Regional Environmental Change (2021)

21:64
https://doi.org/10.1007/s10113-021-01796-w
ORIGINAL ARTICLE

Deforestation and land use change mediate soil carbon changes

in the eastern Brazilian Amazon
Júnior Melo Damian 1 & Mariana Regina Durigan 2 & Maurício Roberto Cherubin 1 & Stoécio Malta Ferreira
Maia & Stephen M. Ogle & Plínio Barbosa de Camargo 5 & Joice Nunes Ferreira6 & Raimundo Cosme de Oliveira
3 4

Júnior 7 & Carlos Eduardo Pellegrino Cerri 1

Received: 25 July 2020 / Accepted: 22 May 2021

# The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2021

Abstract
Deforestation and land use change (LUC) to expand the agricultural frontier in the Brazilian Amazon deplete soil carbon (C)
stocks, and negatively impact climate regulation. The variety of soil types, land-transition options, and management practices
present in the Amazon region require detailed inventories to reduce the uncertainties associated with estimates of soil C change.
Therefore, we conducted a study covering ca. 1 million hectares to estimate the soil C stock changes due to LUC in
Paragominas and Santarém, Pará state, eastern Brazilian Amazon, for the period of 1990–2010. Soil C stocks for 1990 were
modeled based on land cover at the time. In 2010, we carried out a field work taking soil samples to measure soil C stock
changes in 356 transects across contrasting land uses (logged and burnt forest, young secondary forest, intermediate secondary
forest, old secondary forest, pasture, and cropland). The response ratios for the conversion from undisturbed forest to new land
uses were calculated considering the differences in soil C stocks, with the undisturbed forest as reference. Between 1990 and
2010, LUC induced a total loss of 1.51 Tg C year-1 (over an area of 7350 km2). For this period, the uncertainty of estimates
was ± 23.2%. The land transitions to pasture and cropland were the main drivers of soil C losses. Thus, deforestation
contributes to climate change not only through losses of forest biomass but also subsequently soil C losses. These results can
inform national and international climate change initiatives associated with LUC in the eastern Brazilian Amazon.

Keywords Tropicalforest . Landuseintensity . Soilorganiccarbon . Carbonsequestration . Monte Carlomethod . Climatechange

“This article is part of the Topical Collection on Sustainable

management practices to increase soil carbon sequestration:
what are their contributions to climate change mitigation,
adaptation and food security in different ecosystems and regions of
the world?”
Communicated by Communicated by Cornelia Rumpel and accepted by
Topical Collection Chief Editor Christopher Reyer
* Júnior Melo Damian Stephen M. Ogle Stephen.
[email protected] [email protected]
* Carlos Eduardo Pellegrino Cerri Plínio Barbosa de Camargo
[email protected] [email protected]
Mariana Regina Durigan
Joice Nunes Ferreira
[email protected]
[email protected]
Maurício Roberto Cherubin
[email protected] Raimundo Cosme de Oliveira Júnior

Stoécio Malta Ferreira Maia [email protected]

[email protected] Extended author information available on the last page of the
article
 64 Page 2 of Reg Environ Change (2021)
12 21:64

Introduction
Globally, carbon (C) sequestration in the soil is one of the
most effective strategies for curbing the increase of carbon
dioxide (CO2 ) in the atmosphere (Paustian et al. 2016;
Minasny et al. 2017; Cotrufo et al. 2019; Shi et al. 2020).
Furthermore, the Amazon tropical forest is the main global
forest soil C sink with an estimated annual uptake of 0.42–
0.65 Pg C year−1, around 25% of the residual terrestrial C
sink (Pan et al. 2011; Brienen et al. 2015; Soong et al.
2 0 2 0).
Slash and burn tropical forests and the subsequent
conversion of these undisturbed natural ecosystems to human-
modified sys- Fig. 1 Deforestation rate for the nine states in the Brazilian Legal
tems are considered a main cause of anthropogenic CO2 Amazon. AC, Acre; AP, Amapá; AM, Amazonas; MA, Maranhão;
MT, Mato Grosso; PA, Pará; RO, Rondônia; RR, Roraima; TO,
emis- sions to the atmosphere (Brando et al. 2020). However, Tocantins. Source: http://www.inpe.br/
depend-
ing on the system and agricultural management adopted,
deforestation is also thought to be responsible for increasing
these
soil C losses, further contributing to global climate change
areas can be a source or sink of C emissions (Ogle et al.
(Aragão and Shimabukuro 2010; Morgan et al. 2019). Thus,
2005; Cerri et al. 2009; Caviglia-Harris 2018). Soil C stocks
the extraor- dinary importance and complexity of the Amazon
vary with soil texture and bulk density, and associated C inputs
region de- mand scientific research to better understand the
from plant litter or crop residues. These parameters vary
magnitude and impact of human activity on C storage
spatially and thus contribute to the heterogeneity in soil C
(Tyukavina et al. 2017; Silva et al. 2020).
stocks, making it difficult to inventory these C stock changes
Nepstad et al. (2008) estimated that approximately 55% of
across different ecosystems (VandenBygaart 2006; Minasny et
the forests of the Amazon will be cleared, logged,
al. 2017; Gomes et al. 2019). Given the wide variability in
damaged by
soil properties, climatic con- ditions, and land use history,
drought, or burned, emitting 15–26 Pg of C to the
there are many uncertainties associated with estimating soil
atmosphere by 2030. Furthermore, Marengo et al. (2018)
C stock changes (Maia et al. 2010; Ogle et al. 2010; Conant
stated that the
et al. 2011; Patton et al. 2019). Particularly in tropical
impact of human activities and their uncertainties on C
regions, estimates are highly uncertain for most regional
cycles
monitoring applications, and C stocks under native
is one of the major issues that need to be better addressed in
vegetation are also uncertain (Batjes 2 0 11). Therefore, in
studies on the Brazilian Amazon. In addition, reliable data of
order to reveal the accuracy of results and avoid over- or
soil C changes in this strategic region is fundamental to support
underestimations of soil organic C stock changes, it is neces-
the Brazil’s Nationally Determined Contributions (Brasil 2016)
sary to perform an uncertainty analysis (Ogle et al. 2007).
in- cluded in the Paris Agreement, as well as other
This requires knowledge of factors and uncertainties that are
international initiatives, such as the “4 per 1000” Initiative (
asso- ciated with soil C stock changes and consequently
www.4p1000. org). In order to better clarify these issues, we
greenhouse gas (GHG) emissions to the atmosphere.
conducted a field study in the eastern Brazilian Amazon to test
According to Ogle and Paustian (2005), most of the
the hypothesis that the intensification of land use (logged and
analyses for reporting GHG emissions are based on models
burnt forest, secondary forest young, intermediate and old
recommended by the IPCC to conduct national
secondary forest, pasture, and cropland) after conversion from
inventories, although these methods can have high levels
primary undisturbed forest would deplete soil C stocks. Our
of uncertainty, particularly the IPCC Tier 1 methods (Del
objective was to produce inventories of the soil C stock due
Grosso and OgleSM 2011). The IPCC Tier 1 includes
to LUC in Amazon region and calculate the overall C loss at
emission and removal factors and guidance on how to
the sub-state level with the associated uncertainty (Monte Carlo
acquire activity data. Thus, it is important to develop Materials and methods
analysis).
region-specific parameters to represent soil C stock changes
(e.g., Ogle et al. 2007, 2010; Maia et al. 2 0 1 0). Description of the study sites
In Brazil, LUC activities have been identified as a key driver
of anthropogenic GHG emission, especially in the Amazon re-
Two regions were selected to assess the effects of LUC on
gion. Among the nine states that constitute the Brazilian
soil C dynamics in eastern Brazilian Amazon,
Amazon, the Pará state has the largest rates of deforestation in
including
the last few decades (INPE 2021), with an average of 4786 ±
328 km2 year−1 of deforestation rate (Fig. 1). The deforestation
rates are higher due to the fact that forests are being converted
to agricultural lands at a higher rate than in other types of land
uses (Thaler et al. 2 0 1 9). In addition to losses of forest
Reg Environ Change (2021) Page 3 of 12 64
21:64
Paragominas (Lat.: 2° 59′ 50.60′′ S; Long.: 47° 21′ 12.60′′ that the sample data provided a representative assessment of
W) and Santarém (Lat.: 2° 26′ 35.02′′ S; Long.: 54° 42′ general environmental conditions. As an example, if half of
29.99′′ W), which are located within Pará state. the catchment is covered by forest, then this land use
Paragominas and Santarém were selected because they are receives only half of the total number of transects (Gardner
the regions most affected by human activities (e.g., et al. 2013). In total, 184 transects were sampled in
deforestation and expansion of the pasture and agriculture Paragominas, including 54 UF, 46 LB, 4 SY, 11 SI, 6 SO,
activities) in the Pará state. In both regions, the natural 49 PA, and 14 CP, and 172
landscapes have been strongly mod- ified near BR-163 transects were sampled in Santarém including 43 UF, 46 LB,
road, one of the most important routes connecting 8 SY, 15 SI, 21 SO, 24 PA, and 15 CP, which covered an
Santarém (Pará state) to Cuiabá (Mato Grosso state). area of approximately 1 million hectares.
According to the RADAMBRASIL project (1973), these Time-series data
regions can be classified into two major geomorpholog- ical
units known as “Platô do Baixo Amazônas” of the central The time-series data were used in this study to provide one
Amazonia with height of 100 m above mean sea level and inventory for changes in soil C stocks due to LUC in
“Platô Tapajós-Xingú” with height ranging between 120 and Paragominas and Santarém. The information on the LUC in
170 m above mean sea level. The primary soil types are these regions was obtained for the years 1990 and 2010. We
Oxisols and Ultisols (USDA 2014), which represented used this time-series data because the beginning of
87.5% and 7.5% of the sampled soils, respectively. agriculture intensification in those regions occurred in the
The regional climate was classified as tropical hot and hu- mid-1990s and, due to the lack of more information and
mid (Am- Köppen-Geiger climate classification), with aver- studies on the LUC with maps classified after this period.
age temperature between 25 and 28°C, the relative humidity No more information was taken regarding the changes in
is 86%, and the average annual precipitation is 1920 mm soil C stocks and LUC in Paragominas and Santarém
(Instituto Nacional de Meteorologia - INMET, Brazil). recently; however, future surveys are needed to monitor the
Because of the suitable climatic conditions and topography, agriculture expansion in those same regions. In addition, soil
Paragominas and Santarém had a large expansion of mecha- C stock changes induced by LUC between 1990 and 2010
nized agriculture mainly for export of soybean production. can be an important scientific basis (baseline) for future
The primary human-modified landscapes characterizing studies in the eastern Brazilian Amazon. The time-series
the eastern region of Brazilian Amazon were evaluated to data were used to determine the size of areas occupied by
assess the effects of different LUC scenarios on soil C stocks each land use in 1990 and 2010. This in- formation was
and calculate their soil C stock responses ratios (Fig. 2). The obtained through the IMAZON’s classified maps separated
time-series data were taken at seven different land uses clas- by mature forest, secondary forest, non-forest, cloud, and
sified as primary undisturbed forest (UF, reference), logged water. All areas of secondary forest, pasture, or
and burnt forest (LB), young secondary forest (SY), interme- mechanized agriculture were deforested more than 20 years
diate secondary forest (SI), old secondary forest (SO), ago. Information about logging and fire on forests by year
pasture (PA), and cropland (CP). was obtained from manual mapping of logging and fire
Catchment characterization scars. In this study, we consider the classification of land
use to be direct for primary undisturbed forest, pasture, and
Paragominas and Santarém were divided in catchments of cropland, but in some cases, it can be much more difficult to
5000–6000 ha, which were delineated with a digital distinguish between logged and burnt forest and
elevation model and the Soil and Water Assessment Tool secondary forests. Accordingly with Barlow and Peres (
(SWAT) (Winchell et al. 2008) using ARCGIS 9.3 (ESRI, 2008), the forests that suffered multiple fires become more
Redlands, CA, USA). Eighteen catchments in each area similar to secondary for- est, which makes it difficult to
were selected to represent the deforestation ranging between detect the spatial heterogeneity of historical deforestation
10 to 100% re- maining forest cover. The final selection of patterns or intensity of fires. Thus, these two categories
18 catchments was made in each area to ensure satisfactory must be considered general indicators of
representation of age of occupation, types of historical and forest disturbance and state of regeneration.
current land use change, road access, and biophysical For the secondary forests, we subdivided it into three age
variables (e.g., hydrology, geo- morphology, and classes: young (<5 years old), intermediate (5–20 years old),
topography). and old (>20 years old). The secondary forest ageing was
In each catchment, 300-m transects (between 6 and 15) done through visual inspection of a 20-year time-series of
were distributed across the landscape based on a standard satellite images for each transect, calibrated by interviews
density of one transect per 400 m and in proportion to with local farmers (Moura et al. 2013). More details about
the percentage cover of forest and production areas the secondary forests ageing classification can be found in
(pasture and Gardner et al. (2013) and Moura et al. (2013). Finally, the
croplands). This assessment methodology was used to sizes of pastures and croplands were obtained from
ensure MODIS data (https://
 64 Page 4 of 12 Reg Environ Change (2021)
21:64
Fig. 2 Geographical location of
the study sites (a) and
description of the different land
uses evaluated in Santarém and
Paragominas (b)

modis.gsfc.nasa.gov/). To determine the final area values, we 552 and 516 soil samples in Paragominas and Santarém,
integrated all this information following the method respectively.
described by Ferraz et al. (2009). Basically, the Disturbed soil samples were air-dried and sieved with a 2
methodological ap- proach is based on a systematic and mm mesh to remove stones and root fragments. Sub-
automatic land use map- ping that uses a sequence of samples of 10g were sieved with 100 mesh (0.149 mm)
temporal profiles of vegetation indexes (such as the NDVI) and ground to a fine powder. Soil organic C was
built with MODIS. determined by
dry combustion on a LECO CN-2000® elemental
Soil sampling and analyses analyzer (furnace at 1350°C in pure oxygen). Soil bulk
density (BD)
Soil types were grouped combining the soil nomenclature was calculated by dividing the dry soil mass by volume
from the Brazilian Classification System (Santos et al. 2018) of the
and IPCC soil classification (2006), and the two major types core. The BD values were necessary for C stock
(Oxisol and Ultisol) were grouped as low activity clay soils. calculations. However, different land uses can change BD,
Soil sampling for soil C quantifications was carried out in and so, the soil mass that represents a certain soil depth can
2010 in both Paragominas and Santarém. For 1990, the soil also vary. Thus, comparisons of the soil C stocks must be
C was modeled based on land cover at the time. Within each performed through equivalent soil masses by adjusting soil
of the main land use categories, the sampling points were depth based on a reference site (Ellert and Bettany 1995). To
randomly distributed to increase the likelihood of capturing apply this method, it is necessary to determine the depth for
the soil heterogeneity aspects in both forest and production each land use that represents the equivalent soil mass to
areas. In order to reduce the dependence between the tran- determine the soil C stocks using Eq. 1,
sects, the rule of a minimum distance of 1500 m between
the sampling points was established. At each transect, dis- Equivalent depth ¼ ðBDUF=BDNLÞ x RD
turbed soil samples from the 0–10, 10–20, and 20–30 cm ð1Þ
layers were collected from five sampling points spaced 50 m where the equivalent depth is the new depth found in a
apart, providing a total of 2760 samples for soil C quantifica- specific site (cm), BDUF is the BD of for the undisturbed
tions in Paragominas and 2580 samples in Santarém. In the forest (Mg· m−3 ), BDNL is the BD for the new land uses (LB,
central position of each transect, a small trench (30 × 30 × 30 SY, SI, SO, PA, and CP) (Mg·m−3), and RD is the reference
cm) was opened to collect undisturbed soil samples for soil depth (cm).
bulk density analysis, using volumetric cores (100 cm-3) from In this study, we used the 0–30 cm layer to evaluate the
the 0–10, 10–20, and 20–30 cm layers, providing a total Soil C stock
effects ¼ C x land
of different RD uses
x on soil C stocks in the topsoil,
ð2
of BD
as recommended by IPCC (2006). Soil C stocks (Mg haÞ–1)
were calculated according to Eq. 2,
Reg Environ Change (2021) Page 5 of 12 64
21:64
where Soil C stock was expressed in Mg ha-1; C is total In our study, PDFs represent the distributions of uncertain
carbon soil content (%), RD is the reference depth (cm), and input values for the IPCC equation, including reference soil
BD = bulk density (Mg m-3). C stocks, factors of the LUC, and the areas for the different
The response ratios for the conversion from undisturbed land uses (SF, PA, and CP). We did not include the areas
forest to new land uses (LB, SY, SI, SO, PA, and CP) were under logged and burnt forest, as this category represents a
calculated considering the differences in soil C stocks, with transition between areas under undisturbed forest for other
the undisturbed forest as reference, as described in Eq. (3), land uses (e.g., SF, PA, and CP), and have a short transient
period of existence. We simulated 10,000 changes in soil C
stocks for each land use and summed the outputs in order to
CLU
Response ratio ¼ produce an empirical distribution for the entire study region.
ð3Þ From the 10,000 estimates, we used the average to
CUF
approximate the soil C stock change between LUs, and
where CLU = average soil C stock (Mg ha-1) of each land provide 95% confidence intervals based on the posterior
use; CUF= average soil C stock (Mg ha-1) of undisturbed distribution from the Monte Carlo analysis.
forest (UF).

Statistical and uncertainty analyses Results

Descriptive statistics were used to verify the position and There was a reduction in areas with undisturbed forest in
disper- sion of the data. The statistical parameters determined Paragominas and Santarém between 1990 and 2010
were the minimum, mean, maximum, and standard deviation, (Table 1). In Paragominas, the reduction of the undisturbed
in addition to the coefficients of variation (CV%), asymmetry forest area during 1990–2010 was of 5352 km2 (268 km2
(Cs), and kurtosis (Ck). The values of Cs and Ck were used to year−1). The same trend was observed in Santarém, with a
verify the normal distribution of the data. The existence of a reduction of 4870 km2 (243 km2 year−1). However, the in-
central tenden- cy (normality) of the original data was crease in areas with secondary forests in Paragominas and
evaluated using the W-test (p < 0.05), where a p value < 0.05 Santarém indicated that the measures adopted to reduce
was considered to conform to a normal distribution. These defor- estation were effective in the state of Pará. The
statistical analyses were performed using the Statistical increase in secondary forests areas in these two regions was
Analysis System (SAS) 9.4 Software Package for Windows 358 km2 (18 km2 year−1) and 984 km2 (49 km2 year−1)
10 (SAS Inc, Cary, EUA). between 1990 and 2010. In addition, pasture areas had an
To perform the uncertainty analysis, we used the Monte average increase of 2812 km2 (141 km2 year−1), while
Carlo simulation approach, which has advantages of incorpo- cropland showed an in- crease of 192 km2 (10 km2 year−1) in
rating covariances among the input variables and ability to both regions.
propagate large uncertainties through the analysis with mini- In general, the response ratios indicate that LUC from un-
mal assumptions (Smith and Heath 2001; Ogle et al. 2003, disturbed forest reduces soil C stocks. The conversion of un-
2007; Maia et al. 2010). Basically, the uncertainty analysis disturbed forest to logged and burnt forest reduced soil C
(Monte Carlo approach) is a consequence of imprecision in stocks in both Paragominas (Fig. 3a) and Santarém (Fig. 3b),
initial (or reference) values, parameters, inputs, model formu- with 64% of the transects having response ratios below the
lation, and validation data (Ogle et al. 2003). This method reference undisturbed forest (response ratio = 1). Among the
generates a distribution of results from the model based on different land uses, with the conversion of undisturbed forest
randomly selecting model input values from the probability to pasture (Fig. 3i and 3j) and cropland (Fig. 3l and 3m), we
density functions (PDFs) (Smith and Heath 2001; Ogle et al. find the largest losses of soil C with the highest proportion of
2003, 2007; VandenBygaart et al. 2004; Maia et al. 2010). response ratios below to the undisturbed forest (reference),
The PDFs for the soil C stock change factors were derived at 74% and 86%, respectively.
from a linear mixed-effect model using the response ratio The results of the descriptive statistical analysis of soil C
data. The linear mixed-effect model is a type of regression stocks showed that undisturbed forest had the highest soil
analysis that includes both fixed and random effects. While C stocks compared to most land uses in Paragominas and
the fixed effects were used to account for the influence of Santarém (Table 2). In general, the conversion from undis-
native vege- tation type, depth, and time since management turbed forest to the other land uses led to a mean reduction
change, the random-effect variables were used to account for in the soil C stocks of 7% (<4.10 Mg ha−1). The only excep-
dependen- cies in multiple measurements within the same tion to the general finding about conversion of undisturbed
study (Maia et al. 2010). The factors represent the forest to other land uses was for intermediate secondary
management impacts on soil C storage. We estimated forest and old secondary forest in Paragominas, where the
region-specific soil C stock change factors from data soil C stocks for these land uses were on average 10%
collected in different land uses. (>5.18 Mg
 64 Page 6 of 12 Reg Environ Change (2021)
21:64
Table 1 Land use changes in Paragominas and Santarém between 1990 and
2010
Site Land use Area (km2)

2010 2008 2006 2004 2002 2000 1998 1996 1994 1992 1990

Paragominas UF 10944 12967 12886 13241 13586 14340 14351 15270 15119 15771 1629
6
LB 495 270 875 503 535 608 970 334 332 306 271
SF 394 290 305 321 357 326 365 103 127 108 36
PA 4227 3857 4252 3735 3654 2831 3417 2955 2991 2767 2081
CP 574 497 414 339 262 210 88 190 145 364 514
Santarém UF 7992 8112 8072 8227 8282 8373 9183 9684 9455 10696 1286
1
LB 144 98 260 186 372 419 16 11 18 0.02 0.13
SF 1025 802 818 830 1521 1247 1251 473 561 146 41
PA 4690 4725 4544 4329 3947 4181 3818 3925 4078 3188 1212
CP 545 655 694 818 262 172 130 270 255 336 220
UF, primary undisturbed forest; LB, logged and burnt forest; SF, secondary forest; PA, pasture; and CP,
cropland

ha−1 ) and 22% (>11.53 Mg ha−1 ) higher than undisturbed Furthermore, high CV values were found mainly under
forest, respectively. The soil C stocks under undisturbed young secondary forest and pasture, ranging from 31 to
forest did not conform to a normal distribution in the two 63% and from 24 to 36% for Paragominas and Santarém,
sites eval- uated, with coefficients of asymmetry and respectively. For both Paragominas and Santarém, we
kurtosis close to zero. The coefficient of variation (CV) estimated that
showed an average value of 18 ± 9% between land uses in LUC accounted for a total decrease of 1.51 Tg C year −1
the two study sites. [(−0.40 Tg C year −1 )+ (−1.11 Tg C year−1)] (with an
average

Fig. 3 Response ratios of soil C stocks in logged and burnt forest (a and b); young secondary forest (c and d); intermediate secondary forest (c and f); old
secondary forest (g and h); pasture (i and j); and cropland (l and m) for Paragominas and Santarém. The red baseline is the response ratio for
undisturbed forest
Reg Environ Change (2021) Page 7 of 12 64
21:64
Table 2 Descriptive statistics of soil C stocks (Mg ha−1) evaluated in Paragominas and
Santarém
Site Na Land use Minimum Mean Maximum CV (%)b SDc Csd Cke
Wf
Paragominas 46 UF 41.51 64.38 66.22 10.23 1.76 −3.73 13.92 0.32*
LB 40.81 59.09 83.90 18.16 2.86 0.69 1.17 0.96ns
4 UF 41.51 53.31 65.11 25.56 6.81 0 −6 0.73ns
SY 26.04 51.01 98.40 63.38 16.16 1.73 3.21 0.81ns
11 UF 32.53 51.52 66.22 32.35 5.02 −0.28 −2.30 0.70*
SI 40.34 56.70 90.19 29.99 5.13 0.95 −0.30 0.87ns
6 UF 32.53 52.18 66.22 29.38 6.26 −0.19 −2.69 0.80*
SO 44.09 63.70 79.37 21.73 5.60 −0.20 −1.51 0.95ns
49 UF 32.53 49.10 66.22 28.21 1.98 0.28 −1.71 0.75*
PA 20.41 44.74 81.56 35.99 2.30 0.57 −0.63 0.94*
14 UF 41.51 64.38 66.22 10.23 1.76 −3.73 13.93 0.31*
CP 40.81 59.09 83.90 18.16 2.87 0.69 1.17 0.96ns
Santarém 46 UF 37.67 52.94 58.89 16.38 1.67 −1.13 −0.46 0.66*
LB 24.16 52.62 104.90 35.24 3.57 1.22 2.59 0.88*
5 UF 37.67 50.10 58.89 20.99 3.72 −0.49 −2.24 0.72*
SY 27.11 44.29 63.80 30.74 4.81 0.28 −1.65 0.92ns
15 UF 37.77 54.28 58.89 13.52 1.90 −1.70 2.02 0.66*
SI 25.35 52.42 68.45 23.18 3.14 −1.09 0.68 0.89ns
21 UF 52.17 57.61 58.89 4.69 0.59 −1.70 0.98 0.48*
SO 42.23 55.09 77.24 15.09 1.81 0.93 1.13 0.95ns
24 UF 37.67 52.52 58.89 15.80 1.69 −1.09 −0.34 0.71*
PA 27.41 48.48 67.89 24.19 2.39 −0.37 −0.91 0.94ns
15 UF 52.17 57.55 58.89 4.83 0.72 −1.67 0.90 0.50*
CP 32.17 48.74 65.57 16.73 2.11 0.11 0.69 0.98ns

N number of transects, b coefficient of variation; c standard deviation; d coefficient of asymmetry; e coefficient of kurtosis; f Shapiro-Wilk test for
normal distribution, where *significant at levels of p <0.05 and ns not significant. When significant indicates that the hypothesis of a normal
distribution is rejected. UF, primary undisturbed forest; LB, logged and burnt forest; SY, young secondary forest; SI, intermediate secondary forest;
SO, old secondary forest; PA, pasture; and CP, cropland

uncertainty of ± 23.2%) between 1990 and 2010 (Table 3). logged and burnt forest areas between the first (1990) and the
Among the different LUC categories, only the secondary for- last (2010) year of evaluation, we observed an average in-
ests in Paragominas had C gains, which ranges from 0.07 crease of 184 km2 in the areas under these management prac-
to tices in Paragominas and Santarém. According to Brito (
0.17 Tg C year −1 (average uncertainty of ± 39.53% Tg C 2020), deforestation rates in these regions remained high
year−1). In contrast to secondary forests, pasture and cropland until 2008, with a reduction of 45% of the undisturbed
showed similar carbon losses in both regions, with losses forest area in Paragominas.
ranging from 0.33 to 0.67 Tg C year−1 (average uncertainty The areas with secondary forests showed an increase in
of ± 34.53%) and from 0.01 to 0.33 Tg C year−1 (average Paragominas and Santarém during 1990-2010. However,
uncertainty of ± 30.28%) for Paragominas, and from 0.58 to Carvalho et al. (2019) in more recent studies reported that
0.92 Tg C year−1 (average uncertainty of ± 25.26%) and the increase of secondary forests areas in Pará state only oc-
from curred up to 2010. According to these authors, after 2010, the
0.07 to 0.09 Tg C year−1 (average uncertainty of ± 17.27%) areas under secondary forests were converted into new land
in Santarém, respectively. uses such as pastures, cropland, or palm oil. It worth
mention- ing that between 1990 and 2010, the results found
Discussion in our study already indicated an increase in pasture and
cropland areas in Paragominas and Santarém.
Temporal land use changes in Paragominas and
Santarém regions
Response ratios for soil carbon in the eastern
The high losses of the undisturbed forest in Paragominas and Brazilian Amazon
Santarém between 1990 and 2010 may be related to the his-
tory of LUC in both regions (Table 1). According to Nunes Based on results of the response ratios, we observed that the
et al. (2015), between the 1980s and the mid-1990s, forest conversion of undisturbed forest to logged and burnt forest
burning and logging became one of the main management reduces soil C stocks (Fig. 3a and 3b). This effect is mainly
practices in these regions, with a rapid increase of the due to fire which accelerates the mineralization of soil
number of sawmills in this same period. Despite the organic matter and causes loss of C as CO2 and CH4.
fluctuation of the Furthermore, a
 64 Page 8 of 12 Reg Environ Change (2021) 21:64

Table 3 Estimated changes in C stocks (Tg C year−1) in different land uses from 1990 to 2010, along with the land area estimates (km2) for
Paragominas and Santarém

Site Land use Area C change Mean C flux Mean C rate Uncertaint
(km2) (Tg) (Tg C year−1) (Mg C ha-1 year−1) y (± %)

Paragominas SF 358.23 2.33±0.92 0.12±0.05 0.16±0.06 39.53

PA 2145.70 −10.07 −0.50±0.17 −0.12±0.04 34.53
±3.48
CP 59.70 −0.32±0.10 −0.02±0.01 −0.13±0.04 30.28
Total 2563.63 −8.06±2.28 −0.40±0.11 0.14±0.04 28.30
Santarém SF 983.75 −5.23±0.88 −0.26±0.04 −0.13±0.02 16.82
PA 3478.34 −15.54 −0.78±0.20 −0.11±0.03 25.26
±3.93
CP 324.61 −1.52±0 −0.08±0.01 −0.12±0.02 17.27
.26
Total 4786.70 −22.29
±4.03 −1.11±0.20 −0.12±0.02 18.09

Positive values represent increase in soil C. SF, secondary forest; PA, pasture; and CP,
cropland

loss of 30.9 × 106 Mg of total (above + below-ground) live not separated for management practices. Nevertheless,
biomass is estimated with burning of the Amazon many researchers have shown that the introduction of
rainforest during deforestation (Vasconcelos et al. 2013; agriculture with conventional tillage in tropical regions is
Neto et al. 2 0 1 9). a driver of soil C stock loss (Ogle et al. 2012; Ogle et al.
Cerri et al. (2004) using mathematical modeling found a de- 2019; Feudis et al. 2019; Haddix et al. 2020). C losses may
crease in soil C stocks with the forest burning, followed by a be due to the destruction of soil aggregates and
slow increase after this event. The slow increase in soil C consequently faster SOM mineralization (Six et al. 2000).
stocks after burning is consistent with the response ratios found Maia et al. (2010) found similar result with re- sponse
for young secondary forest (Fig. 3c and 3d), intermediate ratios implying less C in soils under convention- al tillage
secondary forest (Fig. 3c and 3f), and old secondary forest practices in Mato Grosso and Rondônia states. This result
(Fig. 3g and 3h). According with the response ratios values, implies that SOM can be maintained or may increase with
those land uses had 50%, 42%, and 39% of the response ratios higher physical protection within soil ag- gregates;
below the reference (undisturbed forest), respectively, for therefore, improving soil structure with less tillage leads
Paragominas and Santarém. Despite the response ratios to larger soil C stocks and lower loss of C to the
indicating soil C stock depletions for land uses compared to atmosphere.
undisturbed forest, an in- crease in soil C is suggested with The lack of sustainable management can also explain
establishment of secondary forest with the order of soil C: the response ratios found for the pasture, which are lower
young secondary forest > inter- mediate secondary forest > old than undisturbed forest (Fig. 3i and 3j). The majority of
secondary forest. This result agrees with Blécourt et al. (2013), pasture areas in Amazonia (50%) are degraded, and/or in
who found a positive correla- tion of soil C concentrations some sort of degradation stage, which is mainly driven
with the tree basal area and total basal area in old secondary by
forest, which suggests a higher bio- mass and consequently lack of control over grazing pressure (Dias Filho 2015).
more C input to the soil through root residues. According to According to Stahl et al. (2017), the high amounts of soil
Carvalho et al. (2019), secondary tropical vegetation is of C storage in Amazonian forest (3.31 Mg C ha-1 year−1 ) can
particular interest for replacing pasture and crop- land due to be partially recovered in pastures only after 24 years of
their value for biodiversity conservation and ecosys- tem implantation, and with the use of sustainable manage-
services, as well as potential as a C sink. ment practices (e.g., avoiding fires and overgrazing, using
The greater soil C stock loss for the cropland (Fig. 3l a grazing rotation plan and a mixture of C3 and C4 spe-
and 3m) was expected because of the intensification of cies). Furthermore, several studies demonstrate the poten-
soil disturbance due to conventional tillage, mainly for tial for pastures to accumulate carbon through intensifica-
cultivation (Fujisaki et al. 2015; Durigan et al. 2017). tion and diversification (e.g., integrated crop-livestock,
Conventional soil tillage practice was largely integrated crop-forest, and crop-livestock-forest systems),
adopted where in some cases, the soil C stocks can be higher than
for the cultivation of the soybean crop since your intro- areas with forests (Grahmann et al. 2020; Silva et al.
duction in the mid-1990s. In this study, the areas were 2017; Vicente et al. 2019; Damian et al. 2 0 2 1 a).
separated only by their different land uses; they were
Reg Environ Change (2021) Page 9 of 12 64
21:64
Impact and uncertainties associated of land use 23.2%) (Table 3). Most of the uncertainty is associated with
changes on soil carbon stock estimates the time series data on LUC and with the diversity of
management practices adopted. Maia et al. (2010) evaluated
Conversion from undisturbed forest to the other land uses the effect of LUC in two states of the Brazilian Amazon,
(secondary forests, pasture, and cropland) reduced the soil C Rondônia, and Mato Grosso, during the 1970–2002 period,
stocks (<7%) (Table 2). Our findings agree with Fujisaki et and found losses ranging from 0.91 to 4.23 Tg C year−1, with
al. (2015) in a study of 52 sites in the Amazon that found an uncertainty of ± 21.9% to ± 41.5%. Limited data on
average loss of 8.5% in soil C stocks with conversion of un- manage- ment practices in the Amazon regions was one of
disturbed forest to others land uses. the key sources of uncertainty in Rondônia and Mato Grosso
Among the different land transitions evaluated, only for reported by Maia et al. (2010). Although the time-series
the conversion from undisturbed forest to intermediate data evaluated in this study does not take into account the
secondary forest and old secondary forest in Paragominas last 10 years (2011–2020), they are very valuable in
resulted in soil C stock gains. This exception may be estimating the impacts of LUC on the Amazon region.
attributed to several initiatives (e.g., Action Plan for However, investments to updating this information, whether
Prevention and Control of Legal Amazon Deforestation in via new field research pro- jects, or even based on remote
the Legal Amazon in 2004 and the creation of the List of sensing, are strongly necessary and encouraged to continue
Priority Municipalities for the Control of Deforestation in this monitoring over time.
2007) for the recovery and sus- tainable use of the soil and Total C losses in Paragominas and Santarém represents
forest resources in this region. In addition, in the secondary approximately 1.63% of the C emission derived from
forests, the characteristics of histor- ical land use (e.g., burns LUC
events and short time periods with pasture) can affect the in Pará state from 1990 to 2010 (92.73 Tg C year−1, Brasil
soil C storage capacity of different regions. Our results are 2017). The C loss estimates on a regional scale found in
consistent with Neumann-Cosel et al. (2011), suggesting this
that intermediate secondary forest and old secondary forest study are consistent with the most comprehensive estimates
in Santarém have not reached their maximum C storage for the entire Amazon performed by Soares-Filho et al. (
capacity, where longer term establish- ment of forest can 2006), who estimates a 40% reduction in Amazon forests,
increase the soil C stocks following inten- sive agricultural and a C emission of 32 ± 8 Pg to the atmosphere due to the
management. According to Guo and Gifford (2002), the LUC by 2050.
changes in the soil occur very slowly and the balance is The secondary forests in Paragominas were the only land
only fully reached after 50 years of conversion. use that had C gains. These results indicated that secondary
The soil C stock data did not follow a normal distribution forests are a strategy to reduce C losses to the atmosphere
in undisturbed forest, and the CV among the land uses was and restore soil functioning, mainly through the high
18%. Likely, this result can be associated with the large accumula- tion of biomass and rapid soil C accumulation
heteroge- neity of soils in Amazon (Batjesm and Dijkshoorn (Wang et al. 2020). In fact, Chazdon et al. (2016) reported
1999) and the large area covered by this study (1 million that secondary forests in Brazilian Amazon have the highest
ha). Koele et al. (2017) also reports that soil C in the topsoil potential for C accumulation in Latin America, with a net C
can vary signifi- cantly over a lateral scale of centimeters to sequestration of
meters with CVs larger than 20%. In this study, high CV 6.04 Pg from 2008 to 2048 period.
values were found mainly for young secondary forest and For both pasture and cropland, the analyses showed
pasture. Neumann- Cosel et al. (2011) also found high similar C losses in Paragominas and Santarém. Our data
variability in soil C stocks for young secondary forest, which corroborate the predictions of Buckeridge (2008), who
was attributed to the intrin- sic soil factors (i.e., clay indicated that conversion of Brazilian forests to pasture
mineralogy), and to the land use history before and cropland in the last 140 years led to a net release of
establishment of secondary forest. Furthermore, the 121 Pg C to the atmosphere. The pasture showed greater
young secondary forest is highly affected for the previous annual losses of C in relation to the cropland due to the
land uses, level of degradation, and time of natural greater area in the two regions. However, the C losses per
restoration. In the case of pasture, Fujisaki et al. (2015) converted area are 8% higher in cropland than pasture.
found that the greater variability in soil C stocks was ex- This effect can be explained by the rapid and complete
plained by the diversity of situations ranging from highly removal of the above-ground biomass and woody roots to
pro- ductivity forage grass to areas with decreased allow for soybean production, with little or no net C
production due to absence of liming and fertilization, accumulation from subsequent agricultural production
overgrazing, erosion, and weed invasion. (Morton et al. 2006). On the other hand, despite the low
For the estimated changes of the C stocks during 1990 C input from aboveground biomass of forages, there is a
and 2010 in Paragominas and Santarém, the LUC accounted greater C input through the root system in addition to
with a decrease of 1.51 Tg C year −1 (average uncertainty of absence of soil disturbance that limit C losses. Based on
± these results, adoption of sustainable management prac-
tices could be recommended by governmental initiatives
 64 Page 10 of 12 Reg Environ Change (2021)
21:64
(e.g., “ABC Program”) to increase C stocks through soil C References
sequestration in areas under cropland and pasture with- out
forest restoration (Matos et al. 2020; Damian et al. 2021b). Aragão LEOC, Shimabukuro YE (2010) The incidence of fire in
Azevedo et al. (2018) estimated that the adoption of Amazonian forests with implications for REDD. Science 328:
1275–1278 http://science.sciencemag.org/content/328/5983/1275
sustainable management practices in Brazil, such as no- Azevedo TR, Junior CC, Junior AB, Cremer MS, Piatto M et al (2018)
till, integrated crop-livestock, and integrated live- stock- SEEG initiative estimates of Brazilian greenhouse gas emissions
forest, can be act as C sinks, with a potential of C from 1970 to 2015. Sci Data 5:1–43.
sequestration of 5.51 to 6.24 Mg CO2eq ha-1 year −1 . https://doi.org/10.1038/sdata.2018.45
Batjesm NH, Dijkshoorn JA (1999) Carbon and nitrogen stocks in the
Because of that, the adoption of management practices soils of the Amazon Region. Geoderma 89:273–286.
such as no-tillage, pasture restoration, and integrated ag- https://doi.org/
ricultural systems is listed among main Brazil’s NDCs to 10.1016/S0016-7061(98)00086-X
tackle global warming and climate changes in the next Batjes NH (2011) Research needs for monitoring, reporting and
verifying soil carbon benefits in sustainable land management and
decades (Brasil 2016). GHG mit- igation projects. In: De Brogniez D, Mayaux P,
Montanarella L (eds) Monitoring reporting and verifications systems
for carbon in soils and vegetation in African, Caribbean and Pacific
countries. European Commission. Joint Research Center, Brussels, pp
27–39 Barlow J, Peres CA (2008) Fire-mediated dieback and
Conclusions compositional cascade in an Amazonian forest. Philos Trans R Soc B
363:1787–
Pará is the state with the highest deforestation rates in the 1794. https://doi.org/10.1098/rstb.2007.0013
Brazilian Amazon, despite the substantial reduction in Brando PM, Soares-Filho B, Rodrigues L, Assunção A, Morton D et al
(2020) The gathering firestorm in southern Amazonia. SciAdv
those rates in the past two decades. However, the recent 6:1– 9 http://advances.sciencemag.org/content/6/2/eaay1632
increase in deforestation in the country is a set- back in the Brasil (2017) Estimativas anuais de emissões de gases de 665 efeito
conservation of the Amazon. In this study, we found that estufa no Brasil. Ministério da Ciência, Tecnologia, Inovações e
the LUC process led to a total soil C loss of 1.51 Tg C Comunicações (MCTIC). https://sirene.mctic.gov.br/portal/
opencms/publicacao/index.html. Accessed 6 June 2020
year −1 (average uncertainty of ± 23.2%). This soil C loss
Brasil (2016) Nationally Determined Contributions. https://www4.
corresponds to 1.63% of the total soil C losses derived unfccc.int/sites/NDCStaging/Pages/Home.aspx.
from LUC in Pará state during 1990 to 2010. Land Brienen RJW, Phillips OL, Feldpausch TR, Gloor E, Baker TR et al
transitions to pasture and agriculture are the main drivers (2015) Long-term decline of the Amazon carbono sink. Nature
of soil C stock losses. Among the different land uses, only 519:44–348. https://doi.org/10.1038/nature14283
Brito B (2020) The pioneer market for forest law compliance in
secondary forests showed increases in soil C stocks after
Paragominas, Eastern Brazilian Amazon. Land Use Policy 94:1–
the conversion from primary/ undisturbed forest. This 10. https://doi.org/10.1016/j.landusepol.2019.104310
highlights the potential of implementing forest Buckeridge MS (2008) Biologia e mudanças climáticas no Brasil.
restoration initiatives to sequester C in Amazon soils. On RiMa, São Carlos
the other hand, the logging and burning of the forest Blécourt M, Brumme R, Xu J, Corre MD, Veldkamp E (2013) Soil
carbon stocks decrease following conversion of secondary forests
reduced the soil C stocks. to rubber (Hevea brasiliensis) plantations. PLoS One 519:1–9.
Our findings suggest that the land use practices https://doi.org/10.1371/journal.pone.0069357
should be changed focusing on the adoption of practices to Carvalho R, Adami M, Amaral S, Bezerra FG, Aguiar APD (2019)
reverse soil C losses and increase soil C sequestra- tion. Changes in secondary vegetation dynamics in a context of
decreas- ing deforestation rates in Pará, Brazilian Amazon. Appl
Thus, reducing deforestation and adopting soil con- Geogr 106: 40–49. https://doi.org/10.1016/j.apgeog.2019.03.001
servation practices (e.g., no-till, integrated crop-live- Caviglia-Harris JL (2018) Agricultural innovation and climate change
stock, integrated livestock-forest) as recommended by policy in the Brazilian Amazon: Intensification practices and the
national (e.g., “ABC Program”) and international (e.g., “4 derived demand for pasture. J Environ Econ Manag 90:232–248.
https://doi.org/10.1016/j.jeem.2018.06.006
per 1000”) initiatives are fundamental to meet the
Cerri CC, Maia SMF, Galdos MV, Cerri CEP, Feigl BJ et al (2009)
greenhouse gas mitigation goals of the Brazilian Brazilian greenhouse gas emissions: the importance of agriculture
government. and livestock. Sci Agrár 66:831–843. https://doi.org/10.1590/
S0103-90162009000600017
Cerri CEP, Paustian K, Bernoux MAL, Victoria RL, Melillo JM et al
(2004) Modeling changes in soil organic matter in Amazon forest
Acknowledgements We would like to thank prof. Silvio Frosini de to pasture conversion with the Century model. Glob Chang Biol
Barros Ferraz for his important help in processing and interpretation of 10: 815–832. https://doi.org/10.1111/j.1365-2486.2004.00759.x
satellite images used in this study. Cotrufo F, Ranalli MG, Haddix ML, Six J, Lugato E (2019) Soil carbon
storage informed by particulate and mineral-associated organic
Funding This study is funded by the CNPq (grant 402992/2013-0) and matter. Nat Geosci 12:989–994.
CAPES (grant 1681809), and FAPESP (grants 2017/15331-3; https://doi.org/10.1038/s41561-019-0484-6
2018/21261-0; 2018/09845-7) who granted scholarship to the first Conant RT, Ogle SM, Paul EA, Paustian K (2011) Measuring and
author. monitor- ing soil organic carbon stocks in agricultural lands for
climate mitiga- tion. Front Ecol Environ 3:169–173.
https://doi.org/10.1890/090153
Chazdon RL, Broadbent EN, Rozendaal DMA, Bongers F, Zambrano
AMA et al (2016) Carbon sequestration potential of second-
Reg Environ Change (2021) Page 11 of 12 64
21:64
growth forest regeneration in the Latin American tropics. Sci Adv Koele N, Bird M, Haig J, Marimon-Junior BH, Marimon BS et al
2: 1–10. https://doi.org/10.1126/sciadv.1501639 (2017) Amazon Basin forest pyrogenic carbon stocks: First
Damian JM, Matos ES, Pedreira BC, Carvalho PCF, Premazzi LM et estimate of deep storage. Geoderma 306:237–243.
al https://doi.org/10.1016/j. geoderma.2017.07.029
(2021a) Predicting soil C changes after pasture intensification and Maia SFM, Ogle SM, Cerri CEP, Cerri CC (2010) Soil organic carbon
diversification in Brazil. Catena 202:1–13. stock change due to land use activity along the agricultural frontier
https://doi.org/10.1016/j. catena.2021.105238 of the southwestern Amazon, Brazil, between 1970 and 2002.
Damian JM, Matos ES, Pedreira BC, Carvalho PCF, Souza AJ et al Glob Chang Biol 16:2775–2788.
(2021b) Pastureland intensification and diversification in Brazil https://doi.org/10.1111/j.1365-2486. 2009.02105.x
me- diate soil bacterial community structure changes and soil C Marengo JA, Souza CM Jr, Thonicke K, Burton C, Halladay K et al
accumu- (2018) Changes in climate and land use over the Amazon region:
lation. Appl Soil Ecol 160:1–13. https://doi.org/10.1016/j.apsoil. current and future variability and trends. Front Earth Sci 6:1–21.
2020.103858 https://doi.org/10.3389/feart.2018.00228
Del Grosso SJ, OgleSM, PartonWJ (2011) Soil organic matter cycling Matos PS, Fonte SJ, Lima SS, Pereira MG, Kelly C et al (2020)
and greenhouse gas accounting methodologies. In: L. Guo A, Linkages
Gunasekara L, Mc Connell LL (eds) Understanding greenhouse among soil properties and litter quality in agroforestry systems of
gas emissions from agricultural management. American Chemical southeastern Brazil. Sustainability 12:1–22.
Society, Washington DC, pp 3-13. https://doi.org/10.3390/ su12229752
Dias Filho MB (2015) Estratégias de recuperação de pastagens Minasny B, Malone BP, McBratney AB, Angers DA, Arrouays D et al
degradadas na Amazônia brasileira. Embrapa Amazônia Oriental, (2017) Soil carbon 4 per mille. Geoderma 292:59–86. https://doi.
Belém org/10.1016/j.geoderma.2017.01.002
Durigan MR, Cherubin MR, De Camargo PB, Ferreira JN, Berenguer E Moura NG, Lees AC, Andretti CB, Davis BJW, Solar RRC et al (2013)
et al (2017) Soilorganicmatter responses Avian biodiversity in multiple-use landscapes of the Brazilian
toanthropogenicforest disturbanceandland use change in Amazon. Biol Conserv 167:339–348. https://doi.org/10.1126/
theeasternBrazilianAmazon. Sustainability 9:1–16. science.1186925
https://doi.org/10.3390/su9030379 Morgan WT, Darbyshire E, Spracklen DV, Artaxo P, Coe H (2019)
Ellert BH, Bettany JR (1995) Calculation of organic matter and Non- deforestation drivers of fires are increasingly important
nutrients sources of aerosol and carbon dioxide emissions across
stored in soils under contrasting management regimes. Can J Soil Sci Amazonia. Sci Rep 9:
75:529–538. https://doi.org/10.4141/cjss95-075 1–5. https://doi.org/10.1038/s41598-019-53112-6
Ferraz RPD, Meirelles MSP, Jobaggy E, Jonathan M, Coutinho HLC Morton DC, De Fries RS, Shimabukuro YE, Anderson LO, Arai E et al
(2009) Utilização de dados MODIS e modelo SEBAL para a (2006) Cropland expansion changes deforestation dynamics in the
estimativa da perda de água por evapotranspiração: Uma southern Brazilian Amazon. PNAS 103:14637–14641.
proposição metodológica para o monitoramento do impacto https://doi. org/10.1073/pnas.0606377103
da Neumann-Cosel L, Zimmermann B, Hall JS, van Breugel M, Elsenbeer
expansão da cultura canavieira na dinâmica hídrica regional. H (2011) Soil carbon dynamics under young tropical secondary
Simpósio brasileiro de sensoriamento remoto. http://marte.sid.inpe. forests on former pastures—a case study from Panama. For Ecol
br/col/dpi.inpe.br/sbsr@80/2008/11.18.12.18/doc/175-182.pdf Manag 261:1625–1633.
? https://doi.org/10.1016/j.foreco.2010.07.023
languagebutton=en. Accessed 20 April 2020 Neto ECS, Pereira MG, Junior EFF, Silva SB, Junior JAC et al (2019)
Feudis M, Cardelli V, Massaccesi L, Trumbore SE, Antisari LV et al Temporal evaluation of soil chemical attributes after slash-and-
(2019) Small altitudinal change and rhizosphere affect the SOM light burn agriculture in the Western Brazilian Amazon. Acta Sci Agron
fractions but not the heavy fraction in European beech forest soil. 41:1–
Catena 181:1–9. https://doi.org/10.1016/j.catena.2019.104091 10. https://doi.org/10.4025/actasciagron.v41i1.42609
Fujisaki K, Perrin AS, Desjardins T, Bernoux M, Balbino LC et al Nepstad DC, Stickler CM, Soares-Filho B, Merry F (2008) Interactions
(2015) From forest to cropland and pasture systems: a critical among Amazon land use, forests and climate: prospects for a near-
review of soil organic carbon stocks changes in Amazonia. Glob term forest tipping point. Philos Trans R Soc B 363:1737–1746.
Chang Biol 21: https://doi.org/10.1098/rstb.2007.0036
2773–2786. https://doi.org/10.1111/gcb.12906 Nunes SS, Barlow J, Gardner TA, Siqueira JV, Sales MR et al (2015) A
Gardner TA, Ferreira J, Barlow J, Lees AC, Parry L et al (2013) A 22 year assessment of deforestation and restoration in riparian for-
social and ecological assessment of tropical land uses at multiple ests in the eastern Brazilian Amazon. Environ Conserv 42:193–
scales: the Sustainable Amazon Network. Philos Trans R Soc B 203. https://doi.org/10.1017/S0376892914000356
368:1–11. https://doi.org/10.1098/rstb.2012.0166 Ogle SM, Breidt FJ, Paustian K (2005) Agricultural management
Gomes LCG, Faria RM, Souza E, Veloso GV, Schaefer CEGR et al impacts
(2019) Modelling and mapping soil organic carbon stocks in on soil organic carbon storage under moist and dry climatic condi-
Brazil. Geoderma 340:337–350. https://doi.org/10.1016/j. tions of temperate and tropical regions. Biogeochemistry 72:87–
geoderma.2019.01.007 121. https://doi.org/10.1007/s10533-004-0360-2
Guo LB, Gifford RM (2002) Soil carbon stocks and land use change: a Ogle SM, Breidt FJ, Easter M, Williams S, Killian K et al (2010) Scale
meta analysis. Glob Chang Biol 8:345–360. and uncertainty in modeled soil organic carbon stock changes for
https://doi.org/10.1046/ j.1354-1013.2002.00486.x US croplands using a process based model. Glob Chang Biol
Grahmann K, Dellepiane VR, Terra JA, Quincke JA (2020) Long-term 16:
observations in contrasting crop-pasture rotations over half a 810–822. https://doi.org/10.1111/j.1365-2486.2009.01951.x
centu- ry: statistical analysis of chemical soil properties and Ogle SM, Breidt FJ, Easter M, Williams S, Paustian K (2007)
implications Empirically-based uncertainty associated with modeling carbon se-
for soil sampling frequency. Agric Ecosyst Environ 287:1–11. questration rates in soils. Ecol Model 205:453–463.
https://doi.org/10.1016/j.agee.2019.106710 https://doi.org/ 10.1016/j.ecolmodel.2007.03.007
Haddix ML, Gregorich EG, Helgason BL, Janzen H, Ellert BH et al Ogle SM, Paustian K (2005) Soil organic carbon as an indicator of
(2020) Climate, carbon content, and soil texture control the envi-
indepen- dent formation and persistence of particulate and ronmental quality at the national scale: monitoring methods and
mineral-associated policy relevance. Can J Soil Sci 85:531–540. https://doi.org/10.
organic matter in soil. Geoderma 363:1–10. 4141/S04-087
https://doi.org/10.1016/ j.geoderma.2019.114160 Ogle SM, Breidt FJ, Eve MD, Paustian K (2003) Uncertainty in
IPCC (2006) Emissions Scenarios-Special Report 3. Intergovernmental estimat- ing land use and management impacts on soil organic
 64 Page 12 of 12 Reg Environ Change (2021)
21:64
for US agricultural lands between 1982 and 1997. Glob Chang Stahl C, Fontaine S, Klumpp K, Picon-Cochard C, Grise MM et al
Biol 9:1521–1542. (2017) Continuous soil carbon storage of old permanent
https://doi.org/10.1046/j.1365-2486.2003.00683.x pastures in Amazonia. Glob Chang Biol 23:3382–3392.
Ogle SM, Swan A, Paustian K (2012) No-till management impacts on https://doi.org/10. 1111/gcb.13573
crop Smith JE, Heath LS (2001) Identifying influences on model
productivity, carbon input and soil carbon sequestration. Agric Ecosyst uncertainty:
Environ 149:37–49. https://doi.org/10.1016/j.agee.2011.12.010 an application using a Forest carbon budget model. Environmental
Ogle SM, Alsaker C, Baldock J, Bernoux M, Breidt FJ et al (2019) Management, New York
Climate and soil characteristics determine where no-till Thaler GM, Viana C, Toni F (2019) From frontier governance to gover-
management can store carbon in soils and mitigate greenhouse nance frontier: the political geography of Brazil’s Amazon transi- tion.
gas emissions. Sci Rep 9:1–8. World Dev 114:59–72. https://doi.org/10.1016/j.worlddev.
https://doi.org/10.1038/s41598-019-47861-7 2018.09.022
Paustian K, Lehmann J, Ogle S, Reay D, Robertson GP et al (2016) Tyukavina A, Hansen MC, Potapov PV, Stehman SV, Smith-Rodriguez
Climate-smart soils. Nature 532:49–57. https://doi.org/10.1038/ K et al (2017) Types and rates of forest disturbance in Brazilian
nature17174 Legal Amazon, 2000–2013. Sci Adv 3:1–15 http://advances.
Pan Y, Birdsey RA, Fang J, Houghton R, Kauppi PE et al (2011) A sciencemag.org/content/3/4/e1601047
large and persistent carbon sink in the world’s forests. Science USDA (2014) Keys to Soil Taxonomy. USDA—Natural Resources
333:988– 993 Conservation Service. https://www.nrcs.usda.gov/wps/portal/nrcs/
https://science.sciencemag.org/content/333/6045/988 detail/soils/survey/class/?cid=nrcs142p2_053580. Accessed 22
Patton NR, Lohse KA, Seyfried MS, Godsey SE, Parsons SB (2019) April 2020
Topographic controls of soil organic carbon on soil-mantled land- VandenBygaart AJ, Gregorich EG, Angers DA, Stoklas UF (2004)
scapes. Sci Rep 9:1–15. https://doi.org/10.1038/s41598-019-42556-5 Uncertainty analysis of soil carbon stock change in Canadian
Santos HG, Jacomine PKT, Anjos LHC, Oliveira VA, Lumbraras JF et crop- land from 1991 to 2001. Glob Chang Biol 10:983–994.
al (2018) Sistema Brasileiro de Classificação de Solos. Embrapa, https://doi. org/10.1111/j.1365-2486.2004.00780.x
Brasília VandenBygaart AJ (2006) Monitoring soil organic carbon stock
Silva ACO, Fonseca LMG, Körting TS, Escada MIS (2020) A spatio- changes in agricultural landscapes: issues and a proposed
temporal Bayesian Network approach for deforestation prediction approach. Can J
in an Amazon rainforest expansion frontier. Spat Stat 9:1–17. Soil Sci 86:451–463. https://doi.org/10.4141/S05-105
https:// doi.org/10.1016/j.spasta.2019.100393 Vicente LC, Gama-Rodrigues EF, Gama-Rodrigues AC, Marciano CR
Silva RO, Barioni LG, Hall JAJ, Moretti AC (2017) Sustainable (2019) Organic carbon within soil aggregates under forestry
intensi- fication of Brazilian livestock production through systems and pasture in a southeast region of Brazil. Catena
optimized pasture restoration. Agric Syst 153:201–211. 182:1–6. https:// doi.org/10.1016/j.catena.2019.104139
https://doi.org/10.1016/j.agsy. 2017.02.001 Vasconcelos SS, Fearnside PM, Graça PMLA, Nogueira EM, Oliveira
Six J, Elliott ET, Paustian K (2000) Soil macroaggregate turnover and LC et al (2013) Forest fires in southwestern Brazilian Amazonia:
microaggregate formation: a mechanism for C sequestration under estimates of area and potential carbon emissions. For Ecol Manag
no-tillage agriculture. Soil Biol Biochem 32:2099–2103. 291:199–208. https://doi.org/10.1016/j.foreco.2012.11.044
https://doi. org/10.1016/S0038-0717(00)00179-6 Wang Y, Ziv G, Adami M, Almeida CA, Antunes JFG et al (2020)
Shi H, Li X, Liu X, Wang S, Liu X et al (2020) Global protected areas Upturn in secondary forest clearing buffers primary forest loss in
boost the carbon sequestration capacity: evidences from economet- the Brazilian Amazon. Nat Sustain 3:290–295. https://doi.org/10.
ric causal analysis. Sci Total Environ 715:1–10. 1038/s41893-019-0470-4
https://doi.org/10. 1016/j.scitotenv.2020.137001 Winchell M, Srinivasan R, Di Luzio M, Arnold J (2008) ArcSWAT 2.1
Soong JL, Janssens IA, Grau O, Margalef O, Stah C et al (2020) Soil Interface for SWAT2005, User’s Guide. Blackland Research and
properties explain tree growth and mortality, but not biomass, Extension
Publisher’s noteCenter, Temple
Springer Nature remains neutral with regard to jurisdic-
across phosphorus-depleted tropical forests. Sci Rep 10:1–13. tional claims in published maps and institutional affiliations.
https://doi. org/10.1038/s41598-020-58913-8
Soares-Filho BS, Nepstad DC, Curran LM, Cerqueira GC, Garcia RA
et al (2006) Modelling conservation in the Amazon basin. Nature
Affiliations
440:520–523. https://doi.org/10.1038/nature04389

Júnior Melo Damian 1 & Mariana Regina Durigan 2 & Maurício Roberto Cherubin 1 & Stoécio Malta Ferreira
Maia & Stephen M. Ogle & Plínio Barbosa de Camargo 5 & Joice Nunes Ferreira6 & Raimundo Cosme de Oliveira
3 4

Júnior 7 & Carlos Eduardo Pellegrino Cerri 1

1 4
Department of Soil Science, “Luiz de Queiroz” College of Natural Resource Ecology Laboratory, Colorado State University,
Agriculture, University of São Paulo, Piracicaba, SP 13418-260, Fort Collins, CO 80523, USA
Brazil 5
Center for Nuclear Energy in Agriculture, University of São Paulo,
2
Department of Entomology and Acarology, “Luiz de Queiroz” Piracicaba, SP 13400-970, Brazil
College of Agriculture, University of São Paulo, 6
Embrapa Amazônia Oriental, Belém, PA 66095-100, Brazil
Piracicaba, SP 13418-260, Brazil
7
3
Núcleo do Médio Amazonas, Embrapa Amazônia Oriental,
Instituto Federal de Educação, Ciência e Tecnologia de
Santarém, PA 68035-110, Brazil
Alagoas, Marechal Deodoro, AL 57160-000, Brazil