A “Dirty” Footprint: Macroinvertebrate diversity in Amazonian Anthropic Soils

Abstract Amazonian rainforests, once thought to be pristine wilderness, are increasingly known to have been widely inhabited, modified, and managed prior to European arrival, by human populations with diverse cultural backgrounds. Amazonian Dark Earths (ADEs) are fertile soils found throughout the Amazon Basin, created by pre‐Columbian societies with sedentary habits. Much is known about the chemistry of these soils, yet their zoology has been neglected. Hence, we characterized soil fertility, macroinvertebrate communities, and their activity at nine archeological sites in three Amazonian regions in ADEs and adjacent reference soils under native forest (young and old) and agricultural systems. We found 673 morphospecies and, despite similar richness in ADEs (385 spp.) and reference soils (399 spp.), we identified a tenacious pre‐Columbian footprint, with 49% of morphospecies found exclusively in ADEs. Termite and total macroinvertebrate abundance were higher in reference soils, while soil fertility and macroinvertebrate activity were higher in the ADEs, and associated with larger earthworm quantities and biomass. We show that ADE habitats have a unique pool of species, but that modern land use of ADEs decreases their populations, diversity, and contributions to soil functioning. These findings support the idea that humans created and sustained high‐fertility ecosystems that persist today, altering biodiversity patterns in Amazonia.


| INTRODUC TI ON
The Amazon basin still contains the largest continuous and relatively well-preserved tract of tropical forest on the planet. However, deforestation rates have been increasing over the last decade, resulting in the loss of an estimated 11.088 km 2 of natural vegetation in 2020 alone (INPE, 2021). Many forested areas have become highly fragmented and may be reaching tipping points where biodiversity and ecosystem functions may be dramatically affected (Barkhordarian et al., 2018;Decaëns et al., 2018), potentially leading to cascading effects that impact ecosystem functioning over a much larger area (Lathuillière et al., 2018;Lawrence & Vandecar, 2015).
But humans have been modifying Amazonian biodiversity patterns over millennia. Native Amazonians created areas with high concentrations of useful trees and hyperdominance of some species, often associated with archeological sites (Levis et al., 2017Ter Steege et al., 2013). Furthermore, occupations of some indigenous societies, beginning at least 6500 years ago, created fertile soils, locally called Amazonian Dark Earths (ADEs) or "Terra Preta de Índio" in Portuguese (Clement et al., 2015;Glaser, 2007;Glaser & Birk, 2012;McMichael et al., 2014;Watling et al., 2018;Figure 1b) that may occupy from 0.1 (Sombroek et al., 2003) up to 3% (McMichael et al., 2014) of the surface area of Amazonia. They appear to be more common along major rivers (Figure 1a) but are also abundant in interfluvial areas (Clement et al., 2015;Levis et al., 2020). ADE sites tend | 4577 DEMETRIO ET al. to have high contents of soil P, Ca, and pyrogenic-C (Glaser & Birk, 2012;Lima et al., 2002;Sombroek et al., 2003), and host particular communities of plants and soil microorganisms (Brossi et al., 2014;Taketani et al., 2013). However, up to now soil animal communities in these anthropic soils are practically unknown, having been the target of only three studies of limited geographic scope (all sites near Manaus), focusing on earthworms (Cunha et al., 2016) and soil arthropods (Sales et al., 2007;Soares et al., 2011).
Soil macroinvertebrates represent as much as 25% of overall known described species , and may easily surpass 1 million species worldwide . However, soil animal communities have been little studied in megadiverse regions such as the Amazonian rainforest (Barros et al., 2006;Franco et al., 2018;Marichal et al., 2014), and these habitats may be home to thousands of described and still undescribed species , particularly smaller invertebrates such as nematodes and mites (Franklin & Morais, 2006;Huang & Cares, 2006) but also macroinvertebrates (Mathieu, 2004). Furthermore, these invertebrates may be particularly susceptible to land-use changes such as deforestation Franco et al., 2018;Mathieu et al., 2005) and can be used as bioindicators of both soil quality and of environmental disturbance (Gerlach et al., 2013;Lawton et al., 1998;Rousseau et al., 2013;Velásquez & Lavelle, 2019).
Hence, the aim of this study was to assess soil invertebrate macrofauna communities and their activity in nine ADEs classified as Anthrosols and nine non-anthropic reference Amazonian Acrisols, Ferralsols and Plinthosols (referred to in this paper as REF soils) under three land-use systems (LUS: old and young secondary forest and recent agricultural/pastoral systems; Figure 1c), to evaluate anthropic effects on soil biodiversity. We predicted that (1) soil biodiversity and soil enrichment in anthropic soils would reflect a unique habitat (explained by a pre-Columbian footprint) but also that (2) animal species richness, biomass, and activity, as well as nutrient contents in these soils, would be determined by present-day land use.

| Study sites
Our study was performed in three regions (central, lower, and southwestern Amazonia) of Brazilian Amazonia, with sampling conducted in Iranduba county in central Amazonia, Belterra county in lower Amazonia, and Porto Velho in southwestern Amazonia ( Figure 1a; Table 1). All regions have a tropical monsoon (Köppen's Am) or without dry season (Köppen's Af) climate, with a mean annual temperature of 24-26.7°C and precipitation between 2000 and 2420 mm year −1 (Alvares et al., 2013). In each region, paired sites with ADEs and nearby non-anthropic REF soils (Figure 1b) were selected under different land-use systems ( Figure 1c): native secondary vegetation (dense ombrophilous forest) classified as old secondary forest when >20 years old, or young regeneration forest when <20 years old, and agricultural systems of maize in Iranduba, soybean in Belterra, and introduced pasture in Porto Velho. The REF sites were located within a minimum distance of 150 m (soybean at Belterra) to a maximum distance of 1.3 km (pasture at Porto Velho) from the ADE sites, and maximum distance between the three sampling locations within a region was 14 km (Embrapa sites to Tapajós National Forest sites in Belterra), totaling 18 sampled sites (3 regions × 3 land-use systems × 2 soil types).
One of the two old secondary forest sites in Belterra was at the Embrapa Amazônia Oriental Belterra Experiment Station, whereas F I G U R E 1 Sampling strategy to assess soil fauna and soil fertility in Central (Iranduba), Southwestern (Porto Velho), and Lower (Belterra) Amazon. (a) Boundary of Amazon Basin (white line), showing municipalities where samples were taken (boundaries in yellow lines), and areas with large occurrence of Amazonian Dark Earths (ADEs, shaded in green), modified from Clement et al. (2015). Amazonia map background: Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus DS, USDA, USGS, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community. (b) Soil profiles of analytically paired ADE and nearby reference (REF) soils. The direction of the arrow shows the increase in soil fertility; Photos G.C. Martins, R. Macedo. (c) Land-use systems sampled in each region, consisting in an intensification/ disturbance gradient including older secondary rainforest (>20 years, undisturbed), young regeneration forest (<20 years old), and recent agricultural systems (pasture, soybean, and maize). The direction of the arrow shows the increase in contemporary anthropogenic disturbance. Photos G.C. Martins, M. Bartz [Colour figure can be viewed at wileyonlinelibrary.com] the other one was at the Tapajós National Forest, a site of previous work on ADEs (Maezumi et al., 2018). The old secondary forests (ADE and REF) in Iranduba were at the Caldeirão Experimental Station of Embrapa Amazônia Ocidental and have been extensively studied in the past for soil fertility and pedogenesis (Alho et al., 2019;Macedo et al., 2017), as well as for soil microbial diversity (Germano et al., 2012;Grossman et al., 2010;Lima et al., 2014Lima et al., , 2015O'Neill et al., 2009;Taketani et al., 2013). Initial and partial results of the earthworm data from the young and old forests, and the maize fields in Iranduba, were presented in an earlier publication (Cunha et al., 2016). ADE formation in Iranduba was estimated to have begun ~1050-950 years bp (Macedo, 2014;Neves et al., 2004) and at Belterra ~530-450 years bp (Maezumi et al., 2018). At Porto Velho, ADE formation began much earlier (~6500 years bp; Watling et al., 2018).
The agricultural fields with annual crops were under continuous (at least 6 years) annual row cropping of maize (Iranduba) and soybean (Belterra) and had been planted <60 days prior to sampling, using conventional tillage (Iranduba) or reduced tillage (Belterra).
The crops received the recommended doses of inorganic fertilizers and pesticides for each crop; all crops were planted using certified commercial seeds. The pastures at Porto Velho were around 9 year TA B L E 1 Land-use system, age of modern human intervention, soil type, and soil category according to IUSS (2015) and location of the sites studied in three regions of Brazilian Amazonia

| Soil macroinvertebrate sampling
We performed field sampling in April (Iranduba) and May ( immediately fixed in 92% ethanol. Earthworms, ants, and termites were identified to species or morphologically different morphospecies (generally with genus-level assignations) by co-authors SWJ and MLCB (earthworms), ACF and RMF (ants), and ANSA (termites), while the remaining macroinvertebrates were sorted into morphospecies within higher taxonomic level assignations (e.g., order and/or family).

| Additional samples for ecosystem engineers
As ecosystem engineers (earthworms, termites, and ants) represent most of the soil macrofauna collected in Amazonian soils (Barros et al., 2006), and we expected them to also be important at the study sites, we performed additional sampling for earthworms, termites, and ants, in order to better estimate their species richness, and termites) were identified to species level or morphospecies level by co-authors as described above.

| Soil physical and chemical attributes
After hand-sorting the soil fauna from each TSBF monolith, 2-to 3-kg soil samples were collected from each depth (0-10, 10-20, and 20-30 cm) for chemical and soil particle size analysis, and although analyzed separately, mean values were calculated over 0-30 cm depth. The following soil properties were assessed using standard methodologies : pH (CaCl 2 ); Ca 2+ , Mg 2+ , and Al 3+ (KCl 1 mol L −1 ); K + and available P (Mehlich-1); total nitrogen (TN) and carbon (TC) by combustion (CNHS). Base saturation, sum of bases (SB) and cation exchange capacity (CEC) were calculated using standard formulae . Soil texture was determined according to the FAO soil texture triangle and the particle size fractions (% sand, silt, and clay) obtained following standard methodologies .
To assess functional differences induced by soil fauna activity in the ADE and REF soils, soil macromorphology samples were taken 2 m away from each monolith ( Figure S1) using a 10 × 10 × 10 cm metal frame. The collected material was separated into different fractions including living invertebrates, litter, roots, pebbles, pottery sherds, charcoal (biochar) fragments, non-aggregated/loose soil, physical aggregates, root-associated aggregates, and faunaproduced aggregates (generally with rounded shapes and darker color than other aggregates) using the method of Velásquez et al.
The earthworms, ants, and termites were also combined into the category of ecosystem engineers (Jones et al., 1994;Lavelle et al., 1997). To calculate the beta (β) diversity index, we removed singleton species (species represented by single individuals, i.e., one individual among all the 8378 individuals collected).

| Statistical analyses
To compare species richness between ADE and REF, we plotted rarefaction and extrapolation curves based on the Chao1 index (Chao, 1984) using the iNEXT package (Hsieh et al., 2016) for total macroinvertebrate, ant, termite, and earthworm morphospecies diversity, using the number of TSBF monolith samples as a measure of sampling effort intensity. The same procedure was used for all earthworm data (9 samples per site), termite data obtained from both the 20 m 2 plots and TSBF monoliths, and ant data obtained from both pitfall traps and TSBF monoliths. Confidence intervals for rarefaction and extrapolation curves were obtained by running a bootstrapping procedure (999 iterations).

| ADEs are distinct ecosystems
The ADEs at all the sites had higher soil pH ( throughout Amazonia Sombroek et al., 2003).
Significantly lower amounts of exchangeable Al were also found in the ADEs (Figure 2f). Soil texture was similar in both ADE and REF soils from each site (Table S1), so the enrichment was not due to differences in clay contents, but was the result of ancient anthropogenic activities Smith, 1980). Some differences in soil fertility among land-use systems were also observed, where plots under annual cropping or pasture use in REF soils had higher Ca and Mg contents (due to liming) than the young regeneration forests (Figure 2b,c), as well as higher K contents and base saturation than in both young and old secondary forests (Table S1) due to fertilization. Total C and N contents were higher in young regeneration forests than in agricultural systems and old forests on both ADEs and REF soils (Figure 2e; Table S1), owing probably to high organic matter deposition in these rapidly regenerating young forests.
Species richness overall was similar in ADEs (385 spp.) and REF (399 spp.) soils, but more species were found in Belterra (314 spp., where two old forests were sampled) than in Porto Velho (238 spp., where both forests were young) and Iranduba (218 spp.). More than 50% of all morphospecies were present in old forests, compared with lower and much lower proportions, respectively, in young regeneration forests and agricultural systems (Figure 3n). From all the monoliths, total species richness of ants, earthworms, spiders, beetles, true bugs, cockroaches, and isopods was also fairly similar in each soil type (Figure 3a,b,d,e,g-i), but termite richness was much higher, and centipede and opilionid richness slightly higher, in REF than in ADE soils (Figure 3c,j,k). On the other hand, richness of both millipedes and snails was higher in ADE than REF soils (Figure 3f,l), possibly owing to the higher soil Ca levels found in ADEs (Figure 2b; Coleman et al., 2004).
The proportion of exclusive morphospecies was high in both soils: 49% in ADEs and 51% in REF soils (Figure 3n) Figure 3o).
Many more species of termites and opilionids were unique to REF soils (24 and 12 spp., respectively) than to ADE soils (5 and 7 spp., respectively), while many more species of millipedes and snails were unique to ADE soils (28 and 10 spp., respectively) than to REF soils (16 and 5 spp., respectively). These trends for ants, earthworms, and termites remained similar even after singleton species were removed (Table S2). Furthermore, among the ecosystem engineers collected, we found a considerable number of species new to science (>20 earthworm, >20 termite, and >30 ant species) that still must be formally described.
ADEs were home to 52 rare (which include doubletons and morphospecies with fewer than 10 ind. over all samples) and to 21 nonrare or abundant macroinvertebrate morphospecies (taxa with ≥10 ind. over all samples) not found in REF soils (Table S2). Interestingly, within the non-rare/abundant taxa, 16 species (of which seven were of ants and five were of earthworms) had greater abundance of individuals in ADEs, while 14 species (half of them ant species) were more abundant in REF soils (Table S2). Overall, very few species were shared between the paired ADE and REF soils at each site, with many species unique to each soil type ( Figure S2).
Based on our results from the monolith samples (n = 45 for each soil type), estimated richness (i.e., that would have been obtained with increased sampling effort) for total macroinvertebrates, for ants, and for earthworms (Figure 4a,b, Figure 4c), and predicted to be attained with 300 samples, that is, more than three times the present sampling effort (90 samples).
These results were confirmed with the additional samples taken for ants, termites, and earthworms, which showed little difference between soil types in the increase in richness of ants and earthworms compared to the monoliths, but large differences for termites The high number of species unique to each soil was reflected in high β-diversity values and species turnover, ranging from 66% to 87% for all of the soil macroinvertebrates, depending on the region, LUS and soil type (Table 2). Interestingly, land-use effects on macroinvertebrate species turnover rates were slightly higher than those of soil type, indicating that species turnover was more affected by land-use change than by soil type (Table 2). Similar results were observed for earthworms, with much higher turnover rates (0.85 and 0.65 within REF and ADEs, respectively) due to LUS than due to soil type, particularly in old secondary forests. Conversely, soil type had a greater impact than land use on termite species turnover, while for ants, the effect of soil type on species turnover was mainly observed in old secondary forests. The species turnover among regions was also very high, especially for overall macroinvertebrates (all taxa) and for earthworms in both soils, implying a high number of macroinvertebrate species (and earthworms) locally endemic to different parts of Amazonia (Table 2). For ants, species turnover was higher in both forest types than in the agricultural systems, implying that agricultural systems include a larger proportion of widespread species common to all three sampling regions. and young forests or agricultural systems on ADEs (Figure 5d,g).

| Ecosystem engineers dominate the soil fauna communities
Also, the abundance of millipedes was higher in young regeneration forests on ADEs than on REF soils (Figure 5e).
Ecosystem engineers represented from 65% to 94% of total soil fauna biomass, with earthworms being the largest component, representing 61%-99% of the engineer biomass and 44%-92% of the total macroinvertebrate biomass (Table S3). In both agricultural systems and in the young regeneration forests, earthworm biomass was higher on ADEs than on REF soils. Furthermore, in the young regeneration forests, ecosystem engineer, millipede, other and total macrofauna biomass were also significantly higher on ADEs than on REF soils (Table S3). On the other hand, in all LUS, termite biomass was significantly higher on REF soils than on ADEs. No other higher taxon of soil animals represented more than 16% of the total macroinvertebrate biomass in any given soil type or LUS (Table S3).

| Soil biota influence ADE soil properties
Soil macromorphology revealed a significantly higher proportion of fauna-produced aggregates in ADEs compared to REF soils

| Modern land use erodes soil biodiversity and function
Modern agricultural systems had lower richness of all major soil animal taxa (except for true bugs and snails in REF soils and beetles in ADEs;   Region: Mean regional effect, presented for each soil type and calculated by averaging all turnovers for each LUS, tested between regions (e.g., old forest at Iranduba vs. old forests at Belterra on REF soil). 2 LUS: Mean effect of all differences in land-use systems, presented for each soil type and within each region, and then averaged across all regions (e.g., both young forests compared with pasture at Porto Velho).   (Barros et al., 2006;Mathieu, 2004;Mathieu et al., 2005).

| DISCUSS ION
We also found that although species richness was similar in ADE and REF soils, these two habitats harbor very different species pools, with few found in both habitats (Figure 3; Figure S2). This high turnover between sites and number of unique species appears to be a prevalent feature of Amazonian rainforest invertebrate communities (Maggia et al., 2021;Mathieu, 2004;Vasconcelos, 2006). Furthermore, although species rarefaction curves were still far from saturation with our current sampling effort, estimated richness showed similar trends, and showcased the wealth of species still to be discovered in both soils (Figure 4).
We believe that anthropic soils represent a major gap in the knowledge of Amazonian biodiversity. Soil animals have been poorly represented in taxonomic surveys in Amazonia (Constantino & Acioli, 2006;Franklin & Morais, 2006;James & Brown, 2006;Vasconcelos, 2006), and ADEs had not previously been sampled for soil macrofauna to this extent. Although ADEs occupy only a small fraction (0.1%-3%) of the Amazonian surface area (McMichael et al., 2014;Sombroek et al., 2003), they are scattered throughout the region (Clement et al., 2015;Kern et al., 2017), representing thousands of localized special habitats for species. The high β-diversity values and species turnovers between different ADEs mean that each of these patches may be home to distinctive soil animal communities, including many new species, judging by the number of new ecosystem engineers found.
Soil provides chemical and physical support for vegetation and, as millennia of human activities created ADEs in the Amazon, patches with higher amounts of nutrients and organic resources were generated throughout a matrix of poorer soils (Kern et al., 2017;Macedo et al., 2019). The formation processes and human management of these soils result in distinct plant and microbial communities (Brossi et al., 2014;Clement et al., 2015;Levis et al., 2018;Taketani & Tsai, 2010), that are a result of disturbance, soil enrichment, and selection processes (both natural and humandriven). Here we show that current soil animal abundance and diversity also reflect the impact of these ancient anthropogenic activities. The ADEs developed a different pool of species compared with REF soils. The former soils tend to favor more animals that recycle organic matter and flourish with higher pH and soil Ca, like earthworms and millipedes, while the latter favor termites, which are particularly sensitive to deforestation and changes in soil moisture and physical conditions (Dambros et al., 2013;de Souza & Brown, 1994;Duran-Bautista, Muñoz Chilatra, et al., 2020;Eggleton et al., 1996). The functional particularities observed in biotic communities of ADEs also mean that ecosystem functioning could be different in these soils, which could imply differences in their ecosystem services to humans, as observed in other human-altered landscapes in Amazonia (Marichal et al., 2014;Rodríguez et al., 2021;Velásquez & Lavelle, 2019 service delivery, on the other, have been mostly indirect (correlation rather than causation), it is well known that larger earthworm populations and improved soil structure owing mainly to fauna-produced aggregates (as occurs in ADE) can alter soil hydraulic properties Hallaire et al., 2000), primary productivity (Brown et al., 1999;Pashanasi et al., 1996), litter decomposition, and nutrient cycling  as well as pedogenetic processes (Cunha et al., 2016;Macedo et al., 2017), and could help stabilize organic carbon in these soils (Cunha et al., 2016;Ponge et al., 2006 Brasil, 1961), but throughout Amazonia they are actively sought out and intensively used for agricultural and horticultural purposes (Fraser et al., 2011;Junqueira et al., 2016;Kern et al., 2017). Intensive annual cropping and extensive livestock production represent a threat to soil macrofauna populations, both in REF and in ADE soils. Macroinvertebrate diversity in both soils decreased dramatically with increasing environmental disturbance (Figures 3 and 5), and negative impacts on some macroinvertebrate populations were higher in ADE than in REF soils. Modern human activity is often associated with negative environmental impacts in the Amazon Franco et al., 2018), but on the other hand, the Pre-Columbian historical human footprints associated with ADE formation processes and their long-term traditional use appear to have "positive" effects on the Amazonian ecosystem (Balée, 2010). For instance, we found that old forests on ADEs were the most diverse LUS in terms of total soil macroinvertebrate morphospecies, and have also been shown to contain numerous useful tree and palm species (Levis et al., 2017. Soil invertebrates are known to display high endemism (Lavelle & Lapied, 2003), and hence high β-diversity values, mainly due to their low dispersal ability (Wu et al., 2011). Still, the high turnover rates between communities of ADE and REF soils suggest that ADEs may represent refuges for large numbers of specialist species that have been overlooked in previous work in the region (Barros et al., 2006;Constantino & Acioli, 2006;Franco et al., 2018;Vasconcelos, 2006), which has not targeted ADEs. This persistent anthropogenic footprint promotes biodiversity (Balée, 2010;Heckenberger et al., 2007) and modifies its distribution patterns in the Amazonian basin, show- special attention and management, to discover and protect their biological resources and promote more sustainable uses of Amazonian soils (Glaser, 2007).

ACK N OWLED G EM ENTS
The study was supported by the Newton Fund and Fundação 18131-6 for Tapajós National Forest was granted by ICMBio.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interests.