Journal of Geochemical Exploration 90 (2006) 183 – 196

www.elsevier.com/locate/jgeoexp

Geochemical mapping using a geomorphologic approach based

on catchments
Massimo Spadoni
CNR–Istituto di Geologia Ambientale e Geoingegneria–Via Bolognola 7–00138 Roma, Italy
Received 5 July 2004; accepted 27 December 2005
Available online 28 February 2006

Abstract

The Extended Sample Catchment Basin (ESCB) mapping technique, discussed in this paper, can be used to display the spatial
distribution of geochemical variables measured in stream sediments taking into consideration the geomorphologic settings and the
hydrographic patterns of surveyed areas. This approach is based on the association of an area of statistical representativeness with
each sample, and on the assumption that the concentrations measured in the stream sediments can be considered as average
reference values for this area.
ESCBs can be easily identified considering the position of the sampling points within the hydrographic network and using the
confluences between the streams of highest rank as break points for representing changes of the geochemical background. This
approach, different from the traditional geostatistical and deterministic ones, does not consider the Euclidean distance among the
sampling points as a measure of geochemical similarity but only refers to their functional relationship along the streams (following
the water and the sediment flow) to measure their proximity.
ESCBs can be seen as a specific development of previous techniques based on catchments and proves to be especially useful for
supporting land planning in a preliminary survey phase while it is not specifically suitable for the identification of point sources of
geochemical anomalies.
Due to the fractal nature of the hydrographic network, all the procedures can be driven in a GIS environment by using digital
terrain models apart from their spatial resolution.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Geochemical mapping; Stream sediments; Catchments

1. Introduction contour maps, have been developed to display the

geochemical characteristics of stream sediments and
Stream sediments are commonly used in environ- their spatial representativeness (Govett, 1983).
mental and exploration geochemistry to measure the The spatial behavior of geochemical variables has
average concentration of chemical elements in soils and been also investigated by means of geostatistical
rocks over wide areas and to identify possible sources of techniques based on autocorrelation principles (Math-
anomalies. A number of mapping techniques, such as eron, 1965; Wackernagel, 1995; Webster and Oliver,
point symbol maps, areal maps, worm diagrams and 2001). Geographically weighted regression techniques
have also been developed to meet the necessity to
E-mail address: [email protected]. partition the study area into more homogeneous sectors
0375-6742/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.gexplo.2005.12.001
184 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196

(e.g. Fotheringham et al., 2000). The efficiency of the importance when facing the problem of a more realistic
above-mentioned techniques is closely dependent on the spatial representation of the geochemical processes
selection criteria of the loadings to be used for especially when considering that the selection of
weighting the contributions of the points where the sampling points placement along the hydrographic
variables have been measured. Some relatively recent network is generally due only to practical considerations
approaches, based on the conversion of spatial distri- (e.g. site accessibility) and the need to follow predefined
bution of variables into the frequency domain, rely on rules of stochastic sampling.
multifractal models to separate anomalies from back-
ground values (Cheng et al., 1994, 1996, 2000; Cheng, 2. Stream sediments and hydrographic network
1999). Fractal geometry has also been used to cha-
racterize geochemical landscapes by means of stochastic 2.1. Stream sediments
simulation techniques in the case of low spatial
resolution of the geochemical data sets available on In exploration and environmental geochemistry, one
stream sediments (Rantitsch, 2001). of the most commonly used method of prospecting is
Even if these mapping techniques prove to be helpful based on the study of active stream sediments. Ac-
to build mathematical models of the spatial distributions cording to the definition given by the Forum of the
of the variables and to identify subtended processes at European Geological Surveys (FOREGS), these are
different scales, they only partially or indirectly face the represented by the fine and medium size fraction of
problem of the geographic origin of stream sediments sediments (b 0.150 μm) carried and settled by second
and of the close dependence of their composition from order streams (Salminen et al., 1998). Stream sediments
the local geomorphologic processes. Catchments have can be considered as averagely representative of the
been used in a number of scientific papers to take into outcropping rocks in the drainage basin, upstream of the
account the provenance area of sediments for develop- sampling point (IGS, 1978; Webb et al., 1978; Meyer et
ing models of geochemical dispersion (Moon, 1999), al., 1979; Bölviken et al., 1986; Lahermo et al., 1996),
mapping anomalous concentrations (Rantitsch, 2004) and their nature is closely linked to a number of local
but also for simply representing geochemical baselines processes, of natural and anthropogenic origin, that can
of selected areas (e.g. Bonham-Carter et al., 1987; Ódor significantly affect their composition (Fig. 1). These
et al., 1997; Spadoni et al., 2003). In a previous work input sources have an inhomogeneous distribution
Spadoni et al. (2004) emphasized the usefulness to within catchments and can be localized in circumscribed
adopt “sample catchment basins” (SCB) as map units. areas (point sources). Conversely, inputs rising from
This geographically based approach produces choro- erosion/deposition processes are widespread in the
pleth maps where catchments areas are assumed to be catchments but act with different intensities according
represented, on average, by the same concentrations of to the local geomorphologic and hydrological features.
the nearest sample downstream. The resulting maps Most of times, during the “strategic” phase of a
prove to be useful in defining the source areas of survey, when the reconnaissance aspects are prevalent,
geochemical concentrations and, in case, of localized but often during the subsequent phases too, an accurate
geochemical anomalies. This technique also avoids modelling of the erosion/deposition processes is miss-
“mathematical interference” between neighbouring ing. The same lack of information affects the distribu-
samples. This is a very important aspect especially tion of many sources of possible anomaly. In these
when neighbouring samples are representative of situations, from a statistical point of view, stream
different catchment systems. However, the SCB ap- sediments collected at each sampling point have to be
proach is not able to produce statistical inference and necessarily assumed as if they were homogeneously
also shows some remarkable inconsistencies at the originated from the whole surface of the SCB upstream.
closing points of sub-catchments. According to this
technique, the concentration value at the sampling point 2.2. Hydrographic network
is considered representative only of the catchment
upstream. As a consequence, sampling points act as In 1983, Mandelbrot introduced the principles of
break points in the maps and as conceptual breaks in fractal geometry to provide an appropriate tool to
illustrating the spatial distribution of the variables. It describe the structure and character of spatial objects
means that the concentration values estimated on the (Mandelbrot, 1983). In particular, when dealing with
two sides of sampling points (upstream and down- geographical objects, jagged irregularity is observed
stream) are different. This incongruence is of primary across a range of scales (in a coastline, in a mountainscape,
 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196 185

Fig. 1. Some of the most important factors affecting the composition of stream sediments. From a practical point of view, tributaries can be also
considered as sources of geochemical inputs.

etc.). Fractal geometry has shown that detailed irregu- Topographic maps are graphic approximation of
larity may resemble coarse irregularity in shape, fractal objects too and isohypses are conventional and
structure and form. This property has been called self- approximate representations of the morphology and of
similarity. In a self-similar object, each part has the same its fractal nature. As a consequence, the hydrographic
nature as the whole. pattern drawn on a map will be closely dependent on the
Self-similar objects also share a very complex map scale and quality, and its shape will be as far from
structure. When measuring a linear, one-dimensional, the reality as the map will be far from the real
object, such as a coastline, we have to consider its topography it represents (Fig. 2).
close scale-dependence. Theoretically we could detail These considerations are particularly important when
our survey to measure the irregularity of each rock and defining a hierarchical level for the stream segments
even of each single grain of the coastline. The more we whose ranks are closely dependent on the characteristics
detail our measure the more the total length of the of the hydrographic pattern used. However, apart from
coastline will increase, stretching towards infinity. the base map accuracy, the general pattern of hydro-
Such an intricate structure comes to fill up more graphic networks is conserved because of its self-
space than a one-dimensional straight line but less similarity properties. For this reason the identification of
space than a two-dimensional area (fractional dimen- the confluence points between the stream branches of
sion between 1 and 2). highest orders is independent of the scale of the
Going back to environmental geochemistry, sedi- topographic base map. Moving into a digital environ-
ments of different origin are in any case transported ment, as in a GIS, the same consideration can be made
downstream by many branches of a hydrographic by using digital terrain models to automatically draw
network and are conventionally collected in streams of and hierarchize hydrography. Even in this case, the
second or higher order (Salminen et al., 1998). identification of the confluences of highest ranks is not
However, the hydrographic network has to be consid- affected by the resolution of the digital model.
ered as a fractal object too. In its path towards the sea,
water movement is driven by the land morphology and 3. Distance between sampling points
the shape of the gravitational field. Theoretically we
could measure and draw all the line of impluvium at In geography there are many ways of considering the
each scale, from the main watershed to the micro-valleys distance between points or objects. It can be calculated
between rock grains. According to this complex in terms of absolute physical distance or including the
structure, when detailing our analysis, the hydrographic cost incurred while traveling rugged terrain, or including
pattern will appear more and more jagged and irregular, the need of traveling along networks, or considering
its length scale dependent, its shape self-similar and it difficulties involving barriers that restrict or prevent
will fill up the space migrating from the one- movements (De Mers, 2000). The simplest way to
dimensional to the two-dimensional space, acquiring calculate a distance is considering the “Euclidean
the properties of a fractal object. distance”, represented on a map by the length of a
186 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196

Fig. 2. Hydrographic patterns drawn from base maps at different scales do not match because of the different accuracy of the topographic
representation. Some inconsistencies are also due to different drawing hands and errors in georeferencing (source: SIRA, Regione Lazio, modified).

straight line linking two points. The estimations with Euclidean distances may lead to rough estimates and
geostatistical or deterministic regressive techniques of errors.
unknown values assumed by a spatial variable are Actually a more correct way to measure the distance
generally calculated by using weighted averages built on between sampling points is based on the “least-cost
the Euclidean distances from the points where the distance” performed by a stream flowing down a
variable has been measured. mountainside (Fig. 3). This “functional” distance
When dealing with mapping of stream sediments obviously has a vector nature, since water directionally
geochemistry, the Euclidean distance approach can be flows downstream. Starting from this point of view, a
misleading and there are some important reasons to feel more efficient way to interpret the logical interaction
unsatisfied with it. First of all the proximity of sampling between sampling points should be based on the
points on a map often does not imply that they share a identification of their sequence along the hydrographic
common source of sediments (Spadoni et al., 2004). pattern.
Sampled streams may belong to different drainage
systems and the sediments on their beds may have 4. Sample catchment basin mapping technique
different origins without any possibility to mix or
interact (Fig. 3). Furthermore, the generally abrupt Sample Catchment Basin (SCB) technique (Spadoni
changes in the nature of outcropping lithologies and, et al., 2004) assumes that the geochemical representa-
more in general, of geological features, especially in tiveness of stream sediment samples can be extended to
hilly or mountainous areas, must be also considered to the geological materials outcropping on the surface of
understand the differences between near sampling the whole drainage basin upstream (Fig. 4). The
points. The different dip of geological layers and the presence of further sampling points upstream defines
path of their limits, the presence of faults with relevant new drainage basins that can be seen as superimposed to
offsets, the spatial heterogeneity of hydrothermal the previous ones. Maps can be read as an overlapping
processes are only few examples of how relatively little of different layers where the top level supplies a more
areas can largely differ in their geochemical nature accurate estimation of the variable value. According to
without showing any gradual transition. Geological this technique, geochemical variables are displayed
nature of soils and rocks deeply affects the intensity of using catchments as map units while every interpolation
erosion processes and, as a consequence, differentiate outside this geomorphologic ambit (e.g. by means of
the pattern, the hierarchical level and the seasonal statistical or mathematical interpolators such as Kriging,
regime of hydrographic networks inside catchments and IDW, polynomial, etc.) is not allowed.
sub-catchments. For all these reasons, the use of the Although SCBs prove to be very useful in
traditional geostatistical mapping techniques based on identifying the source areas of geochemical anomalies,
 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196 187

Fig. 3. An example of functional interaction between sampling points in the Mignone River basin in Central Italy. Some points are very close from an
Euclidean point of view (e.g. samples 150–505, 160–527, 91–526) but, at the scale of our survey, they belong to different catchment systems. Some
other sampling points, belonging to the same catchment system, are no more very close when considering their distance calculated along the stream
network (e.g. samples 150–163).

their usage for land management and planning is literature passing through the concept of “geochemical
limited by the introduction, in cartographic representa- productivity”, re-introduced by Moon (1999), mainly as
tions, of abrupt breaks, passing through the sampling a development of previous studies on dilution (Polikar-
points, at the lower boundaries of each SCB (Fig. 4). pochkin, 1971; Hawkes, 1976; Solovov, 1987). The
Actually, it seems a very unrealistic situation since it is models developed under this concept proved to give
reasonable to suppose that geochemical values right useful insights about stream sediments, downstream
downstream each sampling point does not differ too dispersion and their geochemical interpretation. How-
much from the values at the sampling point itself. ever, this approach showed some limits especially when
Furthermore, the nature of the geographic dependence applied to diluted sources, inhomogeneous sampling
between SCBs, sampling points and cartographic units density, low or inhomogeneous erosion rates in the
also make the final map closely dependent on the catchments and presence of hydromorphic dispersion
spatial distribution of sampling points that, in turn, is a (Moon, 1999). In a more generalized case, during the
function of the sampling scheme. reconnaissance survey phase, when the spatial pattern of
anomalous areas is unknown or when dealing with
5. Extended SCB complex environments where the geochemical setting
has not yet been defined, in case of catchments with very
The geochemical variability of sediments along a inhomogeneous morphology and different erosion rates,
stream and the influence exerted on it by the presence of the application of such a mathematical model can be
geochemical anomalies have been widely discussed in misleading or too demanding for operational purposes
188 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196

Fig. 4. Use of SCB technique in the Mignone River basin. Notwithstanding the heterogeneous distribution of Ba concentrations, SCB technique
allows to identify the average concentrations in geographically defined areas without forcing mathematical relationships between neighbouring
samples. On the other side, sampling points unrealistically act as break points in the identification of catchments with different values of average
concentration.

because of the dependence of stream sediment geo- stream segment, i.e. along that part of the stream
chemistry from a complex series of interaction with an included between two well detectable major sources of
high number of variables. In these cases, the pattern of sediments that can act as “break points” of geochemical
geochemical signals along the streams could be, at a first concentrations in stream sediments.
step, inferred by using mathematical or geostatistic To identify the extension of this stream segment we
interpolators (Fig. 5a–c) assumed as representative of have to consider how stream sediments are com-
the distribution process. However, the use of such a kind pounded of geological materials rising from erosion
of interpolation curves can also be very misleading processes on soils and rock outcropping in the
especially where localized presence of natural or catchment upstream. The interpolation curves shown
anthropogenic sources of anomalies, like variations in in Fig. 5 are only few examples of how variable values,
the transport/deposition energy, local erosion dynamics measured at generic sampling points, could be used to
and differences in the outcropping rocks, produce infer the distribution process in the areas between them.
sudden and persistent concentration changes. Both the observation of Fig. 5 and geological
From the point of view of this work, in these considerations tell us that variable values measured at
situations, the real processes could be interpreted by the sampling points have to be considered as the
introducing breaks and steps in the interpolating curves reference values at least for a certain distance
in order to improve the statistical significance of the downstream; an aspect that cannot be neglected when
model (Fig. 5d). The measured concentrations can be our main goal is to identify a reliable spatial pattern of
considered as having spatial influence along the closest the variable value.
 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196 189

Fig. 5. The trend of a generic geochemical variable H along a stream branch (a) can be plotted in a Cartesian diagram using H concentration values vs.
the distance among the sampling points (b). A number of different interpolation procedures can be adopted to display the progressive variation of H
value along the stream. For instance, in panel c, the interpolation line has been drawn using a radial basis function (spline). When sources of
anomalies or geographic constraints are present you can have more realistic results using a “driven interpolation” with break points introduced “by
hand” at an intermediate point between two consecutive samples.

Geostatistical and deterministic interpolation tech- contrary, the technique proposed in this paper is
niques consider each sample within its functional based on the identification of areas of influence for
relationship with the nearest neighbours. Kriging each sample whereas catchments, along with their
technique, in particular, grounds this assumption to geomorphologic boundaries, are considered as the
the measure of the internal correlation degree between basis for the process of spatial inference.
each single measure of the variable under study and its Considering a couple of samples along the same
measures in the neighbouring points. The knowledge stream branch, their spatial “representativeness” can be
of this spatial autocorrelation value is the starting extended downhill and uphill, to an “intermediate point”
point to estimate the values of the variable in each (IP), as also suggested by the interpolation curves shown
point of the geographic space by fitting an authorized in Fig. 5. This way each sampling point can be
mathematical model on the experimental variogram. In interpreted as the centroid of a pertinence catchment
the field of deterministic approaches, the most used that we will call Extended Sample Catchment Basin
techniques estimate the variable values using an (ESCB). From a geochemical point of view, the latter
Inverse Distance Weighted (IDW) algorithm. In this can be considered as averagely represented by the
case, a cross-validation process can be used to check stream sediment collected at the sampling point itself.
the goodness of the starting assumptions. However, all ESCBs can be ideally considered as divided into two
these techniques take root in measuring the Euclidean sub-catchments: the first upstream of the sampling
distances between sampling points. These are assumed point, from where part of the stream sediment sample
as the only reference to identify the neighbours, with- actually originated, and the second downstream of it,
out taking into account any geomorphologic features or where the variable values has been inferred on the basis
hydrodynamic process. The resulting maps, generally of its spatial proximity with the centroid.
based on isolines (isopleth maps), show gradual In such a kind of mapping the identification of a
transitions between the measured values. On the statistically and geographically significant IP location is
190 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196

of primary importance. If the catchment area between size. This assumption is clearly evident when
two consecutive sampling points (intSCB) was con- considering the similarity of hydrographic patterns
sidered as a simple geometric figures, IP could exactly in these areas. Starting from these assumptions, the
be placed in its geometric centre, allowing to split it confluence points between two streams of equal or
into two parts whose area should be associated with similar rank identify, on average, catchments of
the closest sampling points. However, this approach, comparable size, where geomorphologic and sedi-
deriving from simple geometric consideration, may be mentary processes act at a similar scale and intensity.
misleading if the hydrographic fractal pattern and the Starting from these considerations, each intSCB is
geomorphologic setting of catchments are not consid- split into two equally representative parts using, as IP,
ered. On these basis, a more representative way to split the confluence between the two stream branches with
intSCBs can be based on the hierarchic level of the the highest hierarchical level (Fig. 6). This choice
streams. comes from the consideration that, when the
In areas with similar geomorphologic character- hydrographic networks have a good hierarchization
istics, geological setting and climate we can reason- level, the major input of sediments in the main
ably assume that streams with the same or a similar streams generally rises from the confluences of
rank are associated with catchments of comparable highest rank. For this reason, these confluences are

Fig. 6. The first step when drawing an ESCB map is the identification of Intermediate Points (IP). These are placed at the confluence of highest rank
between two consecutive sampling points (a). It must be noticed that IPs position shows a limited but significant independence from the position of
the sampling points along the stream. As an example, the ESCB marked with oblique lines (b) would have preserved the same shape and extension if
the sampling point D had been shifted in one position between IPs “2” and “3” (e.g. in the positions s1, s2, s3 or s4). In practice, ESCB maps
demonstrate less biased by the position of sampling points or by the possible occurrence of errors in their placement on the map respect to a simple
SCB map (c). ESCB maps also show this ‘resilience effect’ also when compared to maps built using mathematical or geostatistical interpolators based
on the Euclidean distance between sampling points.
 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196 191

Fig. 6 (continued).

the most probable points of geochemical changes for ESCB mapping technique puts together the SCB
stream sediments. It must be also highlighted that approach and the statistical principles of spatial analysis,
sampling points can be seen as geochemical centroids aiming to answer some priority questions concerning the
since they are assumed as point of representativeness correct use of catchments in geochemical mapping and
for each ESCB. their statistical relevance. Resulting maps (Fig. 7) are
Accordingly with this approach, not only the best suitable for the estimation of background concen-
placement of IPs is statistically and geologically well trations and for land use planning while their confron-
founded but, consequently, also the resulting map takes tation with SCB maps shows that they are less accurate
advantage of this procedure, in fact: in defining the source areas of anomalies.

• IP identification procedure is simple and standardized; 6. Example application: the Mignone River basin
• the two resulting sub-catchments (from splitting
intSCB) are averagely expected to have areas of 6.1. Geology
similar size;
• inconsistencies due to the strong dependence of the The Mignone River basin (Latium–Central Italy)
resulting maps from the sampling scheme are was selected as a sample area for experiencing different
significantly reduced; mapping techniques within the “Geochemical Baselines
• sub-catchments can be considered as functional units of Italy Project” of the Italian National Research Council
for the possible reconnaissance and the differentiation (Marini et al., 2003). This basin is characterized by the
of geochemical and geomorphologic processes, presence of a variety of rocks of different volcanic or
including local sources of anomaly of geological or sedimentary origin (Fig. 8; Fazzini et al., 1972). Alkali–
anthropogenic origin. potassic pyroclastic flows and phonolithic tephra are
192 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196

Fig. 6 (continued).

present in the eastern part of the hydrographic basin, ter) were collected with an average sampling density of
whereas acidic lavas outcrop in the North of Tolfa and 1 sample/3.59 km2. However, due to the hard ac-
Allumiere towns (Fig. 8). Pliocene and Pleistocene cessibility of some sectors, sampling density was
sands, clays and conglomerates occur in the North-West, distributed ranging from about 1 sample/km2 in the
whereas in the rest of the area mainly Cretaceous– west plains to about 1 sample/5.4 km2 in the east hilly
Eocene marly flysch and calcarenites outcrop. areas (Fig. 8).
In the past, quarrying and mining activities were The concentration of 37 chemical elements (Ba, Bi,
common in this area. Quarrying mainly consisted in the Ag, Al, As, Au, B, Ca, Cd, Co, Cr, Cu, Fe, Ga, Hg, K,
exploitation of lithoid tuff for masonry while mining La, Mg, Mn, Mo, Na, Ni, P, Pb, S, Sb, Sc, Se, Sr, Te, Th,
activities included excavation of kaolinite, alunite, Ti, Tl, U, V, W, and Zn) was measured by ICP–MS after
marcasite, blende, celestite, and barite. However, at aqua regia extraction (analytical data discussed in this
present time, mining activities are almost completely chapter are available on request).
neglected. Afterwards, three different mapping techniques,
The anthropogenic activities include extensive culti- Inverse Distance Weighted (IDW), SCB mapping and
vations in the west flatlands, stock- and sheep-rearing, ESCB mapping, were tested and compared using GIS
and timberlands. Uncontrolled dumping points of urban functions of spatial analysis.
wastes have also been widely detected. On the whole, ESCB average sizes resulted similar to
SCB ones (3.585 km2 vs. 3.582 km2) but, as expected,
6.2. Methodology and results values are significantly less scattered in the first case
(σ = 3.353 vs. σ = 4.277 of SCBs) as a consequence of
Over a total basin surface of about 495 km2, 138 the resilience effect due to the relative stability of IPs
stream sediment samples (150 μm particle size diame- locations with respect to the given sampling scheme.
 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196 193

Fig. 7. Barium concentrations have been mapped using ESCBs technique.

As an example application ESCB, SCB and Boron abundance is closely linked to the presence of
traditional IDW distribution maps for three selected hydrothermal springs in the south-eastern part of the
elements, As, B, and S, are compared in Fig. 9. basin. For this reason, B concentration in stream
The comparative use of these mapping techniques sediments tends to be uncorrelated with its concentra-
shows different aptitude to describe spatial geo- tion in soils or rocks. For this element, ESCB map
chemical variability. demonstrated more faithful in showing the presence of
Arsenic is notably concentrated in stream sediments boron in the sub-catchments where springs occur (Fig.
of the mineralized area around the core of the Mignone 9-b1), while SCB and IDW showed more widespread
River basin. The pattern of anomalous areas (assumed patterns (Fig. 9-b2 and b3).
as exceeding 90th percentile of As measured concen- The presence of sulfur is mainly due to sulfide
trations) significantly differs in dependence of the mineralizations occurring in south central areas of the
adopted mapping technique. In the ESCB map high Mignone basin. Further presence of S has been recorded
concentration areas appear extended eastward and (N0.2%) in the south-western part of the basin where
westward, this way identifying the actual area where marine Pliocene clay deposits with lens of chalk
stream sediments have likely recorded anomalous outcrop. In the same sector, S shows much lower
concentrations (Fig. 9-a1). On the other side, SCBs concentration in the diluted stream sediments sampled
allowed to better identify the source areas of anomalies in correspondence of the close Holocene alluvial plain
recorded at specific sampling points (Fig. 9-a2). of the Mignone River (b 0.1%). In this situation ESCB
Finally, IDW map demonstrated more readable espe- map shows a better overlapping between sulfur
cially for identifying trends at the entire basin scale concentrations and areas where Pliocene and Holocene
(Fig. 9-a3). sediments outcrop (Fig. 9-c1) with respect to the SCB
194 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196

Fig. 8. Geological sketch map of the Mignone River basin and stream sediments sampling scheme.

map (Fig. 9-c2). The closeness of sampling sites in this proximity of the sampling points where abrupt and
sector (about 1 sample/km2) and their relative weight in illogical spatial discontinuity are produced. When using
the mathematical procedure of estimation erroneously the Extended Sample Catchment Basin (ESCB) map-
lead to an IDW map that overestimates S concentration ping approach, the spatial representativeness of each
over the Holocene plain (Fig 9-c3) and, more in general, sample is weighted accordingly to both the geomorpho-
shows quite high errors when values are checked logic boundaries of catchments and the confluences
through a cross-validation test. between differently ranked streams; the latter seen as the
probable main points of variation of sediment geochem-
7. Conclusions istry along the stream network. In the ESCB mapping
technique sampling points act as centroids of sub-
Stream sediments can be useful to identify the catchments extended up to the confluences of highest
average geochemical composition of bedrocks and rank. The spatial pattern of geochemical variables is,
soils in land areas near the sampling points. Geochem- then, realistically represented by extrapolating the
ical maps drawn by means of the traditional geostatis- values measured in each sample to the corresponding
tical and mathematical interpolators does not take into catchments.
consideration the geomorphologic constraints of the A comparative test between ESCB approach and
watersheds and the functional relationship, in terms of other mapping techniques in the Mignone River basin
transport and deposition processes, between sampling (central Italy) demonstrated ESCB utility to map in a
points along the stream network. On the other hand, the more realistic way the distribution pattern of three
mapping techniques based on the association between selected elements (i.e. As, B and S). However, this
geochemical concentrations in sediments and catch- technique should not be preferred when the main
ments meet this need, helping to identify the source problem is the identification of possible point sources
areas of geochemical anomalies. However, these of anomalies. ESCB maps demonstrated to be especially
mapping techniques show clear lack of accuracy in suited during the reconnaissance phase of geochemical
 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196
Fig. 9. Comparison between ESCB, SCB and IDW mapping techniques applied to As, B and S distribution in the Mignone River basin (Central Italy).

195
196 M. Spadoni / Journal of Geochemical Exploration 90 (2006) 183–196

surveys and, thanks to their close link with catchment Mandelbrot, B.B., 1983. The Fractal Geometry of Nature. Freeman,
morphology, could efficiently be used for land planning San Francisco.
Marini, L., Ottonello, G., Vetuschi Zuccolini, M., 2003. The “archivio
purposes. geochimico nazionale” and its cartographic products: a general
overview. In: Ottonello, G., Serva, L. (Eds.), Geochemical
Acknowledgements Baseline of Italy. Pacini editore. 294 pp.
Matheron, G., 1965. Les Variables Régionalisées et Leur Estimation.
Masson, Paris.
I wish to acknowledge Dr. Mario Voltaggio for his
Meyer, W.T., Theobald, P.K., Bloom, H., 1979. Stream sediment
scientific contribution to critically review the ESCB geochemistry. In: Hood, P.J. (Ed.), Geophysics and Geochemistry
mapping technique. I also thank two anonymous in the Search for Metallic Ores. Geol. Surv. Canada Econ. Geol.
reviewers who gave me the possibility to significantly Rep., vol. 31, pp. 411–434.
improve this paper with their extremely helpful Moon, J.C., 1999. Towards a quantitative model of dowstream
comments. dilution of point source geochemical anomalies. J. Geochem.
Explor. 65, 111–132.
Ódor, L., Horváth, I., Fügedi, U., 1997. Low-density geochemical
References mapping in Hungary. J. Geochem. Explor. 60, 55–66.
Polikarpochkin, V.V., 1971. The quantitative estimation of orebearing
Bölviken, B., Bergström, J., Björklund, A., Lehmuspelto, P., areas from sample data of the drainage system. In: Boyle, R.W.
Lindholm, T., Magnusson, J., Ottesen, R.T., Steenfelt, A., Volden, (Ed.), Transactions 3rd International Geochemical Exploration
T., 1986. Geochemical Atlas of Northern Fennoscandia. Geol. Symposium. Can. Inst. Min. Metall., Spec., vol. 11, pp. 585–586.
Surv., Sweden, Uppsala. Rantitsch, G., 2001. The fractal properties of geochemical landscapes
Bonham-Carter, G.F., Rogers, P.J., Ellwood, D.J., 1987. Catchment as an indicator of weathering and transport processes within the
basin analysis applied to surficial geochemical data, Cobequid Eastern Alps. J. Geochem. Explor. 73, 27–42.
Highlands, Nova Scotia. J. Geochem. Explor. 29, 259–278. Rantitsch, G., 2004. Geochemical exploration in a mountainous
Cheng, Q., 1999. Spatial and scaling modeling for geochemical area by statistical modeling of polypopulational data distribu-
anomaly separation. J. Geochem. Explor. 65, 175–194. tions. J. Geochem. Explor. 82, 79–95.
Cheng, Q., Agterberg, F.P., Ballantyne, S.B., 1994. The separation Salminen, R., Tarvainen, T., Demetriades, A., Duris, M., Fordyce, F.
of geochemical anomalies from background by fractal methods. M., Gregorauskiene, V., Kahelin, H., Kivisilla, J., Klaver, G.,
J. Geochem. Explor. 51, 109–130. Klein, H., Larson, J.O., Lis, J., Locutura, J., Marsina, K.,
Cheng, Q., Agterberg, F.P., Bonham-Carter, G.F., 1996. A spatial Mjartanova, H., Mouvet, C., O'Connor, P., Odor, L., Ottonello,
analysis method for geochemical anomaly separation. J. Geochem. G., Paukola, T., Plant, J.A., Reimann, C., Schermann, O., Siewers,
Explor. 56, 183–195. U., Steenfelt, A., Van der Sluys, J., De Vivo, B., Williams, L.,
Cheng, Q., Xu, Y., Grunsky, E., 2000. Integrated spatial spectrum 1998. FOREGS geochemical mapping field manual. Geological
method for geochemical anomaly separation. Nat. Resour. Res. 9, Survey of Finland, p. 47.
155–163. Solovov, A.P., 1987. Geochemical prospecting for mineral deposits
De Mers, M., 2000. Fundamentals of Geographic Information (Kuznetsov, V.V., Trans.) (Engl. Ed.). Mir, Moscow. 288 pp.
Systems, second edition. Wiley. 498 pp. Spadoni, M., Patera, A., Voltaggio, M., Cavarretta, G., 2003.
Fazzini, P., Gelmini, R., Mantovani, M.P., Pellegrini, M., 1972. Geochemical mapping of the Mignone River basin (Latium,
Geologia dei Monti della Tolfa (Lazio settentrionale; Province di Italy): a pilot project for testing data processing and mapping
Viterbo e Roma). Mem. Soc. Geol. Ital. 11, 65–144. techniques. Proceedings of the 4th European Congress on Regional
Fotheringham, A.S., Brunsdon, C., Charlton, M., 2000. Geographi- Geoscientific Cartography and Information Systems, 632–634.
cally Weighted Regression. Wiley. 265 pp. Spadoni, M., Cavarretta, G., Patera, A., 2004. Cartographic techniques
Govett, G.J.S., 1983. Handbook of Exploration Geochemistry. for mapping the geochemical data of stream sediments: the “sample
Statistics and Data Analysis in Geochemical Prospecting. Elsevier, catchment basin” approach. Environ. Geol. 45 (5), 593–599.
New York. Wackernagel, H., 1995. Multivariate Geostatistics. Springer, Berlin.
Hawkes, H.E., 1976. The downstream dilution of stream sediment Webb, J.S., Thornton, I., Thompson, M., Howarth, R.J., Lowenstein, P.
anomalies. J. Geochem. Explor. 6, 345–358. L., 1978. The Wolfson Geochemical Atlas of England and Wales.
IGS, 1978. Geochemical Atlas of Gt. Britain: Shetland Islands. Natl. Clarendon Press, Oxford.
Environ. Res. Council, London. Webster, R., Oliver, M.A., 2001. Geostatistics for Environmental
Lahermo, P., Väänänen, P., Tavainen, T., Salminen, R., 1996. Scientists. Wiley, Chichester, England.
Geochemical Atlas of Finland: Part 3. Environmental Geo-
chemistry–Stream Waters and Sediments. Geological Survey of
Finland, Espoo.