Factors controlling gully erosion 1369
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Earth Surface Processes and Landforms
Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/esp.1193
Controlling factors of gullying in the Maracujá
Catchment, Southeastern Brazil
L. de A. P. Bacellar,1
* A. L. Coelho Netto2
and W. A. Lacerda3
1
Department of Geology, Ouro Preto Federal University, Ouro Preto, MG, Brazil
2
Department of Geography, Rio de Janeiro Federal University, Ilha do Fundo, Rio de Janeiro, RJ, Brazil
3
COPPE/PEC, Rio de Janeiro Federal University, Ilha do Fundo, Rio de Janeiro, RJ, Brazil
Abstract
Hundreds of gullies (`voçorocas') of huge dimensions (up to 400­500 m long, 150 m wide
and 50 m deep) are very common in the small Maracujá Catchment in southeastern Brazil.
These erosional features, which occur with an uneven intensity throughout the area, started
due to bad soil management practices at the beginning of European settlement, at the end of
the 17th century, and nowadays are still evolving, but at a slower rate. As surface soils are
usually very resistant to erosion, the outcrop of the more erodible basement saprolites seems
to be an essential condition for their beginning. An analysis of well known erosion controlling
factors was performed, aiming to explain the beginning and evolution of these gullies
and to understand the reasons for their spatial distribution. Data shows that geology and,
mainly, geomorphology are the main controlling factors, since gullies tend to be concentrated
in basement rock areas with lower relief (domain 2) of Maracujá Catchment, mainly
at the fringes of broad and flat interfluves. At the detailed scale (1:10 000), gullies are more
common in amphitheatre-like headwater hollows that frequently represent upper Quaternary
gullies (paleogullies), which demonstrate the recurrence of channel erosion. So, gullies
occur in areas of thicker saprolites (domain 2), in places with a natural concentration of
surface and underground water (hollows). Saprolites of the preserved, non-eroded hollows
are usually pressurized (confined aquifer) due to a thick seal of Quaternary clay layer, in a
similar configuration to the ones found in hollows of mass movement (mudflow) sites in
southeastern Brazil. Therefore, the erosion of the resistant soils by human activities, such
as road cuts and trenches (`valos'), or their mobilization by mudflow movements, seem to
be likely mechanisms of gullying initiation. Afterwards, gullies evolve by a combination of
surface and underground processes, such as wash and tunnel erosion and falls and slumps of
gully walls. Copyright  2005 John Wiley & Sons, Ltd.
Keywords: gullying; erosion; human impacts; geomorphologic controls; catchment
Received 5 September 2004
Revised 8 September 2004
Accepted 1 October 2004
*Correspondence to: L. de
A. P. Bacellar, UFOP/EM/DEGEO,
CEP 35400-000 ­ Campus
Universitário, Ouro Preto ­ Minas
Gerais, Brazil. E-mail:
bacellar@degeo.ufop.br
Introduction
Gullies are considered one of the worst environmental problems in crystalline basement rock areas of tropical regions,
where they are frequent and can reach large dimensions (Wells and Andriamihaja, 1993; Xu, 1996; Coelho Netto,
1997). In these areas, gully erosion is responsible for several environmental impacts, such as soil losses, destruction of
properties and natural habitats, silting of reservoirs and groundwater drawdown (Lal, 1990; Xu, 1996; Bacellar, 2000).
Gullies usually appear as a consequence of bad land management practices, but some studies have shown that there is
not always a clear relation between land use and gully intensity, since other factors, such as relief, seem to control
their spatial distribution (Dietrich and Dunne, 1993; Coelho Netto, 1997; Poesen et al., 2003). Although it is not yet
possible to predict gully development, there are some attempts to relate gully location to controlling factors, such as
some relief attributes, such as drainage area and slope gradient and form (see, e.g., Dietrich and Dunne, 1993; Poesen
et al., 2003). Gullies can develop by different erosional processes and mechanisms, such as overland flow erosion and
subsurface flow erosion (tunnel and seepage erosion) as well as several kinds of mass movement (see, e.g., Lal, 1990;
Selby, 1990; Dietrich and Dunne, 1993; Oliveira, 1997; Coelho Netto, 1997). These mechanisms should be known
1370 L. de A. P. Bacellar, A. L. Coelho Netto and W. A. Lacerda
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Figure 1. Location of the Maracujá River Catchment.
when gully prevention and control measures have to be taken. Here, we try to determine the controlling factors and the
evolving mechanisms of gullying in Maracujá Catchment in order to explain its heterogeneous concentration. We hope
that this knowledge will improve future land management practices in this catchment and in other areas with similar
environmental conditions.
Study Area
The Maracujá Catchment, 140 km2
in area, is located in Minas Gerais state, in southeast Brazil (Figure 1), inside a
famous Brazilian geomorphic landscape, the Ferriferous Quadrangle (Door, 1969). This river is a sixth order (Strahler
classification) tributary of the Velhas River, which, in turn, is one of the main tributaries of the So Francisco River,
the third largest in South America (Figure 1). Maracujá Catchment shows a diversified set of lithologies and landscapes
and an uneven distribution of gullies.
Geology
Four geological units can be found in the Maracujá Catchment: crystalline basement, Minas and Rio das Velhas
Supergroups and Quaternary sediments. The crystalline basement rocks (Arquean) outcrop in a structural window,
with a domic form (Baço Complex, Door, 1969; Hippert, 1994), cut in Minas and Rio das Velhas Supergroup rocks
(Figure 1). The Baço Complex is composed of tonalitic, trondhjemitic or granodioritic gneisses, with some intrusions
of granodioritic and granitic granitoids (see, e.g., Hippert, 1994; Endo, 1997). The gneisses frequently show a well
developed compositional banding, with dark micaceous bands alternating with clear quartz­feldspatic ones. In a recent
detailed mapping (Franco, unpublished geological report; Graça, unpublished report; Lopes, unpublished report;
Martins, unpublished report; Salaroli, unpublished geological report; Vilela, unpublished report) three main bodies
of gneisses were mapped (Funil, Amarantina and Praia, Figure 2) ­ with very slight compositional and textural
differences among them.
Factors controlling gully erosion 1371
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Rio das Velhas Supergroup unit (Arquean) is composed in the area by schists, metamafic and metaultramafic rocks
(Endo, 1997). This unit overlies basement rocks, outcropping around the Baço Complex and also inside it, in the
form of long rock slabs, probably placed by shear zones (Figures 1 and 2). Minas Supergroup rocks (Proterozoic)
overlie the previous units (Figure 1). They outcrop on Maracujá Catchment headwaters (Figure 2), in a region of
higher altitudes (`Upper Compartmenť, see below), where they are composed of schists, phyllites and, secondarily, by
quartzites and ferriferous formations (Door, 1969). Some of these quartzites are hard, but there are certain thin veins
of ferruginous quartzite that are very soft and friable. Finally, this region is covered by unconsolidated colluvial and
alluvial sediments (Quaternary), poorly described until now.
The geotectonic evolution of the area is complex, due to the superimposition of several tectonic events from
Arquean to Upper Proterozoic age (Endo, 1997), which resulted in the formation of shear zones, folds and foliations
(schistosity and compositional bandings). Recently, strong evidence has been found of brittle tectonic events during
Tertiary and Quaternary periods that gave origin in certain places to brittle structures (joints and faults) in the
correlative sediments (Saadi, 1991). As these neotectonic phases are not completely understood yet, more studies are
currently being made in order to understand their characteristics and relative chronology.
Figure 2. Geologic map of the Maracujá Catchment (modified from Johnson, 1962; Franco, unpublished report; Graça, unpublished
report; Lopes, unpublished report; Martins, unpublished report; Salaroli, unpublished geological report;Vilela, unpublished report).
1372 L. de A. P. Bacellar, A. L. Coelho Netto and W. A. Lacerda
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Weather and vegetation
Mean annual precipitation in the area ranges from 1024 to 1744 mm, with a maximum between October and March
(Bacellar, 2000). Weather is colder and more humid (CWb, in Köppen classification) on the Upper Compartment
(Minas rocks) than on the lower one, which is hotter and more seasonal (CWa).
Vegetation varies in a similar way, with a dominance of tropical forest on the Upper Compartment and a savannah
type (`campos cerrados') over great part of the Lower Compartment, especially on flatter areas, with a predominance
of oxisols, poor in nutrients. However, even in this compartment there are scattered pockets of riparian tropical forest
along rivers and headwaters hollows or over areas of more fertile soil, as on Rio das Velhas Supergroup areas inside
the Baço Complex (Figure 2).
Historical background and land use
Brazilian native people already occupied this region at least 10 000 years B.P. and the European settlement only began
at the end of the 17th century, as a consequence of a huge gold rush nearby (Burton, 1869; Gutersohn, 1945). Gold did
not occur in the Maracujá Catchment, but this area was a supply centre of cattle and agricultural products for the
neighbouring mining towns (Ouro Preto and Sabará), some kilometres away, at that time economically very active
(Gutersohn, 1945). In fact, compared with its neighbouring regions, the Baço basement window was preferentially
settled because of its relatively lower relief. Land occupation by man has been increasing in the area since the 1970s.
Nowadays, most of the primary vegetation has been cut and the main economic activity is cattle grazing on mismanaged
and overstocked pastures.
Methods
Geomorphology was characterized at 1:25 000 scale, through the integration of the following maps: hypsometric,
slope and relief compartment, the last done following the methodology of Meis et al. (1982). At a larger scale, the
relief was described following Hack and Goodlet (1960).
Quaternary sediments were characterized through an approach adapted to study Quaternary continental sediments of
mountainous regions, the morphostratigraphic approach, initially proposed by Frye and Willman (1962), and further
modified by Meis and Moura (1984) and Mello (1997) in order to couple the morphological description of geomorphic
surfaces with allostratigraphy (NACSN, 1983). A detailed stratigraphic survey was completed on exposures along
road cuttings and gully walls, supported by additional subsurface information (auger boring and GPR data), in order to
identify sedimentary bodies and regional disconformities. Ten representative geological cross-sections were made
across specific areas, mainly at headwaters hollows. This information was always tied to good morphological classification
of colluvial ramp and alluvial terrace surfaces. This involved comparisons of top surface levelling and form
reshaping by either erosional processes (sheet, splash and rill erosion) or colluvial or alluvial deposition through
Upper Quaternary (Bacellar, 2000). Thus, in the recently incised channel valleys typical of this region, the older the
ramp or terrace surface the higher and the rounder it usually is, because their primary form tends to be gradually
reshaped by these geomorphic processes (Martin et al., 1998). Regional correlation and some 14C datings of organic
clay layers and correlated paleosols, as was done by Dietrich et al. (1990), completed this analysis and permitted the
chronological distinction of morphostratigraphic units in the catchment.
Historical recordings (e.g. Burton, 1869) and two sets of aerial photographs (from 1948 and 1996) allowed the
identification of the disturbance caused by humans and the evolution rate, morphology and spatial frequency of gullies
(Figure 3). Finally, gullying controlling factors were analysed, comparing simultaneously several layers of essential
attributes with a gullying distribution map.
Soil erodibility analysis was carried out through field estimates and laboratory tests. Field estimates are based on
the visual inspection of the relative intensity and frequency of erosional features (rills and plunge pools alcoves) on
soil exposures (see, e.g., Alcântara, 1997). Then, soils are classified as of low, intermediate and high erodibility.
Moreover, the following commonly used laboratory tests (Bryan, 1976, 2000; Head, 1986; Lal, 1990) were also
evaluated as erodibility indexes.
* Silt­clay ratio (SCR) ­ this consists of the ratio of silt to clay content, both obtained with grain size tests. It is a
variation of the Boyoucos ratio (Selby, 1990).
* Crumb test (Sherard et al., 1976).
* Penetrometer test, which uses a drop-cone penetrometer (Head, 1986), according to the methodology of Alcântara
(1997). It consists of measurements of the penetration depth of soil at natural moisture content and also in the
Factors controlling gully erosion 1373
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Figure 3. Gullies of the Maracujá Catchment.
water-saturated state. According to Alcântara (1997), the difference of cone penetration depth (DP) between the
water-saturated sample (Psat) and natural moisture content sample (Pnat) is proportional to the erodibility of
tropical soils.
* Aggregate stability, following Yoder (1936) and Kemper and Roseneau (1986).
Maracujá Catchment Characteristics
Geomorphology
The relief varies widely throughout this catchment. Field data, aerial photography interpretation and the superposition
of three maps derived from topography maps (hypsometric, slope and relief compartment) allowed the characterization
of four homogeneous geomorphologic domains in the catchment (Figure 4).
Domain 1 is located on the catchment headwaters (`Upper Compartmenť), on the Minas supergroup rocks. Altitudes
are higher than 1140 metres and the hills are steep, with relief amplitudes between 70 and 140 metres. Hillcrests
are sharp and follow rock foliation, forming a trellis drainage pattern. It is separated from other downstream domains
by some waterfalls (bedrock steps that act as local base levels).
The other domains (2, 3 and 4), located in the medium and low Maracujá River (`Lower Compartmenť), between
1140 and 900 metres in altitude, are composed of basement and minor proportions of Rio das Velhas rocks (Figure 2).
They exhibit an angular-dendritic drainage pattern.
Domain 2, which occurs in three separated areas (2a, 2b and 2c, Figure 4), shows gently sloping hills (<30 per
cemt) and amplitudes lower than 70 metres. On the other hand, a steeper relief, with amplitudes higher than 140
metres, characterizes domain 3. Finally, domain 4 presents an intermediate pattern between domains 2 and 3.
As the lithology of the Lower Compartment domains is similar, the flatter relief in domain 2 is probably due to the
more effective action of local base levels at their channel outlets (Figure 4). These base levels have apparently
prevented deeper river incision during recent geologic time, favouring lateral channel erosion over channel scouring
and allowing wider and thinner Quaternary sedimentation (see below). Deprived of effective local base level control at
their channel outlets, domains 3 and 4 show more incised valleys and narrower Quaternary sedimentation. Studies are
1374 L. de A. P. Bacellar, A. L. Coelho Netto and W. A. Lacerda
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Figure 4. Geomorphologic domains of the Maracujá Catchment.
being performed in order to understand whether local base levels are of tectonic origin. In fact, the alignment of
domain 2a base levels (knick points along a ENE direction, Figure 4) is a good indication of their tectonic origin.
Thicker and well developed soils (oxisols) prevail in the basin, especially on flatter areas (domain 2), while less
developed ones (inceptisols and ultisols) are more common in the steep terrains, mainly in domains 1, 3 and 4.
Entisols are restricted to some areas of domains 1 and 3, over resistant rocks, such as quartzites, schists and some
granitoids.
Quaternary sediments
Unconsolidated sediments of Quaternary age occur throughout the catchment as colluvial ramps or fluvial terraces, up
to 20 metres thick. Such sediments were studied through a morphostratigraphic approach (Frye and Willman, 1962;
Meis and Moura, 1984; Mello, 1997), which allowed the chronological distinction of three main Upper Quaternary
morphostratigraphic bodies in the catchment (Figure 5), very similar to the ones found nearby (Mello, 1997).
T3­R3 sediments. These represent the older alluvial (T3) and colluvial (R3) sediments of the area, occurring with
very well rounded topographic forms, totally reshaped by erosional processes and colluvial or alluvial deposition
(Figure 5). They occur locally on side slopes of the main valleys, usually several metres above the current floodplain.
They are harder and more quartzose than younger sediments and present light grey tones, with reddish and yellowish
mottles (plyntite). Neotectonic joints and faults are very common in some outcrops (Bacellar, 2000).
T2­R2 sediments. Recognized elsewhere in the catchment, they usually lie on valley bottoms, some metres above
the current floodplain. According to some 14C datings on organic clays and paleosols, they are 31 340­7490 years
B.P. and occur discordantly over basement or T3­R3 sediments. The alluvial sediments (T2), composed of pebbly or
clayey sands and organic clays, represent fluvial terraces, which outcrop along low-order channels and in their
unchannelled amphitheatre-like headwaters (hollows), where they frequently fill partially ancient erosional features
(paleogullies; see below). These terraces are partially or totally overlain by reddish-brown, cohesive, massive, sandyclay
sediments, generally with a well developed soil profile (usually oxisols). Stratigraphic and geomorphologic
evidence allow us to interpret them as colluvial ramps (R2), which originated from eroded lateral slopes, provoking
a partial reshaping of T2 terraces (ramped terraces, Figure 5). Joints and strike-slip faults frequently affect these
sediments.
Factors controlling gully erosion 1375
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Figure 5. Idealized geological cross-section depicting the relationship among quaternary sedimentary units of the Maracujá
Catchment.The older the unit the more reshaped it is by subsequent erosional and depositional processes (see the text).
Figure 6. Northward view of typical gullies of Maracuja Catchment. The dotted white line represent the top of the gneiss
saprolite.These gullies (A and B) are located in Holanda monitoring station (see Figure 2).
T1 sediments. There are found as fluvial terraces (T1), incised on T2 and up to 2 metres above the current
floodplains (Figure 5). They are up to 3 metres thick and are composed of sandy clays and also organic clays, but
younger (<5300 years BP) and softer than T2 clays. Because of their younger age, these sediments were not overlaid
by significant colluvial ramps, and their top surface is, therefore, plain (Figure 5). There is no evidence of tectonic
deformation in these sediments.
Erosion
As in other parts of southeastern Brazil, post-settlement bad soil management practices have caused severe sheet and
rill erosion in this region, especially on soils on steep slopes from basement areas, which nowadays are frequently
deprived of the surface organic (A) soil horizons. These kinds of erosion caused progressive reduction of soil fertility
leading to a gradual economic impoverishment of the whole region (Gutersohn, 1945).
Nevertheless, the most astonishing erosional features of the area are the 385 gullies, easily identified by aerial
photography interpretation (Figure 3). They occur with a heterogeneous distribution throughout the catchment, in
some places eroding more than 40 per cent of the ground, as in some parts of geomorphologic domain 2, while they
can be almost absent on other places, as in domain 3 (compare Figures 3 and 4). These gullies are always graded to
the drainage system and are very large (Figure 6), frequently reaching 400 to 500 metres long, 150 metres wide and
50 metres deep. According to historical recordings (Burton, 1869), gullies started due to bad agricultural soil practices
at the beginning of European settlement. The old age of such gullies makes determination of their controlling factors
and starting mechanisms difficult, but their current evolution clearly occurs by several mechanisms (discussed below),
notably slumping, favoured by the headcut undermining and by the high hydraulic gradients in the characteristically
steep gully walls. Nowadays, aerial photography interpretation shows that in the last 50 years few gullies have
appeared and most of the old ones are now totally stabilized or evolving at slow rates (Sobreira, 1998), as many of
them have already reached the interfluves in their headward retreat. This precludes any correlation with the evolution
model of gullying of Oliveira (1997).
1376 L. de A. P. Bacellar, A. L. Coelho Netto and W. A. Lacerda
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Soil erodibility
As is well known, it is very difficult to measure erodibility, because of the great number of intervening factors
(pedologic, geologic, geomorphologic, anthropic, biologic and climatic) and the variety of erosional mechanisms (see,
e.g., Lal, 1990; Bryan, 2000). As there is not yet a universal method to quantify this property, soil erodibility was first
estimated in the field by visual inspection (VI) of road cuts and gully walls, since the magnitude of rills and plungepools
alcoves (Dietrich & Dunne, 1993) on the exposures can be crudely correlated with this property. Based on the
intensity and frequency of these erosional features the soils were classified as of low, intermediate and high erodibility.
Granulometry tests carried out on several samples showed that low and high erodibility soils classified by visual
inspection have silt­clay ratio (SCR) lower than 066 and greater than 144, respectively. This information confirms
that among tropical soils, usually deprived of swelling minerals, the higher the clay content the bigger the cohesion
and aggregation and, consequently, the lower its erodibility (Barths and Roose, 2002).
Thus, soils derived from Minas and Rio das Velhas rocks usually show low erodibility throughout their profiles, due
to their higher content of non-swelling clays (SCR lower than 066). The only exceptions are the soils from some thin
quartzite veins from the Minas group (ferruginous quartzites), which outcrop on catchment headwaters (domain 1), that
produce friable soils and saprolites, with a fine silty texture and SCR > 170, which are very susceptible to gully erosion.
Conversely, basement soil profiles in the Maracujá catchment are usually very heterogeneous, with low erodibility
soils (VI low and SCR < 066) above more erodible saprolites. Indeed, the erodibility of these saprolites apparently
can vary widely, as shown by the VI (intermediate and high erodibility) and by the SCR, which ranges from 200 to
600 (Bacellar, 2000). However, VI and SCR indices could not separate the subtle erodibility differences found
amongst basement saprolites.
In order to determine erodibility more precisely, some traditional geotechnical and agronomical laboratory tests
were carried out on 13 selected samples of Funil gneiss saprolites (Table I), which is the geologic unit most prone to
gully erosion in the Maracujá Catchment. Two samples of soils (oxisol B horizon) were tested as well, to calibrate the
test results (Table I).
Although there is not good agreement with SCR, the VI data indicate two main classes of erodibility among these
Funil gneiss saprolites: intermediate and high (Table I). The first test performed was the crumb desegregation test
(Sherard et al., 1976), which shows poor replicability and no correlation with SCR and VI.
Then, grain size tests without chemical dispersant and strong mechanical agitation were carried out, in order to
obtain some dispersion index through the SCS test (Sherard et al., 1976). However, these samples do not exhibit
particles smaller than 0005 mm with the test procedures (SCS = 0), confirming that these saprolites are not dispersive.
Afterwards, an erodibility test with a drop-cone penetrometer (Head, 1986) was used, following the methodology of
Alcântara (1997). According to this author, the difference of cone penetration (DP) between the saturated sample
(Psat) and natural moisture state sample (Pnat) would be proportional to soil erodibility. The DP results of the selected
samples showed good correlation with visual erodibility, but when low erodibility soil samples were evaluated, this
Table I. Erodibility data from selected Funil gneiss saprolites and soils (oxisol B horizon)
Visual Silt­clay
Sample inspection (VI) ratio (SCR) DP (%) Psat (mm) MWD (mm)
Soil 1 low 0,28 80 8,8 2,08
2 low 0,44 121 6,2 2,2
Saprolite 9 intermediate 4,11 ­ ­ ­
9b intermediate 2,50 28 6,2 0,7
151 intermediate 3,50 15 4,8 1,26
212­7 intermediate 2,54 30 6,0 0,87
212­8 intermediate 2,27 ­ ­ ­
151m high 5,35 143 13,0 0,33
212­1 high ­ ­ ­ ­
212­2 high 6,00 158 16,3 0,22
212­3 high 4,18 137 18,3 0,16
212­4 high 4,45 82 18,0 0,17
212­5 high 6,18 93 12,2 0,13
212­6 high 3,91 158 14,0 0,17
212­10 high 3,64 121 15,0 0,25
­ No results.
Factors controlling gully erosion 1377
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
index failed (Table I). Better agreement can be found when VI for saprolites and soils is compared with the amount of
saturated penetration (Psat, Table I), supporting the idea that there could be some relationship between strength and
erodibility (Bryan, 2000).
The best results were found with the aggregate stability test, measured by the mean weight diameter (MWD index,
following Yoder (1936) and Kemper and Rosenau (1986), which allowed distinction of slight erodibility differences,
confirming previous studies in this area (Parzanese, 1991; Silva, 2000) and elsewhere (Bryan, 1976, 2000; Barths and
Roose, 2002). Funil gneiss saprolites with MWD > 070 mm and <033 mm have low and medium (VI) erodibilities,
respectively (Table I). It is important to say that this test was equally efficient in evaluating erodibility among Funil
gneiss soils (oxisols), which exhibit coarser water-stable aggregates (MWD > 208 mm), due to the iron oxides and
hydroxides as natural cements.
Scan microscope data provided by Morais (2003) showed that Funil gneiss saprolites with intermediate erodibility
(Table I) have either an abnormal particle aggregation, due to some inherited characteristics of the parent rock, usually
coarser and more quartzose (commented on below), or some kind of cement deposition (iron oxides and hydroxides)
during further weathering processes.
Gullying Controlling Factors
As commented previously, gully concentration varies enormously throughout the catchment (Figure 3). In order to
understand the reasons, the usual erosion controlling factors (see, e.g., Xu, 1996) were analysed separately.
Pedologic controls
There is no evident relationship between pedology and channel erosion in Maracujá Catchment, because soils are
systematically less erodible than the saprolites, especially on flatter basement areas, with predominance of oxisols, where
gullies are more frequent. Santos (2001) performed a year-round experiment on field plots, showing that the sheet erosion
rate over local oxisols is very slow under the current volumes of overland flow. Therefore, the exposure of saprolites
seems to be an essential condition for gully initiation, confirming previous studies in this area (Parzanese, 1991).
Anthropogenic controls
Human activities, such as deforestation and the construction of fences, paths, trenches and roads, are usually mentioned
as responsible for the concentration of overland flow and, consequently, for gully initiation in Brazil (see, e.g.,
Parzanese, 1991). About 70 per cent of Maracujá Catchmenťs gullies are directly associated with human activities
(Bacellar, 2000), mainly the boundary ditches (`valos'), which are trenches hundreds of metres long and 2 to 3 metres
depth, dug in the soil by the first settlers to separate properties, instead of fences or walls.
Erodibility tests with the Inderbitzen apparatus (Inderbitzen, 1961) on similar oxisols of southeastern Brazil have
shown that shallow overland flows with velocities > 164 m/s are able to erode them easily (Rego, 1978). Field
observations have showed that during storms some ditches could concentrate torrents up to 3 metres deep. Bacellar
(2000) performed some analytic simulations with the Manning empirical equation for channel flows (Selby, 1990) in
order to estimate the range of flow velocities inside these ditches. Bacellar (2000) has proved that even relatively
shallower flows (30 cm deep) in ditches with a rugosity factor of 0033 and a slope gradient of 20 per cent have
enough erosional power (estimated flow velocity = 56 m/s) to cut the less erodible oxisols (A and B horizon) and
reach the saprolites. It is important to note that Bacellar (2000) adopted conservative parameters in these simulations.
Therefore, basement rock soils are resistant to erosion under the usual conditions of overland flow (Santos, 2001), but
not under conditions of concentrated flow in the ditches. It is important to note that these ancient ditches are frequent
in other parts of Brazil, usually associated with gullying (Furlani, 1980; Augustin, 1994).
However, anthropogenic factors cannot explain by themselves the uneven concentration of gullies, since they are
rare in some parts of the catchment, intensely affected by man, while in other relatively less disturbed parts they are
abundant (Bacellar, 2000). Therefore, although human disturbance is very important, other factors also interfere in
gully propagation.
Geologic controls
Geology plays a fundamental part in gully development, since it directly controls soil profile development and
influences its physical properties, such as erodibility. Gullies, characteristically huge in this catchment, are relatively
1378 L. de A. P. Bacellar, A. L. Coelho Netto and W. A. Lacerda
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
rare in Minas and Rio das Velhas supergroup rocks, as a consequence of their thin and loamy regolith. The only
exception is the friable quartzites (Minas supergroup), from the Upper Compartment, which are characterized by a
thick weathering mantle, composed of silts and fine sands without aggregation.
On the other hand, the basement rock regolith is usually well developed, with thick saprolites of silty or fine sandy
texture, which are very prone to gully erosion. Despite the great variability of weathering and pedological processes,
a good relationship was found between basement rock composition and its saprolite erodibility. As shown by aggregate
stability tests, fine texture and feldspatic gneisses and granitoids from the area produce more erodible saprolites
than coarser and more quartzose varieties.
Thus, basement lithology, with typical abrupt lateral variations (e.g. igneous intrusions and dikes), intensely affects
gully propagation. Lithologic contacts, dikes or even thick compositional bands of gneisses are enough to prevent or
to divert the headward retreat of a gully due to natural differences of their saprolite erodibility (Bacellar, 2000).
A good example of this phenomenon can be observed at the two gullies that occur near Holland creek (Holland
monitoring station, Figures 6 and 7). The eastern gully of this station shows an anomalous pattern of retreat, since it
began to evolve following the local slope gradient (northwest direction), but at a certain moment it shifted abruptly to
a west growing direction (Figure 7), discordant with the slope gradient. Overland erosion processes cannot explain this
behaviour. Detailed field surveys have demonstrated that a W/NW trending ductile shear zone affects the gneiss,
which outcrops in the toe of this gully's north sidewall. The resultant rocks (mylonites) produce thinner and much less
erodible saprolites (Figure 7), because of its aggregates, coarser (with MWD > 07, see above) and more quartzose
than the other ones nearby (with MWD < 033). This contrasting low erodibility explains why the eastern gully has
stopped retreating northwestward, but it is still impossible to understand why it has continued to evolve with an
apparent anomalous pattern, discordant with the slope gradient (westward). Electroresistivity surveys (with an adapted
Figure 7. Map depicting the two gullies (A and B) of Holanda monitoring station (see Figures 2 and 6), with representation of an
apparent resistivity pseudo-section at 45 m deep. Note the area of low apparent resistivity around C. Inset: joint measurements
made at this station.The joint main orientation (NW) coincides with a strong electrical alignment detected by the electroresitivity
survey.
Factors controlling gully erosion 1379
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Figure 8. Rose-diagrams of the west part of the Maracujá Catchment representing the directions of: (a) geologic brittle structures
(joints and faults); (b) gully channels; (c) headwater hollows.
3D-dipole­dipole method, after Nery and Aranha, 1995) showed that its evolution has followed a zone of great
groundwater influx (zone of low apparent resistivity), which occurs just above the head of both gullies (area around
point C, Figure 7). Field structural data suggest that northwest-trending joints (inset of Figure 7), which are common
in this area, possibly explain the low apparent resistivity zones. This relationship between groundwater flow and NW
discontinuities, well known in the region by hydrogeologists (Fernandes, unpublished report), may explain why there
is a statistically good relationship between these structures and gully channels and headwater hollows (see below)
throughout the Maracujá Catchment (Figure 8). So, geology plays either an active role (groundwater preferential
influx by discontinuities) or a passive role (erodibility anisotropy) in gully evolution.
However, lithology cannot explain by itself the uneven concentration of these erosional features, because gullies
can be abundant or not on the same lithology, as, for instance, on Funil gneiss (compare Figures 2 and 3).
Geomorphologic controls
According to the threshold concept, there exists for a given slope gradient a critical drainage area necessary to cause
gully erosion (Poesen et al., 2003). It is almost impossible to determine these topographic parameters in Maracujá
Catchment, because its 385 gullies are usually very large (up to 150 metres wide) and most of them have already
reached the drainage divide. Therefore, they are now in a stable situation or are evolving very slowly (Sobreira, 1998).
So, it is necessary to analyse other topographic parameters in order to explain gully concentration in this catchment.
Gullies are clearly concentrated in geomorphologic domain 2 (Figures 3 and 4), which is the flattest area of the
catchment, with slope gradients < 30 per cent. In fact, domain 2 has 37 gullies/km2
, while the other domains in the
basement rocks (domains 3 and 4) have fewer than 16 gullies/km2
. When areas of gently sloping interfluves (watercourse
interfluves with average slope gradient < 15 per cent, graphically obtained from the slope map) are placed over
the geomorphologic domains (Figure 4), it can be verified that about 70 per cent of the gullies tend to group mainly at
the border of such interfluves (Figure 9), generally on the flatter ones in domain 2. Gullies are rare on areas of this
domain with steeper interfluves as well as on the other domains, even on the surrounding areas of these kinds of
interfluve. Therefore, gullies are more common in the flattest domain, specially on the borders of flat and broad
interfluves. Some field permeability data (Bacellar, 2000) and preliminary hydrological balance of headwater catchments
have shown that these interfluves are probably the best groundwater recharge places of Maracujá Catchment,
due to its smoother relief, with well developed, permeable, porous and deep soils (oxisols). So, the preservation of
thick soil profiles (with erodible saprolites up to 50 metres deep) as well as the larger groundwater availability in these
flatter interfluves seem to be an important control factor of gully distribution in this area.
Gully occurrence is uneven even on a detailed scale (Bacellar, 2000), as they are more frequent inside headwater
topographic hollows (643 per cent) than in noses (137 per cent) and side slopes (220 per cent). This preferential
concentration, described by other researchers in southeastern Brazil (Coelho Netto, 1997) and elsewhere (Montgomery
and Dietrich, 1989; Dunne, 1990), can be explained by the natural convergence of surface and underground water
towards hollows (Hack and Goodlet, 1960).
The geometry of some headwater hollows of Maracujá Catchment support the idea that many of them may represent
old gullies (`paleogullies', Figure 10) that were partially filled by sediments deposited in further depositional
events. These paleogullies are certainly pre-European settlement, as they do not cut historical constructions. They
could have been formed as a consequence of some Upper Quaternary climatic shifts (see, e.g., Suguio, 1999; Bacellar,
2000), as described by Coelho Netto (1997).
It is not always easy to identify these paleogullies, since their primary forms, with scarped walls, have been
gradually reshaped to smoother and flatter ones by a combination of exogenetic processes (mainly by creep, slump and
surface wash) during the Upper Quaternary. Although local factors can affect the current form of paleogullies, as a
1380 L. de A. P. Bacellar, A. L. Coelho Netto and W. A. Lacerda
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
general rule it can be assumed that the greater their reshaping the older they are, which makes it possible to establish
their relative ages. As we can see below, this assumption was further confirmed in the field, because the most easily
identified paleogullies are that ones partially filled by the younger sediments of Maracujá catchment (T2­R2 and T1­
R1 sediments, Figure 10). As current gullies are more frequent in headwater hollows, which usually represent
paleogullies, this phenomenon is clearly recurrent, confirming previous studies in Brazil (Oliveira, 1989; Moreira,
1992; Augustin, 1994; Coelho Netto, 1997) and elsewhere (Smith, 1982; Bothaa et al., 1994). As soon as the saprolites
are reached, usually by concentrated flow in ditches, gullies tend to follow previous routes of gully retreat, due to rock
discontinuities or saprolite erodibility anisotropies. As previously commented, discontinuities guide preferentially the
underground flow (fractured aquifers) and the erodibility variations are able to shift or prevent gully retreat towards
certain directions.
An important question remains unanswered: why are some headwater hollows cut by gullies and others are totally
preserved from erosion? As geological processes can be recorded in sediments, a good way to answer this question is
to study Upper Quaternary stratigraphy. As commented before, a morphostratigraphic approach was chosen to study
Quaternary sediments that occur in the Maracujá Catchment from valley bottoms up to headwaters hollows. This
analysis showed that hollow sediments occur with a relatively uniform stratigraphy throughout Maracujá Catchment,
being composed of T2­R2 and minor amounts of T1­R1 Quaternary sediments (Bacellar, 2000). A typical sequence
consists of alluvial layers, covered up by colluvial sediments (Figures 10, 11 and 12). The alluvial layers comprise
white pebbly or clayey sand, at the base, overlain by black organic clay (T2 clay), which is composed, essentially, of
kaolinite, quartz and organic matter. This 2­6 metres thick clay layer is soft (NSPT varies between 1/45 and 4), plastic
(liquid limit = 47 and plastic limit = 30) and shows an average sensitivity = 60 (Lambe and Whitman, 1969). Seven
14C datings of headwater hollow samples revealed ages between 31 340 and 7490 years B.P., a relatively long time
span, which precludes any correlation with other erosional events identified in southeastern Brazil (Dietrich et al.,
1990; Augustin, 1994; Suguio, 1999).
Figure 9. Map of the Maracujá Catchment depicting the geomorphologic domains and the gently sloping interfluves (<15%). Most
gullies occur in the fringes of these interfluves, mainly in domain 2. Only the larger gullies are represented for better visualization.
Factors controlling gully erosion 1381
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Figure 10. Schematic diagram of typical headwater hollow, which represents a paleogully, filled up by upper Quaternary sediments
(T2/R2 and T1). The channel incision cuts the less erodible sediments (organic clays and colluvial soils) and reaches the more
erodible gneiss saprolite. The further natural convergence of underground and surface water towards the hollows leads to the
development of a new gully. Where the saprolite is thinner (right), deep gullies do not develop.
This alluvial package was overlain by red brown sandy clay, up to 6 metres thick (Figures 11 and 12), composed
basically of quartz, kaolinite and Al or Fe oxides and hydroxides. It shows very good aggregation, with resistant
blocky aggregates with pebbly or coarse sand texture, typical of Brazilian oxisols (Parzanese, 1991). These sediments
were interpreted as pedogenized (oxisol) colluvial ramps, based on local geomorphologic and stratigraphic evidences
(Bacellar, 2000).
The organic clay (T2), with low permeability (k = 56 × 10-7
cm/s), seals an underlying confined aquifer, including
the (pebbly) sand and the top of the gneiss saprolite. Piezometer (Casagrande type, Lambe and Whitman, 1969)
readings in a typical unchannelled amphitheatre-like hollow, preserved from current gully erosion (Condomínio
station, Figure 11), revealed high pore pressures in this aquifer, enough to cause an artesian condition in the hollow
bottom in the rainy season.
Figure 11. Geologic cross-section through a preserved headwater hollow valley (Condominio station, see Figure 2). Note the
soft organic clay (T2) layer under colluvium (R2), both dipping towards the valley bottom. T2 clay is overlain by a younger and
softer organic clay (T1) of Holocene age.Water under T2 clay is pressurized (with artesianism), as can be seen by the piezometric
line.
1382 L. de A. P. Bacellar, A. L. Coelho Netto and W. A. Lacerda
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Figure 12. Some stratigraphic profiles of Quaternary sediments (R2­T2 units) throughout the Maracujá Catchment. In the
geomorphologic domain 2, whose river incision is hampered by local base levels, the sediment thickness is thinner than in the
other domains, apparently deprived of effective local base level control. The thicker sedimentary cover of the other domains
probably difficult channel incision and a widespread gully development in these areas.
Pinhole erodibility tests (Sherard et al., 1976) performed by Silva (2000) and Morais (2003) have shown that the
saprolites of Funil gneiss (Figure 2) are very prone to tunnel erosion (Dunne, 1990). Morais (2003) has demonstrated
through physical modelling that these saprolites are very susceptible to slumping, even under small hydraulic gradients.
As these saprolites are under artesian pressure inside the preserved headwater hollows, slumping and tunnel
erosion by natural discontinuities (joints or animal cavities) seem to be very likely triggering mechanisms of saprolite
erosion, as soon as the cover of less erodible soil (A and B oxisoil horizons) is removed by, for example, concentrated
flow in boundary ditches.
Preliminary studies have demonstrated that the morphology and stratigraphy of headwater hollows of Maracujá
Catchment are very similar to those from basement rock areas of southeastern Brazil (Oliveira, 1989; CETEC, 1992;
Moreira, 1992; Augustin, 1994; Coelho Netto, 1997; Bacellar et al., 2004). Two of these hollows (Vila Albertina and
Vila Barraginha) are very well known in the Brazilian geotechnical literature (CETEC, 1992; Bacellar et al., 2004),
because they were sites of huge, catastrophic mass movements (mudflow type). Both sites are characterized by a
colluvium-buried layer of soft and plastic organic clay, located inside unchannelled headwater hollows, sealing an
underlying confined aquifer, in a situation very similar to the Maracujá hollows (Bacellar et al., 2004). Both accidents
were explained by placement of poorly constructed embankments over these meta-stable hollows. Laboratory data
indicate that the organic clay (T2) of Maracujá Catchment exhibits shear strength (UU type test, Lambe and Whitman,
1969) and sensitivity similar to Vila Barraginha's (Bacellar et al., 2004). This suggests that many Maracujá headwater
Factors controlling gully erosion 1383
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
hollows, which are not cut by current gullies, could be in a meta-stable situation as well, very prone to mudflow
movements. Some Maracujá Catchment gullies seem to be associated with clay rich valley bottoms, with hummocky
forms, a common inherited feature of sites affected by mudflow movements. So, the erosion of upper soil horizons by
concentrated surface flows (conditioned mainly by boundary ditches), and, to a minor degree, by mudflow movements
seem to be the likely triggering mechanisms of gully erosion in this catchment. Both can expose the more erodible
saprolite, leading to further gully development by a combination of secondary mechanisms, mainly falls and slumps
(due to headcut undermining and the high hydraulic gradient), tunnel erosion and wash and rill erosion on the
collapsed or gully wall soils (Bacellar, 2000). These mechanisms (mainly slumps and tunnel erosion) are probably
more effective in the fringes of gently sloping interfluves, where the regolith thickness and groundwater availability
are probably bigger.
The stratigraphy of these headwater hollow sediments varies according to the geomorphologic domains. In domain
2, controlled by bedrock steps, which act as local base levels (Figure 4), the geologically recent fluvial incision was
probably less intense, provoking stronger lateral channel erosion, hampering relief incision and favouring alluvial over
colluvial sedimentation from lateral slopes (colluvial ramps). This explains why in headwater hollows of this domain
the alluvial sediments (T2) are found in shallower deposits, covered by thin colluvial ramps, that thicken up towards
lateral slopes, as shown by several geological profiles (Figures 11 and 12). As a consequence, it is easier to cut the
protective cover of less erodible organic clay and colluvial sediments in this domain and reach the pebbly sands and
saprolites underneath, which show lower erodibility and are under artesian pressure. The opposite occurs in the
headwater hollows of the other geomorphologic domains of basement rock area (domains 3 and 4), which are deprived
of effective local base levels. This situation results in narrower and deeper valley bottoms, with alluvial sediments
(less erodible than the ones from domain 2, due to their coarser grain size) covered by thick layers of colluvial
sediments (Figure 12), which are more available in these steeper relief domains, producing a configuration less prone
to river incision and gully initiation. However, as commented previously, the headwater hollows of these domains
could not be considered stable at all, because they can have the same unstable inner configuration (soft organic clay
layers confining saprolites), which can be destabilized by any process that can cause local channel incision, as showed
by Brierley and Fryirs (1999).
Conclusions
Very large gullies are natural phenomena in the Maracujá Catchment, as indicated by the great frequency of
old erosional features (paleogullies). Current gullying began with the European settlement at the end of the 17th
century, as a consequence of inappropriate land use practices, especially digging of boundary ditches (`valos').
Although triggered by man, gullies grow intensely only in favourable places, dictated by lithology and, mainly,
geomorphology. In fact, large gullies are rare in Minas and Rio das Velhas supergroup areas, as a consequence of the
low erodibility and small thickness of their loamy saprolites. On the other hand, in basement rocks the situation is
more favourable for their development, because of the greater erodibility of its saprolites, specially those consisting of
silty or fine sandy aggregates (Bryan, 1976, 2000; Parzanese, 1991; Silva, 2000). However, lithology does not explain
the uneven concentration of gullying, because its incidence is variable in geologically similar areas. Actually, gullies
are more common on low relief areas (geomorphologic domain 2, with slope gradients < 30 per cent) and, especially,
at the fringes of broad and flat interfluves. The concentration is also uneven at a detailed scale, with gullies growing
mainly in amphitheatre-like headwater topographical hollows. As can be easily seen by aerial photography analysis,
the geometry of many headwater hollows supports the idea that they represent old erosional features (paleogullies),
very similar to the ones identified by Oliveira (1989). Field data have shown that these paleogullies were partially
filled by Upper Quaternary sediments, including layers of organic clay and of colluvial sediments. This clay is soft,
plastic and with low permeability, working as a seal of a confined aquifer on underlying saprolite and alluvial
sediments.
So, the erosion of soils by human activities, such as boundary ditches (`valos'), or the soil mobilization by
mass movements involving the unstable organic clay layer (mudflows) in the hollows, are likely triggering mechanisms
for gullying. When the underlying saprolite is exposed, secondary mechanisms began to operate in gully
retreat, such as wash and tunnel of collapsed or gully wall soils and falls and slumps of gully walls. However,
many more studies are necessary to establish with precision the natural sequence of these primary and secondary
mechanisms.
The uneven distribution of gullying in this and in other areas of basement rocks should be considered in future land
planning or engineering works. More vulnerable areas, such as headwater hollows, should always be well investigated
in order to prevent the possibility of accidents, as mudflows and gullies.
1384 L. de A. P. Bacellar, A. L. Coelho Netto and W. A. Lacerda
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
References
Alcântara MAT. 1997. Aspectos geotécnicos da erodibilidade dos solos. MSc Thesis, EESC/So Paulo University, Brazil.
Augustin CHR. 1994. Amphitheaters and hollows with depositional sequences and their significance on the evolution of tropical landscape.
Proceedings of XIV International Sedimentation Congress, Recife, G5­G6.
Bacellar LAP. 2000. Condicionantes geológicos, geomorfológicos e geotécnicos dos mecanismos de voçorocamento na bacia do rio Maracujá,
Ouro Preto, M.G., D.Sc. Thesis COPPE, Rio de Janeiro Federal University, Brazil.
Bacellar LAP, Lacerda WA, Coelho Netto AL. 2004. Amphitheaterlike headwaters: areas of mudflow hazard in southeastern Brazil Proceedings
of IX International Symposium on Landslides, Rio de Janeiro. Balkema: Leiden.
Barths B, Roose E. 2002. Aggregate stability as an indicator of soil susceptibility to runoff and erosion; validation at several levels. Catena
47: 133­149.
Bothaa GA, Wintleb AG, Vogel JC. 1994. Episodic late Quaternary palaeogully erosion in northern KwaZulu-Ntal, South Africa. Catena 23:
327­340.
Brierley GJ, Fryirs K. 1999. Tributary­trunk stream relations in a cut-and-fill landscape: a case study from Wolumla catchment, New South
Wales, Australia. Geomorphology 28: 61­73.
Bryan RB. 1976. Considerations on soil erodibility indices and sheet wash. Catena 3: 99­111.
Bryan RB. 2000. Soil erodibility and processes of water erosion on hillslope. Geomorphology 32: 385­415.
Burton R. 1869. Explorations of the highlands of the Brazil. Tinsley: London.
CETEC. 1992. Laudo geotécnico sobre o escorregamento de Vila Barraginha, Contagem (MG), Vol. 1. CETEC: Belo Horizonte, Brazil;
1­57.
Coelho Netto AL. 1997. Catastrophic landscape evolution in a humid region (SE Brazil), inheritances from tectonic, climatic, and land use
inducer changes. In Proceedings of IV International Conference Geomorphology Supplement Geogr. Fis. Dinam. Quant. Vol. 3; 21­48.
Dietrich WE, Dunne T. 1993. The channel head. In Channel Network Hydrology, Beven K, Kirkby MJ. (eds). Wiley: London; 175­217.
Dietrich WE, Montgomery D, Coelho Netto AL, Moura JRS. 1990. Evidence for regional aggradation starting in the Early Holocen
in southeastern Brazil and for degradation due to deforestation. Proceedings of American Geophysical Union, Fall Meeting. San
Francisco, Vol. 70; 43.
Door JN. 1969. Physiographic, stratigraphic and structural development of the Quadrilátero Ferrífero, Minas Gerais. US Geological Survey
Professional Paper 641-A: 110.
Dunne T. 1990. Hydrology, mechanics and geomorphic implications of erosion by subsurface flow. Geological Society of America Special
Paper 252: 1­28.
Endo I. 1997. Regimes tectônicos do Arqueano e Proterozóico do Interior da Placa Sulfranciscana: Quadrilátero Ferrífero e áreas adjacentes,
Minas Gerais. Ph.D. Thesis, So Paulo University, So Paulo.
Frye JC, Willman HB. 1962. Morphostratigraphic units in Pleistocene stratigraphy'. American Association of Petroleum Geologists Bulletin
46: 112­113.
Furlani GM. 1980. As boçorocas de Casa Branca e seu significado geomorfológico. Geomorfologia 10: 12­15.
Gutersohn H. 1945. A Regio Central de Minas Gerais (transl. WA Egler). Boletim Geográfico 1: 1­49.
Hack JT, Goodlet JG. 1960. Geomorphology and forest ecology of a mountain region in the Central Appalachians. U.S. Geological Survey
Professional Paper 347: 66.
Head KH. 1986. Manual of soil laboratory testing Vol. 3. Pentech: London.
Hippert JF. 1994. Structures indicative of helicoidal flow in a migmatitic diapir (Baço Complex, southeastern Brazil). Tectonophysics 234:
169­196.
Inderbitzen AL. 1961. An erosion test for soils. Materials Research and Standards July: 553­554.
Johnson RF. 1962. Geology and ore deposits of the Cachoeira do Campo, Dom Bosco e Ouro Branco Quadrangles, Minas Gerais, Brazil.
Geological Survey Professional Paper 341-B: 1­39.
Kemper WD, Rosenau RC. 1986. Aggregate stability and size distribution. In Methods of Soil Analysis Parts 1, 2. American Society of
Agronomy and Soil Science Society of America Madison, 425­442.
Lal R. 1990. Soil erosion in the tropics. Principles and management. McGraw-Hill: New York.
Lambe TW, Whitman RV. 1969. Soil mechanics. Wiley: New York.
Martin L, Flexor JM, Suguio K. 1998. Pleistocene wave-built terraces of the northern Rio de Janeiro State, Brazil. Quaternary of South
America and Antarctic Peninsula 11: 233­245.
Meis MRM, Miranda LHG, Fernandes NF. 1982. Desnivelamento de altitude como parâmetro para a compartimentaço do relevo. Bacia do
Médio-Baixo Paraíba do Sul. XXXII Congresso Brasileiro de Geologica, Anais, Salvador; 1489­1503.
Meis MRM, Moura J. 1984. Upper quaternary sedimentation and hillslope evolution; Southeastern Brazilian Plateau. American Journal of
Science 284(3): 241­254.
Mello C. 1997. Sedimentaço e tectônica cenozóicas no médio vale do rio Doce (MG, sudeste do Brasil) e sua implicaço na evoluço de
um sistema de lagos. Ph.D. Thesis, So Paulo University, Brazil.
Montgomery DR, Dietrich WE. 1989. Source areas, drainage density, and channel initiation. Water Resources Research 25(8): 1907­1918.
Morais F. 2003. Estudo dos processos erosivos subsuperficiais na bacia do rio Maracujá-MG. MSc Thesis, Ouro Preto Federal University,
Brazil.
Moreira VRM. 1992. Fatores condicionantes das voçorocas na sub-bacia do rio Santo Antônio, Bacia do Rio Grande, MG. M.Sc. Thesis,
IGC/Minas Gerais Federal University, Brazil.
Factors controlling gully erosion 1385
Copyright  2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 1369­1385 (2005)
Nery ACF, Aranha PRA. 1995. Uma nova tecnologia em prospecço geoelétrica: Metodologia Cacau. In Anais Congresso Brasileiro de
Geofísica, So Paulo; 996­998.
North American Commission on Stratigraphic Nomenclature (NACSN). 1983. North American Stratigraphy Code. American Association of
Petroleum Geologists Bulletin 67(5): 841­875.
Oliveira MAT. 1989. Erosion disconformities and gully morphology: a three dimensional approach. Catena 16: 413­423.
Oliveira MAT. 1997. Towards the integration of subsurface flow and overland flow in gully head extension: issues from a conceptual model
for gully erosion evolution. South African Journal of Geography 2(79): 120­128.
Parzanese GAC 1991. Gnese e desenvolvimento das voçorocas em solos originados de rochas granitóides da regio de Cachoeira do
Campo, Minas Gerais. M.Sc. Thesis, Viçosa Federal University, Brazil.
Poesen J, Nachtergaele J, Verstraeten G,Valentin C. 2003. Gully erosion and environmental change: importance and research needs. Catena
50: 91­133.
Rego JJV. 1978. Eroso superficial em taludes de corte em solo residual de gnaisse. M.Sc Thesis, COPPE/Rio de Janeiro Federal University,
Rio de Janeiro.
Saadi A. 1991. Ensaio sobre a morfotectônica de Minas Gerais: tenses intraplaca, descontinuidades crustais e morfognese. Belo Horizonte,
MG. Post-Doctoral Thesis, Minas Gerais Federal University, Brazil.
Santos CA. 2001. Comportamento Hidrológico Superficial, Subsuperficial e a Erodibilidade dos Solos da Regio de Santo Antônio do Leite,
Distrito de Ouro Preto ­ MG. MSc Thesis, Tese de Mestrado, Ouro Preto Federal University, Brazil.
Selby MJ. 1990. Hillslope Materials and Processes. Oxford University Press: Oxford.
Sherard JL, Dunnigan LP, Decker RS. 1976. Identification and nature of dispersive soils. Journ. of the Geotech. Engineer., Proc. ASCE 102:
287­301.
Silva TRM. 2000. Caracterizaço e erodibilidade dos solos de uma voçoroca na regio de Ouro Preto­MG. M.Sc. Thesis, COPPE/Rio de
Janeiro Federal University.
Smith BJ. 1982. Effects of climate and land-use change on gully development: An example from Northern Nigeria. Z. Geomorph. N.F.,
Suppl. 44: 33­51.
Sobreira F. 1998. Estudos das eroses da Cachoeira do Campo, MG., report, FAPEMIG, UFOP/EM/DEGEO, Ouro Preto, MG; 1­130.
Suguio K. 1999. Geologia do Quaternário e Mudanças Ambientais. Paulo's Editora: So Paulo.
Wells NA, Andriamihaja B. 1993. The initiation and growth of gullies in Madagascar: are humans to blame? Geomorphology 8: 1­46.
Xu J. 1996. Benggang erosion: the influencing factors, Catena 27: 249­263.
Yoder RE. 1936. A direct method of aggregate analysis of soils and a study of the physical nature of erosion losses. Journal of the American
Society of Agronomy 28(5): 337­351.