International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
Considering the geostatistical methodology presented, four
main steps can be distinguished: a) obtaining spatial variability
functions (variogram functions) for the hard data; b) estimation
of average values of contaminant heavy metals by means of the
ordinary kriging method, c) production of hazard maps based on
simulated images of the spatial dispersion of heavy metals,
using a stochastic simulation technique — direct sequential
simulation and co-simulation, d) optimum sampling desing for
the next campaigns.
2. DATA SET
The study area is a 2 km? region located approximately 10 km
in the South of the Aznalcóllar mine. The information available
was obtained through a soil sampling realized in August 1999,
where 80 samples were collected. For the purpose of this study,
only the samples located inside the study area, i.e. 40 samples,
were used to characterize spatial dispersion of residual
contamination with heavy metals (figure 3). Soil samples were
collected from the topsoil and analytical results of Cu, Pb, Zn,
Cd and As recorded in terms of total concentration (ppm).
Figure 1 shows the position of the sampling locations and the
basic statistics of the metal concentration are summarized in
Table 1.
X
N * eo
ros
{
t
e
{ ;
= J
/ f
/ e
/ e
4
f ./
/. » /
/ .
e^
A 9
p 4A
«1
se .*
e e >
. v;
s * e^
7 . . em
à e e e|
! e 9 }
4
Y » e j
t
;e e À =
' e e
quc 0 ot i d
Figure 3. Sampling locations
Cu Pb Zn Cd As
Mean 242.53 | 686.43 | 1042.5 3.30 | 336.55
0
Std.Dev. | 184.03 | 848.17 | 820.40 2.98 | 448.00
Minimum 47.0 116.0 146.0 0.5 24.0
25" Perc. 104.5 203.5 362.5 1.0 77.0
Median 220.5 | 403.0 | 1031.5 3.0 193.5
75% Perc. 340.0 | 830.5 | 1421.5 40 | 415.0
Maximum | 1074.0 | 5150.0 | 4460.0 17.0 | 2649.0
Range 1027.0 | 5054.0 | 4314.0 16.5 | 2625.0
Table 1. Basic statistics of the available data (in ppm)
3. SPATIAL VARIABILITY CHARACTERIZATION
The spatial variability characterization was one of the main
objectives of this work. This characterization has been made
374
using the variogram functions. Variograms of the metal contents
value were calculated (Figure 4).
h
30000 0) : =
25000
20000 .
wx] Lem :
10000 /
/ Variable Cu:
500017 Struct 1: Exponential
je Range: 450m; Sill: 16552 ppm2
% 500 1000 1500 2000 2500 3000
y (h) "e
250000
200000 . +
150000| pe ten ^
4 .
/
100000 /
/ Variable Pb:
50000 / Struct 1: Nugget effect. Sill: 50000 ppm’
Struct 2: Exponential
Range: 300m; Sill: 111309 ppm“
0 ;
0 500 1000 1500 2000
h (m)
y(h
450000 ©)
400000 .
350000 : .
300000 7777777 nn
250000 m t ot
^
200000 . /
150000 /
i
/
109009): / Variable Zn:
50000| / Struct 1: Exponential
of Range: 750m; Sill: 307762 ppm’
0 500 1000 1500 2000 2500 3000
h (m)
6 v (h)
5 .
ab SPHERE I e Rt Te mal
3 ZU : .
] /
/ Variable Cd:
1 / Struct 1: Exponential
/ Range: 340m; Sill: 4.20 ppm’
0 ———
0 200 400 600 800 1000
h (m)
y(h)
60000 .
50000 : : a
40000 ER vs
PA ?
30000] /
/
20000 Variable As:
Struct 1: Nugget effect. Sill: 20000 ppm’
10000 Struct 2: Exponential
Range: 500m; Sill: 23740 ppm’
© 500 1000 1500 2000
h (m)
Figure 4. Variogram of the experimental data
Internat
The exp
models
distance-
reflects a
Using th
estimatio
probable
The esti!
geostatist
stochastic
from the
improve t
In the o
applied tc
are calcul
analysis.
unbiased
figure 5, :
Fi
5.1 Co-Si
The applie
and cosir
multivariat
sequential
based on tl
(secondary
of another 1
The simul;
reality. Al
occurrence.
following c