2004 
—— 
the 
ues. 
ucía 
the 
each 
evel 
ion, 
rent 
the 
lues 
), 22 
and 
ata 
els: 
Z|Zn 
lean 
Z2Zm 
the 
etals 
Z3Zm 
the 
that 
Z4Zm 
nost 
ding 
X). 
Xo: ™ 
s 10 
) of 
ther 
each 
that 
sory 
sory 
ever 
ent), 
one 
the 
nt is 
u OT 
A 
s of 
ately 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
72% of the study area needed compulsory treatment and 22 % 
of the study area needed a compulsory investigation. 
Risk map A ue E 
ael awn | 
d 
B Scenario i (0.0%) E m p 
Bl Scenario ii (16.990) dur 
ES Scenario iii (30.096) cs i 
  
.—.; Scenario iv (43. 9855. r. "e Y 
500m 
  
Figure 7. Global hazard map for scenario i 
Risk map EN 
| \s 
ME Scenario i (0.0%) en 
BH Scenario ii (6.4%) | WW 
BRR Scenarioii (21.7%) >" / 
T / . 
___ Seenario iv (71.9%) — /; 7 
  
r eue tm fi 
0 500m 
Comes 
  
Figure 8. Global hazard map for scenario iv 
$3 Environmental hazard maps 
Finally, with the hazards maps obtained for the two alternatives 
it is possible to intersect them with an environmental impact 
map. An impact map that classify to areas according its 
environmental importance was made by C.M.A. (C.M.A, 
2000). The maps include 4 different impacts levels. The lowest 
impact level corresponds to an extensive culture occupation that 
occupies 56% of the total study area while the highest 
corresponds to the Guadiamar river margins (27% of the area). 
Intersecting the impact map with the 2 hazards maps (resulting 
from the two alternatives for the scenario iv) 23% and 36% of 
the highest impact area needed treatment for the first and 
second alternative, respectively. 
6. NETWORK SAMPLING DESIGN 
According to the previous results, the residual contamination 
affected to a determined areas. For this reason, a new sampling 
campaign was considered as necessary. The new sampling must 
have a lower grid size in order to reach a better characterization 
of the variability characteristics of the residual contamination 
(local contamination). 
The campaign was made in summer 2001. The new data tried to 
analyse the real situation of the affected area after two years 
from the previous data (1999) and examine the properties that 
have had in the zone the different cleaning tasks (by means of 
the organic and chemistries addition) thus own regenerative 
power of the environment. 
The sampling network design has two principal conditions. 
First, only the north area -from the Aználcollar mine to the 
Doblas Brigde- must be considered due to the elevated 
contained which were detected in this zone. On the other hand, 
the total number of samples must be around 300. 
For the sampling network design geostatistical techniques were 
applied. The method has a base in the fact than the associated 
errors to an estimation process can be calculated a priori inside 
to the geostatistical schema. These errors only depend of the 
variability function (obtained from the structural analysis - 
variogram analysis- of the available data that are considered as 
representative of the variable to sample) and of the position of 
the sampling points being totally independent of the proper 
variable values. Using this approach, it is possible to analyse the 
estimation error (that define the estimation uncertainty) 
obtained for the different sampling network considered in the 
design. 
This analysis has been combined with the land-use and 
environmental hazard level made by the environmental 
technicians of the CMA. Both aspects are fundamental in the 
sampling design network planning. 
The network optimization begins with the determination of the 
variability function that is considered as representative of the 
variable to sample. The used structure (in which the different 
variogram structures of the variables are take into account) are 
composed by a nugget effect (local variability around 20% 
variance) and an exponential model (with 400m range that 
represents 80% variance). 
Three grid sampling design were considered. The sampling 
schemas are shown in figure 9. The used reference distance unit 
-the work unit over the error must be analysed- is 100m 
(divided into 10x10 elements of 10x10m grid size). The grid 
size is variable between 50m x 50m and 500m x 500m, as 
function of the desired information density for a determined 
location. In Table 2, an example of the application of this 
methodology is shown. The final result is obtained as a 
weighted mean of a priori errors considering the surface of the 
different impact units (that must be according with the future 
available information density).