1. Environ Earth Sci
DOI 10.1007/s12665-010-0480-z
ORIGINAL ARTICLE
Geo-environmental mapping using physiographic analysis:
constraints on the evaluation of land instability and groundwater
pollution hazards in the Metropolitan District of Campinas, Brazil
Paulo Cesar Fernandes-da-Silva • Ricardo Vedovello •
Claudio Jose Ferreira • John Canning Cripps •
Maria Jose Brollo • Amelia Joao Fernandes
Received: 13 April 2009 / Accepted: 20 January 2010
Ó Springer-Verlag 2010
Abstract Geo-environmental terrain assessments and land instability and the vulnerability of groundwater to
territorial zoning are useful tools for the formulation and pollution hazards. The implementation incorporated proce-
implementation of environmental management instruments dures for inferring the influences and potential implications
(including policy-making, planning, and enforcement of of tectonic fractures and other discontinuities on ground
statutory regulations). They usually involve a set of proce- behaviour and local groundwater flow. Terrain attributes
dures and techniques for delimitation, characterisation and such as degree of fracturing, bedrock lithology and
classification of terrain units. However, terrain assessments weathered materials were explored as indicators of ground
and zoning exercises are often costly and time-consuming, properties. The paper also discusses constraints on- and
particularly when encompassing large areas, which in many limitations of- the approaches taken.
cases prevent local agencies in developing countries from
properly benefiting from such assessments. In the present Keywords Terrain units Á Satellite imagery Á
paper, a low-cost technique based on the analysis of texture Physiographic compartmentalisation Á Tectonic fracturing Á
of satellite imagery was used for delimitation of terrain Inferential tools
units. The delimited units were further analysed in two test
areas situated in Southeast Brazil to provide estimates of
Introduction
Data about the physical environment (such as rock and soil
P. C. Fernandes-da-Silva (&) Á R. Vedovello Á types, relief, vegetation and natural processes) are essential
C. J. Ferreira Á M. J. Brollo Á A. J. Fernandes to formulate and to implement successful strategies for
˜
Geological Institute, Sao Paulo State Secretariat of Environment,
environmental management. Such data underpin all policy-
˜
Av. Miguel Stefano nr. 3900, Sao Paulo CEP 04301-903, Brazil
e-mail: paulo.fernandes@igeologico.sp.gov.br; making and planning instruments and enforcement regula-
pfernandes_us@yahoo.co.uk tions which usually require geo-environmental terrain
R. Vedovello assessment and territorial zoning in terms of advantages and
e-mail: vedovello@igeologico.sp.gov.br constraints for development of different types (Culshaw
C. J. Ferreira et al. 1990; Zuquette et al. 2004). For regional planning and
e-mail: cferreira@igeologico.sp.gov.br watershed management purposes, such assessments provide
M. J. Brollo advice about the types of land use that would be acceptable
e-mail: mjbrollo@igeologico.sp.gov.br in certain areas but should be precluded in others. Fur-
A. J. Fernandes thermore, ranking of terrain units in terms of the likelihood
e-mail: amelia@igeologico.sp.gov.br and consequences of land instability also enable the iden-
tification, control and mitigation of hazards as well as
J. C. Cripps
provide decision support to contingency actions and/or to
Department of Civil and Structural Engineering, University
of Sheffield, Mappin Street, Sheffield S1 3JD, UK engineering solutions (Cripps et al. 2002; Abella and Van
e-mail: j.c.cripps@sheffield.ac.uk Westen 2008).
123
2. Environ Earth Sci
According to Cendrero et al. (1979) and Bennett and This paper describes an application of the synthetic
Doyle (1997) there are two main approaches to geo-envi- (integrated) approach to a geo-environmental terrain
ronmental terrain assessments and territorial zoning: (a) the assessment and territorial zoning exercise at a semi-regio-
analytical or parametric approach; and (b) the synthetic nal scale. This is exemplified by a case study that explores
approach, also termed integrated, landscape or physio- a low-cost technique comprising physiographic compart-
graphic approach. The parametric approach deals with mentalisation based on the use of satellite imagery for the
environmental features or components individually so that delimitation of terrain units. The resulting map is then
terrain units usually result from the intersection or carto- interpreted in terms of the potential for land instability and
graphic summation of several layers of information. Unit groundwater vulnerability in two test areas situated in the
limits do not necessarily correspond with ground features. ˜
Metropolitan District of Campinas (Sao Paulo State,
In the synthetic approach, the form and spatial distribution Southeast Brazil, see Fig. 2). Key to the success of this
of ground features are analysed in an integrated manner so approach was the incorporation of procedures for inferring
that the land units or divisions correspond with landscape the presence and characteristics of geological structures,
patterns that express interactions between environmental such as fractures and other discontinuities and the assess-
components. ment of these in terms of the potential implications to
Since the advent of airborne and orbital sensors, the ground stability and the flow of groundwater. A general
integrated analysis is based in the first instance, on the description of the physiographic compartmentalisation
interpretation of images and air-photos. In this case, the technique and a discussion of the performance and limi-
content and spatial boundaries of terrain units would tations of the approach are also provided.
directly correspond with ground features. According to
some authors, such correlation and also the recurrence of
particular landscape patterns gives rise to following The physiographic compartmentalisation technique
advantages: (1) facilitation of understanding by non-spe-
cialists and planners (Davidson 1992; Fernandes da Silva Geo-environmental terrain assessments and territorial
et al. 1997); and (2) providing of means of correlating zoning generally involve three main stages, as follows: (1)
known and unknown areas, thus permitting ground condi- delimitation of terrain units; (2) characterisation of units
tions to be reasonably predicted (Finlayson 1984; Moore (e.g. in bio-geographical, engineering geological or geo-
et al. 1993). Terrain units delineated using the physio- technical terms); and (3) evaluation and classification of
graphic approach should hold a genetically linked assem- units.
blage of components such as relief, rocks and soils, The first stage consists of dividing the territory into zones
independent of their sizes. Their definition depends on with respect to a set of pre-determined physical and envi-
climatic, tectonic and lithological criteria, as well as those ronmental characteristics and properties. Regions, zones or
of form (Mitchell 1991). units are regarded as distinguishable entities depending upon
Data collection, derivation from secondary data sources, their internal homogeneity or the internal interrelationships
and integration of data into useful databases are time- of their parts. Some authors argue that such homogeneity is
consuming, costly and difficult tasks to be performed in subjective and small-scale homogeneous units may not exist.
support of a particular project and/or agency function For instance, this has led to the use of fuzzy logic approach
(Nedovic-Budic 2000). In addition, the complexity of GIS (e.g. Zhu et al. 2001; Zhu and Mackay 2001; Shi et al. 2004).
methodology, lack of suitably trained staff and the scarce Although detailed and spatially continuous terrain informa-
organizational resources have been blamed for the under- tion may be attainable through these methods, the required
utilisation of GIS methods (Harris and Weiner 1998; digital data derivation and computing operations tend to be
Vernez-Moudon and Hubner 2000). These difficulties and complex, thus necessitating specialist hard- and software
limitations inhibit both local and regional authorities in that are not always readily available.
developing countries (like Brazil) from properly benefiting The characterisation of terrain units consists of
from geo-environmental terrain assessment outputs in ascribing and surveying relevant properties and character-
planning and environmental management instruments. istics of terrain components that are expected to affect the
From another viewpoint, some authors such as Sahay and ground conditions relevant to the particular application.
ˆ
Walsham (1996); Barton et al. (2002); Camara and Fonseca Such characterisation can be achieved either directly or
(2007) propose that developing countries should ensure indirectly, for instance, by means of (a) ground observa-
that options for using low-cost technology are properly tions and measurements, including in situ tests (e.g. boring,
considered as a way to gain knowledge about the tech- sampling, infiltration tests etc.); (b) laboratory tests (e.g.
nology itself and also in the creation of products that fit grain size, strength, porosity, permeability etc.); (c) infer-
their specific needs. ences derived from existing correlations between relevant
123
3. Environ Earth Sci
parameters and other data such as those obtained from characteristics in satellite images (or air-photos) correspond
previous mapping, remote sensing, geophysical and geo- with specific associations of geo-environmental components
chemical records. (such as bedrock, topography and landforms, soils and
The final stage consists of evaluating and classifying the covering materials) with a common tectonic history and land
terrain units in a manner relevant to the purposes of the surface evolution. Such associations are thought to imply
particular application (e.g. regional planning, transporta- specific ground responses to engineering and other land-use
tion, hazard mapping). This is based on the analysis and actions.
interpretation of properties and characteristics of terrain— The interpretation procedure is a top-down process that
identified as relevant—and their potential effects in terms starts with the whole landscape which is then subdivided
of ground behaviour, particularly in response to human into land parcels. It is assumed that there is a correlation
activities. between image texture and terrain characteristics that are
In order to reduce the fieldwork effort required for the expressed at different scales and levels of compartmental-
delimitation of terrain units, consideration was given to an isation, generally associated with regions or areal domains
increased reliance on remote sensing tools, particularly of decreasing size. The main outcome of this is a single
satellite imagery. The advantages include (a) the genera- cartographic product consisting of comprehensive units
tion of new data in areas where existing data are sparse, delimited by fixed spatial boundaries (that correspond with
discontinuous or non-existent, and (b) the economical ground features). These are referred to as physiographic
coverage of large areas, availability of a variety of spatial compartments or basic compartmentalisation units (BCUs),
resolutions, relatively frequent and periodic updating of which according to Vedovello and Mattos (1998), are the
images (Schmidt and Glaesser 1998; Lillesand and Kiefer smallest units for analysis of geo-environmental compo-
2000; Latifovic et al. 2005; Akiwumi and Butler 2008). nents at the chosen cartographic scale. In other words, there
The physiographic compartmentalisation technique is a relationship between the BCUs and the scales of
(Vedovello 1993, 2000) utilises the spatial information observation and representation, which is governed by the
contained in images and the principles of convergence of spatial resolution of the satellite image or air-photos being
evidence (see Sabins 1987) in a systematic deductive used for the analysis and interpretation.
process of image interpretation. The technique evolved The tracing of limits of textural zones concentrates on
from engineering applications of the synthetic land classi- the analysis of the spatial arrangement of natural align-
fication approach (e.g. Grant 1968, 1974, 1975; TRRL ments of image textural elements, particularly groups of
1978), by incorporating and advancing the logic and pro- contiguous pixels related to the drainage network and relief
cedures of geological-geomorphological photo-interpreta- architecture. Tonal properties are used to help with the
tion (see Guy 1966; Howard 1967; Soares and Fiori 1976), identification and interpretation of linear features. Image
which were then converted to monoscopic imagery as interpretation may also be supported by external sources
proposed by Beaumont and Beaven (1977); Verstappen such as topographic, geological and soil maps.
(1977); Soares et al. (1981) and others. The procedures for the delimitation of units include
Magnitude and variations of light and shade play a key assessment of spatial characteristics of textural zones to
role in the image interpretation, with texture and respective check for internal homogeneity and the degree of similarity
patterns being determined by an interaction between the between zones, particularly their form (spatial distribution)
shapes of surface features and the angle of incidence of and directionality of texture elements (degree of isotropy).
light. In this sense, texture expresses the frequency of tonal Usually, ground checks are carried out to confirm or adjust
(grey-level value) change within an image and arises due to the photo-interpreted boundaries of physiographic units
the distribution and aggregation of minor components (BCUs). Figure 1 shows examples taken from the Campi-
(texture elements) that preserve their own characteristics nas study area presented in this paper, in which two BCUs
(e.g. shape, size, tone) at a determined spatial resolution. are compared in terms of spatial organisation of textural
These unitary elements may be too small to be discerned elements associated with drainage and relief features.
individually on the image, but define a consistent spatial After delimitation, the BCUs are then utilised as a
arrangement that can be described in terms of visual texture module for storage, processing and analysis of geo-envi-
features (Tamura et al. 1978). Image interpretation aims at ronmental data for further land assessments. The organisa-
identifying and delineating textural zones on images tion of data and information in relation to the BCU polygons
according to the properties described in Table 1, wherein in a geo-referenced databank allows optimised procedures
features such as coarseness, roughness, regularity, and of query and production of derived maps. The analysis and
direction are taken into account. evaluation is undertaken up to the (fixed) spatial boundaries
The key assumption proposed by Vedovello (1993, 2000) of the BCUs so that different parameters or attributes can be
is that zones with relatively homogeneous textural used in the subsequent stages of analysis (characterisation
123
4. Environ Earth Sci
Table 1 Description of elements and properties used for recognition and delineation of distinctive textural zones on satellite imagery (Vedovello
1993, 2000)
Textural entities and properties Description
Image texture element The smallest continuous and uniform surface liable to be distinguishable in terms of shape and dimensions and
likely to be repetitive throughout an image. Usual types of image texture elements taken for analysis
include: segments of drainage or relief (e.g. crestlines, slope breaks) and grey tones
Texture density The quantity of textural elements occurring within an area on image. Texture density is defined as the inverse
of the mean distance between texture elements. Although it reflects a quantitative property, textural density
is frequently described in qualitative and relative terms such as high, moderate low etc. Size of texture
elements combined with texture density determines features such as coarseness and roughness
Textural arrangement The form (ordered or not) by which textural elements occur and are spatially distributed on image. Texture
elements of similar characteristics may be contiguous thus defining alignments or linear features on image.
The spatial distribution may be repetitive and it is usually expressed by ‘patterns’ that tend to be recurrent
(regularity). For example, forms defined by texture elements due to drainage expressed in rectangular,
dendritic or radial patterns
Structuring (degree of spatial The greater or lesser organisation underlying the spatial distribution of textural elements and defined by
organisation) repetition of texture elements within a certain rule of placement. Such organisation is usually expressed in
terms of regular or systematic spatial relations, such as length, angularity, asymmetry and especially
prevailing orientations (tropy or directionality)
Tropy reflects the anisotropic (existence of one, two or three preferred directions), or the isotropic (multi-
directional or no predominant direction) character of textural features. Asymmetry refers to length and
angularity of linear features (rows of contiguous texture elements) in relation to an axe or main feature
identified on image. The degree of organisation can also be expressed by qualitative terms such as high,
moderate, low or yet as well- or poorly defined
Structuring order Complexity in the organisation of textural elements, mainly reflecting superposition of image structuring. For
example, a regional directional trend of textural elements that can be extremely pervasive, distinctive and
superimposed to other orientations also observed on imagery. Another example is given by drainage
networks displaying different orders with reference to main stream lines and tributaries (first, second, third
orders)
Fig. 1 Examples of basic
compartmentalisation units
(BCUs) taken from Test Area
T1 with similar codification
CRR: C crystalline basement,
R granitic Gneiss, R large
rolling hills with aligned
crestlines and rectilinear slope
profile. Landsat TM5,
composite image, Bands 3–4–5,
greyscale. a Drainage lines;
b relief lines; c frequency
histograms for azimuth
directions of texture elements
associated with drainage and
relief features. Fourth level of
compartmentalisation of BCUs
expresses the predominant
drainage directions:
CRR2 = ENE ? NW ? NE;
CRR3 = NE ? NNE ? NW
and relief line directions:
CRR2 = NW ? NE;
CRR3 = NNW ? NE in c
frequency histograms
123
5. Environ Earth Sci
and classification of units) while keeping their cartographic visual interpretation and vector format manual digitising
significance and cohesion as unitary entity, i.e. no changes with Erdas Imaging software.
to the boundaries of existing polygons or generation of new The delimitation of units was based on image texture
ones are required (Tominaga et al. 2004). characteristics expressed by groups of contiguous pixels
related to drainage and relief features. For this a minimum
line segment length of 30 m (one pixel) was used. It should
Geo-environmental terrain evaluation: a case study be noted that the dimensions of BCUs directly relate to the
spatial resolution of the image and also to the visibility of
The present study was carried out in two test areas, T1 and ground features, such as drainage and relief lineaments. In
T2 (Fig. 2), located in the Metropolitan District of Cam- the present investigation, the smallest BCU in Area T1 was
pinas (RMC), the State of Sao Paulo, Brazil, which 0.7 km2 (approx. 820 pixels) and the average area was
encompasses 19 municipalities and covers approximately 3.6 km2, whereas in Area T2 their areas were, respectively,
1,800 km2. Area T1 (80 km2) comprises a rugged topo- 1.24 km2 (approx. 1,380 pixels) and 6.17 km2. Visual
graphy with small and large hills and ridges of significant image interpretation was supported by external ancillary
slope steepness, consisted mainly of Pre-Cambrian crys- data concerning bedrock lithology, structural geology,
talline rocks (gneiss and granite). Area T2 (192 km2) con- topography and geomorphology.
sists of Palaeozoic to Tertiary sedimentary and intrusive Depiction of natural linear features is dependent upon
volcanic rocks that form a flatter topography comprising grey-level values that are influenced by the gradient of land
undulating and rolling hills together with Quaternary age surface and its position in relation to sunlight exposure.
alluvial plain deposits. Drainage lines were frequently associated with dark pixel
patches as follows: (a) enriched tonal contrast due to
Terrain Compartmentalisation absorption of energy by surface water and strips of river-
side vegetation in Band 3; (b) dark tonal contrast due to
A Landsat 5 Thematic Mapper (TM) image (path 220, row high moisture content emphasised in Near-IR (Band 4) and
076, captured on 12 September 1997, end of dry season) Mid-IR (Band 5); (c) patches of shading or relatively dark
was selected for this study. Factors influencing this choice tonal contrast as an expression of negative slope breaks
included temporal, spectral, spatial, and synoptic charac- (decreasing slope steepness) in valleys and watercourses.
teristics as well as good availability and lower cost than Relief lines were usually demarcated by subtle limits
other products. The date of image acquisition slightly between contrasting zones of lighting and shading on the
preceded recent major urban and industrial development in image that were defined by relatively bright ground sloping
the region. From the full scene two sub-sets of 250 9 313 towards the direction of sunlight. In areas of low vegetation
and 375 9 500 pixels, corresponding to test areas T1 and density and soil exposure, lower moisture content tended to
T2, respectively, were selected. The BCUs were delimited enhance these contrasts. In many cases, these features
on a geo-referenced composite sub-image—Band 3 (visible corresponded with ridge tops, crestlines and positive slope
wavelength) ? Band 4 (Near-IR) ? Band 5 (Mid-IR)— breaks (increasing slope steepness), whose identification
false RGB colours at 30 m pixel resolution using on-screen was also facilitated by association with drainage heads.
Fig. 2 Location map showing
the study region (Metropolitan
District of Campinas) in the
˜
State of Sao Paulo, Southeast
Brazil, and the Test Areas T1
and T2. Scale bar applies to the
map of the Metropolitan District
of Campinas
0 18 36 km
123
6. Environ Earth Sci
The main characteristics considered for the delimitation Characterisation of units
of BCUs included (a) density of texture elements related to
drainage and relief lines, (b) spatial arrangement of Based on a minimum areal extent of 3 km2, accessibility
drainage and relief lines in terms of form and degree of contiguity of units and the planned structural geological
organisation (direction, regularity and pattern), (c) length analysis, 13 BCUs in each test area were selected for fur-
of lines and their angular relationships, (d) linearity of ther geo-environmental assessments in which both spatial
mainstream channel and asymmetry of tributaries, (e) image characteristics and external data sources were con-
density of interfluves, (f) hillside length, and (g) slope sidered. The areas were verified and complemented with
forms. These characteristics were identified mainly on the ground checks.
basis of image interpretation, but external ancillary data Inferences relating to environmental properties and
were also used to assist image interpretation for the iden- characteristics of geotechnical interest based on correla-
tification of relief-related characteristics, such as slope tions of image properties from remotely sensed data were
forms and interfluve dimensions. particularly investigated. The principle postulated was that
Figure 3 shows the basic compartmentalisation units image texture related to the properties/characteristics of the
(BCUs) delineated for Test Areas T1 and T2, and the imaged target enables reasonable deductions about geo-
respective drainage networks. As illustrated in Fig. 1, units technical-engineering attributes (Beaumont and Beaven
are identified by three-letter codes and one numerical 1977; Beaumont 1985).
character, corresponding, respectively, to (a) physiographic In view of the aims of the study to estimate suscepti-
domain, (b) predominant bedrock lithology and geological bility of land to instability and the vulnerability ground-
structure, (c) geomorphological setting including predom- water to pollution, as well as other factors such as scales of
inant landforms, and (d) specific characteristics such as soil observation and representation, data availability and deri-
profile and erosional and aggradational features. Examples vation, the following attributes were primarily considered
of the codification of UBCs are provided in Table 2. A to be relevant and selected for the characterisation of
summary of relationships between image texture charac- BCUs: (a) bedrock lithology, (b) tectonic discontinuities
teristics, bedrock lithology and relief/landform system is (generically referred to as fracturing), (c) soil profile
presented in Table 3. (including thickness, texture and mineralogy), (d) slope
Fig. 3 Drainage networks
(a, b) and basic compart-
mentalisation units (BCUs) in
Test Areas T1 and T2 delineated
on a Landsat TM5 composite
sub-image—bands 3, 4, 5,
greyscale (c, d). Note greater
density, spatial organisation and
angularity expressed by
drainage network of Area
T1 (crystalline rocks) in
comparison with Area
T2 (predominantly sedimentary
rocks). UTM projection and
coordinates
123
7. Environ Earth Sci
Table 2 Examples of codification of basic physiographic units (UBCs)
UBC Code description
BAA1 Physiographic domain: B (sedimentary basin)
Bedrock lithology: A (sandstone: medium to coarse grained, predominantly massif, quatzose)
System of relief/landforms: A (wide undulating hills, convex to flat top, gentle to moderate slope)
Specific characteristics: 1 (sandy-clayey soil grading to sandy-silty in depth, thickness 1 to 5 m,
predominant uni-directional arrangement of drainage and relief lines)
CLT1 Physiographic domain: C (crystalline basement)
Bedrock lithology: L (laminated gneiss)
System of relief/landforms: T (small rolling hills, sharp and narrow top, aligned crestlines, moderate slope)
Specific characteristics: 1 (sandy to sandy-silty soils, thickness [4 m, concave hillside)
Table 3 Relationships between second (bedrock lithology) and third (relief/landform systems) levels of physiographic compartmentalisation
and image texture characteristics, particularly density and spatial organisation of texture elements associated with drainage and relief features
Image texture characteristics
Bedrock lithology (assigned code)
Granites (S) and granitic-gneisses (R) Drainage and relief alignments that reflect structural geological lineaments at NW and N–S
orientation. High-density (texture density related to) drainage forms ([3 km/km2) with
directional anisotropy expressed by rectangular and oblique patterns
Gneisses: banded (B, O), laminated (L), Moderate to high-density (2–3 km/km2) drainage forms: sub-dendritic, parallel, sub-parallel to
schistose (N) angulated, tendency to bi- or tri-directional anisotropy (one direction mostly associated with
metamorphic foliation)
Sandstone: medium to coarse grained (A) Dendritic drainage forms, locally radial or angulated, low to moderate density (2 km/km2),
and variable tropy (uni-bi, and tri-directional to isotropic)
Mudstones (B), siltstones (G), Moderate to high (2–3 km/km2) density of drainage lineaments with sub-dendritic to angulated
rythmites (B, G) and fine-grained forms, bi- or tri-directional anisotropic arrangements that grade into isotropic (sandy
sandstones (C, F) constituency)
Dolerites (intrusive volcanic rocks) (D) Lineaments associated with positive relief slope breaks of greater amplitude. Drainage forms
tend to be isotropic and low to moderate (2 km/km2) density of drainage lines
Aluvional deposits (no code) Smoother texture bounded by negative slope breaks in association with dense vegetation strips
Relief/landform system (assigned code)
Wide undulating hills (A) Convex hillsides and flat tops characterised by relative scarcity of textural elements related to
drainage. Subtle positive slope breaks. Gentle slopes
Small undulating hills (P, M) Predominant concave hillsides and valleys identified by negative slope breaks, sharp and
narrow ridges, aligned crestlines in some cases, gentle to moderate slope
Large rolling hills (R) Variable and alternate concave and convex hillsides, mostly associated with positive slope
breaks, sharp but wide ridges, aligned crestlines, steep slope
Small rolling hills (T, C) Predominant concave hillsides and valleys identified by negative slope breaks, sharp and
narrow ridges, aligned crestlines transverse to the main ridge top in some cases, moderate
slopes
steepness (as an expression of local topography), and (e) primary (inter-granular) permeability of the unsaturated
water table depth. zone. Thus, this feature directly affects groundwater vul-
nerability. In metamorphic and igneous rocks, which pre-
Bedrock lithology dominate in Test Area T1, secondary permeability (due to
discontinuities) would be more important in terms of
The mineralogy, grain size and fabric of the bedrock and groundwater flow and it is also essential to consider the
related weathered materials, control properties such as weathered materials originating from such crystalline
shear strength, pore water suction, infiltration capacity and rocks. In this sense, Fernandes (2003) suggests that two
natural attenuation of contaminants. According to Vrba and situations should be considered when estimating ground-
Civita (1994), hydraulic accessibility to the saturated zone water vulnerability in crystalline rocks: (a) where weath-
in sedimentary rocks and unconsolidated sediments, which ering cover is thick, the composition of the weathered
predominate in Test Area T2, mainly depends on the materials will strongly influence vulnerability; and (b)
123
8. Environ Earth Sci
where weathering cover is thin or absent, the vulnerability underpinned inferences about major and small-scale
will be conditioned by the occurrence and characteristics of structures, including joints and schistosity.
the discontinuities within the rock mass. As demonstrated by Fernandes and Rudolph (2001) and
Bedrock lithology is also liable to influence land insta- Fernandes da Silva et al. (2005), lineament analysis can be
bility processes depending on the mineralogical composi- integrated with empirical models of tectonic history based
tion, fabric and inherent structures. The orientations, on outcrop scale palaeostress regime determinations to
characteristics and spacings of rock mass discontinuities identify areas of greater density and interconnectivity of
are particularly important in this regard (Hudec 1998). fractures as well as greater probability of open fractures. In
In the present study, the bedrock types were grouped addition, it is possible to deduce angular relationships
according to their fresh (unweathered) state as well as between rock structures (strike and dip) and between these
taking account of any saprolitic and other altered materials and hill slope directions.
where present. Crystalline rocks were grouped as follows: The following assumptions were made in order to
Gr—granites (mostly coarse-grained, massive or foliated); characterise fracturing in the rock or saprolitic soil mass:
Gngr—granitic gneisses (mostly fine-grained, foliated);
(a) Variations of density and connectivity of fractures
B—banded gneisses; X—schistose gneisses and shear zone
could be mapped through lineament analysis by direct
mylonites; Bx—mixed gneisses (including both composi-
correlation, respectively, with density and intersection
tional banding and schistosity); and D—dolerites. Sedi-
of lineaments on images, because in the study area
mentary rocks were grouped into Iam—sandstones
most of the fractures were vertical or sub-vertical so
(medium to coarse grained, mostly massive); Iaf—sand-
they appeared as rectilinear traces at the surface.
stones (fine grained, mostly stratified); IDR—mudstones
(b) Late tectonic events (Cenozoic) control the aperture
with pebbles and laminated rhythmites; FRC—intercalated
of fractures and according to Fernandes and Amaral
sandstones, siltstones, claystones, and mudstones of the
(2002), in most cases, a particular tectonic event gives
weakly consolidated Tertiary age Rio Claro formation.
rise to a generally pervading stress field which
Clay content and its variation through the weathering
controls the orientation and character of fractures in
profile have a particularly significant effect on groundwater
a localised area. Those generated by extensional
vulnerability and erosional processes (Aller et al. 1987;
tectonic stress are of particular interest as they usually
Hill and Rosenbaum 1998). In this regard, lithological
display greater apertures. For instance, water flow
groups B, X and Bx give rise to predominantly clayey
tends to be much faster in the wider aperture fractures
weathered materials that are likely to provide greater
as gravity forces would prevail over capillarity forces
attenuation capacity and reduced hydraulic accessibility to
and soil-matrix hydraulic conductivity in rainy epi-
the saturated zone. On the other hand, groups Gr and Gngr
sodes and nearly saturated conditions (Wang and
would produce sandy materials and greater hydraulic
Narasimhan 1993).
accessibility to the saturated zone. The presence of schis-
tosity and foliation discontinuities within the rock and Lineaments extracted from images were cross-referenced
saprolitic materials would tend to cause slope failure and with field (structural-geological) measurements gathered in
landsliding hazards, depending on the orientation of those the present study and also available from Fernandes (1997).
features with respect to the direction of slopes and also on Density of lineament (km/km2) and lineament intersections
the groundwater conditions. (number/km2) were computed automatically using a com-
Data on bedrock lithology were derived from existing puter script written in MapBasicÒ in a MapInfo package
geological maps which were cross-referenced with image and then cross-referenced with visual inspection and
textural characteristics including density of aligned textural manual counting to check the accuracy of the automated
elements related to drainage and relief in particular (see method.
Table 3). Non-parametric statistical tests (see Fernandes da Silva
et al. 2005; Fernandes da Silva and Cripps 2008) were
Tectonic discontinuities performed in combination with visual analysis of trends of
lineaments on rose diagrams (see Fernandes and Rudolph
Geological structures such as faults and joints in the rock 2001; Fernandes and Amaral 2002) to identify the tectonic
mass, together with their relict structures in saprolitic structures associated with specific tectonic events in each
soils, exert significant influences on shear strength and basic compartmentalisation unit (BCU). Greater probabil-
hydraulic properties of geomaterials (Aydin 2002; Pine ity of occurrence (and frequency) of open fractures was
and Harrison 2003). In the present study, analysis of deduced from BCUs where extensional stress regimes were
lineaments extracted from satellite images and knowledge considered to prevail due to the effect of tectonic event E3-
about the regional tectonic evolution of the area (Table 4) NW (see Table 4).
123
9. Environ Earth Sci
Table 4 Cenozoic tectonic evolution of the region of Campinas according to Fernandes and Amaral (2002)
Age Principal palaeostress directions Shear fracture Extensional fracture Tectonic
(plan view) orientations orientations event
Quaternary σ1 N20–30W and N50–60E N10–30E E5-NNE
σ3
σ1 N30–50W and N30–50E NS E4-NS
σ3
σ1 WNW and NNW–NS N30–60 W E3-NW
σ3
Neogene σ3 N45–65W and N45–65E EW E2-EW
σ1
Cretaceous to paleogene σ1
EW-ENE and NNE-NS NE E1-NE
σ3
Estimates of the potential magnitude of fracturing were drainage conditions. In this regard slope steepness is a
derived from qualitative scores given according to the meaningful and measurable indicator. Similarly, water
following attributes (Table 5): (a) lineament density (per table depth is also controlled by the local topography.
km2); (b) lineament intersections (per km2); and (c) pre- Hence, the assessment of potential for land instability and
dominant tectonic event in each BCU. Such scores were enhanced groundwater vulnerability were based upon soil
assigned in relation to statistical mean values determined profile, slope steepness and water table depth, either solely
for each attribute except for the relevant predominant or in combination.
tectonic event. In this case, maximum score (A) was For instance, infiltration capacity is a function of slope
assigned to tectonic event E3-NW and minimum score (B) steepness and inter-granular (primary) permeability of the
to any other event including E4-NS (also extensional). uppermost layer of the unsaturated zone (Rubin and
Classes of fracturing were derived from the relative pro- Steinhardt 1963, quoted by Fernandes 2003). As observed
portion of these qualitative scores as follows: Class 1: three by Thornton et al. (2001), contaminants are mostly atten-
scores ‘‘B’’; Class 2: one score ‘‘A’’; Class 3: two or three uated by processes of biodegradation and adsorption that
scores ‘‘A’’. These classes were designed to express depend on the mineralogical composition, texture and
increasing magnitude of fracturing and therefore greater thickness of the unsaturated zone materials.
potential influence on ground behaviour. For the present investigation, data on thickness and
texture of soil profiles were assembled to express the
Soil profile, slope steepness and water table depth mineralogy, grain size, structure, strength and density/
degree of compaction of generic soil types. Soil horizons
The development of a particular thickness and type of were characterised using a geotechnical approach as sap-
tropical soil profile depends not only upon the parental rolite, residual, superficial and gravity-transported hori-
materials present but also upon local topography and zons. Primary data were derived from existing soil maps
Table 5 Qualitative scores attributed to the parameters (density of lineaments, density of lineament intersections, predominant tectonic event)
taken for derivation of classes of fracturing
Parameter Density of lineaments (km/km2) Density of lineament intersections (km/km2) Predominant tectonic event
a a b b
Parameter value [3.90 3.90 [2.99 2.99 E3-NW E4-NS Undefined
Score A B A B A B B
a
Range of average values of density of fracturing for BCUs in Test Areas T1 and T2: from 1.26 and 7.97 km/km2
b
Range of average values of density of lineament intersections for BCUs in Test Areas T1 and T2: from 0.06 to 10.97 intersections per km2
123
10. Environ Earth Sci
supplemented with field observations for representative soil classification of units as this was consistent with relatively
profiles in each BCU. Soil profile data were stored as a coarse scale of the cartographic maps (1:50,000), and the
BCU attribute table using the MapInfo package. amount of data available.
Thirteen representative soil weathering profiles, from
sandy to clayey, were identified and three classes of soil Evaluation and classification of units
thickness were defined as follows: (a) 2 m, (b) 2–5 m;
and (c) [5 m. These classes were designed to account for Three steps were required for the evaluation and classifi-
the different impacts of soil thickness on hydraulic acces- cation of units: (a) definition of classes to express the
sibility to the saturated zone as well as to the potential estimated magnitude of the parameter being analysed; (b)
impacts of construction work. definition of classification rules depending on the purposes
Average slope steepnesses were derived both manually of the study; and (c) evaluation and final cartography,
and semi-automatically for each BCU. The manual pro- including the application of classification rules and analysis
cedure involved measurements from existing printed of units for presentation on maps or other forms of output.
topographic 1:50,000 scale maps. The freeware GIS and First, four classes each of groundwater vulnerability and
ˆ
image processing package SPRING (Camara et al. 1996; of susceptibility to land instability were designated as very
INPE 2009) was used for the semi-automated methods. The high, high, moderate and low. Attribute or ‘‘synthesis’’
procedure included (a) digitising of sub-sets of 20-m con- tables were constructed (in GIS MapInfo package) specif-
tour lines from the existing paper record 1:50,000 topo- ically for these purposes. These contained data on the
graphic maps (as no digital maps were available); (b) selected attributes for each BCU, which were allocated in
derivation of heights from 50 9 50-m square numerical fields or columns as follows (see Table 6a, b): (a) BCU:
grids obtained by interpolation from 20-m contour lines. unit code; (b) G_LITO: grouped bedrock lithology; (c)
Generation of numerical grids was performed using a built- CLAS_FRAT: classes of fracturing; (d) TYPE_SOIL:
in weighted mean interpolator based on quadrants and predominant soil type considering the whole weathered
restriction of repeated elevation values. Overlaying oper- profile; (e) THICK_SOIL: average thickness of the whole
ations and use of a computer routine written in LEGALÒ profile; (f) NA: average water table depth; and (g)
(Spatial Language for Geo-processing Algebra) to perform CLAS_DECLIV: class of slope steepness (declivity).
neighbourhood operations over the numerical grids pro- Qualitative and semi-quantitative rules of classification
vided average slope steepness values for each polygon. The were devised from a mixture of empirical knowledge and
following slope steepness classes were used: Low—less statistical approaches and then applied to each BCU. The
than 5°; Medium—between 5° and 10°; High—between classification tool was a spreadsheet-based model that used
10° and 15°; Very high—greater than 15°. nominal, interval and numerical average values as assigned
Tonal contrast in Near-IR (Band 4) and Mid-IR (Band 5) in the synthesis attribute tables, in a two-step procedure to
was used to indicate the presence of water in the sub-sur- produce the required estimates. In the first step, each
face, particularly in the unsaturated zone. However, such selected attribute was analysed and grouped into three
proxy information was insufficient for use in engineering categories shown in Tables 7 and 8, as follows: high (A),
land assessments. Therefore, data on water table depth from moderate (M), and low (B) depending upon their potential
existing borehole and well records were also used and cross- influence on groundwater vulnerability and land instability
referenced with image interpretation. Data on hydrostatic processes. In the second step, all attributes were considered
depth from borehole and well records were plotted on the as having the same relative influence and final classifica-
digitised topographic maps to allow derivation by interpo- tion for each BCU was the sum of the scores (either A, M
lation and extrapolation where water table depth contours or B) respective to each attribute considered. The possible
were assumed to be approximately parallel to the topo- combinations of these are illustrated in Table 9.
graphic contours. However, such derivation approach led to As discussed later in the paper, limitations for deriving
inaccuracies such as major variations of hydrostatic level in information on soil thickness and water table depth as well
similar topographic situations, and in order to reduce them, as to make such information compatible to BCUs prevented
image interpretation and statistical parameters (median and the incorporation of all selected attributes into the classifi-
standard deviation) were combined to determine the trends cation scheme. Therefore, the evaluation/classification of
in hydrostatic depth in top, hillside and valley situations. By units was based upon bedrock lithology, tectonic disconti-
this means contour lines for 5, 10, and 20 m depth were nuities (fracturing), soil type and slope steepness (declivity).
traced with approximately 80% of confidence, where the The classification of units was performed either manu-
confidence level was determined by validation against the ally or semi-automatically through GIS-based operations.
original data. For convenience, the 10-m contour line was The latter involved logical spatial operations to set attri-
then adopted as a criterion in further evaluation and butes into categories high (A), moderate (M) and low (B)
123
11. Environ Earth Sci
Table 6 Summary data
UBC GROUP_ LITO CLAS_FRAT TYPE_SOIL THICK_SOIL N_A CLAS_DECLIV
(a) Test Area T1
CSA1 Gr 2 Sandy [6.5 m 10, [10 Low
CLC1 X 3 No information No information [10 to 10 Medium
CRA1 Gr 3 Sandy [6.5 m [10 to 10 Medium
CNC2 X, Bx 3 Sandy to sandy-silty [4.0 m 10, [10 Medium
CRR3 Bx/GnGr 3 Sandy grading to clayey in depth [3.5 m [10 to 10 Very high
CNC1 B, X 3 Clayey to sandy in depth [1.0 m [10 to 10 Medium
BAC1 IAM, Gr/X 1 Sandy No information 10, [10 Medium
CRR2 GnGr/B 3 No information No information [10 to 10 Very high
CRR5 GnGr/B, Bx 3 Sandy-clayey grading to clayey- 5–10 m [10, 10 High
sandy or sandy-silty in depth;
occurrence of detached blocks
COC3 Bx, B 3 Sandy 3.5 m [10, 10 Medium/high
CSR3 Gr, Bx 3 Occurrence of detached blocks 1–5 m 10, [10 High
CLR3 X, Bx/GnGr 3 No information No information 10, [10 High
CLT1 B, X, Bx/GnGr 3 Sandy to sandy-silty 4.0 m 10, [10 Medium
(b) Test Area T2
BAA1 IAM, IAF 1 Sandy-clayey grading sandy-silty 1–5 m [10, 10 Low/medium
in depth
BBP2 IDR, IAF, D 3 Sandy-silty grading to sandy- 1–5 m 10, [10 Medium
clayey in depth; medium to low
compacity
BAA2 IAF, IAM 1 Clayey-sandy 1–5 m [10 to 10 Medium
BAP1 IAM, D 2 Sandy to sandy-silty grading to [2 m [10, 10 Low/medium
silty-sandy in depth
BBM3 IDR, FRC, D 1 Sandy-clayey; blocky; moderately [2 m 10, [10 Low
compact
BCA1 IAF 1 Clayey-sandy grading to sandy- [1.8 m 10 to [10 Medium
silty in depth; blocky
BGA1 (1) FRC 1 Sandy-clayey grading to sandy- [2 m [10, 10 Low
silty and clayey-sandy in depth
BBP7 IDR, IAF, D 3 Sandy-clayey to sandy-silty [2 m 10 to [10 Low/medium
BCP2 FRC, D 2 Sandy to sandy-silty; massif 5–10 m [10 Low
BFA1 FRC, IAF 1 Sandy-clayey; friable; granular [2 m [10, 10 Low
BGA1 (2) FRC, IAF 1 Sandy-clayey grading to sandy- [2 m [10, 10 Low
silty and clayey-sandy in depth
BDA2 D, IAF 2 Clayey-sandy No information 10, [10 Low
BDA1 D, IAF 1 Clayey-sandy a clayey No information 10, [10 Low
BCU basic compartmentalisation unit code, G_LITO grouped bedrock lithology, CLAS_FRAT class of fracturing, TYPE_SOIL predominant soil
type, THICK_SOIL average soil thickness, NA average water table depth, CLAS_DECLIV class of slope steepness
with mathematical (summation) to produce the final Relatively greater slope steepness and fracturing as well
estimates. as predominant sandy soils in Test Area T1 were associated
The outcomes presented here were achieved manually with a greater number of BCUs classified as to high and
using GIS to display and manipulate results. Tables 10 and very high susceptibility to land instability processes (12 out
11 show the estimated susceptibility to land instability of 13) in comparison with Test Area T2 (just 1 out of 13).
processes and groundwater vulnerability in the two test On the other hand, by reducing rates of infiltration, greater
areas with each attribute considered individually and slope steepness may lower the impact of fracturing on
summed for all the attributes. Figures 4 and 5 show overall groundwater vulnerability, particularly in crystalline and
classifications in spatial map format. less weathered rocks, which predominate in Area T1.
123
12. Environ Earth Sci
Table 7 Attributes used for evaluation/classification of units (BCUs) according to their potential influence (high, moderate, low) on ground-
water vulnerability to pollution hazards
Attributes High (A) Moderate (M) Low (B)
Bedrock IAM, Gr/Xa (coarse-grained Gr (granites) B (banded gneisses)
lithology sandstones ? mix of granites GnGr/B, X, Bxa (mix of granitic and banded X (laminated to schistose
and schists) gneisses ? laminated to schistose gneisses ? banded gneisses)
IAM (coarse-grained mylonitic gneisses) Bx (banded mylonitic gneisses)
sandstones) B, X, Bx/Gngra (banded gneisses ? schistose gneisses ? mix IDR (mudstones with pebbles
of banded mylonitic gneisses and granitic-gneisses) and rythmites)
FRC (Rio Claro Formation—mix of sandy mudstones, D (dolerites)
siltstones, muddy sandstones and rythmites)
IAF (fine-grained sandstones)
Fracturing 3 2 1
Soil type Sandy Sandy-clayey Clayey
Sandy-silty Sandy- silty to sandy-clayey Clayey-sandy
Sandy-clayey grading to sandy- Sandy grading to clayey
silty Sandy-clayey grading to clayey
Sandy to sandy-silty Clayey-sandy grading to sandy-clayey
Clayey-sandy grading to Clayey grading to sandy
sandy-silty
Slope Low Low to medium High
steepness Medium Very high
Medium to high
a
Groups separated by forward slash comprise an undistinguished mixture of bedrock lithologies. The ‘‘comma’’ sign indicates the occurrence of
more than one group of bedrock lithology listed in decreasing order according to their occurrence in terms of areal distribution
Table 8 Attributes used for evaluation/classification of units (BCUs) according to their potential influence (high, moderate, low) on suscep-
tibility to land instability processes
Attributes High (A) Moderate (M) Low (B)
Bedrock lithology Iam, Gr/Xa (coarse-grained sandstones ? mix of granites and IAF (fine-grained IDR (mudstones with
schists) sandstones) pebbles and rythmites)
IAm (coarse-grained sandstones) B (banded gneisses) D (dolerites)
FRC (Rio Claro Formation—mix of sandy mudstones, Bx (banded mylonitic
siltstones, muddy sandstones and rythmites) gneisses)
X (laminated to schistose gneisses)
Gr (Granites)
GnGr/B, X, Bxa (mix of granitic and banded
gneisses ? laminated to schistose gneisses ? banded
mylonitic gneisses)
B, X, Bx/Gngra (banded gneisses ? schistose gneisses ? mix
of banded mylonitic gneisses and granitic-gneisses)
Fracturing 3 2 1
Soil type Sandy Sandy-clayey Clayey
Silty-sandy Sandy grading to clayey Clayey-sandy
Sandy-clayey grading to silty-sandy Clayey-sandy grading to Clayey-sandy grading to
Sandy to silty-sandy sandy-silty sandy-clayey
Sandy-silty to sandy-clayey Clayey grading to sandy Sandy-clayey grading to
clayey
Slope steepness High Medium Low
Very high Low to medium
Medium to high
a
Groups separated by forward slash comprise an undistinguished mixture of bedrock lithologies. The ‘‘comma’’ sign indicates the occurrence of
more than one group of bedrock lithology listed on decreasing order according to their occurrence in terms of areal distribution
123
13. Environ Earth Sci
In Test Area T2, high and moderate groundwater vul- is less than that in the coarse- and medium-grained
nerability (6 out of 13 BCUs) was associated with sand- sandstones.
stone-dominated bedrock lithology, sandy soils and low
slope steepness. In addition, BCUs with greater frequency
and connectivity of fractures were classified as having high Discussion
and moderate groundwater vulnerability despite consisting
of clayey bedrock lithologies such as mudstones and As described in previous sections, a physiographic approach
rhythmites in which inter-granular primary permeability provided basic compartmentalisation units (BCUs) which
were experimentally used for terrain assessments. The
opportunity is taken here to discuss the advantages and
Table 9 Possible combinations of scores ‘‘A’’ (high), ‘‘M’’ (moder-
limitations of the approach taken, particularly how these
ate), and ‘‘B’’ (low) respective to the four attributes (bedrock lithol-
ogy, fracturing, soil type and slope steepness) used for evaluation/ have affected the outcomes, and what can be done to
classification of units (BCUs) and evaluation classes resulting from enhance the results or to overcome difficulties.
these combinations The analysis of fracturing proved that there is good
Combinations of scores Evaluation association between physiographic compartments and
classes homogeneous tectonic domains for which the density and
directional trends were relatively uniform, as proposed by
AAAA Very high
Fernandes and Amaral (2002). In most BCUs it was pos-
AAAM, AAAB, AAMM High
sible to determine particularly significant tectonic events,
AAMB, AABB, AMMM, AMMB, MMMM Medium
for example those of an extensional nature (see E3 and E4
AMBB, ABBB, MMMB, MMBB, MBBB, BBBB Low
in Table 4). In addition, non-parametric statistical tests and
Table 10 Partial susceptibility
Test Area UBC Attributes liable to influence susceptibility Susceptibility class
associated with each individual
attributes and overall Lithology Fracturing Soil type Declivity
susceptibility to land instability
processes resulting from the T1 CSA1 A M A B High
summation of all influential CLC1 A A Ma M High
factors (attributes)
CSA2 A A A M High
CNC2 A A A M High
CRR3 A M M A High
CNC1 M A M M Moderate
BAC1 A B A M High
CRR2 A A Ma B High
CRR5 A A M A High
COC3 M A A A High
a
CSR3 A A A A Very high
CLR3 A A Aa A Very high
CLT1 A A A M Alta High
T2 BDA 1 B B B B Low
BDA 2 B M B B Low
BAA 1 A B A B Moderate
BBP 2 B A A M High
BAA 2 A B B M Moderate
BAP 1 A M A B Moderate
BBM 3 B B M B Low
BCA 1 M B M M Moderate
BGA 1 A B M B Moderate
BBP 7 B A A B Moderate
A, high; M, medium; and B, low BCP 2 A M A B Moderate
a
Soil type deduced according BFA 1 A B M B Moderate
to predominant bedrock BGA 1 A B A B Moderate
lithology with BCU
123
14. Environ Earth Sci
Table 11 Partial vulnerability
Test Area UBC Attributes liable to influence vulnerability Vulnerability class
associated with each individual
attribute and overall Lithology Fracturing Soil type Declivity
groundwater vulnerability to
pollution hazards resulting from T1 CSA1 M M A A High
the summation of all influential CLC1 B A Ba M Low
factors
CSA2 M A A M High
CNC2 B A A M Moderate
CRR3 M M M M Moderate
CNC1 B A M M Moderate
BAC1 A B A M Moderate
CRR2 M A Ma B Moderate
CRR5 M A M B Moderate
COC3 B A A M Moderate
CSR3 M A Ma B Moderate
CLR3 M A Ma B Moderate
CLT1 M A A M High
T2 BDA 1 B B B A Low
BDA 2 B M B A Low
BAA 1 A B A M Moderate
BBP 2 B A M M Moderate
BAA 2 A B B M Low
BAP 1 A M A M High
BBM 3 A B M A Moderate
BCA 1 M B A M Moderate
BGA 1 M B A A Moderate
BBP 7 M A A M High
A, high; M, medium; and B, low BCP 2 M M A A High
a
Soil type deduced according BFA 1 M B M A Moderate
to predominant bedrock BGA 1 M B A A Moderate
lithology with BCU
visual inspection of rose diagrams provided similar reas- allow derivation of 3D relationships. The main aspects to
suringly consistent inferences. This association between be considered include (a) angular and cut-crossing rela-
tectonic domains and physiographic compartments would tionships between different types and sets of structures
probably express the influence of the Cenozoic tectonics (planes of fractures and other discontinuities); and (b)
over the arrangement and structuring of drainage and relief spatial relationships between structures and natural slopes
textural elements on images. Although some variability did (thus taking steepness into account). Consideration of these
exist, the results demonstrated considerable regularity and relationships should enhance the evaluation of BCUs and
persistence of spatial relations held by tectonic structures convey key information to local scale analysis.
across the test areas. These aspects were fully corroborated A general issue of relevance is that monoscopic satellite
by good matching between predominant orientations of images are bi-dimensional representations of land surface
inferred structures and palaeostress regimes as indicated by whilst the intended geo-environmental assessments relate
a regional empirical tectonic model and field observations. to both surface and sub-surface aspects. Thus spatial
Density and interconnectivity of fractures were the key information rather than spectral information needs to be
attributes in the characterisation and evaluation of BCUs in analysed. On the other hand, data on determined attributes
terms of engineering geological and hydrogeological had to be derived from external sources using imagery as a
applications. This empirical tectonic modelling enabled subsidiary tool. Accordingly, textural zones with relatively
both major structures and also small-scale fractures to be high internal homogeneity and fixed spatial boundaries
considered in the analysis where the latter were incorpo- which were observed on images may require practical
rated in the interpretation and evaluation procedures. adaptations to be translated into conceptual classes such as
Additionally, it is suggested that further interpretations comprehensive physiographic units. In this experimental
supported by the use of empirical tectonic models would study such adjustments were incorporated in the later
123
15. Environ Earth Sci
Fig. 4 Maps of susceptibility to land instability processes. Test Areas T1 and T2. UTM projection and coordinates
stages of characterisation and evaluation/classification of profile (thickness), which precluded proper incorporation
units but as explained below, this was not universal. of these attributes in the evaluation/classification of units.
For the sake of the present implementation, some terrain Data on water table depth were derived through an
attributes such as bedrock lithology and related weathered experimental approach that combined hydrostatic depth
materials, degree of fracturing and slope steepness were obtained from borehole and well records with interpolation
selected and taken as proxies for properties and processes and extrapolation of values following topographic contour
including shear strength, permeability, natural attenuation lines. This was manually implemented as semi-automated
capacity, infiltration rates and hydraulic accessibility to procedures based only on spatial data analysis would not
saturated zone. It was assumed that the selected terrain allow direct correlation between interpolated values of
attributes would exert some control over the properties and water table depth and surface contour lines. However, it was
processes. Data on such attributes were derived qualita- found that the manual derivation of data led to unreliable
tively and semi-quantitatively by a combination of means estimates of water table depth, thus resulting in consider-
that included image interpretation, input from existing data able variations at similar topographic conditions. These
and field observations. Shortcomings and inaccuracies may variations may have arisen because the primary borehole
stem from this process of derivation. data were affected by (a) groundwater exploitation in dif-
For instance, in a number of cases, BCUs comprised ferent media and at varied piezometric depths in a same
considerable portions of two or more bedrock lithologies in well (e.g. weathered materials at shallow sub-surface and
which case priority was given to the lithology liable to fractures in fresh rock at depth); (b) heterogeneous
result in greater likelihood of hazard. However, adoption of hydraulic conductivity of the aquifer and of the unsaturated
such criterion may be biased and lead to a greater number zone. These shortcomings suggest that derivation of data on
of BCUs being classified as having higher vulnerability or water table depth from external sources may require more
susceptibility. specific data, particularly on shallow sub-surface layers.
Major difficulties found during the characterisation of Such data could possibly be derived from open pit well
units included the estimation of water table depth and soil measurements, which appears to be more compatible with
123
16. Environ Earth Sci
Fig. 5 Maps of groundwater vulnerability to pollution. Test Areas T1 and T2. UTM projection and coordinates
the characterisation of the unsaturated zone in the shallow depth’’) appeared to produce more accurate results, which
sub-surface and with the physiographic approach itself were then used in the classification process. Inaccuracies
(based on land surface features on images). Data from open observed in the semi-automatic procedure possibly stem-
pit well measurements could be then cross-referenced with med from the averaging process with respect to polygons.
remotely sensed data and topographic maps before extrap- The calculated mean value was meant to be representative
olation of values following topographic contour lines. of slope steepness for the whole BCU. However, slope
Another issue to be considered is data on soil profile steepness was observed to range considerably in some
thickness. In the study these were based on field observa- BCUs, which would affect the interpolated numerical grids.
tions and they were considered to be insufficient for the For instance, in 80% of the area of a BCU slope steepness
intended analysis. In general, difficulties with the charac- ranged between 8° and 10° whilst in 15% of area ranged
terisation of soil profiles in terms of thickness and texture from 15° to 18°, and in 5% of area it ranged between 24°
stem from limited knowledge about the processes that and 27°. The expected representative value would be the
control landscape evolution and soil formation and the 8°–10° range. Nonetheless, since the semi-automatic cal-
ways by which these processes influence image texture. culation took a much greater number of interpolated values
Improved understanding of these issues would allow than the manual procedure, the resulting mean value may
superior correlations and extrapolation of values to be be unnecessarily influenced by outlying values. Further
achieved. Therefore, future work should investigate the investigations into semi-automatic derivation of slope
distribution and the characteristics of soil profiles and steepness data would need to look into ways of restricting
potential correlations of these with image texture due to the range of variation that would be acceptable and con-
relief features, with particular reference to morphometric sidered for calculation of a BCU mean value. For instance,
aspects such as density and amplitude of interfluves (or the calculation procedure could incorporate a priori proba-
ridges) and length of natural slopes. bilities by weighting the resulting value according to the
The manual procedure for derivation of data on slope proportion of the area of a BCU on which slope steepness
steepness (see ‘‘Soil profile, slope steepness and water table intervals were derived.
123
17. Environ Earth Sci
In the stage of evaluation/classification of units, all relationships and enable interpretations the angular rela-
attributes were given equal weight, although the relative tionships between rock structures (strike and dip) and hill
influence of each attribute is not known. The main uncer- slopes to be made. This would greatly enhance the poten-
tainties would be the influence of degree of fracturing tial of the method for engineering applications at a local
(Class 3) on permeability, and the effects of low-slope scale. There is also potential for the development of
steepness and sandy superficial soil on aquifer recharge. automated procedures. For example, for the delimitation of
For instance, in Test Area T1 (see Fig. 4), the high number terrain units based on image classification of spatial prop-
of units (12 out of 13) classified as high to very high erties such as detection of groups of contiguous pixels and
potential for land instability appears to be strongly influ- recognition of line patterns based on length, direction and
enced by steeper slope gradients. Further investigations are angular relations between groups of contiguous pixels.
to consider different weights for each attribute with checks Future and specific investigations should include revi-
on the influence of these on the final classification results. sion of procedures of data derivation from external sources
other than imagery, such as water table depth. Further
implementation of the physiographic compartmentalisation
Conclusions approach for engineering and geo-environmental terrain
assessments are required to evaluate its application in other
In the present study, remote sensing techniques were used geological and geomorphological settings and different
to delimit terrain units and to derive geo-environmental scales of observation, analysis and graphic representation.
data. Data from external sources, including water well logs
and records, existing thematic maps and field studies were Acknowledgments The authors would like to thank Dr. Mara A.
Iritani and Dr. Lidia K. Tominaga for their contribution to data der-
also used. The delimited units were further interpreted in ivation and interpretation, the UK Foreign Commonwealth Office
terms of potential to land instability and vulnerability to (FCO) and the Brazilian National Council for Scientific and Tech-
groundwater contamination at a semi-regional scale of nological Development (CNPq) for their financial support, and the
1:50,000. anonymous reviewers for their helpful advice.
The successful use of low-cost techniques based on
satellite image interpretation, non-commercial software References
package (SPRING) and manual data processing procedures
justified this approach and a wide range of difficulties and Abella EAC, Van Westen CJ (2008) Qualitative landslide suscepti-
limitations liable to be experienced by local and regional bility assessment by multicriteria analysis: a case study from San
agencies in developing countries were addressed. Particular ´
Antonio del Sur, Guantanamo, Cuba. Geomorphology 94:453–
466
limitations such as the need for time-consuming and costly Akiwumi FA, Butler DR (2008) Mining and environmental change in
field mapping and data integration into appropriate dat- Sierra Leone, West Africa: a remote sensing and hydrogeomor-
abases were circumvented and potential problems arising phological study. Environ Monit Assess 142:309–318. doi:
from a lack of hardware and software capabilities, shortages 10.1007/s10661-007-9930-9
Aller LT, Bennet T, Lehr JH, Petty RJ, Hackett G (1987) DRAS-
of trained staff, and scarce organizational resources would TIC—a standardised system for evaluating groundwater pollu-
also be liable to compromise the viability of other approa- tion potential using hydrogeological settings. Report No. 600/2–
ches. The main advantages of utilizing satellite imagery are 87/035. US Environmental Protection Agency, Washington, DC
low-cost coverage of large areas and the large amount data Aydin F (2002) Heterogeneity and behaviour of saprolitic slopes.
Paper presented at 9th International Congress of the Intl. Assoc.
that may be obtained in areas of sparse, discontinuous or Engineering Geology and Environment, Durban, South Africa.
non-existent data. An integrated top-down (regional to site Proceedings, p 846–856
scale) approach was adopted for data analysis and inter- Barton J, Alexander D, Correa C, Mashelkar R, Samuels G, Thomas S
pretation which facilitated a predictive capacity concerning (2002) Integrating intellectual property rights and development
policy. UK Department for International Development, Com-
terrain characteristics including tectonic fracturing, geo- mission on Intellectual Property Rights, London
technical properties and ground conditions. Beaumont TE (1985) An application of satellite imagery for highway
Lineament analysis combined with the use of an maintenance and rehabilitation in Niger. Int J Remote Sens
empirical tectonic model allowed areas with greater density 6(7):1263–1267
Beaumont TE, Beaven PJ (1977) The use of satellite imagery for
and connectivity of fractures, including small-scale ones, to highway engineering in overseas countries. (TRRL—Transport
be delimited and areas of greater probability of occurrence & Road Research Laboratory) England, Supplementary Report
of open fractures were also identified. These aspects were 279, 19 p
considered in the evaluation/classification of terrain units Bennett MR, Doyle P (1997) Environmental geology. Wiley,
Chicester
as they would exert significant influences over land insta- ˆ
Camara G, Fonseca F (2007) Information policies and open source
bility and groundwater vulnerability to pollution hazards. software in developing countries. J Am Soc Info Sci Technol
Further research work should focus on the derivation 3D 58(1):121–132
123