Pre-Columbian raised field agriculture in the tropical lowlands of South America has received increasing attention and been the focus of heated debates regarding its function, productivity, and role in the development of pre-Columbian societies. Even though raised fields are all associated to permanent or semi-permanent high water levels, they occur in different environmental contexts. Very few field-based studies on raised fields have been carried out in the tropical lowlands and little is known about their use and past management. Based on topographic surveying and mapping, soil physical and chemical analysis and OSL and radiocarbon dating, this paper provides insight into the morphology, functioning and time frame of the use of raised fields in the south-western Llanos de Moxos, Bolivian Amazon. We have studied raised fields of different sizes that were built in an area near the town of San Borja, with a complex fluvial history. The results show that differences in field size and height are the result of an adaptation to a site where soil properties vary significantly on a scale of tens to hundreds of metres. The analysis and dating of the raised fields sediments point towards an extensive and rather brief use of the raised fields, for about 100–200 years at the beginning of the 2nd millennium.
The Llanos de Moxos (LM) is one of the largest floodplains in Latin America,
characterized by a prolonged rainy season and a contrasting dry season,
often resulting in either severe flooding or droughts (Hanagarth, 1993).
Nowadays, the LM is sparsely inhabited, soils have been considered
unsuitable for agriculture and the main economic activity is extensive
cattle grazing (Erickson, 2003b). However, the presence of numerous
pre-Columbian earthworks indicates that in the past, humans modified the
landscape in several ways and cultivated crops (Denevan, 2001; Erickson,
2008; Lombardo et al., 2011b; Jaimes Betancourt, 2013; Prümers and
Jaimes Betancourt, 2014; Carson et al., 2015). The number and variety of
earthworks has made the area of the LM one of the most important examples of
pre-Columbian anthropogenic landscapes in the Amazon Basin (Erickson, 2008).
Raised fields are one of the most abundant and impressive types of
earthworks found in the LM. Pre-Columbian raised fields are elevated
agricultural earth platforms. They are also found in a number of other Latin
American countries. Distributed over a wide range of latitudes, spanning
from almost 20
Experimental studies have aimed to assess the productivity of raised fields and, because of their alleged high yields, they have been presented as a “sustainable agriculture” alternative for rural development projects (Erickson, 1992; Stab and Arce, 2000; Saavedra, 2009). However, most of these rehabilitation projects failed due to a number of reasons: overestimation of field productivity (Bandy, 2005), unfavourable structure of modern society (Erickson, 2003a), environmental conditions (Chapin, 1988), non-involvement of local communities due to a top-down approach (Renard et al., 2012), other methodological weaknesses (Lombardo et al., 2011a) and the generalisation of one single technology as a model for all raised fields regardless of their location (Baveye, 2013). As already stated, raised fields are highly diverse in design and exist in very different environmental contexts. Therefore, it has been proposed that the way they were built and their management was probably determined by geographical and/or environmental constraints (Denevan, 2001; McKey et al., 2010; Lombardo et al., 2011a; Baveye, 2013; Rostain, 2013; Rodrigues et al., 2016). However, the reasons behind the regional differences in raised field types is still not clear, as differences in raised fields could also be related to cultural diversity (Erickson, 1995; Denevan, 2001; Walker, 2011; Rostain, 2013).
Understanding the link between environmental variables (e.g. soils, topography, and hydrology) and design of raised fields is key in order to better understand the reasons behind raised field diversity and to infer how they could have been managed in the past. Very little is known with regards to the environmental characteristics of the areas where raised fields were built in lowland South America. Only recently some studies have started to include detailed soil studies providing an insight to the internal structure of raised fields and their relation to the environment (Iriarte et al., 2010; McKey et al., 2010; Lombardo et al., 2011a; McMichael et al., 2014; Rodrigues et al., 2015).
The LM is very suitable for the study of raised fields, as they appear in a variety of forms (ridged fields, platform fields, mound fields, ditched fields) and patterns (e.g. parallel or randomly scattered, with or without embankment and causeways). These different types of fields tend to be present in different parts of the LM and do not normally coexist in the same site (Denevan, 2001; Lombardo et al., 2011b). Nevertheless, recently it has been shown that, in at least two cases in the LM, differences in shape, height and layout can vary considerably within the same cultural area, and are the result of an adaptation to the distinct local edaphology (Lombardo et al., 2011a; Rodrigues et al., 2016).
The present study explores the raised fields located in the savannah on the
western bank of the Rio Maniqui, in the San Borja area (Fig. 1).
Archaeological evidence of occupational sites and raised fields here were
already described by Nordenskiöld (1916) at the beginning of the
20th century followed by Erickson (1980) and again recently by Iriarte
and Dickau (2012). Up until now no data from archaeological excavations have
been available for this area. Here, raised fields of different shapes
co-exist. This study aims to further our understanding of pre-Columbian
agricultural systems in the Bolivian lowlands. It uses topographic surveying
and mapping, soil physical and/or chemical analysis and OSL and radiocarbon dating
to address the following questions:
In which environmental context can raised fields be found and what are their
characteristics (e.g. morphology and soil properties)? Are there links between the dimension and/or shape of raised fields and soil properties? When were these raised fields in use?
The study area is located near San Borja, a town situated in the
south-western part of the Beni department, only a few kilometres away from
the Andean foothills (Fig. 1). The Beni department almost completely
overlaps with the Llanos de Moxos (LM), a seasonally inundated floodplain
drained by three major rivers: Río Mamoré, Río Beni and
Río Iténez. The diverse geomorphology of the LM is shaped by past
and present fluvial dynamics such as alluvial deposition and erosion and
river shifting (Hanagarth, 1993; Dumont and Fournier, 1994; May, 2011;
Plotzki et al., 2011, 2013; Lombardo, 2016), as well as
tectonics (Hanagarth, 1993; Dumont and Fournier, 1994; Lombardo, 2014).
These processes are responsible for changes in the local topography,
determining the flooding dynamics and, in turn, the forest-savannah ecotone
(Mayle et al., 2007). The climate in the LM is controlled by the South
American Summer Monsoon, leading to heavy convective rainfalls in austral
summer and dry conditions in winter (Zhou and Lau, 1998; Garreaud et al.,
2009). The mean annual temperature is 25.8
Soils along the Río Maniqui are in general acidic, with a pH ranging
from 3.86 to 5.11, but they are very heterogeneous in terms of plant
available nutrients, mostly correlated to the particle size of the sediments
(Guèze et al., 2013). Soils in the southern LM are generally loam or
silty loam and silty clay loam (Boixadera et al., 2003); all soils are
subject to hydromorphic processes and are mainly acidic (Boixadera et al.,
2003; Rodrigues et al., 2015), with some exceptions of saline soils (pH of
Our knowledge about the chronology, complexity and evolution of pre-Columbian cultures in the LM is still in its early stages (Prümers and Jaimes Betancourt, 2014). The archaeological landscape is roughly divided by the Río Mamoré; on the western side no raised fields exist, while on the eastern side thousands of hectares of fields have been documented (Denevan, 2001; Lombardo et al., 2011b). Detailed archaeological research has mostly concentrated in the Monumental Mounds Region (MMR) (Prümers, 2008; Bruno, 2010; Dickau et al., 2012; Jaimes Betancourt, 2012; Lombardo and Prümers, 2010; Lombardo et al., 2013), in the south-western LM. Here, no raised fields are present. Unfortunately, almost no chronological data exist for the eastern Llanos, where fields are widespread. Up till now, habitational sites associated to raised fields have only been dated in four locations: the San Juan site (AD 446–613) and the Cerro site (AD 1300–1400), in the northern part of the LM (Walker, 2004), and the Moxitania site (AD 700–1000), Abularach and Carretera Santa Ana sites (AD 900–1100), close to San Ignacio de Moxos (Villalba et al., 2004). Raised fields in Bermeo, close to San Ignacio, have been dated; they were used intermittently from AD 570–770 up to the 14th century (Rodrigues et al., 2015).
Left panel: Google earth image showing mapped palaeo-river features, causeways (red) and raised field areas (yellow). Right panel: digital elevation model (CGIAR_CSISRTM) including same mapped palaeo-river features, causeways (red) and raised field areas (yellow). Higher areas are relict sediment deposits from former rivers.
Study area including anthropogenic earthworks, natural geomorphological features and the two farms Campo España and El Progresso.
The location for the study of raised fields was first predefined with the
help of Google Earth. An area of about 8 ha was selected, where raised
fields of different shapes coexist (Fig. 1). ArcGIS was used to map natural
and anthropogenic features at two different scales (Google Earth, 2002,
2011, provided by the ArcGIS Basmap extension). On a large scale, covering an area of 2500 km
Field work was conducted in August 2012 and again in 2013. The local relief
was measured using a digital level Sokkia D50. A 480 m long topographic
transect perpendicular to the fields was drawn based on measurements taken
approximately every 1 m (Fig. 7). In addition, a specific area where higher
and lower fields lie next to each other was selected for in depth
morphological analysis: 2400 elevation points were measured and a digital
elevation model (DEM) was generated using the 3-D analyst extension of ArcGis
with natural neighbour interpolation (Fig. 7). Four fields were excavated;
trenches were dug from the ridge to the canal (Fields 1–4). Two additional
pits were dug; one in an area away from the raised fields (No Field, NF) and
one in a causeway (CW) (Fig. 1). In total 10 stratigraphic profiles were
prepared and sampled every 10 cm: the NF profile, the CW profile and two for
each Field (1–4), one profile in the ridge and a second one in the adjoining
canal. The description of the horizons and/or layers follow the guidelines of FAO
(IUSS Working Group WRB, 2014). In addition, a virtual grid was applied
onto an area of 450
All the samples have been air-dried. The colour determination, however, was
carried out on moist samples, using the Munsell soil colour charts (1994).
For particle size distribution, organic matter was removed with 30 % H
AMS C analysis on three charcoal samples and two palaeo-soil samples (Table 1) was conducted at the Poznan Radiocarbon Laboratory and LARA AMS Laboratory in Bern, calibrated using Calib 7.1 (Stuiver and Reimer, 1993) and the SHcal13 calibration curve for the Southern Hemisphere (Hogg et al., 2013).
Samples for optically stimulated luminescence (OSL) dating were taken by
pushing steel tubes into the exposed sediment. The concentration of dose
rate relevant elements (Table S1 in the Supplement) was determined using high-resolution
low-level gamma spectrometry, performed on bulk material from the
surrounding sediment. Dose rates have been calculated assuming an average
moisture content of 25–35 % and present-day sediment cover. For
equivalent dose (
Down-profile values of selected geochemical parameters and grain size of all Fields (ridges and canals).
Continued.
The area studied has been shaped by several shifting rivers, and more recently by anthropogenic earthworks. Natural features like palaeo-channels and oxbows, and anthropogenic earthworks, including raised fields and causeways, are illustrated in Fig. 2. Several generations of palaeo-river channels were clearly identified; these share the direction of the modern Río Maniqui, from the south-east to the north-west. One paleo-river can be traced continuously while in the other cases only segments of paleo-channels can be recognized. Channels and oxbows often appear washed out due to enduring erosion, the superimposing of other channels and the construction of earthworks, mostly raised fields. Patches of gallery forest exist along the channels and are recognisable by the topographic differences in Fig. 2b. In total, 370 ha of raised fields were mapped, as well as 52.2 km of causeways (Fig. 2a). All the raised fields are found along palaeo-rivers, on alluvial deposits, and are associated to causeways, but not vice versa. Some causeways were built through the pampa, connecting higher laying areas covered by forest along palaeo rivers (Fig. 2).
On a smaller area, detailed structures like small creeks, point bars,
anthropogenic features, including raised fields, causeways, settlement sites
and ponds, were mapped (Fig. 3). The natural features can be categorised as
continuous meander channels, creeks, oxbows and point bars. The assignment
of the oxbows to the larger channels is not straight forward as they
sometimes display similar channel width and sometimes they differ by several
metres. It is therefore difficult to say if oxbows and larger channels have
been formed by the same river. In some cases oxbows are dried out, but most
of them exist as wetlands. Point bars are found next to the larger channels
and oxbows. During the field survey we mapped four little earth mounds,
locally called
Within this area individual raised fields were mapped, as well as 14 577 m of
causeways (Fig. 3). The longest causeway, crossing the whole area from the
south-west to the north-east, is 2997 m long. From this major causeway several
shorter causeways go north and south, always in connection with raised
fields. The shape of the raised fields, as well as the causeways, are
particularly interesting. Most of the fields (
Profile description: Munsell colour signature is given for each layer,
water table boundaries are illustrated as blue dashed lines, Ab
As field work was conducted twice, in 2012 and 2013, and the rainy season during 2013 was much wetter, the depth of the groundwater table was significantly different each year: at the beginning of August 2012 the water table was 2 m below the surface, whereas at the end of July 2013 it was at a depth of 1 m. Detailed profile descriptions are summarized in Fig. 4.
All profiles share some pedogenic characteristics, but in general they
differ remarkably in a number of aspects. All profiles show intense
mottling, typical of hydromorphic processes. The iron mottles are soft, up
to 1.5 cm in diameter and the colour varies between yellow and orange, which
usually indicates the presence of goethite and lepidocrocite, typically
formed in waterlogged soils (Cornell and Schwertmann, 2003). Black millimetre scale
manganese concretions were also observed. Hydromorphy is present, on
average, up to 35 cm below the top of the ridge, but there is no clear
boundary. Manganese concretions tend to accumulate at the upper limits of
the hydromorphic affected layers. The depth of this diffuse boundary
slightly differs from field to field and is indicated with a blue line in
Fig. 4. With regards to the canals, the profiles can generally be divided
into two major layers: the infilling of the canal after the construction of
the fields and the undisturbed sediments below. This sharp boundary can be
clearly recognized due to the brown infilled canals contrasting with the
yellowish sediments below. This boundary is always present around 30–50 cm
depth (Fig. 4). With regards to the texture, all profiles differ remarkably
(Fig. 5 and Table 1) and distinctive characteristics were observed for each field.
AMS radiocarbon ages of charcoal and soil samples, given both as
Results from the grid sampling shows that most of the sediments contain a
high percentage of silt, meanwhile the content of sand with respect to clay
is very heterogeneous. The proportion of sand varies considerably from 2 to
41 % (compare with Fig. 5). The median particle size of each sample
(d50
Soil texture triangle including all measured samples.
One south–north oriented sand structure which is partly buried by fine sediments is evident in column 13. Two additional sand structures, one in column 11 and the other in column 7, are visible at the border of the sampled area; they seem to have the same south–north orientation. The finest texture of silty clay loam can be found at both west and east ends, in columns 0–3 and 18–20 respectively, where no fields were built (Fig. 6).
For all the fields, down-profile particle size distribution was measured separately and the results are consistent with the results from the grid. The fields with a finer texture of silt loam (Fields 1 and 4) show a coarsening up profile, where in the upper 20 cm there is considerably more sand compared to the bottom. In contrast, the profiles in the middle columns 10–13 show, on top of the sand structures, a fining upward sequence (Field 2 and 3). The highest fields are all built on the coarsest textures (row A, columns 8–15), while the lower smaller fields were mainly built on silty loam (row B–D, columns 3–18). It is striking that in the upper 60 cm of Field 3 the texture changes along its course towards the south-east, starting with sandy loam at the point of the excavation of the raised fields and ending up with silty loam towards the causeway.
Groups of parallel fields with similar heights were identified; these are always separated by a causeway (Fig. 7). The fields we excavated (Fields 1–4 in Fig. 7) were built on a slight slope. A topographic transect going from west to east reveals a downward trend, with a maximum difference of 90 cm (Figs. 6 and 7).
The highest points were measured on the ridge of Field 2 and the causeways
(Fig. 6b). The difference between ridge and canal for most of the fields is
around 25 cm, but in some fields it is up to 60 cm (Field 2). The DEM, which
includes the higher and the lower fields, illustrates the difference in
height and shows that the larger fields are, on average, 50 % taller than
the smaller fields (Fig. 6c). The higher fields are much better preserved
than the lower fields, because the latter have been partially destroyed by
cattle trampling. There are several causeways; one major causeway going from
the south-west to north-east, three perpendicular to it and one cutting all
the others, going from the south-east to north-west. The latter separates the
lower fields from the higher ones (Fig. 6c). The major causeway connects the
area of the raised fields with a settlement area and the
Interpolated grain size distribution of the sampled area using the
median particle size of each sample (d50
Down-profile variations of selected elements and grain size.
The sediments elemental composition is poor in carbonates, with less than
0.1 % of CaO. The sediments are dominated by quartz (SiO
Further soil chemical properties (C
Available phosphorous (P
CEC
A total of four OSL and five radiocarbon ages have been obtained (Table 2, Fig. 11). Two OSL ages were taken from the ridge profile 2 (40 and 140 cm) and two from the adjoining canal (36 and 60 cm). Samples for radiocarbon dating were taken from Palaeosol I, at a depth of 270 cm below Field 2 and from Palaeosol II, below NF, at a depth of 65 cm. The three remaining radiocarbon ages are from charcoal pieces: two extracted from the excavated CW and one from the ridge of Field 2. The oldest age, 5212–4940 cal BC, was found in the Palaeosol I, beneath the canal of Field 2 (270 cm depth). In this case only the humin fraction could be dated, as humates were missing. In contrast, both fractions could be dated in the case of Palaesol II in the NF profile (65 cm). The difference between the age of the humins (5934–5775 cal BC) and the much younger age of the humates (976–816 cal BC) is significant and might point towards a contamination (Walker, 2005). The OSL ages are consistent with the stratigraphy, the two basal ages are 3500–2680 BC for the ridge and 2870–2130 BC for the canal and the top ages are AD 1150–1290 and AD 1320–1430, respectively. The radiocarbon ages of two charcoal pieces extracted from the CW were 797–751 BC and 743–687 cal BC and from Field 2 1438–1262 BC.
Down-profile variations of available cations (CEC), base saturation (BS) and phosphorous (Pav).
Pearson correlation matrix. Values with significance level
Reconstruction of Field 2 illustrating the three suggested different phases: (1) ancient surface before the construction of the fields (short dashed line), (2) original field profile (long dashed line) and (3) present-day profile of abandoned field (solid line).
All the fields mapped in the vicinity of San Borja show two essential characteristics which could help explain why they were built and their shape and size.
Firstly, all the fields were constructed on fluvial deposits, which are naturally higher than the pampa, and are made of relatively coarse, hence better drained, sediments. Secondly, while field height seems to be related to sediment characteristics, distribution pattern seems to depend on the natural landscapes morphology.
Existing studies of raised fields in the LM have similarly shown that the majority of raised fields were built mainly on fluvial levees and on the naturally well-drained areas, often on silty to sandy sediments, with the aim of improving drainage (Walker, 2004; Lombardo et al., 2011a; Rodrigues et al., 2015). It seems that pre-Columbian people took advantage of the natural morphology of the rivers and built raised fields on the levees or point bars, where the coarser sandy sediments are deposited. Areas where surface sediments were too fine were avoided (Fig. 6). It should be noted that no connection exists between the height of the raised fields and the general topography, in fact, the lowest fields (for example Field 4) are located in the lowest lying areas. The soil properties of the raised fields studied differ considerably, probably due to the heterogeneity of the sediments in the area. Frequent changes in soil texture may be explained by the high frequency of crevasse splays and avulsions of the Maniqui River (Lombardo, 2016). The analysis of satellite image and particle size distribution shows that the landscape history of the area is complex, resulting from a combination of several palaeo-river generations of different dimensions and the depositional behaviour of the meandering rivers. On such diverse landscapes sediment properties can therefore vary naturally within the range of metres, in this case resulting in the great variability of sediments comprising the excavated fields.
OSL dating of Field 2 shows three major time phases (Fig. 11): phase 1 comprises the oldest ages, corresponding to the deposition of the fluvial sediments, phase 2 indicates the period during which the field was built/used and phase 3 marks the time when the raised field was abandoned.
The top layers include an age of AD 1150–1290 for the ridge of Field 2 and AD 1320–1430 for the adjoining Canal (Fig. 10, Table 2). The OSL age refers to the moment in which quartz grains were buried, therefore the age AD 1150–1290 probably indicates the moment in which Field 2 was built. In Canal 2, the age AD 1320–1430 marks the time of the canal infilling. This implies that there was no further digging of the canal afterwards, suggesting the abandonment of the fields. The estimated time of abandonment, AD 1320–1430, is consistent with the abandonment of the raised fields studied in Bermeo, around AD 1400 (Rodrigues et al., 2015). The exact time of construction and abandonment of the fields, however, cannot be deduced from these ages, as the fields could have been elevated more than once (Rodrigues et al., 2015). Furthermore, the arrangement of fields in groups separated by causeway indicates that fields were most probably built separately and could therefore have different ages. However, the OSL ages show that the raised fields were in use during a rather short time span, about 100–200 years at the beginning of the 2nd millennium. This is consistent with the raised fields in Bermeo (Rodrigues et al., 2015), where fields were used for short periods, and with the relatively brief occupation of settlements associated to raised fields in the northern LM close to Santa Ana (Walker, 2004). The age below the infilling of Canal 2, 2870–2130 BC, is consistent with the age obtained for the base of the ridge profile (3500–2680 BC), indicating that the sediments below 60 cm in Canal 2 were not reworked for the construction of the raised fields.
The suggested original depth of Canal 2 is consistent with these results and it may be reasonably assumed that the canal was originally 50 cm deeper (Fig. 11). Hence, the difference between the canal and the ridge was at least 150 cm, without taking into account the eroded material from the ridge (Fig. 11). Most probably all fields were much steeper, as in the case of Bermeo (Rodrigues et al., 2015). Compared to the platform fields in the northern part of the Llanos de Moxos (Walker, 2004; Lombardo, 2010), the large fields of San Borja are about three times higher. In addition, the raised fields in this area are associated to causeways which most probably were needed to reach the fields during high water season. It is important to note that causeways do not embank the fields, providing open runoff. There is no evidence that causeways were built to hold back water in the raised field's area or to extend the time of flooding in the area, as has been proposed for canals in the Apere Region (Erickson and Walker, 2009). During the year 2012, the canals adjoining the causeways were already completely dried out in July. Most causeways found in the adjacent settlement area, away from raised fields, must have served as a form of transportation and communication between settlements during the wet season (Erickson and Walker, 2009), suggesting that even the most elevated parts got flooded (Fig. 3). Because of its location close to the Andean Piedmont, San Borja gets on average 400 mm more precipitation compared to the northern LM (Hijmanns et al., 2005) and ground water levels during the dry season are 2 to 3 m below the surface, up to 3 times higher compared to the northern LM (Hanagarth, 1993). Because of this, during the rainy season sediments get saturated quickly. This combination makes flooding in the southern LM much more pronounced compared to the northern part of the LM and may be an important reasons why fields in the southern part of the LM are commonly much higher.
Charcoal derived from the CW (765–471 BC; 566–998 BC) and from the ridge profile of Field 2 (1438–1262 BC) are much older than the construction of the fields, probably predating the occupation of the area. While in the CW charcoal was plentiful, in the area of raised fields only one single piece has been found. In comparison, charcoal in the raised fields of Bermeo was more abundant, found in the canals and specific layers in the elevated fields (Rodrigues et al., 2015). In Bermeo the raised fields are under dense forest and it has been shown that the fields were in use during at least two different periods. This suggests that fire could have been used to clear the forest between periods of field use. The use of fire for raised field management in the northern LM has been also suggested by Whitney et al. (2014) and Erickson and Balée (2006). The use of fire however stands in contrast to the fact that charcoal was not directly found in the fields in San Borja and there is no evidence of the fields having been used during several different periods and managed with fire.
The age of the Palaeosol I (5212–4940 cal BC) below Field 2 is consistent with the age of the sediments that cover it (3500–2680 BC) and with the general chronology. There is an important difference between the ages from the two fractions of Palaeosol II in the NF (5934–5775 cal BC for the humin fraction and 976–816 cal BC humate fraction), this might point towards significant contamination. In theory, the humin fraction (residues) is considered to be more stable and represents the oldest age, whereas the humate (humic acid) fraction gives crucial information about the degree of contamination and, consequently, its reliability (Pessenda et al., 2001). The fact that the sample in the Palaeosol II of the NF was taken relatively near the surface (at a depth of 65 cm) could point towards contamination of the humate fraction with modern carbon from the surface (Walker, 2005). In addition, the fact that the Palaeosol I is chronologically consistent with the OSL ages above it, suggests that the age of the humine fraction might be more reliable. Taking the humin ages of the two palaesols does suggest that these two could have belonged to the same palaeosurface. Thus, the topography was much steeper than today and the depressions were later filled up with sediments. Palaeosol I could have been covered by a crevasse splay, while Palaeosol II could have been covered more slowly, with fine flood sediments from the surrounding area or the overflow of a distal river.
Hydromorphic characteristics present in all raised fields show that they are highly influenced by high water tables. The average depth of the water table can be derived from the Fe / Al ratio (McQueen, 2006), showing the in situ accumulation of iron which in some profiles is clearly expressed with a peak. Field observations show that for most profiles manganese tends to accumulate some centimetres above the iron oxides. This is common in hydromorphic soils, because of the different solubility of iron and manganese (Lindbo et al., 2010). The depth of these oxides and the Fe/Al ratio suggest that the depth of the modern water table in the ridge profiles of Fields 1 and 2 normally oscillates between 40 and 50 cm below the surface, while in the lower Fields 3 and 4 the depth of the water table is between 30 and 35 cm respectively. The hydromorphic features formed at the time of field construction have probably been erased, as hydromorphic features are forming continuously and the fields were not used for a very long period of time. As already mentioned, no connection exists between the height of the raised fields and the topography. For raised fields studied in the northern part of the LM it has been shown that these were built higher on finer sediments, as coarse sediments provided better drainage (Rodrigues et al., 2016). Surprisingly, in San Borja the opposite can be observed: smaller fields are built on finer sediments and the higher fields are built on the coarser sediments. There are several possible explanations for this apparent contradiction.
When comparing the relative depth of the water table below the ridge in the high fields vs. the low fields, we can see that there is a small difference of about 10–20 cm. Surprisingly, the water table below the ridge in Field 3, which has a similar height but much coarser sediments than Field 4, is almost at the same depth as in Field 4. This suggests that the drainage is not significantly better in the sandy area. This might be explained by the fact that the regional water table in San Borja is very close to the surface (Hanagarth, 1993; Miguez-Macho and Fan, 2012), hindering vertical drainage. Furthermore, as seen in Fig. 6, the sandy areas on which the fields were built seem to be enclosed by fine sediments, hence also hindering lateral water movement.
Besides the hydrology, another important factor determining the height of the fields seems to be the sedimentary characteristics of each field. In the case of the lower/smaller fields, sediments become finer towards the bottom, with a high percentage of clay in the lower layers (Fig. 4). The sediments in these deeper layers have a similar clay content (up to 28 %) to the sediments in the “no fields area” (NF). Such soils are generally avoided in agriculture because of their poor physical properties (e.g. low permeability and poor soil structure) and workability. It is therefore conceivable that due to the limited availability of coarser sediments in these deeper layers only the uppermost layers were used to raise the fields. In contrast, in the area of the large fields this impediment does not exist and fields could be built higher. In addition, coarser sediments are much easier to work, which could also explain why these fields are higher.
Taking into account the very different soil properties and fertility status of the sediments, people could have brought the more sandy sediments from the large fields area in order to improve the soil conditions of the area with clay-rich sediments, and vice versa. However, there is no evidence of this. On the contrary, as we can see in Field 3, the northern part of the field is composed of sandy loams, whilst further south the field is formed by silty loams and there is no evidence of attempting to improve the field's soil structure by mixing the finer and coarser sediments (Fig. 6).
In general, the geochemical and physical properties of the soils here are similar to those of other soils studied along the Maniqui (Guèze et al., 2013). This area has unfertile, acidic sandy-loamy soils. The exceptional high pH of Palaeosol II in profile NF is a result of the considerably higher amount of Ca and Na within the same layer (Fig. 6 and Table 1). Such saline soils, with accumulation of carbonates in the subsoil, have been described by other authors (Boixadera et al., 2003; Hanagarth, 1993). The Ca and Na could have been supplied through capillary rise (Boixadera et al., 2003). Similarly, as shown by Boixadera et al. (2003), the present profile is poorly drained and clay-rich. The layer just above Palaeosol II, with considerably more clay, could further prevent the outwash of the bases. Whether the Ca and Na come from capillary rise or are relict features from Palaeosol II, is beyond the scope of this study.
In general there is a clear relationship between particle size and
CEC
In the ridge profile of the coarser Field 3, the leaching of cations, geochemical changes and enhanced hydromorphism all suggests that, in the long run, the construction of raised fields could accelerate soil weathering. These processes should be taken into account when considering raised fields as a model for sustainable agriculture today.
It has been suggested that manure or muck grown in the canal could have been
used to fertilise the fields, allowing continuous production without the
need of fallow periods (Erickson, 1994; Lee, 1997; Barba, 2003; Saavedra,
2006). Nevertheless, in order to produce green manure the canals would have
had to retain water during the dry season, which is not the case in the San
Borja fields nor in other fields studied in the LM (Lombardo et al., 2011a;
Rodrigues et al., 2015). It is possible, however, that earth from the canals
could have been reused to raise the fields. If we compare the soil
properties from the infilling of the canal and the raised field, the former
has slightly higher content of organic matter, P
Up till now raised fields studied in the LM have not shown evidence of intensive manuring but rather of extensive agricultural practices (Lombardo, 2010; Lombardo et al., 2011a; Rodrigues et al., 2015' 2016). However comparing the fertility status of the raised fields studied here with the ones in the northern LM, the results show that the soils in San Borja are considerably more fertile.
The soils in the northern LM are much older and much more weathered (Rodrigues et al., 2015). On the other hand, density of fields in the northern LM is much higher than in San Borja. This high density of raised fields has been interpreted as the result of intermittent use by small groups of mobile people which were shifting their fields over a period of hundreds of years (Rodrigues et al., 2016).
Even though the fertility status of the soils in San Borja is better, raised fields here would similarly have needed fallow periods, especially those built on the sandy sediments. It is probable that the fields built on the more fertile soils could have been cultivated for longer periods, with shorter fallows. If we assume a similar scenario of small groups of mobile people in the present study area, the better soil quality could explain the much lower density of raised fields here compared to the northern LM.
Another reason for the lower density could be related to the widespread availability of more elevated well-drained areas. These areas normally do not get flooded and similarly, as shown for the region of Bermeo, raised fields could have been constructed to overcome periods of more severe and frequent flooding (Lombardo et al., 2011a; Rodrigues et al., 2015). On the contrary, the northern LM is affected by ponding water on a regular basis, because of relatively impermeable soils, and raised fields were needed to improve the drainage (Rodrigues et al., 2016). Consequently, the density of fields in the different parts of the LM could also be related to the frequency of use, with raised fields in the southern LM used only during periods of extreme events while in the northern part fields were used annually.
However, as almost no data exist about the size and timing of pre-Columbian occupations such scenarios are difficult to prove.
Furthermore, as argued by McKey and Rostain (2014), the raised fields may have possibly been complemented by other subsistence systems which also could have been different for each region.
Nevertheless, information about soil properties are important in order to understand the development of societies (McNeill and Winiwarter, 2006), differences in agricultural strategies and in distribution and density of people living in the LM. Up till now, raised fields studied in the LM all show that fields were constructed with the main purpose of drainage. The fact that some of the raised fields studied here were constructed on the highly unfertile sands supports the idea that drainage was the first priority. There is no clear evidence suggesting that raised fields were more productive compared with similar soils which are naturally drained. Hence, there is no indication which suggests that the construction of raised fields was a highly productive strategy which could sustain dense populations.
As already proposed with the first description of raised fields in 1916 by Erland Nordenskiöld (Denevan, 2009), they must have played a crucial role in protecting the crops from the floods. The abundant causeways, even on the more elevated area, further suggest that in San Borja flooding used to be a frequent problem.
As similarly suggested for the raised fields in Bermeo, the period of use in San Borja coincides with a period of higher ENSO activity, which has been reported to be an important factor responsible for extreme floods and droughts during the past 2500 years in South America (Meggers, 1994; Markgraf and Díaz, 2000; Moy et al., 2002; Rein et al., 2005). Some major floods in the LM have been associated with the negative ENSO phase La Niña, where rainfall is above normal in the Basin (Aalto et al., 2003). Moy et al. (2002) reported higher frequency of extreme ENSO events occurring between 1000 and 2000 years ago, with its maximum around AD 800, which coincides with the time when fields were in use. However, as reported for the extreme event in 2014, severe flooding can as well occur in absence of ENSO, as a result of tropical and subtropical changes in South Atlantic Sea Surface Temperature (SST) (Espinoza et al., 2014). Today losses of harvest due to flooding are frequently reported in the LM (UNDP, 2011).
We analyse raised fields of different sizes which were built in an area, near San Borja, with a complex fluvial history. Different generations of palaeo rivers, partly overlapping each other, coexist in the area, resulting in a heterogeneous depositional environment. This is reflected in the great variability of sediment particle size of the excavated raised fields. The results show that differences in field size and height are the result of an adaptation to this heterogeneous depositional environment. The dimension of the fields is related to particle size. Only coarse, silty to sandy sediments were used for the construction of the raised fields. The height of the fields depends on how deep the coarse sediments are: fields are relatively small where the coarse sediments are limited to the surface, whilst in areas where the subsoil was also made of coarse sediments these could be used to build larger fields. Areas with exclusively fine clay rich sediments were not used for the construction of raised fields. Raised fields were built by piling up the sediments taken from the excavation of the adjacent canal; there is no evidence of other agricultural strategies such as mixing of sediments or intensive manuring. Geochemical changes along the stratigraphic profiles show that the construction of fields might accelerate the weathering process in the long term, calling into question the idea of reintroducing raised fields as a very productive model of sustainable agriculture for today. The raised fields in the area are always associated to causeways. There is no evidence that causeways were built to manage the floodwaters; they were more likely used to reach the fields and to connect settlement areas. Although the construction of raised fields did not directly improve soil fertility, leading to higher productivity, it was of major importance to protect crops during severe flooding.
Leonor Rodigues and Umberto Lombardo conceived and designed the study. Leonor Rodrigues, Umberto Lombardo, Perrine Huber and Heinz Veit performed field work. Leonor Rodigues, Perrine Huber and Sandra Mohr carried out laboratory analyses. Mareike Trauerstein conducted OSL measurements. Heinz Veit secured funding. Leonor Rodigues prepared the manuscript with contributions from all co-authors.
The present study has been funded by the Swiss National Science Foundation (SNSF), grant no. SNF 200020-141277/1, and performed under authorisation N_017/2012 issued by the Unidad de Arqueología y Museos (UDAM) del Estado Plurinacional de Bolivia. We thank M. R. Michel López from the Ministerio de Culturas and our Bolivian counterpart J. M. Capriles for their support. A special thanks to the owners of the farms El Progresso and Campo España and their workers for their logistical support in the field and for allowing us free access to their land. Fieldwork assistance by B. Vogt, L. M. Salazar and C. Welker is gratefully acknowledged. We thank D. Fischer for technical support in the laboratory. X-ray fluorescence spectroscopy (XRF) was measured at the Geological Institute of the University of Fribourg. A special thanks to E. Canal for improvement of the manuscript. Edited by: O. Evrard Reviewed by: C. Prat and one anonymous referee