Modern humans have been present in Borneo for over 50,000 years [33] and in East Kalimantan for over 10,000 years [34,35,36]. The earliest evidence of rice consumption in Borneo dates to c. 4500 BP [37], though rice phytoliths dating before 7000 BP have been reported from Sarawak [33]. Regular clearing of significant areas of forests as a land-use system is unlikely to predate the advent of iron tools [38]. Suitable technologies likely reached Borneo’s coasts around 2000 BP but made slow and uneven progress into the interior [39,40]. While 19th century European explorers to Borneo’s interior predominantly found rice cultivators using iron tools, stone tools remained in use in some communities until the mid-20th century [41,42,43,44].

Pre-iron forest clearance would likely have focused on small scale cultivation of root and tuber crops [33,45,46]. Some modern groups still rely on a mix of rice and tubers [47,48]. Only among certain iron possessing groups was swidden rice cultivation extensively developed, contributing to these groups’ demographic, cultural and military prominence. These groups expanded their influence, conquering, assimilating, replacing and influencing others [46,49,50,51]. During the last millennium a number of these swidden rice farming groups (including Kenyah, Kayan and, in our field area, the Merap) expanded into the interior of northern East Borneo [52]. The resulting settlements included hundreds to thousands of inhabitants and were organized to defend their territories, especially the limited land suitable for cultivation [53].

All Malinau’s ethnic groups have been mobile—fleeing from wars, persecution and disease, or invading, or invited as part of new alliances [49,53,54,55]. Most Malinau communities (though not all of the Merap) say that they came from other watersheds within the last two centuries (various pers. comm. by community members to DS and IB). Other ethnic groups such as the Bulusu (or Berusu), were driven away [53]. Other evidence, such as the genetics of local rice cultivars, also indicates considerable migration and mixing in recent times [56].

Ethnographic accounts suggest that settlements in Malinau were traditionally located in strategic sites for protection against inter-tribal warfare and head-hunting raids [53] and leading to locations comparable to Denevan’s “bluff model” for pre-Colombian Amazonian settlements [57]. Villages relocated within their territory four or more times each century when there was a shortage of good farm land nearby. Catastrophes (floods, epidemics) and inter-tribal warfare, also led to frequent village relocation and to a relatively dynamic history of settlement in the region [53]. Ethno-historical accounts suggest that all contemporary communities have relocated their settlements many times in the last two centuries [58]. Since the early 20th century inter-tribal warfare ceased and communities have settled in accessible riverside locations [53].

Old village sites continue to be visited due to their high density of native fruit trees [59]. As the Malinau River possesses few riverside locations suited to settlement, most such sites were used at some point in the last half century (IB and DS unpublished interviews).

Swidden rice farming groups are not the only inhabitants. Much of the forested interior is unsuited to agriculture and remains the domain of sparsely distributed nomadic hunter-gatherers known as “Punan” [60]. The Punan often formed close associations with cultivators such as the Merap, who found the Punan useful as a means to exploit and guard resources in their territory [60]. While the Punan had been familiar with agriculture for centuries, and often encouraged fruit trees around their camps, few developed significant gardens until the last three decades when the government encouraged them to settle [47].

Oxisols were found at 145 sites, Ultisols at 24, while Entisols, Alfisols and Inceptisols were observed at 13, 10 and 8 sites, respectively. We found no Andisols. Seventeen sites showed poor drainage and associated high levels of C (and are excluded when analyses refer to “dryland”). In the remaining 183 sites, soil nutrient levels were consistently very low, with low pH, limited CEC, low organic C, P, and base saturation. Soils showed marked acidity. Half of the forest locations had a pH-water lower than 4.4. Ratios of Calcium (Ca) to Aluminum (Al) (i.e., Ca²⁺/Al³⁺) were low, with median values of 0.15; 0.09; 0.23 for the Oxisols, Ultisols, and Inceptisols respectively, suggesting possible aluminum (Al) toxicity [61]. The best soils, in terms of nutrients required for plant growth, occurred on flood plain sites and in swampy areas. These soils were often visibly darker and richer in organic matter. Both Merap and Punan expressed a preference for these dark soils for agriculture compared to the more common red and yellow soils.

Local informants had a rich vocabulary and understanding of soil properties and agricultural potential. As might be expected from their contrasting agricultural and hunter-gatherer cultures, the Merap had a richer vocabulary and conceptual framework for describing soils than the Punan. The Merap used 20 terms to describe 79 plots, while the Punan used 14 terms to describe 84 plots (other plots were excluded as Kenyah and Putuk language terms were used, for a more detailed account of local soil terminology and criteria according to Malinau’s Merap and Punan people see [62]). All seven communities considered black or dark soils (tiem in Merap, punyuh in Punan) the best for agriculture, even though these soils were rare and localized. In our initial survey 43 out of 200 sites were identified by local residents as such “black soil(s)”. Of these, seven were on poorly drained sites.

Relative levels of soil C in the dryland sites were well predicted by soil texture (Pearson’s coefficient = 0.407, n = 183, p < 0.001) but were lower than expected (mean = 0.57% suggesting that topsoils in Malinau on average have lower soil C after correction for texture, than the soils in Sumatra where the equations were calibrated). There is no marked difference between forest and other land-use types, or between dark and non-dark soils (see Figure 1).

Black soils made up 36 of 183 (19%) of the dryland sites (Table 1). Black soil sites were commonly associated with human modified sites and comparatively scarce in old-growth forest sites (a null assumption that these frequencies are equal gives a χ² = 4.72, p = 0.030). Among the old village sites nearly half were black soils (the likelihood that this is the same proportion as for forest sites, χ² = 10.65, p = 0.0011).

Two soils possessed high P (Table 1). Both were from relatively flat riverside locations and both surpassed the criterion of ≥65 mg kg⁻¹ elemental P for ADEs in the Amazon [21]. [Note that Kämpf et al. [21] used Mehlich-1 (a standard approach in Brazil), whereas we used Bray-1 (the standard approach in Indonesia where Mehlich-1 is not used) for P assessment. These methods give similar but not equivalent results. One study in an Amazonian oxisol suggests that Bray-1 can yield slightly higher estimates than Mehlich-1 [63]. Because both soils had much higher P-levels than mentioned by Kämpf et al. [21], we suggest they fit with the criterion of anthrosols]. One site in the territory of Gong Solok village (P concentration = 92.4 mg kg⁻¹ at 0–20 cm) was under a fruit tree garden and recognized as “tana tiem” or “black soil” in Merap. The other old village site in the territory of Laban Nyarit village (P = 206.1 mg kg⁻¹ at 0–20 cm), was also in a fruit garden and included both black and non-black soils, and was called “tano bulah” or “mixed soil” in Punan. The Laban Nyarit site possessed higher P-levels at 20–40 cm (418.7 mg kg⁻¹), suggesting ancient P-enrichment, while the Gong Solok site has lower deep P and is likely less ancient (P = 59.8 mg kg⁻¹ at 20–40 cm). Selecting seven other sites in similar riverside settings judged at roughly similar likelihood of flooding (at least 10 m but not more than 50 m from any river and all on level ground a few meters above river levels) revealed much lower levels of available P (mean 8.1 mg kg⁻¹ at 0–20 cm), not different from the non-black soils. These localized black soils cannot be classified as Hortic Anthrosols as they lack the obvious hortic, irragric, plaggic or terric horizons required by the WRB soil classification system [64]. It remains striking, though, that farmers classify and value these black soils, even though our soil chemical analysis did not indicate any special properties.

Carbon levels and texture (clay and sand) of black soils were not significantly different from those of the other soils. The two sites with high P levels also did not show elevated levels of organic matter. Charred material was uncommon in the soil profiles. Char was seen in only 12 soil samples only four of which were black soils. The recognizable char observed in these profiles comprised small crumb-like grains and particles occasionally reaching 10 mm in diameter. These appeared to be charred wood. Some pieces were soft and brittle, while others were hard, and all shed black-dust. Char was observed in both fruit orchards with high P (Gong Solok and Laban Nyarit). All three primary forest sites that showed char were on hills near Long-Jalan—but none appeared to be black soils. Although char levels were not quantified, observations that neither C levels nor CEC were elevated in the two high P fruit orchard sites suggest only limited char deposition. Calcium levels were also not elevated in either site (Table 1).

With guidance from local respondents we selected three additional well-drained black soil sites in three communities (Table 2). Distance from the Malinau River ranged from 200 to 300 m and the three sites were not considered at risk of regular flooding. Observations of the profiles, sample locations and the wider vicinity indicated that local soils were neither Histosols nor Andisols. Each was under dryland rice cultivation at the time of sampling. All three sites, including the surroundings, had been cleared and burned in the previous 4 to 6 months, but ash was no longer visible on the soil surface. The Gong Solok site had been used as a trial coffee garden in 1981 before being converted to rice. The Lio Mutai and Long Jalan sites (far upriver—in areas traditionally used by both Punan and Merap) had been cultivated as dryland rice swidden fields in the past, but a full history was unavailable. Local people indicated that no chemical fertilizer had been used on these sites. We compared five paired soil samples for each site (black versus non-black).

All three sites had somewhat elevated P levels (Table 3), although one site was below the 200 mg kg⁻¹ P criterion [21]. The black soils at two sites had significantly higher C levels (nearly an order of magnitude more abundant) than the surrounding soils. Soil C:P ratios were around 500–600 for black soils and around 150–200 for the adjacent soils in two sites, suggesting that the increased C levels were due to inputs with very high C:P ratios (e.g., wood). On the other hand soil C:P ratios were 100 in black soils in the third site (Gong Solok) and around 200 for the adjacent soil (see Table 3) suggesting input of P-rich materials (possibly guano, and see discussion below). In that site a major part of the P was associated with the Ca-P fraction, rather than as iron (Fe) and Al (hydr-)oxides as in the two other sites (Figure 2).

Our first survey indicated that most soils were Oxisols with low levels of organic matter and plant nutrients. According to local classifications around 20% of our dryland sites possessed black/dark soils, most of which appeared unexceptional. Nonetheless two had distinctly elevated P levels (with unremarkable C levels). The second survey, which focused on three dark agriculturally productive sites, found that two possessed markedly higher P and C levels than surrounding soils.

Various characteristics of these four sites (two each from the first and second survey) are associated with ADEs: color, elevated levels of C and P, river-side locations, spatial extent, farmer-preferences and apparent extended human presence. Most strikingly, available soil P was substantially above the 65 mg kg⁻¹ in the two sites from the first survey, and total soil P was about 200 mg kg⁻¹ in the two sites from the second survey [21]. If these sites are anthropogenic should they also be considered as Terra Preta? We recognize several arguments against this:

Terra Preta soils possess significant charred organic matter. In our Borneo sites direct observations of char were limited—again, though this in itself is not conclusive, and further study is required [65].

Terra Preta sites are characterized by high HA:FA ratios and consequently high CEC. The two sites in our first survey did not have elevated CEC. Similarly Terra Preta sites have high levels of base cations and high base saturation. Again, this was not observed except in Gong Solok, with its high levels of calcium phosphates (Figure 2).

Terra Preta, like many Anthrosols [66], is characterized by high available (Mehlich-1, Bray-1) and total P, often in association with Ca, due to input from bones or guano. Two of our sites showed elevated available P (one also at depths below 20 cm, suggesting P enrichment took place a longer time ago), and two sites had total P consistent with the criterion of 200 mg kg⁻¹. However, Kämpf et al. [21] used that criterion as a substitute only in cases where available P pool (was depleted by recent crop production. Sites in the second survey were only recently under intensive agriculture, so the application of that criterion (based on the consideration that P was only assessed that way) remains questionable. Classification of the black soils from the second survey as anthropogenic, though plausible, remains unproven.

Terra Preta soils are often associated with ceramics and other artifacts. Ceramics have not been found in our Malinau sites. This result is inconclusive due to the recognized differences in material cultures between regions. In contrast to Amazonia buried ceramics of any kind are, with specific exceptions, scarce precious imports in Malinau [67].

Sombroek (1966: 175) [68] proposed to subdivide the Amazonian Dark Earths into Terra Preta proper (black earth) and Terra Mulata (brown earth). Terra Mulata soils generally lack human artifacts and are of a lighter color. Others have expanded upon that distinction and suggested different origins for these soils—with Terra Preta being kitchen refuse “middens” and Terra Mulata being agricultural soils [69]. These distinctions remains somewhat theoretical as Amazonian Dark Earths reveal a continuum of intermediate forms [70]. Malinau’s soils cannot (yet) be accommodated in these schemes—more important for now are their origins.

Contemporary farmers seek out dark soils [62]. The fact that eight of 17 samples from abandoned villages assessed during our first survey were black soils also indicates some relationship with human settlements.

While sustained human presence and cultivation were previously considered as prerequisites for the generation of ADEs, evidence from southern Africa and temperate Australia indicates that even non-agricultural aboriginal peoples can create such soils [31,71]. We require an open minded examination of soil formation processes and outcomes.

We are not in a position to judge the degree to which our P rich Borneo soils are a result of simple agricultural practices or other aspects of human presence (e.g., depositional processes associated with middens) or both. The variation in soil chemical signatures suggests mixed histories of agricultural production and settlement, and a possible continuum between slash-and-burn and slash-and-char practices. For the agricultural sites, rather than an absolute contrast between the char and burn practices, we speculate that practices have resulted in partial burns that added both ash and char to soils (see the woody debris in Figure 3). Thus repeated burning of fallows generates darker soils with enhanced fertility and production. Such sites could contribute to more sedentary life styles. Soil heating, such as occurs with burning, can bring about other soil changes that remain incompletely understood [25,72]. In any case manure and other waste inputs would add P. Manure’s low N:P ratio allows accumulation of P. Such practices may have occurred in the past, as evidenced by the very high P levels in the 20–40 cm soil layer at one site in Laban Nyarit, and may still be occurring. An alternative and not necessarily exclusive hypothesis, particularly for sites high in both P and Ca, is that they result from human settlements.

The occurrence of dark soils on steep slopes (the Lio Mutai and Long Jalan sites in our second survey) is notable. This is unlike anything reported from the Amazon region and may relate to the presence of hill-top settlements in the past—though this requires investigation. None of the soils in the first or second survey were Histosols or Andisols—this makes it less plausible that high P or organic matter might have a natural origin.

We do not believe natural factors explain Malinau’s P rich black soils. Similar to Terra Preta, natural explanations for such soils have been proposed that suggest that people preferentially settled at such sites rather than improved them (Katzer, 1903: 68) [74]. Carbon rich peat soils do not currently occur in the Malinau Valley. Peats are generally low in mineral nutrients, do not form or persist in well-drained locations, and they decompose (and burn) when subjected to drainage, cultivation and fire [75,76]. Swampy organic-carbon rich sites may have been drained by shifting river courses or the geological warping and up-thrust that have occurred across the region [77] but we would not expect such organic matter to persist in aerated and drained soils under cultivation. We also excluded sites with poor drainage from our evaluation of C-rich or P-rich soils. We therefore reject explanations that such soils are natural rather than human-made. For similar reasons we consider it unlikely that organic matter has been carried from poorly drained areas and applied to the fields: such material would have low persistence, would not explain the chemical attributes (P enrichment), and would require labor inputs we judge unlikely, especially given the lack of such practices in the region, though such practices in the past cannot be excluded.

Naturally occurring vegetation fires are unlikely to be localised to restricted riverside sites. We conclude that if burning has played a role in generating these soils it is predominantly human in origin. It is worth noting that, according to our data, land use has had little obvious impact on organic carbon (C_org) [78]. If farmers select sites with higher C_org this may somewhat balance any reduction caused by forest conversion but there is also a possibility that, as seen in anthropogenic soils elsewhere, such effects may not always be negative.

We can speculate on the factors leading to specific soils. Some sites with increased P-levels also had high C suggesting the addition of organic material. Wood addition would have resulted, at least initially, in a very high C:P ratio, and in two sites the black soils had a C:P ratio that was around three times as high as the adjacent soils. In contrast, the black soils at the third site (Gong Solok) had a C:P ratio that was lower than that of the adjacent soil, suggesting addition of P-rich material. Refuse from a settlement would be one possible explanation. There are caves in the wider area making guano another possibility that could also explain the high levels of Ca in this soil (Table 2). In the first survey, the high P-levels in two sites suggest other origins, as they were not accompanied by higher C or Ca levels.

High P can occur in locations possessing an underlying phosphate-rich geology, but none of our observations suggest such an origin. There are no active volcanoes in Borneo [79] and several of our sites are on raised ground. Work on repeated swidden cycles elsewhere in Kalimantan indicates that under some circumstances, P availability increases, presumably related to deep rooting of the fallow vegetation and increased storage in organic matter [25], but such increases are far less than in the two soils of the first survey. The likely explanation is that the P results directly from human activity.

Whatever the sources of nutrients, there are potentials for feedback loops. Farmers in Malinau select cultivation sites based on soil color, surrounding vegetation, and “soil stickiness” [62]. Soils that are favored and further cultivated may improve over time if conditions are right and char accumulates. Furthermore, once a site has been cleared for cultivation it is more likely to be cleared and cultivated again. Clearing of fallows requires less labor than the clearing of intact rain forest (likely even more critical prior to iron-tools) and working close to settlements was safer in the days of inter-villages warfare [80].

Local people in Malinau, as in the Amazon, are unaware of how locally valued anthropogenic soils have formed. Though Borneo never suffered the catastrophic depopulation and cultural loss that occurred following European contact in the Americas, local continuity is lacking in Malinau. Large-scale population movements, limited archaeological information and changing land-use practices limit our ability to describe local cultural histories in any detail.

Modern evidence of former human presence is seen in the fruiting and other useful trees in both the Amazon and in Malinau [22,54]. In Malinau, groves of fruit trees are found associated with abandoned village and camp sites often on ridges or mountain saddles inland from the main rivers. These sites are estimated to be as old as 250 years, a result consistent with oral histories and genealogical data, though the age of the trees themselves have yet to be determined. There are also old trees in and around extant villages and regular forest camps used for hunting and gathering trips, which indicate an extended human presence of at least a century and perhaps longer [54]. In any case, we are confident that farming people, though not necessarily rice farmers, have lived in Malinau for several centuries [51].

Our headwater soil sites are especially interesting due to their hill-slope locations. The short history of Punan cultivation makes it very unlikely that they are responsible if cultivation is the cause. More likely would be peoples whose lifestyles are more clearly determined by warfare and localized cultivation. While most Punan now farm to some degree, until recent decades most lived as hunter-gatherers, also trading forest products for rice and other goods. The limited soil-centered vocabulary of the Punan when compared to the Merap reflects their relative inexperience with agriculture. Merap judgments of soil properties are substantially closer to laboratory based measurements than those of the Punan [62]. The historical role of the Merap or other such farming peoples in these sites requires further exploration.

Could these soils result from some other practices? We know that Punan have maintained extensive Imperata grasslands by regular burning elsewhere in the region [54] but such practices seem unlikely to have contributed to localized P rich riverside black soils where such grasslands are absent. The ovens associated with Dark Earths in Australia resulted in a specific pattern of localized mounds and soil changes [71] that seem inapplicable in the Malinau context. Some combination of land preparation, cultivation, and waste disposal, whether intended or inadvertent, seems the most likely explanation for the Dark Earths in our Malinau sites.

Are Malinau’s dark anthropogenic soils relics as in the Amazon or are they still being created? Given how villagers clear, burn and char woody vegetation while preparing their gardens, and live nearby for years at a time, soil improvement and enrichment may be ongoing. Soils may be in various stages of transformation.

Research was carried out within the Malinau River catchment in Malinau District, East Kalimantan, Indonesia (Figure 4). This region possesses rugged terrain, few roads and is judged to have low potential for agricultural intensification [62]. The region escaped the droughts and forest fires of 1997–1998 [81,82] and there are no reports, recent or historical, of large forest fires [83]. The area maintains dense rich rain forest, with small areas of human cultivation and fallows. Soils are diverse but no deep peats occur [62].

Population densities in the upper Malinau are less than one person per square kilometer [55,84]. Local people use land for a combination of shifting cultivation, gardening, permanent fruit orchards, hunting, fishing and the collection and trade of forest products [62,85,86]. Upland rice is now the predominant staple [54,62]. Modern-day settlements are all next to moderately sediment-rich rivers and most permanent cultivation of gardens and fruit orchards (agroforests) occurs on the restricted areas of relatively level riverside land even though these locations are often vulnerable to floods that occur every few years [86,87]. Swidden rice fields are found along the river’s edge as well as on steeper slopes inland, often alongside smaller streams and tributaries. Many farmers cultivate larger areas on the better soils near the river and smaller “insurance areas” away from the river that are less vulnerable to flood damage [62].

When clearing fields, local farmers generally cut and gather smaller woody debris into piles before burning. These piles tend to be haphazard, though often placed around the base of larger trees and stumps or on the lower slopes of hillsides. Larger material is left where it falls. As there is no pronounced dry season, people burn as soon as the material is flammable due to worries that heavy rains may otherwise intervene. Burning is often incomplete. Sometimes unburned material is gathered into piles and burned a second time—even so pieces of charred green wood and roots, and charcoal survive (IB and DS unpublished observations and interview data). Woody roots in the soil profile are especially likely to char [72]. Rice planting involves making holes using pointed sticks. Rice is sowed into the holes which are then filled with loose surface material including ash and charcoal (Figure 4 [88]).

We conducted two surveys. The first explored the variation in soils, and people’s perceptions of them, with seven different communities across the upper half of the catchment of the Malinau River—an area of around 2,200 km². We did not explicitly seek dark or anthropogenic soils. The second survey focused on three selected sites with locally valued dark soils. In both surveys burned and charred material on the soil surface (generally crumbling fragments of blackened wood) was avoided during the sampling.