Science of the Total Environment 856 (2023) 158955

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Soil type determines the magnitude of soil fertility changes by forest-to-

pasture conversion in Western Amazonia
⁎
Fernando Igne Rocha a,b, , Ederson da Conceição Jesus b, , Wenceslau Geraldes Teixeira c,
⁎⁎
José Francisco Lumbreras c, Eliane de Paula Clemente c, Paulo Emilio Ferreira da Motta c, Ana Carolina Borsanelli d,
Iveraldo dos Santos Dutra e, Aline Pacobahyba de Oliveira c
a
Department of Soil Science, Federal Rural University of Rio de Janeiro (UFRRJ), Seropédica, RJ, Brazil
b
National Agrobiology Research Center, Embrapa Agrobiologia, Seropédica, RJ, Brazil
c
National Soil Research Center, Embrapa Solos, Rio de Janeiro, RJ, Brazil
d
Department of Veterinary Medicine, Federal University of Goiás (UFG), Goiânia, Brazil
e
Departament of Production and Animal Health, São Paulo State University (UNESP), Araçatuba, Brazil

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Little is known about how land-use affects

soils of different pedogenesis.
• We covered a gradient of soil fertility
across the Western Amazonia.
• Soil type defined the magnitude of impact
of land use on soil fertility.
• The sum of bases was the most sensitive
indicator of land-use change.
• Actinobacteria:Proteobacteria ratio reflects
anthropogenic changes in soil fertility.

A R T I C L E I N F O A B S T R A C T

Editor: Manuel Esteban Lucas-Borja The deforestation of tropical forests raises environmental concerns worldwide. Removing the pristine forest impacts
the soil, consequently affecting the environmental services it provides. Within this context, the main goal of
Keywords: this study was to determine how the conversion of the tropical rainforest to pasture affects soil fertility across an
Pedodiversity extended range of soil heterogeneity, including different soil types. We sampled 13 sites, among forests, recent pastures
Amazonia
(≤7-year-old), and old pastures (≥10-year-old), on Acrisols, Ferralsols, Plinthosols, and Luvisols, across a ± 800 km
Bioindicator
Land-use change
geographical range in the Western Brazilian Amazon. Soils were classified taxonomically, and their superficial layer's
Land clearing chemical and physical properties (0–10 cm) were analyzed. Furthermore, we tested the sensibility of Actinobacteria and
Deforestation Proteobacteria to detect changes in these soil properties based on their ecological habitat. An inter-regional gradient of
soil fertility was observed, and the sampling sites were clustered mostly by soil type and associated land use than by
spatial distance. The Sum of bases, Ca + Mg, base saturation, Al saturation, and pH were consistently affected by
land use, increasing after conversion to pasture, at different degrees and with a more pronounced effect on oxidic
soils. The Sum of bases was the only property that increased significantly among the study sites (Radj = 0.860,
p < 0.001), being able to detect the effect of anthropic land use on a larger coverage of soil types. Finally, the
Actinobacteria:Proteobacteria ratio was also sensitive to the impact of forest-to-pasture conversion, with a higher ratio

⁎ Correspondence to: F.I. Rocha, Department of Soil Science, Federal Rural University of Rio de Janeiro (UFRRJ), Seropédica, RJ, Brazil.
⁎⁎ Correspondence to: E. da C. Jesus, Embrapa Agrobiologia, BR 465, km 7, s/n, Ecologia, Seropédica, 23891-000, RJ, Brazil.
E-mail addresses: fernando.igne@hotmail.com (F.I. Rocha), ederson.jesus@embrapa.br (E.C. Jesus).

http://dx.doi.org/10.1016/j.scitotenv.2022.158955
Received 31 July 2022; Received in revised form 18 September 2022; Accepted 19 September 2022
Available online 22 September 2022
0048-9697/© 2022 Elsevier B.V. All rights reserved.
F.I. Rocha et al. Science of the Total Environment 856 (2023) 158955

observed in pasture systems, and it was positively correlated with soil pH (rho = 0.469, p < 0.001). Our results con-
sistently show that the forest-to-pasture conversion leads to strong alterations in the soil environment, with varying
intensities depending on soil type.

1. Introduction 8°44′26″S, 67°23′3″W, elevation 99 m). and Manicoré (state of Amazonas,

5°48′34″S, 61°18′2″W, elevation 32 m). The region has an ‘Am,’ according
The soil as a natural resource performs essential services of environmen- to the Köppen system, with tropical monsoon rain and a dry period from
tal, economic, and sociocultural importance, such as nutrient cycling, nutri- June to August. The annual average rainfall varies between 2200 and
ent provision for plants, and carbon storage, besides being a biodiversity 2800 mm, and the average annual temperature varies between 24 and
repository (Wubie and Assen, 2020). Nevertheless, the indiscriminate 26 °C (Alvares et al., 2013). The soil parent materials are mixed-textured
development of human activities has threatened this resource, with delete- Tertiary and Quaternary fluvial sediments of Andean origin (Smyth, 1996).
rious consequences at the local and global scales (Delgado-Baquerizo et al., Samples for soil fertility and microbiological analysis were taken in
2020). In this context, tropical forests have been undergoing a process of August 2017 on five farms: two in Boca do Acre (BAC1 and BAC2), one in
biodiversity loss, with hitherto little-known effects on the soil ecosystem Bujari (BUJ), and two others in Manicoré (MAN1 and MAN2). Each of
(de Lima et al., 2020; Ponge, 2015). Furthermore, soils in these environ- these farms had a recent and old pasture (≤7 and ≥10-year-old, respec-
ments usually have poor fertility and low pH (Hartemink, 2002) and are tively) and an adjacent forest, except for Manicoré, whose areas are of
susceptible to degradation after deforestation due to their intensive use recent colonization, with no old pastures. Old pastures in BAC and BUJ,
and heavy rainfalls. Thus, understanding how land-use change affects soil as well as recent pastures in MAN, were formed after slash-and-burn,
properties in these biodiversity hotspots becomes vital to forecasting the although the fire has not been used to reform the pastures in BAC and
consequences of deforestation (Dias-Filho et al., 2001; Numata et al., 2007). BUJ. Young pastures in BAC and BUJ were formed without the use of
The Amazonia rainforest is one of these tropical hotspots and is consid- fire. The grasses found in young and old pastures are Panicum spp. And
ered one of the most oligotrophic forests in the world (Braz et al., 2013). Urochloa spp., respectively. In Amazonia, Panicum maximum is usually the
That condition is driven mainly by geological aspects associated with the first grass to be used, and it is substituted by Urochloa spp. in subsequent
high weathering degree of the predominant soils and nutrient leaching years, a practice that has been carried out since a long ago (Serrão et al.,
(Aprile et al., 2013). However, the conversion of the primary rainforest to 1979) and that continues in recent years (Zu Ermgassen et al., 2018). No
agriculture or pasture land affects soil fertility significantly, leading to a liming and fertilization have been carried out in any of these sites. Forest
significant increase in pH and a reduction of exchangeable aluminum con- sites have pristine native rainforest vegetation, except for one of the farms
centrations by the liming-effect of the deposited ashes from the combusted in BAC, where the adjacent forest was a secondary forest formed from
forest biomass (Alfaia et al., 2004; Moreira et al., 2011). This alteration is native vegetation.
not permanent, and soil fertility may be compromised in the long turn. A total of 13 sites were sampled, six in BAC, three in BUJ, and four in
Moreover, modifications in the natural plant cover are also reported to MAN. A transect was sampled in each of the 13 sampling sites. The transects
change the diversity, activity, functions, and biomass of microbial groups, were 200 m long with five points distant 50 m between them (Fig. S1).
which are the main factors responsible for releasing nutrients into the soil Three soil subsamples (0–10 cm) were taken in each of the five points
(Kaschuk et al., 2011), although it is generally time-dependent (Farella and mixed to generate a composite sample per point. See Table S1 for
et al., 2007), and variable by soil type (de Moraes et al., 1996; Moreira more details about landscape and sample conditions.
et al., 2011). Nonetheless, although this rainforest encompasses a large
diversity of soil types (Quesada et al., 2010), studies have been carried 2.2. Soil classification and characterization
out mainly on highly weathered soils such as Ferralsols and Acrisols,
which cover most of the Amazonia basin (Schaefer et al., 2017). Increasing A trench for soil classification was dug in each of the 13 sampling sites.
information on how less studied soil types, such as naturally fertile Luvisols, The morphological characterization and horizon soil sampling were carried
are affected by land-use conversion is vital to improve soil conservation out according to Santos et al. (2005), and the soil classification was made
practices and to assist decision-making. following the Brazilian System of Soil Classification (dos Santos et al.,
Therefore, the main goal of this study was to characterize how land use 2018) and the World Soil Reference Base (WRB, 2015). The soil physical
change in the Amazonia rainforest affects soil fertility across different soil and chemical analyzes were performed as described in the Manual of Soil
orders. Forests and pastures with varying ages since their conversion Analysis Methods (Teixeira et al., 2017), which are standard methods
were selected along a natural soil fertility gradient throughout the Western used in the laboratories of the National Center for Soil Research/Embrapa
Amazonia. We tested the hypothesis that the magnitude of changes in soil Soils, Rio de Janeiro, Brazil, where all the analyzes were performed. Soil
fertility properties due to land use varies with soil type. Additionally, we physical attributes (particle size distribution, clay dispersed in water, and
evaluated the Actinobacteria:Proteobacteria ratio as an indicator of these flocculation degree) were determined by the sedimentation method and
transformations. These two phyla are considered dominant in copiotrophic reading by densimeter from the sample dispersion with 0.1 mol L−1 sodium
(i.e., high nutrient availability) and oligotrophic (i.e., low nutrient avail- hydroxide solution. The chemical analyses consisted of pH in water and KCl
ability) habitats, respectively (Clivot et al., 1996; Fierer et al., 2012). We 1 mol L−1, determined potentiometrically, in the soil: 1:2.5 solution with
hypothesize that they are suitable to reflect changes in soil fertility due to 1 h of contact and agitation of the suspension at the time of reading.
these contrasting responses to nutrient availability. We took advantage of Exchangeable sodium and potassium (Na+ and K+) were extracted with
the ongoing deforestation scenario in the region as a possible indicator of HCl 0.5 mol L−1 + H2SO4 0.0125 mol L−1 (Mehlich−1) in the proportion
what might occur in other rainforest environments. of 1:10 and determined by photometry of flame emission. The measure-
ment of exchangeable calcium and magnesium (Ca2+ and Mg2+) was
2. Material and methods performed by atomic absorption spectroscopy and exchangeable aluminum
(Al3+) by titration after extraction with KCl 1 mol L−1 in the proportion of
2.1. Sampling and experimental design 1:10. The determination of potential acidity (H + Al) was carried out by
titration after extraction with calcium acetate 0.5 mol L−1 in the proportion
This study was carried out in the Brazilian Western Amazonia, within a 1:10 and pH 7.0. The organic carbon was determined by titration of the
geographical range of ±800 km, covering sites in Bujari (state of Acre, remaining potassium dichromate with ammoniacal ferrous sulfate after
9°49′22″S, 67°56′51″W, elevation 196 m), Boca do Acre (state of Amazonas, oxidation. The calculation of derived correlations, i.e., sum of bases

2
F.I. Rocha et al. Science of the Total Environment 856 (2023) 158955

Fig. 1. Ordination of soil variables across Amazonian sites and their altered land uses by the forest-to-pasture conversion. A) Principal component analysis (PCA) including all
observations among study sites in the Brazilian Western Amazonia and (B–F) individual PCAs for each one of the study localities displaying the five most important variables
based on their contribution value. BUJ: Bujari/state of Acre, BAC: Boca do Acre/state of Amazonas, including BAC1 and BAC2 localities, MAN: Manicoré/state of Amazonas,
including MAN1 and MAN2 localities.

(SB = Ca2+ + Mg2+ + K+), base saturation index [BS% = 100 × SB/ taxonomy was assigned for each amplicon sequence variant (ASV),
total cation exchange capacity (T-CEC)], and Al saturation index [m% = assessing the Silva taxonomic training (database v132) (Quast et al.,
(mmolc (Al3+) dm−3 × 100)/(mmolc (effective CEC) dm−3)]. Finally, 2012). As explained below, Actinobacteria and Proteobacteria sequences
the micronutrients Cu, Zn, Fe, and Mn were extracted with the Mehlich-1 were extracted from the dataset and used for further analyses. For details
solution (HCl 0.05 mol L−1 + H2SO4 0.0125 mol L−1) and determined about next-generation sequencing parameters, sequence datasets, accession
after the mechanical stirring of the filtrate solution for 5 min by spectrom- numbers, and bioinformatic pipelines, see Rocha et al. (2021). R packages'
etry of optical emission with inductively coupled plasma (ICP-OES). dada2' v.1.14.0 (Callahan et al., 2016) and ‘decipher’ v.2.14. (Wright et al.,
2012) were used in the R 3.6.1 environment (Team, 2018).
2.3. DNA extraction and sequencing analysis
2.4. Statistical analysis
We extracted soil DNA using the standard DNeasy PowerSoil kit proto-
col (MO BIO Laboratories Inc.). The amplification and sequencing of the A principal component analysis (PCA) on the correlation matrix was
16S rRNA gene were performed using barcoding DNA and the Illumina used to select the most important soil variables based on their component
MiSeq technology (Caporaso et al., 2012) at the Argonne National Lab loadings. We first performed a PCA using all sampling points and measured
Core Sequencing Facility, USA, following the Earth Microbiome Project variables. This PCA will be called “inter-regional PCA” since it compares
protocol. The sequence data were further processed, aligned, and catego- sites from the different sampled locations. Then, all variables with a contri-
rized using the DADA2 microbiome pipeline (https://github.com/ bution larger than the cutoff of 3.85 % [i.e., 100 × (1/26)] were selected
benjjneb/dada2) by recommended parameters with quality filtering of following Abdi and Williams (2010), where 26 is the number of considered
sequence length over 250 base pairs (Callahan et al., 2016). Further, the soil variables. The contribution of a variable for a given PC was obtained by

3
F.I. Rocha et al. Science of the Total Environment 856 (2023) 158955

Fig. 2. Trends in the gradient of soil fertility across land-use change profiles. A) Inter-regional (between sites) and land-use systems (within sites) comparisons; B) linear
correlation between PC1 scores for the overall PCA and total sum of bases (SB) (equation: y = −0.79 + 0.31x); localities were colored based in their site/municipality
(i.e., orange: BUJ, blue: BAC, gray: MAN). Statistical differences were determined by two-way ANOVA (p < 0.05). Significant by t-test at p < 0.001 (⁎⁎⁎), n.s. non-significant.
The fitted values for each model are represented by the black line and their standard errors are indicated by the shaded area. BUJ: Bujari/state of Acre, BAC: Boca do Acre/
state of Amazonas, including BAC1 and BAC2 localities, MAN: Manicoré/state of Amazonas, including MAN1 and MAN2 localities.

the ratio of the squared component loading of the variable by the eigen- BAC and MAN are predominantly Ferrasols, while other soil classes were
value associated with the PC. The variables of each study location with found in BUJ (Table S1). The PC1 scores for BUJ differed statistically in com-
the highest contributions to PC1 were merged into a single subset to be parison to the scores from other localities (Fig. 2A; χ2 = 37.59, p < 0.001),
further used in a two-way ANOVA to test for differences among sites and they were correlated with higher values of total sum of bases (Fig. 2B;
(inter-regional) and between land-uses within each site (intra-regional) Radj = 86 %, p < 0.001). Nevertheless, a clear land-use effect could be
after meeting the statistical assumptions for this analysis. The PCAs were observed within each of the studied localities (Fig. 1B to F). A higher fertility
conducted using ‘factoextra’ v.1.0.7 R package (Kassambara and Mundt, was observed in pasture soils (Figs. 1B to E; 2A), which has been demon-
2018). Finally, a heatmap was used to visualize possible clusters among strated for other pastures throughout the Amazonia region (Braz et al.,
samples within each factor (i.e., site, land-use, and soil classes) and 2013; Machado et al., 2017). A higher dissimilarity between pastures and for-
measured variables using ‘pheatmap’ v.1.0.12 R package. ests of low-fertility soils from BAC and MAN has also been observed, showing
To test their use as an ecological indicator, Actinobacteria and that the effects of forest-to-pasture conversion on soil fertility are more pro-
Proteobacteria had their abundances compared in STAMP v.3.0 (Parks nounced under this fertility level and, thus, dependent on the predominant
et al., 2014). The q-values were calculated using two-sided Welch's t-test soil type. Forest soils in BUJ have higher natural fertility, which shows the
with Benjamini–Hochberg false discovery rate corrections (Benjamini and clear influence of naturally fertile Luvisols. These soils are a patch of naturally
Hochberg, 1995). Spearman's rho test tested the correlations between the eutrophic soils (Bernini et al., 2013) and were formed predominantly from
relative abundances of these two phyla and between their ratio the weathering of sedimentary rocks generated by the Andean orogeny and
(Actinobacteria:Proteobacteria) and soil pH. the sediment flux into the lowland (Quesada et al., 2011). BAC and MAN
are predominantly covered with highly weathered soils, such as Acrisols
3. Results and discussion and Ferralsols, developed from sandstones and claystones, and mainly
formed in remnants of ferralitic and convex plateaus (Shinzato et al., 2015;
3.1. Natural and anthropogenic factors build the inter and intra-regional gradient Souza et al., 2018). Thus, the soil fertility reflects both the inherent pedoge-
of soil fertility netic characteristics as well as the processes mediated by land-use conversion,
which increases the dissimilarity among variables related to soil acidity
The PCA with all sites explained 59.4 % of the data variability (Fig. 1A). (i.e., H+, H + Al, m%, Al) and those related to high soil fertility, i.e., pH,
Despite their geographical proximity, BUJ and BAC were ordered separately. base saturation (BS%), Sum of bases (SB), and total cation exchange capacity
BAC showed more similarities to MAN, a result highlighting the predominant (T-CEC). It is important to highlight that the pastures have never been limed
influence of pedogenetic soil characteristics over land use to determine the or fertilized, which also explains that natural soil conditions have a greater
fertility of the soil's superficial layer under the studied conditions. Soils in influence on determining fertility levels.

4
F.I. Rocha et al. Science of the Total Environment 856 (2023) 158955

Fig. 3. Correlation heatmap of soil variables. Horizontal dendrograms gather correlated soil classes, sites, and land-uses, based on the soil variables clustering in the vertical axis.
BUJ: Bujari/state of Acre, BAC: Boca do Acre/state of Amazonas, including BAC1 and BAC2 localities, MAN: Manicoré/state of Amazonas, including MAN1 and MAN2 localities.

Fig. 4. Comparison between forests and pastures for selected soil variables. Soil variables (0–10 cm) were extracted as important in the principal component analysis among
different study localities in Western Amazonia. Significant differences between each land-use were determined by two-way ANOVA (p < 0.05); Significant by t-test at
p < 0.001 (⁎⁎⁎), p < 0.01 (⁎⁎), p < 0.05 (⁎), n.s. non-significant. Error bars indicate the ± standard error (SE) (n = 5). BUJ: Bujari/state of Acre, BAC1 and BAC2: Boca
do Acre/state of Amazonas, MAN1 and MAN2: Manicoré/state of Amazonas.

5
F.I. Rocha et al. Science of the Total Environment 856 (2023) 158955

The evaluated systems are divided into two groups, as shown by a

hierarchical dendrogram based on the fifteen most crucial soil variables
(Fig. 3). The first group was correlated with variables linked to higher soil
fertility (higher pH, Mn, BS%, SB, and Ca + Mg), clustering the BUJ's land
uses and their respective soil types (i.e., Luvisol, Stagnic Plinthosol, and
Xanthic Acrisol). The second cluster has an intermediate subgroup compris-
ing both a pasture and a forest from BAC, which were more positively corre-
lated with clay, C, N, and T-CEC, as well as soil acidity variables, i.e., H+,
H + Al, Al, and Aluminum saturation (m%). The largest subgroup includes
land uses from BAC and MAN, where the forests are positively correlated
with soil acidity variables and the C:N ratio, which were more pronounced
in those systems. In turn, the prominent influence of land management
decreased soil acidity variables in pastures, although those systems are on
weathered soils of the Amazonian region, as pointed out above.
A two-way ANOVA between forests and pastures shows that only the Ca
+ Mg and SB differ between forests and pastures in all study locations
(Fig. 4). Soil pH, m%, and BS% differed statistically between land uses in
BAC and MAN but not in BUJ. Mainly, the BUJ land uses showed the lowest
m% values among the studied locations. The Aluminum saturation in the
BUJ forest was 79 % lower than the average m% of forests in MAN (second
lowest m% among forests), and 12 % lower in BUJ pastures than MAN pas-
tures. The average values and deviations of the soil physical and chemical
variables for each land use demonstrate how soil formation processes
and, consequently, the different soil types are essential issues to be consid-
ered. Despite the high similarity between BUJ's land uses regarding soil
variables linked to soil fertility indices (i.e., pH, m%, and BS%), this does
not indicate that the effect of converting forest to pasture is unable to
exert drastic changes in edaphic properties. Instead, the results show only
that these changes are less pronounced due to the soil type, which gives
them a greater buffering capacity and natural fertility. In that case, a
more sensitive indicator of land-use change must be used to detect varia-
tions between soils under different land uses but with high base saturation
in the first horizon. As already stated, the Ca + Mg and SB were the best
indicators, differentiating all sites independent of their soil types.
Although the soil phosphorus (P) is recognized for its low mobility,
mostly in tropical soils, this variable was essential to differentiate BUJ's
land uses, being higher in pasture soils (forest: 2.0 mg kg−1, pasture: 5.8
± 3.18 mg kg−1; χ2 = 9.94, p < 0.01). Moreover, we considered the stan-
dardized effect size of the comparison between forest and pasture for the in-
dicators of soil fertility (i.e., H + Al, m%, pH, and SB) in each individual Fig. 5. Standardized effect size of the forest-to-pasture conversion on soil variables.
locality, as well as the overall effect among sites (Fig. 5). The analysis Coefficient estimates from linear models are plotted with 95 % confidence intervals,
demonstrated that the m% is representative of most of the forest systems, where the dots represent the standardized effect size for the model between land-
whereas the soil pH and SB are highly representative of pastures. The use and soil variables; The overall effect represent the magnitude of the effect
Sum of bases was the most sensitive variable to discriminate the effect of considering all study sites in the model; Statistical differences were determined by
two-way ANOVA (p < 0.05). Significant by t-test at p < 0.001 (⁎⁎⁎), p < 0.01 (⁎⁎),
land-use change at all sites, which suggests it as a suitable indicator of
p < 0.05 (⁎), n.s. non-significant; BUJ: Bujari/state of Acre, BAC1 and BAC2: Boca
these effects.
do Acre/state of Amazonas, MAN1 and MAN2: Manicoré/state of Amazonas.
The forest-to-pasture conversion also resulted in in-depth modifications
(Fig. S2, Table S3). Characterizing soil profiles allows us to understand
edaphic characteristics based on the description of each horizon, which soil base saturation among pastures over time. Moreover, the Al saturation
surpasses the understanding of changes in the composition of soil variables (m%) and exchangeable aluminum (H + Al) were lower in older pastures
only on the soil surface. Especially for m%, SB, and BS%, transformations (Figs. 6, S3).
between the relative comparison of horizons A and B were perceived The changes in the soil environment after forest-to-pasture conversion
among pasture soils, with a relatively higher proportion of the variables through slash-and-burn are commonly accompanied by the release of
related to the increased exchangeable cations in horizon A over B. The basic cations previously stored in the forest biomass, which increase the
total and saturation levels for aluminum (Al and m%, respectively) evi- soil pH and base saturation (Béliveau et al., 2015; Cochrane and Sanchez,
denced that for BUJ, regardless of land use, there is a considerable 1982). This effect has been thoroughly demonstrated in several studies be-
contrast between the two horizons, with values higher than 95 % on the cause of ash deposition, soil heating, and increased rates of organic matter
B horizon. This reflects the aluminic character of most of the soils in this decomposition (Palm et al., 1996; Smyth, 1996), and models explaining
study, corroborating previous reports for the state of Acre (Lips and these changes have been long reported (Giardina et al., 2000). However,
Duivenvoorden, 1996). A relative decrease from 97 % to 51 % in m% some authors consider the ash-deposition model an oversimplification of
between forest and pasture soil and from approximately 80 % to 40 % for what happens after slash-and-burn and proposed a more complex model
Al was observed for site MAN1. This may be due to the direct effect of in which other factors, such as soil heating, water erosion, and ash drift
changes in the soil surface after conversion from forest to pasture and sub- by wind (Giardina et al., 2000). For example, heating may release nutrients
sequent management of the extensive system for livestock production. stored in the soil biomass and make available hitherto unavailable nutrients
Sharing the same trends as Müller et al. (2004), Braz et al. (2013), and in the mineral fraction (Giardina et al., 2000; Ketterings and Bigham,
Krainovic et al. (2020), our results did not clearly indicate a decrease in 2000). As time goes by, nutrients are depleted, and these attributes tend

6
F.I. Rocha et al. Science of the Total Environment 856 (2023) 158955

Fig. 6. Influence of pasture age since forest conversion on soil fertility indicators in Amazonian localities. A) Bar plots between each pasture age and soil variables, error bars
indicate the ± standard error (SE) (n = 5); B) Standardized effect size for the model between pasture age and soil variables, coefficient estimates from linear models are
plotted with 95 % confidence intervals; Statistical differences were determined by two-way ANOVA (p < 0.05;, Significant by t-test at p < 0.001 (⁎⁎⁎), p < 0.01 (⁎⁎), p <
0.05 (⁎), n.s. non-significant; BUJ: Bujari/state of Acre, BAC1 and BAC2: Boca do Acre/state of Amazonas.

to return to their previous state (Juo and Manu, 1996; Palm et al., 1996). Overall, our results show that the relative abundance of Proteobacteria and
However, our old pastures (>10 years old) still have fertility levels like Actinobacteria are negatively correlated (rho = −44.2 %, p < 0.001),
those observed in recent pastures, highlighting a possible effect of soil with similar findings in most study locations. The discrepancy in the relative
type. In line with our results, Numata et al. (2007) in the state of Rondônia, abundances found in the MAN2 site is likely reflecting some of its inherent
also in the Brazilian Western Amazonia, did not find a significant influence characteristics, as seen in Fig. 4, since no statistical differences were found
of the land-use change on soil fertility when considering different pasture between its forest and pasture for the soil variables: C, H+, H + Al, P, T-
ages since the forest removal. Instead, the study also found that the magni- CEC, N, and Zn. Moreover, the highest content of the coarse sand fraction
tude of the effects varies significantly among soil types, with low amplitude associated with high levels of C:N ratio predominantly found in this locality
in Luvisols. Contrary to previous studies (de Moraes et al., 1996; Moreira (see Fig. 1A) can drive the reduced variability in the relative abundance of
et al., 2009), our results exhibit an overall standardized effect reflecting Actinobacteria between the land uses since this bacterial group is commonly
the soil pH increase over time since forest conversion. No statistical differ- correlated with copiotroph environments (Fierer et al., 2012).
ences were observed between old and recently formed pastures for any Nonetheless, the Actinobacteria:Proteobacteria ratio was sensitive to rep-
indicator of soil fertility of BUJ, which is also explained by the buffering resenting the gradient of soil fertility across sites and land uses for most of
effect of its soils, which reduces the intensity of chemical transformations the evaluated localities, following a positive correlation with soil pH
in the topsoil. (rho = 46.9 %, p < 0.001). The soil environment intrinsically shapes
microbial communities, especially the topsoil, which encompasses the
3.2. Actinobacteria:Proteobacteria ratio as a biotic indicator of the gradient of rhizosphere activity, complex biological interactions, organic matter
soil fertility decomposition, and food webs, which drives the soil chemical complex
(Suleiman et al., 2013; Tripathi et al., 2018). Considering this, the present
A general trend found among the study locations shows that the pre- biological indicator may be helpful for microbial ecologists interested in
dominance of Proteobacteria in the forest system clearly reflects the natural measuring the fingerprints of land-use changes in tropical soil environ-
conditions of its topsoil, despite the differences found between the soil attri- ments. A similar approach has already been applied to phyla Acidobacteria
butes in the regions and soil classes covered in this study (Figs. 7, S4). In and Proteobacteria, which have been proposed as indicators of soil trophic
turn, despite their highest relative abundance among the evaluated level (Smit et al., 2001) and used as such in the tropical forest literature
pastures, Actinobacteria did not differ statistically (p > 0.05) between land (Nemergut et al., 2010; de Carvalho et al., 2016).
uses of BAC1, MAN1, and MAN2.
Petersen et al. (2019), in a metanalysis that coupled several tropical 4. Conclusions
systems under land-use change, found that soil pH is a critical variable in
driving changes in the soil microbiota after forest-to-pasture conversion. This study showed that the level of changes in soil fertility by forest-to-
Besides the increase in bacterial alpha diversity, they pointed out that the pasture conversion depends mainly on the soil type, which regulates how
relative abundance of Actinobacteria increases in converted pastures soil wide the difference will be. Properties related to base saturation and soil acid-
and that Proteobacteria is a representative group in tropical rainforests. ity were the most affected but with considerable site-specific influence.

7
F.I. Rocha et al. Science of the Total Environment 856 (2023) 158955

Fig. 7. Actinobacteria:Proteobacteria ratio as a biological indicator of land-use conversion. A) Significant differences were determined by two-way ANOVA (p < 0.05); Significant by
t-test at p < 0.001 (⁎⁎⁎), p < 0.01 (⁎⁎), p < 0.05 (⁎), n.s. non-significant; Error bars indicate the ± standard error (SE) (n = 5). Spearman's rho correlation between B) Actinobacteria
and Proteobacteria phyla, and C) Actinobacteria:Proteobacteria ratio and soil pH. The fitted values for each model are represented by the black line and their standard errors are
indicated by the shaded area. BUJ: Bujari/state of Acre, BAC1 and BAC2: Boca do Acre/state of Amazonas, MAN1 and MAN2: Manicoré/state of Amazonas.

Among the soil attributes, the sum of bases was the most sensitive to discrim- Acknowledgments
inate between forest and pasture under the different soil types. The
Actinobacteria:Proteobacteria ratio was sensitive to reflect the gradient in soil We acknowledge the USAID and the National Academies of Sciences,
fertility derived from the impact of land-use change, being a potential micro- Engineering, and Medicine of the United States (NAS) for funding our
biological indicator of forest-to-pasture conversion in tropical ecosystems. research under PEER project 4-299, USAID agreement AID-OAA-A-11-
Supplementary data to this article can be found online at https://doi. 00012. Any opinions, findings, conclusions, or recommendations expressed
org/10.1016/j.scitotenv.2022.158955. here are those of the authors alone and do not necessarily reflect the views
of USAID or the NAS. We also thank CNPq, Brazil, for the research fellow-
CRediT authorship contribution statement ships provided to Ederson da Conceição Jesus (project 475168/2012-7)
and Fernando Igne Rocha (165571/2017-9). FIR was also supported by
AO, EJ, ID Conceptualization; FR, WT, JL, EA, AO Data curation; FR, AO CAPES, Brazil (PDSE call no 41/2018).
Formal analysis; EJ, ID, AO Funding acquisition; FR, EJ, WT, JL, EA, PM,
AB, ID, AO Investigation; EJ, WT, JL, EA, PM, AO Methodology; AO, EJ References
Project administration; EJ, ID, AO Resources; FR, AO Software; EJ, WT,
AO Supervision; AO, WT, JL, EA Validation; FR Visualization; FR, EJ, WT,
Abdi, H., Williams, L.J., 2010. Principal component analysis. Wiley Interdiscip. Rev. Comput.
AO Roles/Writing – original draft; FR, EJ, WG, JL, EA, PM, AB, ID, AO Stat. 2, 433–459.
Writing – review & editing. Alfaia, S.S., Ribeiro, G.A., Nobre, A.D., Luizão, R.C., Luizão, F.J., 2004. Evaluation of soil fer-
tility in smallholder agroforestry systems and pastures in western Amazonia. Agric.
Ecosyst. Environ. 102, 409–414.
Data availability Alvares, C.A., Stape, J.L., Sentelhas, P.C., Gonçalves, J.D.M., Sparovek, G., 2013. Köppen's cli-
mate classification map for Brazil. Meteorol. Zeitschrift 22, 711–728. https://doi.org/10.
Data will be made available on request. 1127/0941-2948/2013/0507.
Aprile, F., Siqueira, G.W., Darwich, A.J., dos Santos, V.C., Ribeiro, A.A., 2013. Concentration
of nutrients in litter as a function of soil type, climate and forest composition in Amazon.
Declaration of competing interest Agric. Sci. Dev. 2, 59–66.
Béliveau, A., Davidson, R., Lucotte, M., Lopes, L.O.D.O.C., Paquet, S., Vasseur, C., 2015. Early
effects of slash-and-burn cultivation on soil physicochemical properties of small-scale
The authors declare that they have no known competing financial inter-
farms in the Tapajós region, Brazilian Amazon. J. Agric. Sci. 153, 205.
ests or personal relationships that could have appeared to influence the Benjamini, Y., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and pow-
work reported in this paper. erful approach to multiple testing. J. R. Stat. Soc. Ser. B 57, 289–300.

8
F.I. Rocha et al. Science of the Total Environment 856 (2023) 158955

Bernini, T.de A., Pereira, M.G., Fontana, A., Anjos, L.H.C.dos, Calderano, S.B., Wadt, P.G.S., Müller, M.M.L., Guimaraes, M.F., Desjardins, T., Mitja, D., 2004. The relationship between
Moraes, A.G.de L., Santos, L.L.dos, 2013. Taxonomia de solos desenvolvidos sobre pasture degradation and soil properties in the Brazilian Amazon: a case study. Agric.
depósitos sedimentares da Formação Solimões no Estado do Acre. 72. Bragantia, Ecosyst. Environ. 103, 279–288.
pp. 71–80. Nemergut, D.R., Cleveland, C.C., Wieder, W.R., Washenberger, C.L., Townsend, A.R., 2010.
Braz, A.M.S., Fernandes, A.R., Alleoni, L.R.F., 2013. Soil attributes after the conversion from Plot-scale manipulations of organic matter inputs to soils correlate with shifts in micro-
forest to pasture in Amazon. Land Degrad. Dev. 24, 33–38. bial community composition in a lowland tropical rain forest. Soil Biol. Biochem. 42
Callahan, B.J., McMurdie, P.J., Rosen, M.J., Han, A.W., Johnson, A.J.A., Holmes, S.P., 2016. (12), 2153–2160.
DADA2: high-resolution sample inference from Illumina amplicon data. Nat. Methods Numata, I., Chadwick, O.A., Roberts, D.A., Schimel, J.P., Sampaio, F.F., Leonidas, F.C., Soares,
13, 581–583. https://doi.org/10.1038/nmeth.3869. J.V., 2007. Temporal nutrient variation in soil and vegetation of post-forest pastures as
Caporaso, J.G., Lauber, C.L., Walters, W.A., Berg-Lyons, D., Huntley, J., Fierer, N., Owens, a function of soil order, pasture age, and management, Rondônia, Brazil. Agric. Ecosyst.
S.M., Betley, J., Fraser, L., Bauer, M., Gormley, N., Gilbert, J.A., Smith, G., Knight, R., Environ. 118, 159–172.
2012. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and Palm, C.A., Swift, M.J., Woomer, P.L., 1996. Soil biological dynamics in slash-and-burn agri-
MiSeq platforms. ISME J. 6, 1621–1624. https://doi.org/10.1038/ismej.2012.8. culture. Agric. Ecosyst. Environ. 58 (1), 61–74.
Clivot, H., Pagnout, C., Aran, D., Devin, S., Bauda, P., Poupin, P., Guérold, F., Numata, I., Parks, D.H., Tyson, G.W., Hugenholtz, P., Beiko, R.G., 2014. STAMP: statistical analysis of tax-
Chadwick, O.A., Roberts, D.A., Schimel, J.P., Sampaio, F.F., Leonidas, F.C., Soares, J.V., onomic and functional profiles. Bioinformatics 30, 3123–3124.
Müller, M.M.L., Guimaraes, M.F., Desjardins, T., Mitja, D., Schaefer, C., Lima, H.N., Petersen, I.A.B., Meyer, K.M., Bohannan, B.J.M., 2019. Meta-analysis reveals consistent bacte-
Teixeira, W.G., Vale Junior, J.F., Souza, K.W., Corrêia, G.R., Mendonça, B.A.F., Amaral, rial responses to land use change across the tropics. Front. Ecol. Evol. 7, 391.
E.F., Campos, M.C.C., Ruivo, M.L.P., Hoorn, C., Wesselingh, F.P., Ter Steege, H., Ponge, J.-F., 2015. The soil as an ecosystem. Biol. Fertil. Soils 51, 645–648.
Bermudez, M.A., Mora, A., Sevink, J., Sanmartín, I., Sanchez-Meseguer, A., Anderson, Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P., Peplies, J., Glöckner, F.O.,
C.L., Figueiredo, J.P., Quesada, C.A., Lloyd, J., Anderson, L.O., Fyllas, N.M., Schwarz, 2012. The SILVA ribosomal RNA gene database project: improved data processing and
M., Czimczik, C.I., Dias-Filho, M.B., Davidson, E.A., Carvalho, C.J.R., Wubie, M.A., web-based tools. Nucleic Acids Res. 41, D590–D596.
Assen, M., Demattê, J.A.M.L.I., Demattê, J.A.M.L.I., De Souza Braz, A.M., Fernandes, Quesada, C.A., Lloyd, J., Anderson, L.O., Fyllas, N.M., Schwarz, M., Czimczik, C.I., 2011. Soils
A.R., Alleoni, L.R.F., 1996. Fertilidade e sustentabilidade de solos Amazônicos. Agric. of Amazonia with particular reference to the RAINFOR sites. Biogeosciences 8,
Ecosyst. Environ. 118, 33–38. https://doi.org/10.1007/s40808-019-00660-5. 1415–1440.
Cochrane, T.T., Sanchez, P.A., 1982. Land resources, soils and their management in the Am- Quesada, C.A., Lloyd, J., Schwarz, M., Patiño, S., Baker, T.R., Czimczik, C., Fyllas, N.M.,
azon region: a state of knowledge report. In: Hecht, S. (Ed.), Proceedings of the Interna- Martinelli, L., Nardoto, G.B., Schmerler, J., 2010. Variations in chemical and physical
tional Conference on Amazonian Agriculture And Land-use Research. Series 03E-382. properties of Amazon forest soils in relation to their genesis. Biogeosciences 7,
CIAT, Cali, Colombia, pp. 137–209. 1515–1541.
de Carvalho, T.S., Jesus, E.D.C., Barlow, J., Gardner, T.A., Soares, I.C., Tiedje, J.M., Moreira, Rocha, F.I., Ribeiro, T.G., Fontes, M.A., Schwab, S., Coelho, M.R.R., Lumbreras, J.F., da Motta,
F.M.D.S., 2016. Land use intensification in the humid tropics increased both alpha and P.E.F., Teixeira, W.G., Cole, J., Borsanelli, A.C., Dutra, I.dos S., Howe, A., de Oliveira,
beta diversity of soil bacteria. Ecology 97 (10), 2760–2771. https://doi.org/10.1002/ A.P., Jesus, E.da C., 2021. Land-use system and forest floor explain prokaryotic
ecy.1513. metacommunity structuring and spatial turnover in Amazonian forest-to-pasture conver-
Delgado-Baquerizo, M., Reich, P.B., Bardgett, R.D., Eldridge, D.J., Lambers, H., Wardle, D.A., sion areas. Front. Microbiol. 12, 657508. https://doi.org/10.3389/fmicb.2021.657508.
Reed, S.C., Plaza, C., Png, G.K., Neuhauser, S., 2020. The influence of soil age on ecosys- Santos, R.D., Lemos, R.D., Santos, H.D., Ker, J.C., Anjos, L.D., Shimizu, S.H., 2005. Manual de
tem structure and function across biomes. Nat. Commun. 11, 1–14. descrição e coleta de solo no campo. Sociedade Brasileira de Ciência do Solo, Viçosa,
de Lima, R.A.F., Oliveira, A.A., Pitta, G.R., de Gasper, A.L., Vibrans, A.C., Chave, J., Ter p. 15.
Steege, H., Prado, P.I., 2020. The erosion of biodiversity and biomass in the Atlantic For- Schaefer, C., Lima, H.N., Teixeira, W.G., Vale Junior, J.F., Souza, K.W., Corrêia, G.R.,
est biodiversity hotspot. Nat. Commun. 11, 1–16. Mendonça, B.A.F., Amaral, E.F., Campos, M.C.C., Ruivo, M.L.P., 2017. Solos da região
de Moraes, J.F.L., Volkoff, B., Cerri, C.C., Bernoux, M., 1996. Soil properties under Amazon Amazônica. Pedol. dos biomas Bras. Viçosa, MG Soc. Bras. Ciência do Solo, pp. 111–175.
forest and changes due to pasture installation in Rondônia, Brazil. Geoderma 70, 63–81. Serrão, E.A.S., Falesi, I.C., de Veiga, J.B., Teixeira Neto, J.F., 1979. Productivity of cultivated
Dias-Filho, M.B., Davidson, E.A., Carvalho, C.J.R., 2001. Linking biogeochemical cycles to cat- pastures on low fertility soils in the Amazon of Brazil. In: Sanchez, P.A., Tergas, L.E.
tle pasture management and sustainability. Biogeochem. Amaz. Basin. Oxford Univ. (Eds.), Pasture Production in Acid Soils of the Tropics, pp. 195–226.
Press, New York, pp. 84–105. Shinzato, E., Teixeira, W.G., Dantas, M.E., 2015. Principais classes de solos. Embrapa Solos-
dos Santos, H.G., Jacomine, P.K.T., Dos Anjos, L.H.C., De Oliveira, V.A., Lumbreras, J.F., Capítulo em livro científico.
Coelho, M.R., De Almeida, J.A., de Araujo Filho, J.C., de Oliveira, J.B., Cunha, T.J.F., Smit, E., Leeflang, P., Gommans, S., van den Broek, J., van Mil, S., Wernars, K., 2001. Diversity
2018. Sistema brasileiro de classificação de solos. Embrapa, Brasília, DF 2018. and seasonal fluctuations of the dominant members of the bacterial soil community in a
Farella, N., Davidson, R., Lucotte, M., Daigle, S., 2007. Nutrient and mercury variations in wheat field as determined by cultivation and molecular methods. Appl. Environ.
soils from family farms of the Tapajós region (Brazilian Amazon): recommendations for Microbiol. 67 (5), 2284–2291.
better farming. Agric. Ecosyst. Environ. 120, 449–462. Smyth, T.J., 1996. Manejo da fertilidade do solo para a produção sustentada de cultivos na
Fierer, N., Lauber, C.L., Ramirez, K.S., Zaneveld, J., Bradford, M.A., Knight, R., 2012. Compar- Amazônia. ALVAREZ VENEGAS, VH; FONTES, LEF; FONTES, MPF O solo nos grandes
ative metagenomic, phylogenetic and physiological analyses of soil microbial communi- domínios morfoclimáticos do Brasil e o desenvolvimento sustentado. Viçosa: SBCS/
ties across nitrogen gradients. ISME J. 6, 1007–1017. UFV, pp. 71–94.
Giardina, C.P., Sanford, R.L., Døckersmith, I.C., Jaramillo, V.J., 2000. The effects of slash Souza, J.L.L.de S., Fontes, M.P.F., Gilkes, R., Costa, L.M.da, Oliveira, T.S.de, 2018. Geochem-
burning on ecosystem nutrients during the land preparation phase of shifting cultivation. ical signature of Amazon tropical rainforest soils. Rev. Bras. Ciênc. Solo 42.
Plant Soil 220 (1), 247–260. Suleiman, A.K.A., Manoeli, L., Boldo, J.T., Pereira, M.G., Roesch, L.F.W., 2013. Shifts in soil
Hartemink, A.E., 2002. Soil science in tropical and temperate regions—some differences and sim- bacterial community after eight years of land-use change. Syst. Appl. Microbiol. 36,
ilarities. Adv. Agron. 77, 269–292. https://doi.org/10.1016/S0065-2113(02)77016-8. 137–144. https://doi.org/10.1016/j.syapm.2012.10.007.
Juo, A.S., Manu, A., 1996. Chemical dynamics in slash-and-burn agriculture. Agric. Ecosyst. Team, R.C., 2018. R: A Language And Environment for Statistical Computing. R Foundation
Environ. 58 (1), 49–60. for Statistical Computing, Vienna, Austria Version 3.6.1.
Kaschuk, G., Alberton, O., Hungria, M., 2011. Quantifying effects of different agricultural land Teixeira, P.C., Donagemma, G.K., Fontana, A., Teixeira, W.G., 2017. Manual de métodos de
uses on soil microbial biomass and activity in Brazilian biomes: inferences to improve soil análise de solo. Embrapa, Rio Janeiro 573 pp.
quality. Plant Soil 338, 467–481. Tripathi, B.M., Stegen, J.C., Kim, M., Dong, K., Adams, J.M., Lee, Y.K., 2018. Soil pH mediates
Kassambara, A., Mundt, F., 2018. Factoextra: Extract And Visualize the Results of Multivariate the balance between stochastic and deterministic assembly of bacteria. ISME J. 12,
Data Analyses. 2017. R Packag. Version 1. 1072–1083.
Ketterings, Q.M., Bigham, J.M., 2000. Soil color as an indicator of slash-and-burn fire severity WRB, I.W.G., 2015. World reference base for soil resources 2014, update 2015: international
and soil fertility in Sumatra,Indonesia. Soil Sci. Soc. Am. J. 64 (5), 1826–1833. soil classification system for naming soils and creating legends for soil maps. World Soil
Krainovic, P.M., Bastos, R.P., de Almeida, D.R., Junior, A.F.N., Sampaio, P.de T.B., de Souza, Resour. Reports No. 106.
L.A.G., de Souza Falcão, N.P., 2020. Effect of rosewood plantation chronosequence on soil Wright, E.S., Yilmaz, L.S., Noguera, D.R., 2012. DECIPHER, a search-based approach to chi-
attributes in Central Amazonia. Geoderma 357, 113952. mera identification for 16S rRNA sequences. Appl. Environ. Microbiol. 78, 717–725.
Lips, J.M., Duivenvoorden, J.F., 1996. Regional patterns of well drained upland soil differen- https://doi.org/10.1128/AEM.06516-11.
tiation in the middle Caquetá basin of Colombian Amazonia. Geoderma 72, 219–257. Wubie, M.A., Assen, M., 2020. Effects of land cover changes and slope gradient on soil quality
Machado, M.R., Camara, R., Sampaio, P.de T.B., Pereira, M.G., Ferraz, J.B.S., 2017. Land in the Gumara watershed, Lake Tana basin of North-West Ethiopia. Model. Earth Syst. En-
cover changes affect soil chemical attributes in the Brazilian Amazon. Acta Sci. Agron. viron. 6, 85–97. https://doi.org/10.1007/s40808-019-00660-5.
39, 385–391. Zu Ermgassen, E.K.H.J., Alcântara, M.P., Balmford, A., Barioni, L., Neto, F.B., Bettarello, M.M.,
Moreira, A., Fageria, N.K., Garcia y Garcia, A., 2011. Soil fertility, mineral nitrogen, and mi- De Brito, G., Carrero, G.C., Florence, E.D.A.S., Garcia, E., Gonçalves, E.T., Da Luz, C.T.,
crobial biomass in upland soils of the Central Amazon under different plant covers. Mallman, G.M., Strassburg, B.B.N., Valentim, J.F., Latawiec, A., 2018. Results from on-
Commun. Soil Sci. Plant Anal. 42, 694–705. the-ground efforts to promote sustainable cattle ranching in the Brazilian Amazon. Sus-
Moreira, F.M.de S., Nóbrega, R.S.A., Jesus, E.da C., Ferreira, D.F., Pérez, D.V., 2009. Differen- tainability 10 (4), 1301. https://doi.org/10.3390/su10041301.
tiation in the fertility of Inceptisols as related to land use in the upper Solimões river re-
gion, western Amazon. Sci. Total Environ. 408, 349–355.