Lidar reveals pre-Hispanic low-density urbanism in the Bolivian Amazon

During the Late Holocene, pre-Hispanic agriculturalists in the Llanos de Mojos (Bolivia) transformed vast seasonally flooded savannahs into productive agro-aquacultural landscapes with diverse sociopolitical organization, water-control systems and economies. In the southeast sector, slightly elevated, nutrient-rich Andean-derived soils favored agriculture due to a mid-Holocene sedimentary lobe. In this context, the Casarabe culture developed between about AD 500–1400, spreading across roughly 4,500 km2. Previous reconnaissance and remote sensing identified numerous monumental mounded sites (lomas), smaller sites, and extensive canals and causeways, and excavations show these were year-round inhabited agrarian settlements with maize as a staple and hunting/fishing for protein. However, detailed mapping of mounded civic-ceremonial architecture and regional organization was limited by dense tropical forest. To address this gap, the authors conducted airborne laser scanning (lidar) over six targeted areas (totaling about 204 km2) to document settlement architecture and regional connectivity, aiming to evaluate whether Casarabe settlement represents low-density urbanism in Amazonia and to clarify hierarchical organization and landscape engineering.

The study situates Casarabe settlements within global discussions of low-density, agrarian-based urbanism documented in Southeast Asia, Sri Lanka, and Central America, including lidar-revealed urban landscapes at Angkor and in the Maya lowlands. Prior work in the Amazon includes reports of interconnected settlements in southern Amazonia and mound-village systems in SW Amazonia. In the Llanos de Mojos, earlier research documented monumental mounded sites, canals, causeways, and evidence for diversified agriculture and managed fisheries. Nonetheless, prior to this study, the scale, architectural elaboration, and regional organization of major Casarabe sites were poorly understood due to mapping challenges in forested settings. The paper builds on and contrasts these earlier findings to assess whether Casarabe represents a distinct form of tropical low-density urbanism.

Airborne lidar mapping: Six unconnected target areas (10–85 km2 each; total ~200–204 km2) within the Casarabe culture area were surveyed. A Riegl VUX-1 scanner with Trimble APX-15 UAV GNSS was mounted on a Eurocopter AS350. Laser PRR: 200 kHz; flight altitude: ~200 m AGL; airspeed: ~45 knots; flight lines: 200 m parallel strips with 50% overlap. Raw point cloud densities were typically ~18 million points per km2 (range 13–20 million). Data processing (RiAnalyze) addressed an 18 s time offset between laser and trajectory and height differences between adjacent tracks through manual and semi-automated iterative corrections; residual vertical discrepancies up to ~50 cm remained. Filtering initially considered only last returns and single-return points; macros for pre-classification were iteratively tuned to terrain conditions. DEM LAS files produced with mean point spacing ~0.3 m; DEMs rendered at 0.5 m/pixel using natural neighbours in ArcMap. Visualizations included hillshade, slope, and relief visualization toolbox (RVT) methods to optimize archaeological visibility. Site hierarchy classification: Based on lidar-derived measures of (1) base platform dimensions; (2) elaboration of civic-ceremonial architecture; (3) number and extent of outer polygonal enclosures; (4) number and extent of straight, raised causeways; and (5) investment in water-management infrastructures (canals/reservoirs). Volume calculations: For core areas (base platforms plus platforms and truncated/conical pyramids) at major sites, volumes were computed in ArcGIS by raster extraction of platform areas, calculating elevation (Z) per pixel, computing per-pixel volumes, and summing via zonal statistics. Reported base platform metrics include: El Cerrito 61,970 m2 area, 323,988 m3 volume; Landívar 99,795 m2 area, 276,030 m3 volume; Cotoca 159,649 m2 area, 384,228 m3 volume. Overall core-area investment estimates include Landívar ~276,000 m3 and Cotoca ~570,690 m3. Radiocarbon chronology: 144 radiocarbon dates underpin Casarabe chronology, though from a limited number of sites; stratigraphic control at four sites (Loma Alta de Casarabe, Mendoza, Salvatierra, Pancho Román). For PABAM sites, 94 dates (46 Mendoza; 48 Salvatierra) were analyzed; two outliers rejected. Bayesian modeling used OxCal v4.4.2 with SHCal20. An additional 50 dates from several regional sites were compiled. Twenty-two dates from Chocolatelito Unit 3 are only published as 1σ ranges, precluding integration; two dates (AD 0–220) appear unusually early and require further assessment. Mapping and analysis: Integration of lidar with prior reconnaissance and remote sensing to map site distributions, measure inter-site distances/densities, identify causeways, canals, enclosures, and quantify settlement clusters and catchments.

• Lidar revealed two exceptionally large Casarabe settlement sites: Cotoca (147 ha) and Landívar (315 ha), each exhibiting massive civic-ceremonial architecture and multiple concentric defensive ramparts with moats; some sections include double walls.
• Four-tier settlement hierarchy across the Casarabe culture area (~4,500 km2): • Primary centres: Cotoca and Landívar with extensive artificial terraces (up to ~6 m high) supporting U-shaped and rectangular platform mounds and conical pyramids rising >20 m above the savannah; buildings oriented uniformly to the north-northwest, likely reflecting cosmological principles. • Secondary centres (e.g., El Cerrito; Salvatierra sites 106, 186, 193, 195): base platforms 2–6 ha; single visible polygonal enclosure of ~21–41 ha; multiple platform mounds. • Tertiary centres (e.g., sites 189, 192): base platform ~0.5 ha with single platform; circular ditch enclosing up to ~2.5 ha. • Fourth tier: small elevated forest islands (~0.34 ha average), likely temporary campsites or specialized activity locales; a possible fifth tier of small hamlets without mounded architecture may exist but is not detectable by lidar.
• Regional integration and density: Across the 4,500 km2 area, an average of ~10 sites (primary to tertiary) lie within a 10-km radius (~2-hour walk) of any settlement. The eastern sector shows higher densities with mean inter-site distances of ~1,800–3,970 m. Sites form spatial clusters interconnected by causeways and canals spanning ~100–>500 km2.
• Connectivity and access control: Straight, raised causeways radiate from primary centres for several kilometres, linking to lower-tier sites, rivers, and lakes. Small platforms (~20 × 25 m, up to 2 m high) at strategic points on causeways and at enclosure gaps suggest controlled access.
• Water-management infrastructure: Extensive canal and reservoir systems integrate with settlement layouts. A ~7-km canal supplied water from Laguna San José to Cotoca, indicating large-scale landscape engineering and labour mobilization.
• Labour investment and monumentality: Estimated core-area construction volumes are ~276,000 m3 at Landívar and ~570,690 m3 at Cotoca; the latter is roughly ten times the earth moved for Tiwanaku’s Akapana (53,546 m3), underscoring extraordinary labour investment relative to well-known Andean monuments.
• Correlation observed between the height of civic-ceremonial architecture and base platform size. Evidence of earth extraction zones around central terraces suggests construction borrow pits that seasonally hold water.
• Spatial roles: Cotoca appears to anchor a ~500 km2 area (roughly half forest, half savannah) including 18 other monumental sites (three secondary, two tertiary, plus clusters of fourth-tier sites). In regions lacking a primary centre, secondary centres can function as central nodes for surrounding lower-tier sites.

The lidar evidence demonstrates that the Casarabe culture developed a highly integrated, dispersed yet connected settlement network with monumental civic-ceremonial cores, defensive enclosures, engineered causeways, and sophisticated water-management systems. This addresses long-standing debates by providing concrete, landscape-scale evidence that western Amazonia hosted substantial pre-Hispanic populations organized in low-density urban configurations. Architectural orientations suggest shared cosmological principles, while the arrangement of causeways and platformed checkpoints implies regulated movement and socio-political control. The hierarchical network—primary centres linked to secondary and tertiary sites, with fourth-tier forest islands—reveals planned spatial organization and regional integration within clusters spanning up to several hundred square kilometres. Comparisons of labour investment and spatial extent indicate that Casarabe settlements were on a scale comparable to some Andean cultural achievements and exceeded previously known interconnected settlements in southern Amazonia that lack equivalent monumental civic architecture. These findings contribute to broader discussions of the diversity of early urban forms globally and challenge simplistic binaries of city versus village in tropical lowlands.

The study identifies the Casarabe settlement system as a distinct form of tropical, agrarian, low-density urbanism—the first such documented case in the tropical lowlands of South America. The scale and monumentality of civic-ceremonial architecture, extensive water-management infrastructure, and dense, hierarchical regional networks refute notions of sparsely populated western Amazonia in pre-Hispanic times. The results expand the recognized spectrum of early urban diversity and invite a rethinking of the categories used to describe Amazonian societies. Future research should refine chronology and functional variability across tiers, expand lidar coverage to detect potential non-mounded hamlets, integrate additional excavations and paleoenvironmental data, and further model labour investments and demographic implications.

• Spatial coverage: Lidar mapping was limited to six unconnected areas totaling ~200–204 km2 within a larger ~4,500 km2 cultural area, leaving many regions unsurveyed.
• Data quality: Despite extensive post-processing, residual vertical discrepancies up to ~50 cm remained in the lidar data, potentially affecting fine-grained measurements.
• Detection bias: Potential fifth-tier small hamlets without mounded architecture are likely invisible to lidar; forest cover and taphonomic processes may obscure features.
• Chronology: Radiocarbon dates are concentrated at a limited number of sites; some published dates (e.g., Chocolatelito Unit 3) are only available as 1σ ranges and could not be fully integrated; two unusually early dates (AD 0–220) require further validation.
• Preservation: Inner defensive structures at some sites (e.g., Cotoca) are only partially preserved, complicating complete reconstructions.
• Functional/temporal variability: Variability of civic-ceremonial architecture within tiers may reflect chronological and functional differences that remain to be clarified.