Auler Et Al 2022

Geomorphology 398 (2022) 108068
Contents lists available at ScienceDirect
Geomorphology
journal homepage: www.elsevier.com/locate/geomorph
Silica and iron mobilization, cave development and landscape evolution

in iron formations in Brazil
Augusto S. Auler a,⁎, Hazel A. Barton b,c,d, Barbara Zambelli e, John Senko b,c,d, Ceth W. Parker f, Ira D. Sasowsky d,
Tatiana A.R. Souza a, Diego Pujoni g, Jorge Peñaranda h, Reed Davis c
a
Instituto do Carste/Carste Ciência Ambiental, R. Barcelona 240/302, Belo Horizonte, MG, Brazil
b
Department of Biology, University of Akron, Akron, OH, United States
c
Integrated Bioscience, University of Akron, Akron, OH, United States
d
Department of Geosciences, University of Akron, Akron, OH, United States
e
Rua Espírito Santo 980/802, Belo Horizonte, MG, Brazil
f
Planetary Protection Center of Excellence, NASA Jet Propulsion Laboratory, Pasadena, CA, United States
g
Laboratório de Limnologia, Ecotoxicologia e Ecologia Aquática, Universidade Federal de Minas Gerais, Belo Horizonte, MG, Brazil
h
CEPAS, University of São Paulo, São Paulo, SP, Brazil
a r t i c l e i n f o a b s t r a c t
Article history: Iron formations display some of the oldest and most enigmatic landforms on Earth, presently restricted to few
Received 31 August 2021 areas of the planet. The original rock, the Banded Iron Formation (BIF) was formed in the Proterozoic and has
Received in revised form 1 December 2021 been continuously weathered through a complex interplay of chemical and biological processes involving the
Accepted 1 December 2021
massive mobilization of the main elements silica and iron, resulting in silica depleted (but friable) high-grade
Available online 6 December 2021
iron bodies and a Fe-rich surficial duricrust known as canga. This study presents new data and includes a compre-
hensive review of the existing literature, incorporating extensive data from unpublished reports and new find-
Keywords:
Iron formations ings. It aims to quantitatively analyze the morphology and development of porosity in iron formations both at
Landscape evolution micro and macro scales, assessing the geobiological mechanisms responsible for iron and silica mobilization
Speleogenesis and their role in the dynamics of this ever-evolving landscape. Silica leaching is the initial process of porosity gen-
Iron caves eration at depth in the phreatic zone and results in a productive aquifer that contributes to the removal of solute
Iron porosity and the generation of initial low-density zones. The development of numerous pores and voids is favored by the
Iron reduction permeability contrast at the shallow contact zone between the canga and the weathered ore. Caves tend to occur
at the base of scarps at the limit of plateaus and ridges, and tend to scale down in size away from the borders. The
landscape evolves through the retreat of scarps, mostly through the collapse of cave passages. Caves exhibit a
characteristic morphology that intercalates larger rooms and smaller connecting passages, suggesting that
macropores started initially as isolated voids and occur either entirely in canga or in the weathered BIF, but com-
monly are associated with the contact between these rocks. Caves in the Amazonian Carajás region display the
largest dimensional values, regardless of bedrock context. Longer caves show a larger number of connections be-
tween rooms, suggesting that the coalescence of rooms mostly through slope interflow processes is a key player
in fostering the development of caves. Fractal Dimension (FD) calculated for the larger caves demonstrates that
there is no relationship between lithology and FD, although values of FD vary between regions, with Carajás dis-
playing the highest ones.
Geomicrobiological processes play a major role in the mobilization of iron through Fe(III) reducing bacteria. In
particular, it promotes the long-term stabilization of voids through the generation of a hardened surface that pro-
tects the walls of caves and voids from collapse. The interplay of chemical and bacterial processes allows for the
generation of a highly integrated network of pores and caves that represent outlets for the removal of silica and
mobilization/concentration of iron. These voids are the result of geochemical leaching and evolve synchronously
with the landscape, mostly at scarp level. Such processes connect isolated pores, promote evacuation of solutes
and allow for the slow lateral degradation of the landscape. This geologically slow process is now heavily affected
by mining operations, which reinforces the need for searching for mitigation and/or regeneration pathways in
order to preserve this unique geo-biological ancient landscape.
© 2021 Elsevier B.V. All rights reserved.
⁎ Corresponding author.
E-mail address: [email protected] (A.S. Auler).
https://doi.org/10.1016/j.geomorph.2021.108068
0169-555X/© 2021 Elsevier B.V. All rights reserved.
A.S. Auler, H.A. Barton, B. Zambelli et al. Geomorphology 398 (2022) 108068
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Geology, landforms and environment of if regions in Brazil. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. Carajás . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2. Iron Quadrangle (IQ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3. Southern Espinhaço Range (SE). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4. Weathering profiles and canga . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4. Geological and physiographical control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.1. Microporosity characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.2. Cave-landscape relationship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.3. Geological controls on cave dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5. Morphology of if caves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.1. Fractal Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.1.1. Lithological, geographical and morphological controls on FD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.2. Underground landforms and sedimentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
6. Generation of porosity and speleogenesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
6.1. Initial porosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
6.2. Cave evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
7. Iron mobilization and landscape evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Declaration of competing interest. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1. Introduction commented on these features throughout the world, although detailed

studies did not start until Simmons' (1963) research in Brazil. Caves,
Landscapes in iron formations represent some of the oldest land- being underground features, allow for unique access to key geological
forms and longest exposed rocks on Earth (Vasconcelos et al., 2019). features, which otherwise would tend to be masked by soil or largely
The formation of massive iron deposits took place mostly during the Ar- obliterated by weathering at the surface (Plotnick et al., 2015). In
chean, with the largest volumes being precipitated between 3.5 and the mostly flat canga-capped surface of IF regions, where canga
2.5 Ga (Klein, 2005). Presently, iron formation regions occur in most cli- limits the occurrence of outcrops of subjacent rocks, caves allow for a
matic settings, with major deposits in Brazil, Australia, the Soviet Union, three-dimensional perspective of stratigraphic relationships that is
Canada and eastern/southern Africa (Bekker et al., 2010). Of note is the largely unavailable at the surface. Furthermore, caves represent the
frequent occurrence of an iron-rich duricrust known as canga, which macro dimensional analogue of small-scale (micro) porosity, and as
blankets the landscape, behaving as a protective cover that inhibits de- such provide a window into a vast interstitial habitat characterized by
nudation and preserves, for long periods, the mostly friable subjacent highly heterogeneous pores and vugs, in which a complex interplay of
iron formation bedrock. The peculiar geochemistry of these iron-rich chemical and bacterial processes take place (Levett et al., 2016; Parker
environments results in very long surface exposure ages and low ero- et al., 2013b, 2018; Gagen et al., 2019b).
sional rates (Monteiro et al., 2018a, 2018b) and thus these landforms Besides their unequivocal economic and geomorphological impor-
usually stand as high elevation ridges and plateaus capped by canga. tance, IF ecosystems are home to a unique faunal and floral biodiversity
Iron formations (IF) are here defined following Klein (2005), as a (Jacobi et al., 2007). Canga is a highly porous rock that hosts a diverse
broad term that denotes not only the original rock, the Banded Iron For- sub-surface habitat colonized by fauna adapted to this subterranean en-
mation (BIF), but also a complex suite of weathering byproducts which vironment (Ferreira et al., 2015; Auler et al., 2019; Trevelin et al., 2019).
result from various diachronic and long-lasting phases of BIF weathering. The cave-adapted fauna in iron landscapes is more biodiverse than
The initial stage involves silica/carbonate-removal, which may preserve found in other rock types (Silva et al., 2011), and also comprise a re-
the original rock banding, moving gradually towards a friable and porous markably rich microbiological community (Gagen et al., 2018; Parker
rock due to extensive supergene and hypogene alteration processes et al., 2018; Calapa et al., 2021; Lemes et al., 2021). This porous environ-
(Rosière and Stacey, 2018). These weathered profiles can be as thick as ment includes also humanly accessible voids, here defined as caves.
600 m (Rosière and Stacey, 2018) but are normally in the range of Thousands of caves have been recorded within and under the canga,
50–100 m deep (Ribeiro, 2003). Their development involves iron mobili- representing an as yet little known speleological geo-ecosystem (Piló
zation and commonly the concentration of iron oxyhydroxides, resulting et al., 2015). Surface ecosystems are represented by endemic plants es-
in high grade hematite ore deposits (>66% Fe; Dorr, 1965). The enigmatic pecially adapted to tolerate a high metal content environment (Skirycz
canga cap is a later iron-rich conglomerate and various genetic interpre- et al., 2014; Schaefer et al., 2016; Jacobi et al., 2007, 2015; Zappi et al.,
tations have been proposed (Simmons, 1960; Dorr, 1964; Spier et al., 2019). These ecosystems display higher species richness when com-
2019). Recently, it has been demonstrated that canga evolves through a pared with areas of similar dimensions in other rock types (Gomes
“self-healing” mechanism (Monteiro et al., 2018a), in which iron mobili- et al., 2019). They also host numerous archaeological and paleontologi-
zation and precipitation take place continuously, resulting in a dynami- cal sites (Jacobi et al., 2015; Bittencourt et al., 2015) and have outstand-
cally regenerated crust that protects the friable subjacent IF. ing geoheritage value, comprising geographical landmarks as peaks,
The existence of pores and caves in IFs has been noticed since the historical mining sites and early colonization structures (Ruchkys and
19th century (Pissis, 1849; Henwood, 1871) and numerous studies Machado, 2013; Jacobi et al., 2015). Despite such unequivocal impor-
(e.g. Dixey, 1920; Axelrod et al., 1952; Campana et al., 1964; tance, iron-rich surface and subsurface environments are currently
Marescaux, 1973; Bowden, 1980; Twidale et al., 1985) have later among the most threatened ecosystems in Brazil (Jacobi et al., 2015).
2
The commodity global market has seen a marked increase in iron 2.1. Carajás
prices since the late 2000's and presently most of the IF rocks are eco-
nomically classified as “ore”. In Brazil and elsewhere this has resulted The giant Carajás iron deposit is in the Equatorial Amazon Rainforest
in a dramatic increase in mining activities. The two largest IF regions biome. It comprises a Neoarchean metavolcanosedimentary sequence,
in Brazil, Iron Quadrangle (IQ) in southeastern Brazil and Carajás in mainly composed of BIF represented by jaspilites belonging to Carajás
Amazonia, have been subject to major impacts to canga ecosystems. Formation of the Grão Pará Group, with a minimum age of 2.7 Ga
The IQ has witnessed a significant loss of canga areas, estimated at (Trendall et al., 1998). This sequence lies unconformably over gneisses
50% (Salles et al., 2019) while Carajás has lost approximately 20% of and granitoids of the Xingu Complex. The iron formations are covered
its canga areas since 1973 (Souza-Filho et al., 2019). The remaining by mafic rocks of Parauapebas Formation. The jaspilites show an inter-
non-mined areas are largely either owned by mining companies or in- calation of light and dark bands of iron oxide and silica (jasper and
cluded in future mining plans (Jacobi et al., 2015). Although large chert) of millimetric to centimetric thickness alternating with minor do-
patches of iron outcrops are essential for maintaining plant diversity lomitic rocks and irregular mafic rock intrusions (Lobato et al., 2005).
(Salles et al., 2019), most preserved areas form disconnected and The thickness of this sequence is up to 400 m (Klein and Ladeira,
fragmented zones. Landscape and ecosystem degradation due to iron 2002). As a result of hydrothermal and weathering processes, the iron
mining goes well beyond the immediate vicinity of open cast mines. It content of jaspillite is locally increased to over 64% (Lobato et al.,
has been estimated that iron mining effects in the Carajás area may ex- 2005), forming the iron ore deposits. The most prominent structure in
tend up to 70 km beyond mining lease boundaries and represents ap- the area is the Carajás transcurrent fault, running in the E–W direction,
proximately 9% of all forest loss in the Amazon rainforest (Sonter which divides the iron deposits into North and South ridges (Serra
et al., 2017). Considering the short life span of these mining operations Norte and Serra Sul) (Pinheiro, 1997) besides nearby East Ridge (Serra
(a few decades in most areas — Jacobi et al., 2015), post-mining scenar- Leste) (Fig. 1B).
ios hold the potential to be ecologically catastrophic. Climate in the Carajás region is of the Aw type in the Köppen classi-
There is, thus, a strong need for understanding and protecting this fication, with an annual mean rainfall of ~2000 mm (Alvares et al.,
highly threatened environment. Studying the macroporosity of IF 2014), with 80% occurring between December and May (Silva Junior
areas is crucial to setting up robust conservation and mining strategies et al., 2017). Mean annual temperatures are approximately 26 °C
(Nola and Bacellar, 2021). In Brazil, caves are protected by the constitu- (Alvares et al., 2014). The rupestrian vegetation over the canga plateaus
tion (Brasil, 1988), together with a protection buffer zone (Brasil, 1990; comprises metal tolerant species adapted to restrictive edaphic condi-
Auler and Piló, 2015). Due to the ubiquitous occurrence of caves in IF re- tions (Schaefer et al., 2016). The canga is bordered by dense Amazonian
gions, this represents often the most straightforward way of protecting rainforest growing over iron scree and soils derived mostly from the ig-
adjacent IF landscapes (Auler, 2016), even though up to 20% of these hy- neous rocks. The Carajás region comprises a series of canga-capped pla-
pothetically protected buffer zones have been impacted in the IQ teaus of various sizes and shapes. Canga also occurs at the lower slopes,
(Gomes et al., 2019). Environmental compensation related to caves sometimes associated with fluvial valleys. The low relief top of the pla-
has also resulted in the protection of significant IF areas at both the teaus shows undulating surfaces and a few lakes filling depressions
state and federal level, including the creation of new National Parks alike shallow dolines. The original area of canga-capped surfaces in
(Mota et al., 2018; Souza-Filho et al., 2019; Salles et al., 2019). Carajás was 144 km2 in 1973, prior to the beginning of mining opera-
This research provides a review of the mechanisms of silica leaching tions (Souza-Filho et al., 2019).
and iron mobilization within IFs and the generation and development of
porosity at both the micro and macro scale and how these processes af- 2.2. Iron Quadrangle (IQ)
fect the evolution of IF landscapes in the three most significant IF areas
in Brazil, the Carajás area in northern Brazil, the Quadrilátero Ferrífero The iron deposits of the IQ, located in the mountainous highlands of
(Iron Quadrangle — IQ) and the Southern Espinhaço Range (SE) areas southeastern Brazil, have been subject to scientific studies since the
both in southeastern Brazil (Fig. 1). We have incorporated unpublished 19th Century (Eschwege, 1822) which detailed the stratigraphy
data about hundreds of caves, and provided an integrated model for (Derby, 1910; Dorr, 1969) and provided the first insights on the geo-
their genesis, involving both geochemical and microbiological mecha- morphology of IF (Dorr, 1964, 1965, 1969). The area is characterized
nisms. Our review also provides a link between IF porosity and the evo- by a complex geological and tectonic evolution that involved hypogene
lution of iron-rich plateaus in Brazil, which are directly connected with and supergene iron mobilization (Rosière et al., 2018) and presently dis-
the evolution of high grade iron ore. The three study areas constitute plays a series of ridges formed by the more resistant IFs (Fig. 1D). The
over 90% of the iron production in Brazil (ANM, 2020) and are home iron-rich beds belong to the Cauê Formation and are locally known as
of 2979 IF caves, approximately 95% of the total number of IF caves re- itabirites. The BIF was deposited between 2.52 and 2.42 Ga (Babinski
corded in Brazil's official database (CANIE, 2021). A better understand- et al., 1995) and comprises interbedded quartz and hematite bands
ing of these processes is important not only for conservation and (with occasional carbonate), but also dolomite, magnetite, amphibole
management purposes, but also for providing insights on the dynamics and lenses of marble and phyllite (Dorr, 1969). The thickness of the
of iron mobilization and enrichment that generates high-grade iron iron formation is highly variable, originally reaching approximately
deposits. 250–300 m (Spier et al., 2006) but now may be as much as 1400 m
due to tectonic processes (Dorr, 1964).
Climate in the IQ is humid subtropical, Cwa to Cwb according to
2. Geology, landforms and environment of if regions in Brazil Köppen and mean annual rainfall ranges between 1300 and 1600 mm
(Alvares et al., 2014) concentrated during the austral summer. Shrub-
The three major IF regions in Brazil (Fig. 1), Carajás, Iron Quadrangle dominated vegetation over IFs exhibits a floristic identity, with
(IQ), and Southern Espinhaço Range (SE), display canga deposits lying extremely high local and regional diversity (Carmo and Jacobi, 2016).
over friable altered IFs and are the focus of this work. Environmental Unlike Carajás, there is no sharp vegetational transition away from the
studies for iron mine expansion have led to the identification of thou- canga-capped ridgetops, since early European colonization has re-
sands of caves and catalyzed research on the genetic relationship be- moved much of the original forested cover. Besides covering the ridge
tween canga, weathered BIF and associated porosity. Although sharing tops, IFs also occur in lower and less steep flanks. Anthropogenic im-
geological similarities, these areas are presently located under distinct pacts (mining, groundwater exploitation, agriculture, urbanization and
climate regimes, which result in marked surface and subsurface geo- deforestation) have profoundly altered the IQ landscape. The original
morphic and ecological differences. exposure of IFs in the 1960's in the IQ was 186.4 km2 (Salles et al., 2019).
3
Fig. 1. (A) Location of the three study areas (IF outcrops in red) and sites mentioned in the text. (B) Carajás; (C) Southern Espinhaço and (D) Iron Quadrangle. Key for Carajás: 1. South
Ridge; 2. North Ridge; 3. N1 (a plateau that belongs to North Ridge); 4. Bocaina plateau (part of South Ridge); 5. East Ridge.
2.3. Southern Espinhaço Range (SE) 2.4. Weathering profiles and canga
The Southern Espinhaço Range IFs (Fig. 1C) are located close to the The IFs as exposed presently in all three regions are the result of
IQ, and thus share many of the same climatic and vegetational charac- massive long term weathering processes and groundwater-related
teristics. The Fe-rich deposits comprise two distinct sequences sepa- leaching of constituents (Rosière et al., 2008; Rosière and Stacey,
rated by an erosional unconformity. They belong to the Serra do Sapo 2018). These processes have considerably modified the original BIF, re-
Formation of the Serra da Serpentina Group, and the Canjica Formation sulting in a heterogeneous and non-coherent rock, colloquially termed
of the São José Group, with depositional ages respectively at 1.99 and “hard” and “soft” ores, which relates not only to its friability but also
1.66 Ga (Rolim et al., 2016). Both BIF formations exhibit similar charac- to whether it is dominantly hematite or goethite/other Fe-hydroxides
teristics with mm to cm thick alternate hematite-rich and quartz-rich (Rosière et al., 2008). Particularly in the IQ, leaching of carbonates and
banding. The Serra do Sapo Formation has an average iron content of silica are believed to have resulted in low density collapse-prone
32.5 wt%, but contains some higher grade ore bodies (Rolim et al., zones and the deepest weathering profiles (up to 400 m) tend to
2016). The thickness varies from a lower value of 15 m in the southern- occur beneath the high grade iron bodies (Ribeiro et al., 2021). In
most portions to up to 350 m in the northern sector (Rosière et al., Carajás the thickest weathering profiles are associated with the
2018). As in the IQ, the SE deposit outcrops mostly at the top and flanks doline-like depressions occupied by lakes (Ribeiro, 2003).
of irregular roughly N-S trending ridges. Total area of iron outcrops, as Characteristic of all three areas (and elsewhere in Brazil and several
estimated from Rolim et al. (2016) is 34.2 km2. Cattle and dairy farming IF areas in the world) is the occurrence of canga, an iron-rich duricrust
have altered the original vegetation. Iron mining operations in the SE that caps and protects the underlying friable rock from erosion. The
started in the early 2010's, thus most of the IF deposits are still overlying canga in average contains over 90% Fe2O3 and is composed
preserved. of fragments inherited from the BIF, cemented by goethite, minor
4
gibbsite and rare manganese oxides and phosphates (Spier et al., 2019). 1831 caves). This data set is restricted to caves developed entirely in
Thickness varies from a few centimeters to more than 30 m, averaging IFs. Caves at the contact with other rock types (such as mafic rocks com-
1–2 m (Dorr, 1964). Dorr (1964) realized that canga must be continu- mon in Carajás) were discarded as well as caves in Fe-rich soil classified
ally renewed, otherwise it would have been removed over geological as ferricrete. For each cave, besides a detailed survey map, general infor-
time. This has been confirmed by 40Ar/39Ar and (U/Th)/He dating of mation about the dimensions (length, depth, area, volume), local geol-
canga in Carajás and IQ which has yielded stratigraphically non- ogy, cave pattern, cave geological features were provided, allowing for
coherent ages (Monteiro et al., 2014; Spier et al., 2006; Monteiro et al., a quantitative assessment of the relationship between these parame-
2018a, 2018b) showing that iron mobilization and reprecipitation has ters. For cave morphometry we have defined a 50 m minimum length
occurred over millions of years (Vasconcelos and Carmo, 2018; Levett for extracting meaningful morphometric information, which resulted
et al., 2020a, 2020b, 2020c). Erosion rates in the weathering profiles in 238 caves, being 185, 36 and 17 respectively in Carajás, IQ and SE.
are in the range of 0.20–0.24 m/Ma for Carajás (Monteiro et al., Statistical analyses were performed using Microsoft Excel® and R
2018b) and 0.17–0.31 m/Ma for three ridges in the IQ (Monteiro et al., software (v.3.6.1) (R Core Team., 2019). Relations between morphome-
2018a), which represent extremely low values, over two orders of mag- tric variables were quantified and tested by the calculation of Pearson
nitude lower than the world's average based on drainage basins correlation index. In addition, a linear regression was performed in
(Sasowsky, 2007), helping explain the geomorphic resilience of IF pla- order to model the relation between cave area and volume. Percentilic
teaus and ridges. confidence intervals were calculated through a quantile regression.
Several hypotheses have been proposed to explain the genesis of Fractal analysis was applied to the 238 caves larger than 50 m. The
canga (see review in Simmons, 1960 and Dorr, 1964), involving per aim was to investigate how the Fractal Dimension (FD) relates to the
ascensum or downward/laterally water flowing processes; in situ hydra- cave's morphology, lithology, size (length and area), and location. For
tion of hematite and partial leaching of quartz; replacement of former this, the first step was to perform a 2D fractal investigation in order to
soil horizons, among others. The role of bacteria, first commented by calculate the Fractal Dimension. Caves were analyzed using two differ-
Dorr (1964) in the IQ but also by McFarlane and Twidale (1987), has ent software, ImageJ with FracLac plugin (https://imagej.nih.gov/ij/
been confirmed by Parker et al. (2013a, 2013b) and Levett et al. plugins/fraclac/FLHelp/Introduction.htm), and Fractal Analysis 3.4.7
(2016), showing that microbially mediated redox processes are an im- (http://cse.naro.affrc.go.jp/sasaki/index-e.html). Since this software
portant catalyst of Fe mobilization and recycling. The presence of plant only accepts binary and grayscale bitmap files as input, it was necessary
structures preserved within canga shows that vegetation can also signif- to perform the conversion of all cave maps, originally in jpg or pdf for-
icantly influence the evolution of canga (Gagen et al., 2019b; Paz et al., mat, to an 8-bit image (where white would represent the space inside
2020). The complex, and still little known, interaction of biological and the cave – voids – and black everything else – pillars, rocks, cave
chemical processes is critical for the understanding of canga dynamics, walls). This conversion was performed with Adobe Illustrator for vector
high-grade ore preservation and landscape evolution in IF regions. images (pdf) and with Power Point and Paint for non-vector ones. The
process consists in erasing details inside the caves such as drainage,
3. Methods speleothems, blocks and contours, and leaving only the external walls.
If the cave has more than one level, its projected maximum extension
This review article made use of extensive data from consulting and was adopted. After the conversion, all the images were saved in the
mining companies, besides speleological studies made by professional same size and resolution (512 × 256 points, 150 dpi) for standardized
and amateur caving clubs. The integration between these comprehen- analysis, since images with different formats, sizes or resolution would
sive archives of geological and speleological data allowed for insights result in diverse Fractal Dimensions. An example of this output is
on the genesis and relevance of caves to both the long term geomorphic shown in Fig. 2.
evolution of IF regions and the generation of high grade ore. The software (ImageJ/FracLac and Fractal 3.4.7) use boxcounting
Several 51 mm diameter, approximately 30 cm long, cores were (box sizes range from 2 to 1024 pixels) and further linear regression
drilled perpendicular to the surface in canga areas in both the IQ and to calculate FD. For FracLac, the number of boxes was fixed, while for
SE areas using a Shaw portable core drill (www.backpackdrill.com), to Fractal 3.4.7, the number of boxes varied from cave to cave and was
allow for the measurement of permeability. A total of 24 cores were se- not clear in the calculations. The number of boxes is important because
lected and cut both perpendicular and parallel to the surface. Permeabil- it directly affects the FD calculation. After calculating FD, statistical anal-
ity measurements were made with a TinyPerm air permeameter yses were performed using Microsoft Excel to verify possible correla-
(http://www.ner.com/site/systems/item/27-tinyperm.html) from tions. The relationship involving cave length, area and FD were
CEPAS, University of São Paulo. Five measurements were performed in investigated by means of Pearson correlation.
each sample, totaling 125 measurements. We imaged the BIF, canga and cave wall samples using a Nikon XTH
For our cave data set we have adopted Brazil's official definition of 320 CT system (Nikon Metrology Inc., Brighton, MI, USA), housed in the
cave, an anthropocentric concept that restricts caves to natural voids National Center for Education and Research on Corrosion and Materials
enterable by humans (Palmer, 2007). A consensual 5 m minimum Performance (NCERCAMP) at The University of Akron. Settings used for
length has been applied for most consulting works in Brazil. Although the scan can be found in Table 1. We reconstructed the CT images of
this leaves a two order of magnitude gap between the pores analyzed each sample using NRecon (v.1.7.4.2, Bruker, Kontich, Belgium) by fol-
through core drilling and the caves available for direct inspection, a di- lowing the manufacturer's instructions to import a non-proprietary
mensional continuum can be observed. We treat the larger caves as dataset and saved the output as 16-bit TIFF images to maintain the
macro dimensional analogues resulting from the development of range of grey values in the images. Following reconstruction of the CT
smaller voids. Such transition can be observed inside caves, which are images, we loaded the images into DataViewer (v. 1.5.6.2, Bruker) to
subject to the same processes of iron mobilization. align the reconstructed sample vertically. This step ensures that pores
The Brazilian National Cave Database (CANIE — https://www. are not measured obliquely during morphometry and provides more
icmbio.gov.br/cecav/canie.html) was used for obtaining general loca- accurate data. Following alignment, we employed CTAnalyser
tional information about IF caves. Much of our analyses was based on (v. 1.19.11.1, Bruker) to crop a volume of interest (VOI) near the center
speleological reports by the consulting company Carste Ciência of each sample. We did this to ensure equal sample size and to minimize
Ambiental. These reports have been archived in the national or state en- the effects of artifacts near the edges of the samples. The final size of the
vironmental agencies and in public domain websites. Data from Carajás VOI for each sample was 3.44 × 2.67 × 10.13 mm. Using these new
can also be accessed in Jaffé et al. (2016, 2018). Our data set comprised a VOIs of each sample, we gathered morphometric data of the rock and
total of 1155 caves in Carajás, 405 in the IQ and 271 in the SE (total of pore phases of the sample using built-in plugins of CTAnalyser.
5
Fig. 2. (A) Original map (pdf format) of a cave. (B) 8-bit grayscale resulting image used to perform the fractal analysis.
These morphometric data included total volume (TV), rock volume with flow being controlled by relict banding and foliation (Bertachini
(RV), percent rock volume (RV/TV), rock surface area (RS), rock surface et al., 2018). Permeability values of 10−8 m/s have been reported
to volume ratio (RS/RV), pore diameter (PD), pore separation (PS), (Mesquita et al., 2017), and values can be as much as three times higher
number of closed pores (CP), volume of closed pores (CV), volume of along the banding (Mourão, 2007). The upper boundary of the friable
open pores (OV), percent open pores (OV/TV), percent closed pores ore is mostly at high elevations, entirely in the vadose zone, and nor-
(CV/TV), and percent total porosity ([OV + CV] / TV, TP). The specific mally does not store water, except temporarily during rainfall events.
tasklist used was determined through extensive trial and error testing Canga, being a later byproduct, does not display banding or foliation,
and can be found in the Suppl. Table 1. Three-dimensional (3D) renders although it is common to show irregular unloading joints (sensu Allaby,
of the rock and the pore systems were visualized using Dragonfly 2013) parallel to the surface. Canga is mostly impermeable at the sur-
(v. 4.1.0.647, Object Research Systems, Montréal, Canada). face (Dorr, 1964) but is highly porous, displaying large and irregular
vugs that form a continuum towards larger caves and represent a habi-
4. Geological and physiographical control tat for subterranean adapted fauna (Auler et al., 2019). Our cores show
the preferential concentration of vugs along specific horizons, probably
4.1. Microporosity characterization controlled by redox fronts or perched water levels. Unloading also fa-
vors the expansion of surface-parallel voids. Permeability measure-
The two types of BIF weathering products, canga and the friable bed- ments in the IQ and SE (Supplementary Material Table 2) show a high
rock underneath, are highly porous, but in distinct ways. Due to the degree of anisotropy (Fig. 3), with the largest values being along
mostly unconsolidated nature of the friable iron bedrock, porosity and horizons with preferential concentration of vugs. Measurements
aquifer storativity increases with degree of weathering, but permeabil- perpendicular to these horizons were so low as to prevent obtaining
ity tends to decrease (Bertachini et al., 2018). Porosity values between meaningful data. Average values of hydraulic conductivity in the IQ
24 and 29% (Costa and Sá, 2018) and 14 and 24% (Dias and Bacellar, (8.00 × 10−4 m/s), although in the same order of magnitude, are
2021) have been reported, depending on the type of canga and area. twice as large as in the SE (4.04 × 10−4). This may reflect local differ-
The friable ore behaves as a heterogeneous and anisotropic aquifer, ences between drilling sites. Values between 2.31 × 10−6 and
2.80 × 10−4 have been reported for Carajás (Bertachini et al., 2018).
Table 1 The high permeability of canga is comparable to unconsolidated rocks
Scan parameters used for CT imaging. such as gravel, highly fractured rocks or even karstified carbonate
(Fig. 3), but unlike in those rocks, it is developed through the dynamic
Scan parameter Scan setting
generation of vugs by chemical and biological processes.
Source voltage 220 kV The extreme permeability contrast between the highly porous canga
Source current 240 μA
Filter 1.0 mm tin
and the less permeable lower friable ore results in the temporary stor-
Pixel size 15.98 μm age of water in this contact zone (Bertachini et al., 2018). The hydraulic
Rotation step 0.11° gradient tends to increase towards the borders of plateaus and ridges,
Exposure 708 ms and thus the horizon represented by the canga/ore contact is a focal
Total sample rotation 360°
point for water movement and solute removal.
6
surrounding rocks is marked by small scarps. IF scarps are ubiquitous

landforms and represent sharp breaks in the topography (Figs. 4, 5B).
Scarps are typically between 2 and 10 m high, run perpendicular to
the slope, marking the ridge/plateau limit, are irregular both vertically
and longitudinally and can extend laterally for several hundreds of me-
ters. Scarps in Carajás tend to be more extensive and higher than in the
other two areas. Scarps are undoubtedly among the most important
geomorphological features of IF areas. They host the vast majority of IF
caves, denoting a clear association between caves and plateau/ridge
boundaries (Fig. 4). Generally, there are three situations in which
caves are not associated with scarps: (i) when located at inner lake bor-
ders as in the South Ridge of Carajás; (ii) when located in the interior of
plateaus and ridges, normally linked to the collapse of canga cave ceil-
ings or associated with minor outcrops; (iii) when associated with flu-
vial valleys in the lower slopes. These alternative locations tend not to
include the larger and more representative caves.
Scarps are also important because they represent the dynamic
boundary of plateaus and ridges (Vasconcelos and Carmo, 2018;
Fig. 3. Boxplot of hydraulic conductivity values obtained from 125 measurements in canga Vasconcelos et al., 2019) since the landscape evolves, albeit very slowly,
samples from the IQ (left) and SE (right). Values are highly heterogeneous as they reflect through scarp disintegration and retreat. Scarps commonly mark the
the concentration of vugs along specific horizons. For the complete set of values see
contact between the permeable canga and the less permeable altered
Supplementary Material Table 2.
Range of values for other rock types based on Freeze and Cherry (1979). ore, which act as a focal horizon for iron evacuation through discharge
of a temporary perched aquifer (Bertachini et al., 2018). The vast major-
ity of caves in all three areas open at the scarp's face (Fig. 4). Collapse of
4.2. Cave-landscape relationship cave ceiling at the scarp can result in the arcuate retreat of the scarp, a
feature especially common in Carajás, termed as amphitheater
The canga cap that overlies the plateaus and ridges tends to be con- (Fig. 5C). Of note is the frequent occurrence of previously non-
tinuous, generating an undulating landscape with few protruding out- described crescent-shaped “cusps” (Fig. 5A) at the border of Carajás
crops. In the IQ and SE the IFs are in contact with quartzites, a plateaus. They are largely responsible for the irregular pattern of some
geochemically resistant rock that is also subject to very low (<3 m/ plateaus. These ubiquitous features, although an order of magnitude
Ma) denudational rates (Salgado et al., 2008). The quartzites help sup- larger, bear resemblance to cave amphitheaters. The possibility that
port the ridge topography typical of these areas. In Carajás, the canga cusps are formed through coalescence of smaller amphitheaters
plateaus are surrounded by igneous rocks, which weather quickly would place cave processes as major triggers of IF landscape evolution,
under the equatorial climate. This results in sharp changes in elevation a topic that deserves further research. Soil washed from the caves
(approximately 300 m) between plateau tops and the surrounding low- generates a small vegetation-rich flat zone at the scarp front,
lands. In all three areas, the contact between the canga cover and the representing a break in the steep profile of the slope.
Fig. 4. Aerial view of N1 Plateau, North Ridge, Carajás with the canga outcrop (light grey) surrounded by Amazonian rainforest. The image shows that the location of cave entrances (yellow
dots) are nearly always associated with scarps at the limit of the canga cap. Larger (>50 m long) caves are marked with a red dot. Image from Google Earth and cave locations from the
Brazilian cave database (CANIE).
7
Fig. 5. (A) Sequential “cusps” commonly mark the boundary of canga plateaus as shown in this oblique SE trending view of the South Ridge in Carajás. They may represent macroscale
analogues of cave amphitheaters. Google Earth image with 3× vertical exaggeration. (B) A typical scarp in the IQ. Arrow points to a person inside the cave. (C) Cave amphitheater
associated with collapse of an entrance room in Carajás.
4.3. Geological controls on cave dimension
Table 2 updates a previous study (Auler et al., 2019) and allows for a
dimensional analysis of caves in terms of geological context for the Table 2
Dimensional data for caves in Carajás, IQ and SE.
study areas. Caves can occur (i) entirely within canga, (ii) within weath-
ered BIF, or (iii) at the contact between these two IF rock types. Our cave Carajás
data set exhibits a large dominance of small caves (median length of n Length Depth Area Volume Area/length Volume/length
15.9 m), with the longest cave being only 372 m long. All three areas dis- (m) (m) (m2) (m3)
play the same distribution pattern, with a limited number of caves being Canga 552 26.13 3.53 102.9 183 3.94 7.00
over 100 m long (Fig. 6). Mean/median values for Carajás, IQ and SE are Contact 255 52.08 6.49 205.1 404 3.94 7.76
respectively 32.5 m/18.5 m, 21.0 m/12.4 m and 18.3 m/12.2 m. Alt. BIF 348 28.35 3.80 102.8 204 3.63 7.20
The Carajás region has the larger caves in all three dimensional pa- Total 1155 32.53 4.27 125.4 239 3.85 7.35
rameters (length, area and volume) (Table 2, Fig. 7). In Carajás, the
Iron Quadrangle
area/length and volume/length ratios are significantly higher than in
the IQ and SE, which display similar values. The high values for cave vol- n Length Depth Area Volume Area/length Volume/length
(m) (m) (m2) (m3)
ume in Carajás caves are particularly conspicuous, with volume per unit
length over twice as much (mean of 7.3) as in the other two areas Canga 170 20.19 2.51 51.0 57 2.53 2.82
Contact 79 27.47 3.02 66.2 82 2.41 2.99
(mean of 3.1 and 2.8 respectively for IQ and SE). The distribution of
Alt. BIF 155 18.65 4.08 48.4 68 2.60 3.65
caves regarding lithological context (Table 2) shows that dimension Total 404 21.02 3.21 53.0 66 2.52 3.14
positively correlates with the contact between canga and the weathered
BIF (Fig. 8). This lithological contact presents the largest caves in all Southern Espinhaço
three areas, although in Carajás and IQ it does not hold the largest num- n Length Depth Area Volume Area/length Volume/length
ber of caves. Dimensional difference is especially significant in Carajás, (m) (m) (m2) (m3)
with mean values being nearly 100% larger for caves located at the con- Canga 57 14.38 2.21 34.1 34 2.37 2.36
tact, when compared to caves entirely in weathered BIF or canga Contact 109 22.04 3.89 57.8 53 2.62 2.40
(Table 2). Caves in canga and in weathered BIF display similar mean Alt. BIF 105 16.44 3.31 44.5 58 2.71 3.53
Total 271 18.27 3.31 47.7 51 2.61 2.79
values. This may be explained by the fact that the canga cap is of limited
8
contact. This is probably due to the fact that the canga layers tend to fol-
low the slope and these caves normally display inclined passages (see
Section 5.1 and Fig. 11).
An analysis of cave maps shows that the cave's plan configuration is
spatially linked to the scarps. Caves normally develop at a right angle
relative to the scarp line, but never extend beyond 100 m perpendicular
(up-slope) from the scarp. This is a clear indication of a genetic relation-
ship between caves and scarps as will be discussed later.
5. Morphology of if caves
Most caves in our data set are too small to reveal characteristic mor-
phological features or even a more general plan pattern. We have thus
selected caves which are longer than 50 m, totaling 238 caves for this
analysis. Examining the plan map of these caves it becomes clear that
they do not conform with the widely used patterns for carbonate
caves which are largely controlled by groundwater recharge character-
Fig. 6. Cumulative frequency graph (log scale) of cave length at Carajás, IQ and SE showing
that small caves predominate in all areas. Light colored upper horizontal line represents istics and dominant initial porosity (Palmer, 1991). IF caves are mostly
mean values and lower green line the median. extremely irregular, even when predominantly rectilinear. Most long
caves are organized as a series of highly irregular rooms connected by
narrow passages (Fig. 9). Piló and Auler (2005) have termed this pattern
as “globular”, but in reality these rooms do not adhere to a spherical
thickness, somewhat restricting the dimensional expansion of caves, globe-like or bulbous morphology. Instead, they are commonly asym-
while the lower altered IF is less cohesive and may not be stable enough metric and display a series of small conduits and projections at floor
to sustain large continuous voids. and walls that result in a highly jagged perimeter. Cave plan morphol-
In Carajás and SE the difference in elevation between the highest and ogy appears to comprise variants of only one situation, represented by
lowest point in the cave (depth) is more pronounced for caves at the the coalescence of larger rooms and narrow connecting passages. They
Fig. 7. Dimensional analysis of caves in Carajás, IQ and SE. Carajás has the largest caves (high volume/area ratio — see also Table 2). Linear regression for each region is shown in the
corresponding color. Equations are y = 2.1135x − 26.606 for Carajás; y = 1.3629x − 6.294 for IQ and y = 1.556x − 23.18 for SE. Blue and red dashed lines represent respectively
99% and 95% percentilic confidence intervals.
9
vertical collapse shafts, being originally devoid of a natural entrance.

In Carajás approximately 50% of caves display an amphitheater at the
entrance, showing that scarp retreat intercepted a room (Table 3 and
Fig. 5C). On the other hand, narrow entrances are much more common
in the IQ and SE areas (Table 3).
We have determined the number of rooms and narrow passages for
each cave for our data set of 238 caves longer than 50 m, according to
the example provided in Fig. 10. Data for the three areas (Table 4)
shows that Carajás caves have, by far, the largest mean room area, al-
though they present, overall, fewer rooms. Again, IQ and SE show ap-
proximately similar values, being slightly larger in SE. Caves located at
the contact between canga and weathered iron formation always
show smaller rooms, regardless of the region (Table 4). Caves in weath-
ered BIF or canga show similar values.
Longitudinal profiles of IF caves display strong geological control
(Fig. 11). Caves in the contact are dominantly characterized by ascend-
ing passages (Fig. 12) that follow the interface of canga (at ceiling) and
weathered BIF (floor). These caves may display sections entirely in
canga or in weathered BIF, and still retain their steep profile. Caves lo-
Fig. 8. Mean length of caves according to geological situations, showing that the contact
cated entirely in weathered BIF tend to show a more level profile.
between canga and weathered BIF favors longer cave development. Caves entirely in canga will conform to the local angle of the canga layer.
may develop both along one axis (normally following the slope gradi- 5.1. Fractal Dimension
ent) displaying a rough rectilinear pattern that comprises the longitudi-
nal connection of a series of connected longitudinal once independent Fractal is a multiscale geometry element used to describe forms in na-
rooms (Fig. 9B) or through lateral connections (Fig. 9A). Caves in ture that are too complex and irregular to be described using simple
canga display a more irregular pattern (Fig. 9C) both in plan and profile. Euclidean forms. Fractals were first described by Mandelbrot (1977)
In caves with multiple rooms, the narrow connection, if along the slope, and further discussed in Mandelbrot (1983). Since then, many authors
is usually characterized by a steeper passage, while the rooms tend to be have used Fractal Dimension (FD) to analyze natural morphological fea-
flat floored. The alternation between rooms and narrow passages also tures such as hillslopes, river patterns and cave morphology (Curl, 1986;
occurs at entrance level. Many caves show restricted entrances at the Laverty, 1987; La Barbera and Rosso, 1989; Chase, 1992; Sagar and
base of scarps (Fig. 9D), while others can only be accessed through Chockalingam, 2004; Kambesis et al., 2016; Pardo-Igúzquiza et al., 2018).
Fig. 9. Examples of cave plan patterns. (A) Cave in the North Ridge (Carajás) developed at the contact between canga and iron formation showing connection of multiple rooms by smaller
passages both laterally and perpendicular to the scarp. This is the longest cave known in IFs with its 372 m in length. (B) A rectilinear cave in Bocaina plateau, Carajás, shows coalescence of
rooms along one main axis. (C) A highly irregular cave in canga also at Bocaina ridge in Carajás. (D) An example of a cave in the IQ with very narrow distal opposite entrances. E — entrance.
10
Table 3 Table 4
Relative occurrence of entrance types (narrow entrances or amphitheaters/bisected Relative proportion of rooms in each geological context for the three IF study areas. Room
rooms) in each geological context and IF region. Situations where the entrance is neither area represents the maximum area, because the calculations include the entire area of the
a clear amphitheater nor narrow are described as “Other”. cave, since there is considerable imprecision in removing smaller passages (which how-
ever represent a minor portion of the total area) from cave maps.
Amphitheater Narrow Other Total
Region Geological Mean Mean Mean Mean Mean
# % # % # %
context cave number room room number
Carajás 95 51% 68 37% 22 12% 185 area of area area per of
IQ 10 28% 19 53% 7 19% 36 (m2) rooms (m2) region rooms
SE 5 29% 9 53% 3 18% 17 (m2) per
Total 110 46% 96 40% 32 13% 238 region
Carajás Canga 469.8 4.8 111.6 97.9 5.5

Contact 475.3 6.1 86.7
BIF 398.3 5.1 101.8
The results obtained for 2D Fractal Dimension (FD) of caves are IQ Canga 228.5 7.8 34.8 33.8 7.4
shown in Table 5, segregated by lithology, region and software. For Contact 169.9 8.9 23.5
FracLac, FD ranges from 1.3475 to 1.7653. The mean FD value obtained BIF 211.4 6 44.2
is 1.66 (rounded), regardless of the lithology. However, the FD range SE Canga 331.9 8 41.5 44.5 6.6
Contact 236.2 7.1 34.2
is lower for canga caves (0.2726) and larger for caves in BIF (0.4121).
BIF 303.6 5.2 67.9
When comparing the three different regions, Carajás has the highest
FD (mean of 1.6724) and the lowest amplitude of measurements
(0.2726), while SE shows the lowest FD (mean of 1.5943) and the larg- character (rugosity) of the walls does not show clear interaction with
est range (0.3546). IQ values lie in between (1.6469 — mean, 0.3045 FD, nor does with the presence/absence of pillars. Caves with a preva-
range). FD results obtained by Fractal 3.4.7 are slightly different, ranging lence of large rooms tend to have a higher FD than the ones with dom-
from 1.4278 to 1.7981. As for Fractal 3.4.7, the mean FD value obtained is inance of narrow passages. Pearson correlation was performed and the
1.67 (rounded) regardless of lithology. Similarly, the largest FD range results are displayed in Table 6. According to the results, it is not possi-
relates to caves in BIF (0.3471), although the lowest variation is ob- ble to relate directly the size of the cave to the FD.
served at caves developed in the contact of Canga and BIF (0.2128). Con-
cerning the studied regions, Carajás showed the highest FD (mean of
5.1.1. Lithological, geographical and morphological controls on FD
1.6781), while the IQ and SE show similar values with FD (mean) of
The 2D Fractal Dimension (FD) represents how a shape (in this case
1.6499. The FD ranges are the lowest for the IQ (0.2471) and the largest
a cave layout) adheres to a predefined space (2 × 1 proportion,
for SE (0.3115).
512 × 256 pts). On one hand, the closer to 1 FD, the more likely that
Caves that develop in only one direction (linear caves) have lower
the shape will resemble a straight line. On the other hand, the closer
FD than caves that develop in two or more directions. The irregular
to 2 FD, the space is more effectively filled by the cave layout (voids).
The results show that the lithology where the cave develops does
not define its FD. On the contrary, the FD distribution is similar among
the three rock types (Fig. 13). This is probably because iron caves are
highly irregular, thus the lithology does not play a major role in defining
the cave's fractal morphology.
On the other hand, the location, whether in Carajás, IQ or SE does
play a role in its size and shape, hence in its FD (Fig. 14). Even though
it has the largest sample, the FD variation for Carajás was the smallest
among the three regions.
The role of cave morphology in the FD is shown in Fig. 15. The lowest
and the highest FD values are shown at the left column. The right col-
umn shows three different caves, from the same region (Carajás),
same lithology (contact) and with similar FD. It is possible to infer by
Fig. 10. (A) Plan view showing an example of the distinction between rooms, marked with Fig. 11. Distribution of cave profiles in relation to geological context for the three areas
“O”, and narrow connecting passages (X) for a cave in the IQ. (B) Schematic profile view. combined. There is a clear dominance of steep ascending profiles in caves developed in
E — entrance. the contact zone. I, III and V — inclined profile. II, IV and VI — flat profile.
11
Fig. 12. Two representative ascending profiles. (A) Contact cave in SE, showing ascending profiles in both divergent passages. (B) Cave in the contact zone in IQ, showing ceiling collapse in
the initial lower section.
analyzing this figure that caves with very different plan patterns can dis- the rock banding (in caves in weathered BIF) are common. The bed-
play similar FD. It is thus not possible to use FD alone to describe pattern rock floor of IF caves is usually covered by sediment, but where ex-
or specific genetic processes taking place in IF caves. posed, can be very irregular, with jagged pinnacles (Fig. 16B) and
Cave dimension does not affect the FD as shown by the Pearson cor- basins (Fig. 16C). At the ceiling, the presence of pendants and pil-
relation values (Table 6). This could be associated with the fact that frac- lars is characteristic, notably in canga (Fig. 17) which results in ex-
tal is a multiscale geometric element, meaning that the same shape tremely irregular roofs. Pillars and pendants can be up to 3 m high,
would result in the same FD, regardless of its size. More important being limited by the cave height. Pillars do not display the common
than the cave area itself, is how this area is distributed in a given space. hourglass morphology as in sandstone caves (Aubrecht et al., 2011;
Sauro, 2014; Bruthans et al., 2014) and can actually be inclined
5.2. Underground landforms and sedimentation (Fig. 17D). The shallow nature of IF caves allows for the occurrence
of ceiling collapse, which sometimes results in the opening of ver-
Much of the highly irregular pattern of IF caves, represented by tical entrances (skylights) (Fig. 12B). Roots commonly penetrate
elongations in the cave perimeter, is due to the existence of numer- through the ceiling or entrance. Cave collapse is the major source
ous small conduits (canaliculi), normally at floor level, but also oc- of cave sedimentation, which is mostly autogenic. As mentioned
curring along the walls and ceiling (Piló and Auler, 2005; Calux earlier, the ascending (or flat) profile of IF caves does not favor
et al., 2019). These channels (Fig. 16A) may represent an earlier the input of surface sediments, except near the entrance. Another
evolutionary stage of the narrow junctions, not wide enough to major source of sedimentation is represented by the small
allow human access. IF caves show no evidence of dissolutional fea- channels. Some display a sediment fan at their mouth. The iron-
tures (speleogens) which are common in caves formed by flowing rich sediment coming out of these small conduits is highly sorted
water in most rock types. Rock projections oriented according to at fine-grained size, but may be soft and goo-like, especially in
Table 5
Fractal Dimension for each lithology and region, calculated by FracLac and Fractal 3.4.7.
Fractal Dimension (FD)
FracLac Fractal 3.4.7
Min Mean Max Range Min Mean Max Range
Lithology Canga 1.4927 1.6677 1.7653 0.2726 1.5090 1.6731 1.7981 0.2891
Contact 1.4551 1.6601 1.7537 0.2986 1.5498 1.6717 1.7646 0.2148
Iron Formation 1.3475 1.6626 1.7596 0.4121 1.4278 1.6709 1.7749 0.3471
Region Carajás 1.4927 1.6724 17,653 0.2726 1.5090 1.6781 1.7981 0.2891
Iron Quadrangle 1.4551 1.6469 1.7596 0.3045 1.5066 1.6499 1.7537 0.2471
Southern Espinhaço 1.3475 1.5943 1.7021 0.3546 1.4278 1.6499 1.7393 0.3115
12
Table 6 for the complex transformation of BIF into a more porous Fe-rich rock
Pearson correlation for cave dimensional variables (length and area) and Fractal Dimen- and their relative importance may vary according to the environment
sion. Statistically significant values are shown with an asterisk.
in which they take place. The competing or interdependent role of
Correlations Length Area Area/Length FracLac FRACTAL these processes may bear relation to the evolving geochemical gradients
(m) (m2) (m) 3.4.7 that necessarily occur during the long history of the evolution of the IF
Length (m) 1 deposits. Nonetheless, it is clear that the observed micro porosity and
Area (m2) 0.79* 1 caves are the result of a complex suite of interrelated and still little
Area/length 0.06 0.61* 1
known mechanisms that evolved with various intensities over vast
(m)
FracLac 0.06 0.17 0.20 1 timescales. The most relevant to void generation will be discussed in
FRACTAL 3.4.7 −0.01 0.14 0.27 0.80* 1 detail.
6.1. Initial porosity

orifices in the walls or ceiling. Due to the absence of soil at the
canga surface, this sediment can only originate from processes The initial porosity may occur at depth and is not associated with the
within the rock, as will be discussed in Section 6.2. surficial canga deposits. In the IF regions of Brazil, entranceless voids
A peculiar feature of many IF caves is the occurrence of a crust coat- have been routinely intercepted by drilling (Maurity and Kotschoubey,
ing the bedrock along the floor and walls. The crust smooths the cave 1995) sometimes at depths of up to 600 m below the present surface
contours and gives the walls a polished, lustrous appearance (Fig. 18). (Rios and Santos, 2015). This deep origin of IF caves has led to analogies
Mineralogical studies have revealed that the crust coating is mostly with carbonate hypogene cave morphologies (Auler et al., 2014). IFs
comprised by Fe(III) oxides and phosphates (Albuquerque et al., generally contain productive aquifers suggesting porosity, and much
2018). The coating can be found throughout the caves although it usu- of the available hydrogeological data comes from large scale dewatering
ally predominates at the distal portions, where environmental condi- projects required for the operation of deep open cast mines. Median
tions tend to be stable, with relative humidity close to 100%. These outflow of springs is 400 m3/day reaching as much as 12,000 m3/day
distal passages are sometimes the hibernaculum of bat colonies, which (CPRM, 2004). A dual porosity aquifer, IFs have components of a granu-
has led to the association of phosphate crusts to reaction of bedrock lar (in porous IFs) and fractured aquifer (in compact or little weathered
with bat guano (Simmons, 1964; Albuquerque et al., 2018; Figueira BIF) (Chiste-Costa, 2017; Gonçalves et al., 2018), which results in het-
et al., 2019). However, as will be demonstrated in Section 6.2, the iron erogeneous hydrogeological behavior. In BIF and hardened IFs, a corre-
oxyhydroxide crust has a complex origin and its development is linked lation between fracture frequency and hydraulic conductivity has been
to the growth and stabilization of the cave itself. observed (Carneiro et al., 2016). This anisotropy is also expressed ac-
cording to the direction of the banding or foliation, with values of hy-
6. Generation of porosity and speleogenesis draulic conductivity 3× higher along these structures (Mourão, 2007).
Preferential groundwater flow is thus focused along enlarged horizons
There are a number of physical, chemical and biological processes represented by the rock banding and fractures (Mourão, 2007;
that may account for porosity generation in IFs. These processes may Bertachini et al., 2018).
take place over long timescales and depths, affect diachronically differ- The extreme values of hydraulic conductivity are due to porosity at
ent portions of the rock unit and may act under both oxic and anoxic depth, which imply the generation of voids (Amorim et al., 2001). Car-
conditions. Rosière et al. (2008) point out that gradation from the orig- bonate facies are known to occur in certain portions of BIF deposits in
inal BIF to the high grade ore involves long term hypogene processes the IQ (Dorr, 1964; Spier et al., 2007) and Carajás (Guedes et al., 2013)
starting in the Proterozoic followed by supergene enrichment processes and the removal of the soluble carbonate has been invoked as an impor-
that evolved to the present day. These processes include remobilization tant mechanism of porosity generation (Simmons, 1963; Rosière et al.,
of iron and leaching of gangue minerals (e.g. quartz, carbonate, amphi- 2008; Monteiro et al., 2018a). Ribeiro (2003) has calculated that the net
bole), resulting in an increase in rock porosity (Ribeiro et al., 2021). Sev- dissolution of carbonates, and then silica, can result in the creation of
eral distinct geochemical and microbiological mechanisms can account low density zones with up to 40% overall decrease in rock unit volume;
Fig. 13. Distribution of caves according to their geology and FD (Fractal 3.4.7).
13
Fig. 14. Distribution of caves according to the region and FD (Fractal 3.4.7).
Fig. 15. Diverse cave morphologies in 8-bit. (A) Lowest FD (by FracLac and Fractal 3.4.7). (B) Highest FD by FracLac. (C) Highest FD by Fractal 3.4.7. (D), (E) and (F) represent caves with
similar FD but very different shapes.
14
Fig. 16. Morphological features of IF caves. (A) Small channels at the floor level in a cave in the IQ. (B) Jagged floor, with pinnacles and projections in a cave in Carajás. (C) Sequence of
rounded basins which fill intermittently in a cave in Carajás.
Fig. 17. (A) Pillars and pendants in a cave in IQ. Photo by Ataliba Coelho. (B) and (C) Pillars in caves in Carajás. (D) Excavated pillar in a cave in IQ.
15
Fig. 18. (A) Coating by iron oxyhydroxides and phosphates produces a crust that smooths the irregular cave floor as in this cave in SE. Photo by Luciana Alt and Vitor Moura. (B) A polished
section of cave wall in IQ, showing induration due to the development of crust (whitish layer) at the outer (lower) surface of the sample.
however, in the majority of the IF deposits, silica and iron are the major soluble ferrous Fe(II) is a critical step in promoting the mobility (and
constituents and carbonates are largely absent. Silica dissolution is a later the reprecipitation) of iron in the shallow canga profiles, where
well-known process (see review in Crundwell, 2017 and references most micropores and caves occur. Extensive microbially mediated
herein) responsible for generation of large scale cave and karst features transformations clearly occur in IF cave systems, including Fe(III) reduc-
worldwide (Wray and Sauro, 2017; Auler and Sauro, 2019). Silica dissolution (Parker et al., 2018; Gagen et al., 2019a, 2019b; Calapa et al., 2021)
tion is particularly favored under high temperature and pH (Lunevich, and Fe(II) oxidation (Levett et al., 2020a, 2020b, 2020c). The presence of
2019) along with organic leaching, which may explain the abundance oxygen prevents microbially mediated Fe(III)-reduction (Weber et al.,
of silica karst in warmer areas (Wray and Sauro, 2017). Being a slow pro- 2006) preventing iron-reduction from occurring on surfaces within
cess, regions under long term geomorphic evolution, especially in wet cli- the caves. Instead, Fe-reduction is occurring behind the wall (the Fe-
mates, commonly display developed silica dissolution landforms (Auler oxide crust) that is present on cave surfaces (Fig. 18). The iron-oxide
and Sauro, 2019). Silica dissolution does not necessarily imply surface crust prevents diffusion of oxygen, and the material in the caves behind
(meteoric) sources of alkalinity due to its positive correlation with tem- this crust is soft (with a toothpaste-like consistency) and contains a high
perature. As such, dissolution could occur at depth aided by geothermally (~1 × 107 cells/g) microbial cell count (Calapa et al., 2021). This material
heated groundwater, which has been associated with the generation of is termed sub muros (Latin for ‘behind wall’) and appears to be the pri-
caves and dissolutional features deep below the surface in quartzite and mary site of Fe-reducing activities, dramatically reducing the amount of
sandstone (Sauro et al., 2014; Suchý et al., 2017). Fe(III)-oxides and increasing porosity (Fig. 19, Parker et al., 2018, Calapa
The preferential chemical leaching of silica, when in association with et al., 2021). The importance of bacterially mediated redox reactions in
iron, is a key process in porosity generation in IFs. Numerous studies the canga profile has been stressed by Parker et al. (2013a, 2013b,
have stressed that silica solubility in groundwater can be substantially 2018), Levett et al. (2016), Monteiro et al. (2018a), Gagen et al. (2018,
enhanced in the presence of iron, and iron solubility (and Fe(II) stabil- 2019a) and Paz et al. (2020).
ity) can be enhanced when complexed with silica (Reardon, 1979; The growth of microorganisms behind the wall appears to be fueled
Morris and Fletcher, 1987; Baumgartner and Faivre, 2015; Kinsella by organic carbon brought in from the landscape above by meteoric
et al., 2016 and references herein). Conversely, the presence of quartz water, which percolates into the sub muros zone (Parker et al., 2018).
can also mediate the removal of iron from solution (Sasowsky et al., Behind the oxide crust, and in the absence of oxygen, the microbial com-
2000). Thus, silica removal and iron mobilization may take place abiot- munity uses Fe(III) as the primary electron acceptor for growth, with
ically under phreatic conditions in the early stages of porosity genera- the generation of soluble Fe(II) (Parker et al., 2018). This process is sup-
tion. The various silica phases behave differently in regard to ported by the observed microbial communities found in the sub muros,
geochemical mobilization. Abreu et al. (2016) have shown through pet- which are enriched in microorganisms encoding iron-binding oxidases
rographical studies in Carajás that silica dissolution follows a sequential (Parker, 2018). Nonetheless, the production of excess Fe(II) in solution
path involving initially chert dissolution and then chalcedony/jasper, could cause passivation of the Fe(III) oxides and limit reduction. To
progressing from the center towards the border of the silica bands, cre- test this, Calapa et al. (2021) investigated the effect of punctuated
ating voids that will eventually be enlarged through coalescence. The waterflow, as would be encountered during the wet/dry seasons
same study showed that less soluble minerals (including hematite) found in Brazil, on dissolution. These studies, which used columns of
cause deviations in the dissolutional route. Abiotic transformative pro- powdered canga, demonstrated that periodic flushing of the system
cesses also take place within the Fe constituents. In particular, it has dramatically enhances dissolution and increases porosity (Calapa
been noted in the IQ that the transformation of magnetite into hematite et al., 2021).
or martite in oxidizing environments may promote the dissolution of The Fe(II) released by microbial Fe-reduction may remain in solution
iron oxyhydroxide minerals, resulting in volume reduction within the in the anaerobic sub muros, or alternatively, will react with the cave air,
IF rock mass (Lagoeiro, 1998; Rosière and Stacey, 2018). A review of auto-oxidizing back into Fe(III). This turnover of Fe (Levett et al., 2016;
chemical reactions in Brazilian IFs involving the various Fe and Si phases Gagen et al., 2019b; Levett et al., 2020b, 2020c) may also be responsible
is presented in Monteiro et al. (2018a) and more comprehensive gen- for the observed microscale “turnover” of canga, while the generated
eral reviews can be found in Cornell and Schwertmann (2003) and oxides may enhance its durability (Monteiro et al., 2014; Monteiro
Baumgartner and Faivre (2015). et al., 2018a, 2018b; Levett et al., 2016; Gagen et al., 2019b; Levett
Void generation also takes place in the vadose near-surface environ- et al., 2020b, 2020c). Additionally, microbial activities have been
ment. The transformation of the mostly insoluble ferric Fe(III) into the
16
geochemical and erosional agents. Canga is highly permeable, in

marked contrast with the low permeability of the weathered ore. Verti-
cal infiltration of water in canga can occur both through fissures or the
matrix (Dias and Bacellar, 2021) and leads to the formation of a perched
aquifer during the wet season (Bertachini et al., 2018), with
downgradient flow along the canga/weathered BIF interface. The down-
slope hydraulic gradient can be significant, due to the often steep angle
of canga slopes (see cave profiles in Fig. 12). Subsurface flow (interflow)
along the canga/weathered BIF contact represents a powerful mecha-
nism for the generation and enlargement of pores and caves. The inte-
gration of existing porosity with slope and surface hydrological
processes is a key component of the later vadose evolution of caves
(Auler et al., 2014) because it will supply the canga/weathered BIF con-
tact not only with abundant organic matter, fostering dissolution and
bacterially mediated processes, but also with erosional power, enabling
the evacuation of insoluble material.
The concentrated austral summer rainfall in the IF regions in Brazil
leads to temporary hydrological saturation of the canga pores, which
will drain through vertical infiltration towards existing caves (Dias
and Bacellar, 2021). Parker et al. (2013a) has shown that pH of infiltra-
tion water in IF caves can be as low as 2, and the amount of organic dis-
solved carbon is much higher than in carbonate caves (Parker et al.,
2018). Besides mechanical removal processes downslope, the seasonal
perched canga aquifer favors geochemical and bacterially mediated
processes.
The peculiar sequential combination of larger rooms and smaller
connecting passages (bead-on-a-string morphology) is highly sugges-
tive of coalescence of previously isolated (at least in terms of human ac-
Fig. 19. Computed Tomography (CT) scan of the cave wall (oxide crust) with the sub muros cess) rooms (Auler et al., 2014). When the connection occurs between
layer behind it. The Fe(III)-oxides (rock in grey and voids in purple) are shown in the full rooms located along the scarp (i.e. not uphill following the slope), the
CT scan; however, when the rock is subtracted, leaving only the voids highlighted (purple) caves may grow laterally. The larger caves always display both lateral
the increased porosity of the sub muros zone is clearly visible.
– and along the slope – connections (see Fig. 9). The basic, initial ele-
ment of IF caves is, thus, the room, with the smaller connections and
elongations representing later additions. We have quantitatively ana-
proposed as a means to reclaim canga damaged by mining activities lyzed the relationship between narrow junctions and rooms for our
(Levett et al., 2020a, 2021). data set of 238 longer caves (Fig. 21). There is a positive trend between
The extensive microbial Fe(III) reduction has been proposed as a number of rooms and mean length, showing that the coalescence of
mechanism for the formation of caves in IF (Parker et al., 2013a, rooms plays a role in cave development (Fig. 21B). There appears to
2013b, 2018; Parker, 2018). In this scenario, microbial activity continues be a limit to the number of rooms, with most caves having between 4
to enlarge the sub muros, eventually the weight of the sub muric material and 6. This may be controlled by the elimination of rooms due to me-
or weakening of the Fe(III)-oxide crust is sufficient for the crust to give chanical scarp breakdown, generating amphitheaters.
way, allowing the sub muric material to fall into the cave, enlarging the Caves located at the contact always display, regardless of the region,
void. The exposed Fe(II)-rich sub muros is then exposed, auto-oxidizing smaller room area than in the other geological situations, even when the
and re-forming the crust. The formation of this crust re-establishes the mean cave area is larger (as in Carajás — Table 4). We interpret this as
anoxic conditions within the sub muros, allowing Fe-reduction to re- due to the modification of rooms by the interflow along the contact. In
sume and the process to repeat (Fig. 20). This process could lead to the other situations there will be less mechanical power to mask the
the bead-on-a-string morphology of the cave passages, which form original room morphology.
through a series of voids that collapse in on themselves, creating con- As mentioned in Section 6.1, mobilization of iron through Fe(III) re-
nected voids (Fig. 20). This hypothesis is supported by the oxide crust ducing bacteria (Parker et al., 2013b, 2018; Calapa et al., 2021) is a pow-
fragments found on the floors throughout these caves. The process erful mechanism that aids in the formation and expansion of pores and
may demonstrate a more substantial redistribution of Fe than the mi- caves. Of particular interest is the precipitation of iron at the cave inner
croscale cycling of Fe in canga, and the presence of these caves indicates boundaries (ceiling, walls and floor), generating a coating layer
the possibility of more extensive weathering of IF than previously ap- (Fig. 18). Similar crusts in non-IF caves are formed by precipitation of
preciated, including a contribution to the supra iron ore bodies solid phases from aqueous solutions originating from chemically satu-
themselves. rated infiltration water within the cave. Since coatings in IF caves
cover most surfaces, including overhanging sections, and flowing
6.2. Cave evolution water is observed infrequently over them, a normal percolation origin
can be discarded. The bacterially driven mode of cave genesis, wherein
Macro pores and caves hosted entirely by fresh or weathered BIF, microbial Fe-reducing activity behind the walls causes collapse and wall
without canga association can have an origin linked to the ancient and retreat (Fig. 20) benefits from the episodic hydrological saturation of the
long lasting geochemical processes outlined previously, as evidenced canga profile during the rainy season, allowing for ample input of or-
by the existence of voids intercepted by deep drilling. Later long-term ganic carbon and removal of passivation by Fe(II) (Calapa et al., 2021).
vertical uplift and denudation will eventually bring these caves closer Caves are thus protected by the iron oxyhydroxide crust and will
to the surface, even considering the very low denudational rates of IF re- mainly evolve through rounds of collapse of portions of ceiling and
gions. At shallower depths, voids will be under the influence of a walls (Fig. 20). This will lead to further oxidation and crust precipitation
weathering front controlled by a combination of microbiological, in the newly exposed section. This may explain why the crust tends to
17
Fig. 20. The model for IF cave formation. Stage I. The autooxidation of Fe(II) exposed to oxygen in the cave passage creates Fe(III)-oxides that cement the passage walls into an ~2–3 cm
thick crust, allowing anaerobic conditions to be established. Water percolating through the canga brings organic carbon-rich water to this anaerobic, sub muric zone, providing the reducing
potential for FeRM to reduce the insoluble goethite cements to soluble Fe(II), which is removed by water flow. Stage II. Eventually, the loss of the goethite cements in the canga leads to a
loss of structural integrity in the sub muric zone and the iron crust is no longer strong enough to hold back the wall, which collapses into the cave void. Stage III. Following collapse,
exposure of Fe(II) fluids to oxygen reforms the Fe(III)-oxide crust, restoring anaerobic conditions behind the wall and the process repeats. Over time, repeated collapse leads to wall
retreat and formation of the cave passages and observed morphology.
be thicker on the floor or on less inclined surfaces (see Fig. 18) because IFs terrains is slow due to the protective role of canga (Vasconcelos
these areas are less prone to gravitational collapse. The sub-muros depo- et al., 2019). However, scarp retreat is a dynamic process (Vasconcelos
sition of the crust is of paramount importance to cave preservation in and Carmo, 2018) and, although canga can be regenerated within the
the same way as surface canga promotes the long term resilience of weathering profile (Monteiro et al., 2018a), scarp rupture tends to be
iron surfaces. Without sub-muros coating, caves would likely have col- mostly a mechanical process, controlled in IF areas by vertical (aligned
lapsed due to low mechanical stability of the mobile and sometimes with the scarp) and surface-normal unloading joints frequent in nearly
soft iron constituents of the bedrock. This represents a novel mechanism all caves and scarp zones. In this regard, caves in Carajás are in a more
of cave formation that enables IF caves and related porosity to survive advanced dismantling stage, in which rooms are being exposed through
over exceptionally long periods in the landscape. scarp disintegration, generating amphitheaters (~50% of all Carajás
As shown in Table 4, the elementary volumetric unit of IF caves – the caves), while in IQ and SE, the “original” configuration, in which connec-
room – has been shown to be significantly larger in Carajás than in the tion with the surface occurs through narrow passages, still prevails
IQ and SE. The fact that there are still narrow connections between (Table 3).
rooms, even in Carajás, demonstrates that they have not yet achieved The genesis of pendants and pillars, always associated with canga
their maximum possible volume, which would lead to coalescence ceilings (Fig. 17), cannot be linked to classical dissolutional mechanisms
and elimination of the connections. Larger rooms tend to be more unsta- as in caves in other rock types. Gravitational percolation of water
ble and thus a positive growth feedback can be created in Carajás, through the permeable canga may create vertically oriented redox
allowing for more frequent collapse and faster cave volumetric expan- zones within the rock, cementing clasts, and generating hardened “fin-
sion. The ratio connections/rooms (approximately 1) is the same re- ger like” zones which are more resistant than the surrounding rock
gardless of study area, showing that these parameters may not mass (Fig. 22). This process can be described as a vertically oriented
necessarily represent different stages of evolution. The existence of sub-muros which will serve as an avenue to further catalyze oxidation
larger caves in Carajás can be explained by: (i) larger plateau areas and induration. In addition to gravitational infiltration, roots, tubules
which will provide a more extensive catchment area for interflow hy- and plants can also serve as templates for iron oxide precipitation
drological processes than in the less extensive ridges of the IQ and SE; (Gagen et al., 2019b) and can help in the creation of these vertically ori-
(ii) much warmer climate, enhancing porosity development and cave ented redox fronts. The conditions for generating pillars and pendants
formation; (iii) wetter climate which results in more water availability precede the formation of the caves, the pillars and pendants being “ex-
and increased runoff, favoring both chemical and mechanical removal humed” due to the sub-muros sequential collapse. This mechanism
of IF constituents and (iv) the availability of organic carbon under bears resemblance to finger flow phenomena common in coarse soil
Carajás equatorial conditions which will catalyze redox and microbio- profiles (Liu et al., 1994; Allaire et al., 2009; Wang et al., 2018).
logical interactions. Autogenic well sorted fine-grained sediment supplied by the small
The persistence of IF caves in the landscape also bears relation with channels are ubiquitous in IF caves. As pointed out, due to the absence
the geomorphic evolution of the IF plateaus and ridges. Denudation in of soil above ground, they must result from microbiological reactions
18
Fig. 21. (A) Combined distribution of caves per number of rooms for the three study areas. Length distribution in each room class (marked by a red line) follow a very asymmetric
distribution similar to a Pareto (hyperbolic) distribution, typical for the statistical ordering of caves (Curl, 1986). Caves with 5 rooms predominate. (B) A similar analysis, but separated
by area. Mean values show that cave length tends to increase with number of rooms in all areas.
within the rock mass and represent the non-soluble phase resulting 7. Iron mobilization and landscape evolution
from the process of porosity generation. Although Piló and Auler
(2005) have suggested that piping processes could be responsible for IF landscapes evolve through a complex combination of geochemical
sediment evacuation, this is not strictly the case, as Fe and Si mobiliza- and microbiological processes. The deep iron-rich profiles, generated by
tion and transport are related to a geo-biochemical front that advances long term hypogene and supergene processes (Rosière et al., 2008),
towards the inner portion of the IFs plateaus and ridges. have been able to defy erosion not only due to their low geochemical
In caves with limited or no soil it is possible to observe a highly jag- rates of dissolution, but also due to the mobile and self-regenerating
ged floor displaying pinnacles, deep cut basins and sharp projections canga blanket (Dorr, 1969). The two major original constituents of the
(Fig. 16B, C), especially if there is active infiltration. This may be related unweathered BIF bedrock, silica and iron, follow distinct geochemical
to highly acidic dripping as demonstrated in one such cave (Parker et al., paths throughout the long IF evolutionary history. Silica, relatively mo-
2013a). Scherer (2017) hypothesized that, in Carajás, caves that display bile in tropical temperatures, is the first to be removed by groundwater,
such jagged floor are located close to, and at lower elevations than, lakes resulting in the concentration of iron. Several authors (Ribeiro, 2003;
and swamps, which could have favored the input of acidic and organic- Spier et al., 2003; Rosière et al., 2008; Rosière and Stacey, 2018;
rich infiltration water. This relationship, however, is not observed in Ribeiro et al., 2021) have pointed out that silica (and if present, carbon-
similar caves in the SE. Albuquerque et al. (2018) suggest that phospho- ate) leaching represents the first step, both for hypogene and supergene
ric acid derived from bat guano can also account for dissolution (and processes, towards porosity generation and, consequently, iron concen-
later precipitation) of iron oxyhydroxides. Bat guano has an important tration. The early existence of a porous rock, even if not displaying voids
role in precipitating phosphate speleothems (Simmons, 1964; large enough to be considered as caves, is an important condition for
Albuquerque et al., 2018; Figueira et al., 2019) but their possible influ- generating the highly transmissive aquifers typical of IFs (Amorim
ence in cave expansion, if significant, will likely be obliterated by the as- et al., 2001; Mourão, 2007) which allows for groundwater circulation
sociated constructional processes. and further Si leaching. The occurrence of deep weathered zones
19
Fig. 22. Hypothesized finger flow genesis of pillars and pendants in IF caves. (A) Vertically oriented gravitational infiltration of water creates redox finger-like fronts. (B) Following
generation of void (cave), exposure will lead to oxidation. These finger zones will be more resistant to collapse. Coating of sub-muros material will smooth the external surface of pillars.
along relatively fresh (unweathered) BIF (Rosière et al., 2008), shows in particle size (Fig. 23A). The in situ induration of the previously uncon-
the selective nature of the geochemical alteration front, which is con- solidated iron-rich sediment results in what we name as “secondary
trolled by groundwater flow and specific geological structures such as canga” because it was formed later than both surface canga and caves,
major faults and fractures, besides the more soluble avenues repre- and can be positioned stratigraphically below the weathered BIF.
sented by the (silica) bands (Rosière and Stacey, 2018). In parallel Rates of canga generation at the contact zone between canga/weath-
with silica and carbonate leaching, the low Fe content characteristic of ered BIF must be somehow in equilibrium with the very slow rates of
IF springs (Mourão, 2007; Carvalho, 2012; Chiste-Costa, 2017) indicates surface denudation (Monteiro et al., 2018a), otherwise canga would
limited net loss of iron, suggesting that Fe tends to be remobilized within have been gradually eliminated from the landscape, exposing the friable
the system leading to iron concentration and the gradual generation of IF to erosion. The dynamic equilibrium between denudation and soil
canga profiles and high-grade iron bodies. generation has been demonstrated for thick weathering profiles
Canga is a key component of the IF landscape. Although normally (Dosseto et al., 2008, 2011). The canga/weathered BIF contact zone
interpreted as being generated through surface-related processes will thus migrate downwards in time, in parallel with the slow disman-
(Simmons, 1960; Dorr, 1964; Spier et al., 2019), observations inside tling of canga at the surface. The fact that the unconsolidated weathered
caves have demonstrated that the weathering front related to canga for- BIF has survived for hundreds of millions of years attests to the long
mation occurs also at depth and involves chemical, microbiological and term existence of canga. The coating layer associated with the sub-
mechanical processes as suggested by Monteiro et al. (2018a). The grad- muros can be interpreted as an underground, void (and rock) protective
ual change between a rock with leached silica, but with still recogniz- cover, which promotes the stabilization of the canga profile, protracting
able relict banding structure, towards canga described by Dorr (1964), mechanical breakdown. Caves can, thus, evolve for long periods since
can be observed at various transitional stages (Fig. 23), generating their inception in the deep phreatic zone, much likely beyond the oldest
what has been named by Dorr (1964) as “structured canga”. This in- ~70 Ma ages obtained from the self-healing canga profiles in the IF areas
volves initially the fragmentation of the iron bands. Then, through col- of Brazil (Vasconcelos and Carmo, 2018). These ever evolving Fe
lapse and translocation, the fragments lose their orientation leading to oxyhydroxide fluid migration zones will help indurate surficial and un-
the chaotic arrangement of clasts typical of canga. Canga matrix can derground canga, as well as cave/pore outer surfaces, besides fostering
also vary from hardened to soft paste-like oxyhydroxide, depending the selective hardening of heterogeneous sections of the canga profile.
on its redox state. In a few situations (as through collapse openings at Both canga and sub-muros layers represent sinks for the mobile iron
ceiling or when the cave has an up gradient entrance) there may and are genetically associated with high-grade supergene iron deposits.
occur local transportation of sediment towards the caves. This sediment Scarp disintegration appears to be a major controlling process in the
is easily distinguished from both canga and the fine-grained sediment evolution of IF ridges and plateaus (Dorr, 1969). As shown in Fig. 4,
derived from small channels due to its imbricated nature, and variation caves are dominantly located at the scarp front and seldom extend
20
Fig. 23. Examples of canga/BIF transitions observed in caves. (A) Cave wall in Carajás displaying a later iron-rich sedimentary deposit of allochthonous origin (1) with some horizontal
chemical Fe oxyhydroxide layers (2) overlying the folded weathered BIF (3). The original unlithified sediment was probably carried towards the cave and was indurated through Fe
oxidation, resulting in “secondary” canga. (B) BIF bands in Carajás grading towards canga. The original banding structure, although fragmented, can still be recognized. This has been
named as “structured canga” by Dorr (1964). (C) Sharp contact between BIF (below) and structured canga (top) in an IQ cave, with relict banding still preserved in the latter.
beyond it. They must evolve synchronously with the scarp retreat be- of IF ridges and plateaus. IF landscapes comprise a dynamic eco-
cause otherwise they would have been removed from the landscape geosystem that has been active for millions of years and evolves mostly
due to the dismantling and upslope migration of the scarp line. Caves through a complex interplay of biogeochemical processes at the
tend to occupy over half of the vertical canga profile creating a mechan- subsurface.
ically unstable zone at the scarp front. Collapse leads to the generation
of arcuate scarp lines, which will give rise to parallel (curved) vertical 8. Conclusions
unloading joints that will further promote scarp dismantling. Through
this positive feedback, coalescence and amplification of amphitheaters Silica (and carbonate) leaching, iron mobilization/concentration,
may lead to cusps (Fig. 5A), resulting in major asymmetrical indenta- and porosity generation are closely associated processes in IF regions.
tions in ridges and plateaus. Such large cusps can exert control over Together, they play a major role as large scale and long term geomor-
the hydraulic gradient, forcing convergence of interflow lines and phic agents, controlling the evolution and degradation of the ancient
channeling sediment/solute delivery towards the lower and more pro- IF landscapes. The fresh BIF is seldom found in near surface environ-
nounced cuspate scarp line. ments, as silica/carbonate leaching and iron mobilization alters its
IF landscapes are nearly always located at higher elevation and so chemical composition and texture. Hypogene processes that took
there is no input of allogenic material. They are also mostly devoid of place since the deposition of BIF in the Proterozoic are responsible for
soil. Since there is no input of Fe from external sources and limited net initial porosity generation, as more mobile elements are removed
iron loss takes place, the entire IF landscape is progressively leached of from the system. The resulting rock usually displays relict banding and
more mobile constituents, evolving towards an endpoint in which variable degrees of cohesiveness. It is a productive aquifer and
most silica will have been removed from the system, yielding the high dewatering for deep open cast iron mines has always proved to be chal-
grade porous ore characteristic of the IF regions of Brazil. Caves are lenging. Supergene processes controlled by groundwater flow and fur-
thus biogeochemical outlets and act as major players in iron mobiliza- ther removal of constituents result in a dual porosity aquifer with
tion processes. Caves are also determinant in the geomorphic evolution major transmissive horizons controlled by fractures and relict banding.
21
The common interception of voids at depths of several hundreds of Declaration of competing interest
meters attests to the existence of early porosity in the deep phreatic
zone. The authors declare that they have no known competing financial
The friable nature of the weathered BIF renders it highly suscep- interests or personal relationships that could have appeared to influ-
tible to erosion if not for being capped by canga, the iron-rich super- ence the work reported in this paper.
gene duricrust that protects the subjacent rock. Canga display high
millimeter to centimeter scale porosity that allows for fast infiltra- Acknowledgements
tion of meteoric water towards the contact with the less permeable
weathered BIF. This paper results from many years of research in most IF regions in
Thousands of caves have been identified in the three major IF re- Brazil. Sampling was performed through SISBIO permit 35992. Several
gions of Brazil: Carajás, Iron Quadrangle and Southern Espinhaço. mining companies supported our research along the years, especially
They can occur in three different geological situations, entirely in Anglo American, Vale and Gerdau. We thank the help of the staff of
canga or in weathered BIF, or along the contact between these Carste Ciência Ambiental during field work and data processing. Luciana
two rock types. Caves at the contact tend to be the more well- Alt, Vitor Moura and Ataliba Coelho provided photographs. Dr. Renato
developed. Caves are nearly always associated with the scarp Paes de Andrade (University of São Paulo) allowed the use of the
lines of IF plateaus and ridges, seldom extending more than permeameter. Insightful reviews by Jo De Waele and Carlos Spier im-
100 m beyond it. Morphology of caves show a highly irregular com- proved the manuscript. This work was supported by funding from the
bination of rooms and narrow passages, suggesting that they National Science Foundation Geobiology and Low Temperature Geo-
evolve through the coalescence of individual rooms. Microbial me- chemistry program (NSF# 1645180).
tabolism is an important controller of IF weathering and stability.
Coupled Fe(III) reducing and Fe(II) oxidizing processes can lead Appendix A. Supplementary data
to canga formation and enhance its stability, while Fe(III) reducing
activities distal to the cave walls, in the sub-muros, lead to IF disso- Supplementary data to this article can be found online at https://doi.
lution and cave formation. Caves persistence in the landscape is fa- org/10.1016/j.geomorph.2021.108068.
vored by the sub-muros-covering Fe(III) oxyhydroxides, in the
same way as with canga, both internally due to vertically oriented
References
iron infiltration routes, and at the surface.
The heterogeneous and porous canga layer is the result of complex Abreu, S.S., Macambira, J.B., Cabral, E.S., 2016. Geologia e petrografia de formações
and dynamic processes involving multiple chemical and biological in- ferríferas com ênfase na gênese de cavidades em Serra Norte, Carajás-PA. Espeleo-
Tema 27, 33–44.
teractions. Canga is a unique rock that exhibits self-healing characteris-
Albuquerque, A.R.L., Angélica, R.S., Gonçalves, D.F., Paz, S.P.A., 2018. Phosphate
tics and is formed continuously. The subsurface transformation of speleothems in caves developed in iron ores and laterites of the Carajás Mineral Prov-
weathered BIF into canga occurs approximately in equilibrium with de- ince (Brazil) and a new occurrence of spheniscidite. Int. J. Speleol. 47, 53–67. https://
nudation at the surface, allowing for the long term resilience of IF ridges doi.org/10.5038/1827-806X.47.1.2135.
Allaby, M., 2013. A Dictionary of Geology And Earth Sciences. Oxford University Press,
and plateaus.
Oxford.
The generation of caves and micro porosity is critical to the geo- Allaire, S.E., Roullier, S., Cessna, A.J., 2009. Quantifying preferential flow in soils: a review
chemical dynamics of IF regions. They are major players of long lasting of different techniques. J. Hydrol. 178, 179–204. https://doi.org/10.1016/j.jhydrol.
leaching of more mobile constituents, yielding high grade iron deposits. 2009.08.013.
Alvares, C.A., Stape, J.L., Sentelhas, P.C., Gonçalves, J.L.M., Sparovek, G., 2014. Köppen's cli-
The scarp front is the locus of both solute and sediment leaching, with
mate classification map for Brazil. Meteorol. Z. 222, 711–728. https://doi.org/10.1127/
caves acting as geochemical outlets. Caves also play a major role in 0941-2948/2013/0507.
scarp retreat, through collapse of cave passages creating indentations Amorim, L.Q., Granchamp, C.A.P., Bertachini, A.C., 2001. Dewatering of iron ore mines and
in the scarp line and gradually decreasing the total area of canga- capability to predict environmental impacts – the experience of MBR. Proceedings
International Mine Water Association Symposium. Belo Horizonte . https://www.
capped surfaces. imwa.info/docs/imwa_2001/DewateringofIron.pdf.
The mobile character of iron allows for the possibility of rehabilita- ANM (Agência Nacional de Mineração), 2020. Anuário Mineral Brasileiro. Principais
tion of IF regions through canga restoration. This is of paramount impor- Substâncias Metálicas. (25 pp.)ANM, Brasília. https://www.gov.br/anm/pt-br/
centrais-de-conteudo/publicacoes/serie-estatisticas-e-economia-mineral/anuario-
tance as severe degradation is inherently associated with iron mining
mineral/anuario-mineral-brasileiro/amb_2020_ano_base_2019_revisada2_28_09.
operations (Sonter et al., 2017). This represents an area of active re- pdf. (Accessed 20 August 2021).
search (Gagen et al., 2019b; Levett et al., 2021) that should provide fur- Aubrecht, R., Lánczos, T., Gregor, M., Schlögl, J., Šmida, B., Liščák, P., Brewer-Carias, C.,
ther information on rates and processes of bacterially mediated iron Vlček, L., 2011. Sandstone caves on Venezuelan tepuis: return to pseudokarst? Geo-
morphology 132, 351–365. https://doi.org/10.1016/j.geomorph.2011.05.023.
cycling.
Auler, A.S., 2016. Cave protection as a cave conservation tool in the environmentally sen-
Approximately 50% of IF landscapes in IQ and 20% in Carajás sitive Lagoa Santa Karst. Acta Carsologica 45, 131–145. https://doi.org/10.3986/ac.
have already been severely impacted (Salles et al., 2019; Souza- v45i2.4429.
Filho et al., 2019). This represents an irreversible loss to a unique Auler, A.S., Sauro, F., 2019. Quartzite and quartz-sandstone caves of South America. In:
White, W.B., Culver, D.C., Pipan, T. (Eds.), Encyclopedia of Caves. Academic Press,
ecosystem that comprises several subterranean species adapted Cambridge, pp. 850–860.
to subsurface canga habitat (Auler et al., 2019; Trevelin et al., Auler, A.S., Parker, C.W., Barton, H.A., Soares, G.A., 2019. Iron formation caves: genesis and
2019) as well as endemic metal tolerant plant species (Skirycz ecology. In: White, W.B., Culver, D.C., Pipan, T. (Eds.), Encyclopedia of Caves. Aca-
et al., 2014; Zappi et al., 2019). IF areas, among the oldest land- demic Press, Cambridge, pp. 559–566.
Auler, A.S., Piló, L.B., 2015. Caves and mining in Brazil: the dilemma of cave preservation
forms of the planet (Vasconcelos et al., 2019), have remarkable within a mining context. In: Andreo, B., Carrasco, F., Durán, J.J., Jiménez, P.,
geo and biodiversity value and should be the focus of increased LaMoreaux, P. (Eds.), Hydrogeological And Environmental Investigations in Karst Sys-
conservation efforts. Caves represent the only geomorphic element tems. Springer, Berlin, pp. 487–496.
that is intrinsically protected by the Brazilian constitution and as Auler, A.S., Piló, L.B., Parker, C.W., Senko, J.M., Sasowsky, I.D., Barton, H.A., 2014. Hypogene
cave patterns in iron ore caves: convergence of forms or processes. In: Klimchouk, A.,
such, improving the knowledge about caves and their protection Sasowsky, I.D., Mylroie, J.E., Engel, S.A., Engel, A.S. (Eds.), Hypogene Cave Morphol-
zone may be, at present, the most straightforward way towards ogies. Special Publication 18. Karst Waters Institute, Leesburg, pp. 15–19.
the creation of conservation areas in canga. Understanding the im- Axelrod, J.M., Carron, M.K., Milton, C., Thayer, T.P., 1952. Phosphate mineralization at
Bomi Hill and Bambuta,Liberia, West Africa. Am. Mineral. 37, 883–909.
portant role of iron and silica mobilization is key to encourage fur-
Babinski, M., Chemale Jr., F., van Schmus, W.R., 1995. The Pb/Pb age of the Minas Super-
ther research and protection of a remarkable dynamic environment group carbonate rocks, Quadrilátero Ferrífero,Brazil. Precambrian Res. 72, 235–245.
with few equivalents on Earth. https://doi.org/10.1016/0301-9268(94)00091-5.
22
Baumgartner, J., Faivre, D., 2015. Iron solubility, colloids and their impact on iron Eschwege, W.L., 1822. Geognostisches Gemälde von Brasilien und wahrscheinliches
(oxyhyd)oxide formation from solution. Earth Sci. Rev. 150, 520–530. https://doi. Muttergestein der Diamanten. Landes-Industrie-Comptoir, Weimar.
org/10.1016/j.earscirev.2015.09.003. Ferreira, R.L., Oliveira, M.P.A., Silva, M.S., 2015. Biodiversidade subterrânea em
Bekker, A., Slack, J.F., Planawsky, N., Krapez, B., Hofmann, A., Konhauser, K.O., Rouxel, O.J., geossistemas ferruginosos. In: Carmo, F.F., Kamino, L.H.Y. (Eds.), Geossistemas
2010. Iron Formation: the sedimentary product of a complex interplay among man- Ferruginosos Do Brasil. Instituto Prístino, Belo Horizonte, pp. 195–231.
tle, tectonic, oceanic, and biospheric processes. Econ. Geol. 105, 467–508. https://doi. Figueira, R.L., Horbe, A.M.C., Aragón, F.F.H., 2019. Exotic sulfate and phosphate
org/10.2113/gsecongeo.105.3.467. speleothems in caves from eastern Amazonia (Carajás. Brazil): crystallographical
Bertachini, A.C., Pereira, B.A., Bertachini, D.P., De Felippo, M.A., Beale, G., 2018. Hydrogeol- and chemical insights. J. S. Am. Earth Sci. 90, 412–422. https://doi.org/10.1016/j.
ogy of soft iron ores and associated rocks. In: Martin, D., Stacey, P. (Eds.), Guidelines jsames.2018.12.007.
for Open Pit Slope Design in Weak Rocks. CSIRO, Clayton, pp. 291–295. Freeze, R.A., Cherry, J.A., 1979. Groundwater. Prentice Hall, Englewood Cliffs.
Bittencourt, J., Gomide, A., Carmo, F., Buchmann, F.S., 2015. Registro paleontológico em Gagen, E.J., Levett, A., Shuster, J., Fortin, D., Vasconcelos, P.M., Southam, G., 2018. Microbial
caverna desenvolvida em formações ferríferas na Serra do Gandarela. In: Ruchkys, diversity in actively forming iron oxides from weathered banded iron formations sys-
U.A., Travassos, L.E.P., Rasteiro, M.A., Faria, L.E. (Eds.), Patrimônio Espeleológico em tems. Microbes Environ. 33, 385–393. https://doi.org/10.1264/jsme2.ME18019.
Rochas Ferruginosas. Sociedade Brasileira de Espeleologia, Campinas, pp. 192–209. Gagen, E.J., Zaugg, J., Tyson, G., Southam, G., 2019a. Goethite reduction by a neutrophilic
Bowden, D.J., 1980. Sub-laterite cave systems and other pseudo-karst phenomena in the member of the alphaproteobacterial genus Telmanospirillum. Front. Microbiol.
humid tropics: the example of the Kasewe Hills, Sierra Leone. Z. Geomorphol. 24, https://doi.org/10.3389/fmicb.2019.02938.
77–90. https://doi.org/10.1127/zfg/24/1984/77. Gagen, E.J., Levett, A., Paz, A., Gastauer, M., Caldeira, C.F., Valadares, R.B.S., Bitencourt, J.A.P.,
Brasil, 1988. Constituição da República Federativa do Brasil de 1988. Presidência da Alves, R., Oliveira, G., Siqueira, J.O., Vasconcelos, P.M., Southam, G., 2019b. Biogeo-
República. http://www.planalto.gov.br/ccivil_03/constituicao/constituicao.htm. chemical processes in canga ecosystems: armoring of iron ore against erosion and
(Accessed 20 August 2021). importance in iron duricrust restoration in Brazil. Ore Geol. Rev. 107, 573–586.
Brasil, 1990. Decreto Federal 99.556. Presidência da República. http://www.planalto.gov. https://doi.org/10.1016/j.oregeorev.2019.03.013.
br/ccivil_03/decreto/1990-1994/D99556.htm. (Accessed 20 August 2021). Gomes, M., Ferreira, R.L., Ruchkys, U.A., 2019. Landscape evolution in ferruginous
Bruthans, J., Soukup, J., Vaculikova, J., Filippi, M., Schweigstillova, J., Mayo, A.L., Masin, D., geosystems of the Iron Quadrangle, Brazil: a speleological approach in a biodiversity
Kletetschka, G., Rihosek, J., 2014. Sandstone landforms shaped by negative feedback hotspot. SN Appl. Sci. 1, 1102. https://doi.org/10.1007/s42452-019-1139-3.
between stress and erosion. Nat. Geosci. 7, 597–601. https://doi.org/10.1038/ Gonçalves, J.A.C., Pereira, P.H.R., Vieira, E.M., 2018. Sistemas aquíferos: Hidrogeologia da
ngeo2209. porção nordeste do Quadrilátero Ferrífero na região de Itabira (MG). Proceedings
Calapa, K.A., Mulford, M.K., Rieman, T.D., Senko, J.M., Auler, A.S., Parker, C.W., Barton, H.A., XX Congresso Brasileiro de Águas Subterrâneas . https://aguassubterraneas.abas.
2021. Hydrologic alteration and enhanced microbial reductive dissolution of Fe(III) org/asubterraneas/article/view/29349.
(hydr)oxides under flow conditions in Fe(III)-rich rocks: contribution to cave- Guedes, S.C., Rosière, C.A., Barley, M.E., 2013. The association of carbonate alteration of
forming processes. Front. Microbiol. 12, 1932. https://doi.org/10.3389/fmicb.2021. banded iron formation with the Carajás high-grade hematite deposits. Appl. Earth
696534. Sci. 112, 26–30. https://doi.org/10.1179/037174503225011234.
Calux, A., Cassimiro, R., Salgado, A., 2019. Caves in iron formations in the Quadrilátero Henwood, W.J., 1871. On the gold mines of Minas Geraes. Trans. R. Geol. Soc. Corn. 8,
Ferrífero, Minas Gerais, southeastern Brazil: lithological, morphological and hydro- 168–370.
logical setting and speleogenesis. Z. Geomorphol. 62, 125–144. https://doi.org/10. Jacobi, C.M., Carmo, F.F., Vincent, R.C., Stehmann, J.R., 2007. Plant communities in iron-
1127/zfg/2019/0475. stone outcrops: a diverse and endangered Brazilian ecosystem. Biodivers. Conserv.
Campana, B., Hughes, F.E., Burns, W.G., Whitcher, I.G., Muceniekos, E., 1964. Discovery of 16, 2185–2200. https://doi.org/10.1007/s10531-007-9156-8.
the Hamersley iron deposits. Proc. Aust. Inst. Min. Metall. 210, 1–30. Jacobi, C.M., Carmo, F.F., Carmo, F.F., Campos, I.C., 2015. Iron geosystems: priority areas for
CANIE (Cadastro Nacional de Informações Espeleológicas), 2021. https://www.icmbio. conservation in Brazil. In: Tibbett, M. (Ed.), Mining in Ecologically Sensitive Land-
gov.br/cecav/canie.html. scapes. CSIRO Publishing, Clayton South, pp. 55–77.
Carmo, F.F., Jacobi, C.M., 2016. Diversity and plant trait-soil relationships among rock out- Jaffé, R., Prous, X., Zampaulo, R., Giannini, T.C., Imperatriz-Fonseca, L., Maurity, C., Oliveira,
crops in the Brazilian Atlantic rainforest. Plant Soil 403, 7–20. https://www.jstor.org/ G., Brandi, I.V., Siqueira, J.O., 2016. Reconciling mining with the conservation of cave
stable/43872630. biodiversity: a quantitative baseline to help establish conservation priorities. Plos One
11 (12), 0168348. https://doi.org/10.1371/journal.pone.0168348.
Carneiro, S.R.C., Marques, E.A.G., Chiste-Costa, J.P., 2016. Evaluation of hydraulic conduc-
tivity of itabirites of the Alegria Centro and Alegria Sul open pits, Samarco Mineração Jaffé, R., Prous, X., Calux, A., Gastauer, M., Nicacio, G., Zampaulo, R., Souza-Filho, P.W.M.,
S.A., Minas Gerais state,Brazil. Mine Water Environ. 35, 136–144. https://doi.org/10. Oliveira, G., Brandi, I.V., Siqueira, J.O., 2018. Conserving relicts from ancient under-
1007/s10230-015-0328-6. ground worlds: assessing the influence of cave and landscape features on obligate
iron cave dwellers from the Eastern Amazon. PeerJ 6, e4531. https://doi.org/10.
Carvalho, J.B., 2012. Avaliação isotópica e hidroquímica na porção noroeste do Aquífero
7717/peerj.4531.
Cauê, Quadrilátero Ferrífero, Brasil. Centro de Desenvolvimento da Tecnologia Nu-
Kambesis, P.N., Larson, E.B., Mylroie, J.E., 2016. Morphometric analysis of cave patterns
clear, Belo Horizonte (MSc Dissertation).
using fractal indices. In: Feinberg, J., Gao, Y., Alexander, E.C. (Eds.), Caves And Karst
Chase, C.G., 1992. Fluvial landsculpting and the fractal dimension of topography. Geomor-
Across Time. Geological Society of America Special Papers. 516, pp. 67–86. https://
phology 5, 39–57. https://doi.org/10.1016/0169-555X(92)90057-U.
doi.org/10.1130/2016.2516(06).
Chiste-Costa, J.P., 2017. Numerical Groundwater Flow Modeling Applied to Dewatering
Kinsella, A.S., Jones, A.M., Bligh, M.W., Pham, A.N., Collins, R.N., Harrison, J.J., Wilsher, K.L.,
Prediction of Cauê Aquifer in Alegria Mine, East of Quadrilátero Ferrífero, Minas
Payne, T.E., Waite, T.D., 2016. Influence of dissolved silicate on rates of Fe(II) oxida-
Gerais. University of São Paulo, São Paulo (MSc Dissertation).
tion. Environ. Sci. Technol. 50, 11663–11671. https://doi.org/10.1021/acs.est.
Cornell, R.M., Schwertmann, U., 2003. The Iron Oxides: Structure, Properties, Reactions,
6b03015.
Occurences And Uses. Wiley, Weinheim.
Klein, C., 2005. Some Precambrian banded iron-formations (BIFs) from around the world:
Costa, T., Sá, G., 2018. Soft iron ores: geotechnical characteristics. In: Martin, D., Stacey, P.
their age, geological setting, mineralogy, metamorphism, geochemistry, and origin.
(Eds.), Guidelines for Open Pit Slope Design in Weak Rocks. CSIRO Publishing, Clayton
Am. Mineral. 90, 1473–1499. https://doi.org/10.2138/am.2005.1871.
South, pp. 273–285.
Klein, C., Ladeira, E.A., 2002. Petrography and geochemistry of the least altered banded
CPRM, 2004. Projeto APA Sul RMBH: hidrogeologia, mapa hidrogeológico escala 1:50.000. iron formation of the Archean Carajás Formation, northern Brazil. Econ. Geol. 97,
SEMAD, CPRM, Belo Horizonte. 643–651. https://doi.org/10.2113/gsecongeo.97.3.643.
Crundwell, F.K., 2017. On the mechanism of the dissolution of quartz and silica in aqueous La Barbera, P., Rosso, R., 1989. On the fractal dimension of stream networks. Water
solution. ACS Omega 2, 1116–1127. https://doi.org/10.1021/acsomega.7b00019. Resour. Res. 25, 735–741. https://doi.org/10.1029/WR025i004p00735.
Curl, R.L., 1986. Fractal dimensions and geometries of caves. Math. Geol. 18, 765–783. Lagoeiro, L.E., 1998. Transformation of magnetite to hematite and its influence on the dis-
https://doi.org/10.1007/BF00899743. solution of iron oxide minerals. J. Metamorph. Geol. 16, 415–423. https://doi.org/10.
Derby, O.A., 1910. The iron ores of Brazil. The Iron Ore Resources of the World. 2, 1111/j.1525-1314.1998.00144.x.
pp. 813–822. Laverty, M., 1987. Fractals in karst. Earth Surf. Process. Landf. 12, 475–480. https://doi.org/
Dias, J.C.S., Bacellar, L.A.P., 2021. A hydrogeological conceptual model for the groundwater 10.1002/esp.3290120505.
dynamics in the ferricretes of Capão Xavier, Iron Quadrangle,Southeastern Brazil. Ca- Lemes, C.G.C., Villa, M.M., Felestrino, E.G., Perucci, L.O., Assis, R.A.B., Cordeiro, I.F., Fonseca,
tena 207, 105633. https://doi.org/10.1016/j.catena.2021.105663. N.P., Guerra, L.C.C., Caneshi, W.L., Moraes, L.A.G., Carmo, F.F., Kamino, L.H.Y., Vale,
Dixey, F., 1920. Notes on laterization in Sierra Leone. Geol. Mag. 57, 211–220. https://doi. P.N.C., Guima, S.E.S., Setubal, J.C., Salgado, A.R.R., Moreira, L.M., 2021. 16S rRNA gene
org/10.1017/S0016756800101530. amplicon sequencing data from the Iron Quadrangle ferruginous caves (Brazil)
Dorr, J.N., 1964. Supergene iron ores of Minas Gerais,Brazil. Econ. Geol. 59, 1203–1240. shows the importance of conserving this singular and threatened geosystem. Divers
https://doi.org/10.2113/gsecongeo.59.7.1203. 13, 494. https://doi.org/10.3390/d13100494.
Dorr, J.N., 1965. Nature and origin of the high-grade hematite ore of Minas Gerais,Brazil. Levett, A., Gagen, E., Shuster, J., Rintoul, L., Tobin, M., Vongsvivut, J., Bambery, K.,
Econ. Geol. 60, 1–46. https://doi.org/10.2113/gsecongeo.60.1.1. Vasconcelos, P., Southam, G., 2016. Evidence of biogeochemical processes in iron du-
Dorr, J.N., 1969. Physiographic. stratigraphic and structural development of the ricrust formation. J. S. Am. Earth Sci. 71, 131–142. https://doi.org/10.1016/j.jsames.
Quadrilátero Ferrífero, Minas Gerais, Brazil. U.S. Geological Survey Professional 2016.06.016.
Paper 641-A. Levett, A., Gagen, E.J., Zhao, Y., Vasconcelos, P.M., Southam, G., 2020a. Biocement stabiliza-
Dosseto, A., Turner, A., Chappell, J., 2008. The evolution of weathering profiles through tion of an experimental-scale artificial slope and the reformation of iron-rich crusts.
time: new insights from uranum-series isotopes. Earth Planet. Sci. Lett. 274, Proc. Natl. Acad. Sci. U. S. A. 117, 18347–18354. https://doi.org/10.1073/pnas.
359–371. https://doi.org/10.1016/j.epsl.2008.07.050. 2001740117.
Dosseto, A., Buss, H., Suresh, P.O., 2011. The delicate balance between soil production and Levett, A., Vasconcelos, P.M., Gagen, E.J., Rintoul, L., Spier, C., Guagliardo, P., Southam, G.,
erosion, and its role on landscape evolution. Appl. Geochem. 26, 524–527. https://doi. 2020b. Microbial weathering signatures in lateritic ferruginous duricrusts. Earth
org/10.1016/j.apgeochem.2011.03.020. Planet. Sci. Lett. 538, 116209. https://doi.org/10.1016/j.epsl.2020.116209.
23
Levett, A., Gagen, E.J., Vasconcelos, P.M., Zhao, Y., Paz, A., Southam, G., 2020c. Biogeochem- Plotnick, R.E., Kenig, F., Scott, A., 2015. Using the voids to fill the gaps: caves, time and
ical cycling of iron: implications for biocementation and slope stabilization. Sci. Total stratigraphy. In: Smith, D.G., Bailey, R.J., Burgess, P.M., Fraser, A.J. (Eds.), Strata And
Environ. 707, 136128. https://doi.org/10.1016/j.scitotenv.2019.136128. Time: Probing the Gaps in Our Understanding. 404. Geological Society, London, Spe-
Levett, A., Gagen, E., Paz, A., Vasconcelos, P., Southam, G., 2021. Strategising the bioreme- cial Publications, pp. 233–250.
diation of Brazilian iron ore mines. Crit. Rev. Environ. Sci. Technol. https://doi.org/10. R Core Team., 2019. R: A Language And Environment for Statistical Computing. R Founda-
1080/10643389.2021.1896346. tion for Statistical Computing, Vienna, Austria. https://www.R-project.org/.
Liu, Y., Steenhuis, T.S., Parlange, J.Y., 1994. Formation and persistence of fingered flow in Reardon, E.J., 1979. Complexing of silica by iron(III) in natural waters. Chem. Geol. 25,
coarse grained soils under different moisture contents. J. Hydrol. 159, 187–195. 339–345. https://doi.org/10.1016/0009-2541(79)90066-4.
https://doi.org/10.1016/0022-1694(94)90255-0. Ribeiro, D.T., 2003. Enriquecimento supergênico de formações ferríferas bandadas:
Lobato, L.M., Rosière, C.A., Silva, R.C.F., Zucchetti, M., Baars, F.J., Seoane, J.C.S., Rios, F.J., estruturas de colapso e desordem. Universidade Federal do Rio de Janeiro, Rio de
Pimentel, M., Mendes, G.E., Monteiro, A.M., 2005. A mineralização hidrotermal de Janeiro (DSc Thesis).
ferro da Província Mineral de Carajás – Controle estrutural e tectônico na evolução Ribeiro, D., Moraes, I., Kwitko-Ribeiro, R., Braga, D., Spier, C., Santos, P., 2021. From fresh
metalogenética da província. In: Marini, O.J., Queiroz, E.T., Ramos, B.W. (Eds.), itabirites and carbonates to weathered iron ore: mineral composition, density and
Caracterização de Depósitos Minerais em Distritos Mineiros da Amazônia. ADIMB, porosity of different fresh and altered rocks from the Quadrilátero Ferrífero,Brazil.
Brasília, pp. 25–92. Minerals 11, 29. https://doi.org/10.3390/min11010029.
Lunevich, L., 2019. Aqueous silica and silica polymerization. In: Farahani, M.H.D.A., Rios, F.R., Santos Jr., A.E.A., 2015. Petrografia e caracterização de espeleotemas no
Vatanpour, V., Taheri, A.H. (Eds.), Desalination. Challenges And Opportunities. Intech testemunho de sondagem N4WS-1495 em formações ferríferas bandadas, Serra
Open, London https://doi.org/10.5772/intechopen.84824. Norte, Província Mineral de Carajás. Proceedings 33° Congresso Brasileiro de
Mandelbrot, B.B., 1977. Fractals: Form, Chance And Dimension. Freeman, San Francisco. Espeleologia, Sociedade Brasileira de Espeleologia, Eldorado, pp. 603–614.
Mandelbrot, B.B., 1983. The Fractal Geometry of Nature. Freeman, New York. Rolim, V.K., Rosière, C.A., Santos, J.O.S., McNaughton, N.J., 2016. The Orosirian-Statherian
Marescaux, M.G., 1973. Les grottes du Gabon Nord-oriental. Un “karst” dans l'oxyde de fer banded iron formation-bearing sequences of the Southern border of the Espinhaço
et la silice. Bull. Assoc. Géogr. Franç. 410, 607–618. https://doi.org/10.3406/bagf.1973. Range, Southeast Brazil. J. S. Am. Earth Sci. 65, 43–66. https://doi.org/10.1016/j.
4718. jsames.2015.11.003.
Maurity, C.W., Kotschoubey, B., 1995. Evolução recente da cobertura de alteração no Platô Rosière, C.A., Heimann, A., Oyhantçabal, P., Santos, J.O.S., 2018. The iron formations of the
de N1 – Serra dos Carajás-PA. Degradação, pseudocarstificação, espeleotemas. Bol. South American Platform. In: Siegesmund, S., Basei, M.A.S., Oyhantçabal, P., Oriolo, S.
Mus. Para. Emilio Goeldi. Série Ciências da Terra 7, 331–362. (Eds.), Geology of Southwest Gondwana. Springer, Cham, pp. 493–526.
McFarlane, M., Twidale, C., 1987. Karstic features associated with tropical weathering pro- Rosière, C., Spier, C.A., Rios, F.J., Suckau, V.E., 2008. The itabirites of the Quadrilátero
files. Z. Geomorphol. (Suppl.64), 73–95. Ferrífero and related high-grade iron ore deposits: an overview. In: Hagemann, S.,
Mesquita, D.C., Dantas, J.C.M., Paula, R.S., Guerra, K.J., 2017. Estudo dos parâmetros Rosière, C.A., Gutzmer, J., Beukes, N.J. (Eds.), Banded Iron Formation-related High-
hidrodinâmicos obtidos em ensaios de campo em itabiritos brandos da porção grade Iron Ore. Society of Economic Geologists, Littleton, pp. 223–254 https://doi.
sudoeste do Quadrilátero Ferrífero, MG. Geonomos 25, 12–19. https://doi.org/10. org/10.5382/Rev.15.09.
18285/geonomos.v25i2.1077. Rosière, C., Stacey, P., 2018. Soft iron ores: geology. In: Martin, D., Stacey, P. (Eds.), Guide-
Monteiro, H.S., Vasconcelos, P.M., Farley, K.A., Spier, C.A., Mello, C.L., 2014. (U-Th)/He geo- lines for Open Pit Slope Design in Weak Rocks. CSIRO Publishing, Clayton South,
chronology of goethite and the origin of Cangas. Geochim. Cosmochim. Acta 131, pp. 270–272.
267–289. https://doi.org/10.1016/j.gca.2014.01.036. Ruchkys, U.A., Machado, M.M.M., 2013. Geological and mining heritage of Iron Quadran-
Monteiro, H.S., Vasconcelos, P.M.P., Farley, K.A., 2018a. A combined (U-Th)/He and cos- gle. Minas Gerais – characterisation and strategies for education and geotourism. Bol.
mogenic 3He record of landscape armoring by biogeochemical iron cycling. Parana. Geoci. 70, 120–136.
J. Geophys.Res. Earth Sci. 123, 298–323. https://doi.org/10.1002/2017JF004282. Sagar, B.D., Chockalingam, L., 2004. Fractal dimension of non-network space of a catch-
Monteiro, H.S., Vasconcelos, P.M.P., Farley, K.A., Lopes, C.A.M., 2018b. Age and evolution of ment basin. Geophys. Res. Lett. 31. https://doi.org/10.1029/2004GL019749.
diachronous erosion surfaces in the Amazon: combining (U-Th)/He and cosmogenic Salgado, A., Braucher, R., Varajão, A.C., Colin, F., Varajão, A.F.D.C., Nalini Jr., H.A., 2008. Re-
3He records. Geochim. Cosmochim. Acta 229, 162–183. https://doi.org/10.1016/j.gca. lief evolution of the Quadrilátero Ferrífero (Minas Gerais, Brazil) by means of (10Be)
2018.02.045. cosmogenic nuclei. Z. Geomorphol. 52, 317–323. https://doi.org/10.1127/0372-8854/
Morris, R.C., Fletcher, A.B., 1987. Increased solubility of quartz following ferrous-ferric iron 2008/0052-0317.
reactions. Nature 330, 558–561. https://doi.org/10.1038/330558a0. Salles, D.M., Carmo, F.F., Jacobi, C.M., 2019. Habitat loss challenges the conservation of en-
Mourão, M.A.A., 2007. Caracterização hidrogeológica do Aquífero Cauê, Quadrilátero demic plants in mining-targeted Brazilian mountains. Environ. Conserv. 46, 140–146.
Ferrífero, MG. Universidade Federal de Minas Gerais, Belo Horizonte (DSc Thesis). https://doi.org/10.1017/S0376892918000401.
Mota, N.F.O., Watanabe, M.T.C., Zappi, D.C., Hiura, A.L., Pallos, J., Viveros, A.M., Giuliette, Sasowsky, I.D., 2007. Clastic sediments in caves – imperfect recorders of processes in
A.M., Viana, P.L., 2018. Amazon canga: the unique vegetation of Carajás revealed by karst. Acta Carsologica 36, 143–149. https://doi.org/10.3986/ac.v36i1.216.
the list of seed species. Rodriguésia 69, 1435–1488. https://doi.org/10.1590/2175- Sasowsky, I.D., Foos, A., Miller, C.M., 2000. Lithic controls on the removal of iron and re-
7860201869336. mediation of acidic mine drainage. Water Res. 34, 2742–2746. https://doi.org/10.
Nola, I.T.S., Bacellar, L.A.P., 2021. Multi-criteria analysis for mapping susceptibility to iron 1016/S0043-1354(00)00019-1.
formation caves development in the Gandarela mountain range (MG), southeast Sauro, F., 2014. Structural and lithological guidance on speleogenesis in quartz–
Brazil. Int. J. Speleol. 50 (2), 173–187. https://doi.org/10.5038/1827-806X.50.2.2376. sandstone: evidence of the arenisation process. Geomorphology 226, 106–123.
Palmer, A.N., 1991. Origin and morphology of limestone caves. Geol. Soc. Am. Bull. 103, https://doi.org/10.1016/j.geomorph.2014.07.033.
1–21. https://doi.org/10.1130/0016-7606(1991)103<0001:OAMOLC>2.3.CO;2. Sauro, F., De Waele, J., Onac, B.O., Galli, E., Dublyansky, Y., Baldoni, E., Sanna, L., 2014.
Palmer, A.N., 2007. Cave Geology. Cave Books, Dayton. Hypogenic speleogenesis in quartzite: the case of Corona 'e Sa Craba Cave (SW Sar-
Pardo-Igúzquiza, E., Dowd, P.A., Durán, J.J., Robledo-Ardila, P., 2018. A review of fractals in dinia, Italy). Geomorphology 211, 77–88. https://doi.org/10.1016/j.geomorph.2013.
karst. Int. J. Speleol. 48, 11–20. https://doi.org/10.5038/1827-806X.48.1.2218. 12.031.
Parker, C.W., 2018. Microbial Iron Reduction in the Development of Iron Formation Caves. Schaefer, C.E.G.R., Corrêa, G.R., Candido, H.G., Arruda, D.M., Nunes, J.A., Araújo, R.W.,
PhD DissertationUniversity of Akron, Akron. Rodrigues, P.M.S., Fernandes Filho, E.I., Pereira, A.F.S., Brandão, P.C., Neri, A.V., 2016.
Parker, C.W., Auler, A.S., Senko, J.M., Sasowsky, I.D., Piló, L.B., Smith, M., Johnson, M., The physical environment of rupestrian grasslands (campos rupestres) in Brazil: geo-
Barton, H.A., 2013a. Microbial iron cycling and biospeleogenesis: cave development logical, geomorphological and pedological characteristics and interplays. In:
in the Carajás Formation, Brazil. In: Filippi, M., Bosak, P. (Eds.), Proceedings of the Fernandes, G.W. (Ed.), Ecology And Conservation of Mountaintop Grasslands in
16th Congress of Speleology. 1. Czech Speleological Society, Prague, pp. 442–446. Brazil. Springer, pp. 15–53.
Parker, C.W., Wolf, J.A., Auler, A.S., Barton, H.A., Senko, J.M., 2013b. Microbial reducibility Scherer, R.S., 2017. Ocorrência de espeleotemas fosfáticos e feições morfológicas raras em
of Fe(III) phases associated with the genesis of iron ore caves in the Iron Quadrangle, cavernas ferríferas da Serra de Carajás no Pará. Instituto Tecnológico Vale, Belém.
Minas Gerais, Brazil. Minerals 3, 393–411. https://doi.org/10.3390/min3040395. Silva Junior, R.O., Queiroz, J.C.B., Ferreira, D.B.S., Lima, A.T., Souza Filho, P.W.M., Guimarães,
Parker, C.W., Auler, A.S., Barton, M.D., Sasowsky, I.D., Senko, J.M., Barton, H.A., 2018. Fe(III) J.T.F., Rocha, E.P., 2017. Estimativa de precipitação e vazões médias Para a bacia
reducing microorganisms from iron ore caves demonstrate fermentative Fe(III) re- hidrográfica do rio Itacaiúnas (BHRI). Amazônia Oriental,Brasil. Rev. Bras. Geogr. Fis.
duction and promote cave formation. Geomicrobiol J. 35, 311–322. https://doi.org/ 10, 1638–1654. https://doi.org/10.26848/rbgf.v.10.5.p1638-1654.
10.1080/01490451.2017.1368741. Silva, M.S., Martins, R.P., Ferreira, R.L., 2011. Cave lithology determining the structure of
Paz, A., Gagen, E.J., Levett, A., Zhao, Y., Kopittke, P.M., Southam, G., 2020. Biogeochemical the invertebrate communities in the Brazilian Atlantic Rain Forest. Biodiver. Conserv.
cycling of iron oxides in the rhizosphere of plants grown on ferruginous duricrust 20, 1713–1729. https://doi.org/10.1007/s10531-011-0057-5.
(canga). Sci. Total Environ. 713, 136637. https://doi.org/10.1016/j.scitotenv.2020. Simmons, G.C., 1960. Origin of certain cangas of the “Quadrilátero Ferrífero” of Minas
136637. Gerais,Brazil. Bol. Soc. Bras. Geol. 9, 37–59.
Piló, L.B., Auler, A.S., 2005. Cavernas em minério de ferro e canga de Capão Xavier. Simmons, G.C., 1963. Canga caves in the Quadrilátero Ferrífero, Minas Gerais,Brazil. Bull.
Quadrilátero Ferrífero,MG. O Carste 17, 92–105. Nat. Speleol. Soc. 25, 66–72.
Piló, L.B., Auler, A.S., Martins, F., 2015. Carajás National Forest: iron ore plateaus and caves Simmons, G.C., 1964. Leucophosphite, a new occurrence in the Quadrilátero Ferrífero,
in southeastern Amazon. In: Vieira, B.C., Salgado, A.A.R., Santos, L.J.C. (Eds.), Land- Minas Gerais,Brazil. Am. Mineral. 49, 377–386.
scapes And Landforms of Brazil. Springer, Dordrecht, pp. 273–283. Skirycz, A., Castilho, A., Chaparro, C., Carvalho, N., Tzotzos, G., Siqueira, J.O., 2014. Canga
Pinheiro, R.V.L., 1997. Reactivation History of the Carajás And Cinzento Strike-slip System. biodiversity, a matter of mining. Front. Plant Sci. 5, 653. https://doi.org/10.3389/
Amazon. Brazil. University of Durham, Durham (PhD Thesis). fpls.2014.00653.
Pissis, N.A., 1849. Mémoire sur la position géologique des terrains de la partie austral du Sonter, L.J., Herrera, D., Barrett, D.J., Galford, G.L., Moran, C.J., Soares-Filho, B.S., 2017. Min-
Brésil et sur les soulèvements qui, à diverses époques, ont changé le relief de cette ing drives extensive deforestation in the Brazilian Amazon. Nat. Comm. 8, 1013.
contrée. Mém. Inst. Fr. 10, 353–413. https://doi.org/10.1038/s41467-017-00557-w.
24
Souza-Filho, P.W., Giannini, T.C., Jaffé, R., Giulietti, A.M., Santos, D.C., Nascimento Jr., W.R., Trevelin, L.C., Gastauer, M., Prous, X., Nicácio, G., Zampaulo, R., Brandi, I., Oliveira, G.,
Guimarães, J.T.F., Costa, M.F., Imperatriz-Fonseca, V.L., Siqueira, J.O., 2019. Mapping Siqueira, J.O., Jaffé, R., 2019. Biodiversity surrogates in Amazonian iron cave eco-
and quantification of ferruginous outcrop savannas in the Brazilian Amazon: a chal- systems. Ecol. Indic. 101, 813–820. https://doi.org/10.1016/j.ecolind.2019.01.
lenge for biodiversity conservation. Plos One 14 (1), e0211095. https://doi.org/10. 086.
1371/journal.pone.0211095. Twidale, C.R., Horwitz, R.C., Campbell, E.M., 1985. Hamersley landscapes of Western
Spier, C.A., Levett, A., Rosière, C.A., 2019. Geochemistry of canga (ferricrete) and evolution Australia. Ver. Géol. Dynam. Géogr. Phys. 26, 173–186.
of the weathering profile developed on itabirite and iron ore in the Quadrilátero Vasconcelos, P.M., Carmo, I.O., 2018. Calibrating denudation chronology through 40Ar/
Ferrífero, Minas Gerais,Brazil. Miner. Depos. 54, 983–1010. https://doi.org/10.1007/ 39Ar weathering geochronology. Earth Sci. Rev. 179, 411–435. https://doi.org/10.
s00126-018-0856-7. 1016/j.earscirev.2018.01.003.
Spier, C.A., Oliveira, S.M.B., Rosière, C.A., 2003. Geology and geochemistry of the Águas Vasconcelos, P.M., Farley, K.A., Stone, J., Piacentini, T., Fifield, L.K., 2019. Stranded land-
Claras and Pico iron mines, Quadrilátero Ferrífero, Minas Gerais, Brazil. Miner. scapes in the humid tropics: Earth's oldest land surfaces. Earth Planet. Sci. Lett. 519,
Depos. 38, 751–774. https://doi.org/10.1007/s00126-003-0371-2. 152–164. https://doi.org/10.1016/j.epsl.2019.04.014.
Spier, C.A., Oliveira, S.M.B., Sial, A.N., Rios, F.J., 2007. Geochemistry and genesis of the Wang, Y., Li, Y., Wang, X., Chau, H.W., 2018. Finger flow development in layered water-
banded iron formations of the Cauê Formation, Quadrilátero Ferrífero, Minas Gerais, repellent soils. Vadose Zone J. 17, 170171. https://doi.org/10.2136/vzj2017.09.0171.
Brazil. Precambrian Res. 152, 170–206. https://doi.org/10.1016/j.precamres.2006.10. Weber, K.A., Achenbach, L.A., Coates, J.D., 2006. Microorganisms pumping iron: anaerobic
003. microbial iron oxidation and reduction. Nat. Rev. Microbiol. 4, 752–764. https://doi.
Spier, C.A., Vasconcelos, P.M., Oliveira, S.M.B., 2006. 40Ar/39Ar geochronological con- org/10.1038/nrmicro1490.
straints on the evolution of lateritic iron deposits in the Quadrilátero Ferrífero, Wray, R.A., Sauro, F., 2017. An updated global review of solutional weathering processes
Minas Gerais, Brazil. Chem. Geol. 234, 79–104. https://doi.org/10.1016/j.chemgeo. and forms in quartz sandstone and quartzites. Earth Sci. Rev. 171, 520–557. https://
2006.04.006. doi.org/10.1016/j.earscirev.2017.06.008.
Suchý, V., Sýborová, I., Zachariáš, J., Filip, J., Machovič, V., Lapčák, L., 2017. Hypogene fea- Zappi, D.C., Moro, M.F., Walker, B., Meagher, T., Viana, P.L., Mota, N.F.O., Watanabe, M.T.C.,
tures in sandstone: an example from Carboniferous basins of Central and Western Lughadha, E.N., 2019. Plotting a future for Amazonian canga vegetation in a campo
Bohemia, Czech Republic. In: Klimchouk, A., Palmer, A.N., De Waele, J., Auler, A.S., rupestre context. Plos One 14 (8), e0219753. https://doi.org/10.1371/journal.pone.
Audra, P. (Eds.), Hypogene Karst Regions And Caves of the World. Springer, Cham, 0219753.
pp. 313–328.
Trendall, A.F., Basei, M.A.S., de Laeter, J.R., Nelson, D.R., 1998. SHRIMP zircon U-Pb con-
straints on the age of the Carajás formation, Grão Pará Group, Amazon Craton. J. S.
Am. Earth Sci. 11, 265–277. https://doi.org/10.1016/S0895-9811(98)00015-7.
25

Auler Et Al 2022

Uploaded by

Copyright:

Available Formats

Auler Et Al 2022

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Auler Et Al 2022

Uploaded by

Copyright:

Available Formats

Geomorphology 398 (2022) 108068

Contents lists available at ScienceDirect

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

Silica and iron mobilization, cave development and landscape evolution

1. Introduction commented on these features throughout the world, although detailed

surrounding rocks is marked by small scarps. IF scarps are ubiquitous

4.3. Geological controls on cave dimension

vertical collapse shafts, being originally devoid of a natural entrance.

Carajás Canga 469.8 4.8 111.6 97.9 5.5

Fractal Dimension (FD)

FracLac Fractal 3.4.7

Min Mean Max Range Min Mean Max Range

6.1. Initial porosity

geochemical and erosional agents. Canga is highly permeable, in

You might also like