tamarugo for tool making is intriguing, considering that this species was widely used for these purposes and as a building material in later periods (44–46). We hypothesize that sparing tamarugos might be related to conceiving them as places, or part of the giving environment (47).

The Tree Grove as a Possible Congregation Area.

QM32 and QM35 are two large archaeological sites that lie close to each other (1 km). They share similar chronologies and the pattern of living among or close to trees. QM35 was in the center of a grove. QM32 is in the fringes of the same grove toward the southeast, maybe within a smaller cluster of trees. In terms of their cultural assemblages QM32 and QM35 have similarities and differences. Both sites yielded materials from the Pacific Ocean, particularly shells, which have not been dated (10). QM32 yielded a wide array of other foreign materials, including cordage made of camelid fiber, rodent fur, and human hair, as well as the piece of Ceiba sp. wood. Stable isotope studies made on the camelid and rodent textile pieces revealed abundant C₃ plants and medium values of δ¹⁵N in their diets, which was interpreted by Santoro et al. (10) to mean that the animals came from the highlands. Even if these textiles are not foreign, they are unique to QM32, and the techniques used for their manufacture are also exceptional for this time (10, 26, 48). QM35 shows few highland cultural markers, while a projectile point called Escallonia is the most abundant and unique to this site regionally and at the Atacama Desert at large (SI Appendix, Fig. S12A) (49). In addition, this area, particularly QM40, contains a large collection of tools morphologically dedicated to wood processing. QM32, in contrast, is a more spatially confined camp (~0.1 km²), with a high frequency of very thin unifacial tools and no Escallonia projectile points.

Two distinct groups of hunter-gatherers may have frequented the area between ~11,900 and 11,200 cal BP. One camped inside the grove (site QM35), the other spent their time on the tree margins (QM32). QM32 could reflect shorter visits, something that would be reflected in the size of the occupation areas (~0.1 km² for QM32; ~1.3 km² for QM35). Radiocarbon dating resolution precludes us from knowing whether these potentially different groups were contemporaries. The phase models and cultural materials indicate that these camps were residential, recurrent occupations, and since they were so close to each other, we are inclined to think that people made efforts to overlap at the place, for social interactions that could have included a wide range of activities. Ethnographic records of different bands of arid and semiarid hunter-gatherers such as the Aranda (40) and the Enxet and Enhlet (50), show that they inhabited different spaces, congregating around water sources and groves for special occasions, particularly during seasons of resource abundance. We propose that due to the hyperaridity of the region, the combination of a wetland and a grove, particularly one composed of S. tamarugo, a such well-adapted, nitrogen-fixing tree constituted oases that would have promoted this kind of aggregation.

Conclusion

Five archaeological camps found on the alluvial fans of the Maní and Guatacondo drainages were settled during the Late Pleistocene–Early Holocene transition in the hyperarid core of the Atacama Desert. Increased rainfall at this time in the Andes led to the development of a wetter and greener landscape than today. With a robust radiocarbon dataset and modeling, we have clarified the chronological and cultural relationship between these camps and established a relationship between hunter-gatherer groups and their local environments, particularly with its trees. Our interpretations were enabled by an exceptional record of preserved tree stumps, unique in global late Pleistocene records, which we now know to be contemporary and spatially coincident with some of the archaeological sites.

People first inhabited the QM12 site, located at the proximal end of the QM alluvial fan, before 12,000 cal BP. 12,000 cal BP seems to have been a turning point between human exploration of the PdT to more residential, long-term occupations. During the early phase of occupation of QM12 (~12,900 to 12,500 cal BP), people likely settled an area specifically because it was protected by one or two S. tamarugo trees but did not use it as a source for fuel or tools. Around 11,600 cal BP, before the area became drier and the floodplain stopped supporting lush vegetation (16), people at QM12 abandoned the site. They may have moved westward to a more biotically productive location around QM35, in the distal section of QM’s alluvial fan. Here, the water table was high, and a wetland formed, promoting vegetation and modern fauna.

When people began to inhabit QM35 ~11,900 cal BP, we do not know whether there was an already dense grove of S. tamarugo and S. molle trees, or just a few trees. We do know that by ~11,600 cal BP the grove was established, and people lived among the trees until ~11,200 cal BP. Around this grove and wetland, a local cultural tradition likely developed, expressed in the projectile point morphology called Escallonia (49). QM32, located 1 km southeast of QM35, represents a “contemporaneous” residential occupation between ~12,000 to 11,200 cal BP. The site displays a cultural assemblage different from that at QM35. The artifact record includes a mixed unifacial-bifacial stone tool industry, animal and plant fiber cordage, and a piece of wood brought in or traded from the tropics east of the Andes. The contrasting contents of the two sites suggest that two distinct groups inhabited this wetland-grove environment, but whether these groups met cannot be determined. However, hunter-gatherer ethnographies from other world deserts suggest that the people aggregated at QM32 and QM35.

Based on the archaeological data presented here, we propose that people tended to prefer S. molle over S. tamarugo for fuel and as raw material for tools, even though S. tamarugo was more abundant and provided good quality wood. The human occupation at QM35 coincided with a grove of S. tamarugo and S. molle. Tamarugo wood was usually not used for tool production or as fuel by these people, which may indicate a region-wide tendency to spare tamarugo trees in favor of their structural benefits. In contrast, there was preference for S. molle and other less common woods because of their suitability for fuel and toolmaking. These patterns of use testify to intimate knowledge of the benefits of tamarugos, which are nitrogen-fixing, soil improving, highly adapted trees (30). In fact, later villages, and public spaces in this part of the Atacama were constructed within S. tamarugo/Neltuma sp. forests (44), indicating their importance not only as sources of wood but as places. Our results have important broader implications for future archaeological hunter-gatherer studies, in that we need to find ways to research human-nature interactions beyond our conception of nature as an expendable resource exclusively (51). We then pose a question: how can we understand an archaeological record of something that was not exploited, but preserved?

Materials and Methods

Survey, Mapping, and Collecting Tree Stumps.

UTM coordinates for known tree stumps in southern PdT were compiled in a database. With this database and the aid of geological maps, we developed survey areas around archaeological sites PR5, PR7, PR8, QM12, QM32, QM33, QM34, QM35, and QM40. Additional survey areas were created based on surface age (≥Pleistocene), elevation (700 to 1,250 masl) and position within the hydrological system (distal part of alluvial fans). Geopositioning of tree stumps was accomplished with a handheld GPS device. Tree stumps were photographed as found, cleaned with a brush, excavated around to check that they were in an upright, in situ position (Fig. 2C). Three samples were collected on each stump: two for radiocarbon dating (bark or outermost section, and center) and one larger sample (3 × 3 cm) for the taxonomic analysis.

Laboratory Analyses.

Radiocarbon dating.

Five types of materials were dated for this study: wood, charcoal, rodent and camelid bone, camelid fiber, and soil organic matter (SI Appendix, Table S6). For charcoal and wood,^‡ we followed a standard ABA pretreatment to remove contaminants. Animal hair samples were fragile and scarce. After experimentation, we decided to treat them only with acid (10 mL of 1N HCl at 80 °C for 20 min). In the case of bones, because collagen is usually degraded in our sites, we chose to extract and date the bioapatite using a protocol for weathered bones (52). Pretreatment for soils followed Honke et al. (53). Combustion and graphitization followed in-house University of Arizona protocols. Graphite targets were submitted to the University of Georgia AMS Laboratory for ¹⁴C determination. All dates were calibrated and modeled using OxCal 4.4 with the SHCal20 calibration curve and are reported in thousands of calendar years before 1950 (cal BP) and rounded to the nearest decade. δ¹³C was calculated using a gas-ratio mass spectrometer (Finnigan MAT 252) at the environmental Isotope Laboratory, University of Arizona.

Taxonomic identifications.

Tree stumps and charcoal were identified to the lowest taxonomic level possible. We used Joly’s reference collection for the Atacama’s woody taxa (18) to identify both tree stumps and archaeological charcoal. Definition of anatomical features was carried out under an episcopic microscope on three anatomical planes: transversal, radial, and tangential. Faunal remains were identified taxonomically and anatomically based on reference skeletal collections, comprising Lama guanicoe and Vicugna vicugna, and osteological guides (54, 55). Osteological terminology is based mainly on Smuts and Bezuidenhout (56) and Hershkovitz (57).

Statistical analysis and modeling.

To explore our chronological data, we created phase models in OxCal 4.4, a function enclosed within the Sequence command. We used our raw radiocarbon determinations as input, and these were calibrated at 95.4% (SHCal20). We only used dates on terrestrial material, from known provenance and in situ collected samples. OxCal’s Bayesian Analysis uses a Monte Carlo Markov Chain algorithm to evaluate the probability that each calibrated range for a ¹⁴C determination is related to another in a sequence and allows the researchers to add prior knowledge to assess whether and how dates are related to each other. Since our sites do not provide good stratigraphy, and materials tend to be mixed within the whole span of Late Pleistocene/Early Holocene occupation, we first produced summed probabilities graphs to observe the behavior of the data in terms of gaps, valleys, and peaks, both by site as well as a complete assemblage of ¹⁴C determinations. This allowed us to determine, along with our knowledge of the sites based on lithic and zooarchaeological materials, how many occupational phases were possible to explore with the phase models and establish boundaries. These possibilities were evaluated by observing the Agreement Indices (Overall Agreement and the Model Agreement), as well as the reliability of each boundary. We used the Span command to estimate possible periods of occupation for each phase. Only in one case we observed that two different phase models were equally possible for a site (QM12), and thus we implemented a simulation in OxCal with randomized radiocarbon determinations within the same time span to observe if one of the phases was an artifact of the calibration curve.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.