Mapping and monitoring of residual heavy metal contamination and acidification risk after the Aznalcóllar mining accident (Andalusia, Spain) using field and airborne hyperspectral data T. Kempera and S. Sommera a
European Commission, Joint Research Centre, Institute for Environment & Sustainability 21020 Ispra (Va), Italy, email:
[email protected]
ABSTRACT A spectral mixture modelling approach applied to field and airborne hyperspectral data was implemented to map residual contamination after a mining accident. Test case was the Aznalcóllar Mine (Southern Spain) accident, where heavy metal bearing sludge from a tailings pond was distributed over large areas of the Guadiamar flood plain. Although the sludge and the contaminated topsoils have been removed mechanically in the whole affected area, still high abundance of pyritic material remained on the ground. During dedicated field campaigns in two subsequent years soil samples were acquired for geochemical and spectral laboratory analysis and spectral field measurements were carried out in parallel to data acquisition with the HyMap sensor. A Variable Multiple Endmember Spectral Mixture Analysis (VMESMA) tool was used providing possibilities of multiple endmember unmixing, aiming to estimate the quantities and distribution of the remaining tailings material. A spectrally based zonal partition of the area was introduced to allow the application of different submodels to the selected areas. Based on an iterative feedback process, the unmixing performance could be improved in each stage until an optimum level was reached. The sludge abundances obtained by unmixing the hyperspectral spectral data were confirmed by the field observations and chemical measurements of samples taken in the area. The semi-quantitative sludge abundances of residual pyritic material could be transformed into quantitative information for an assessment of acidification risk and distribution of residual heavy metal contamination based on an artificial mixture experiment. The unmixing of the second year images allowed identification of secondary minerals of pyrite as indicators of pyrite oxidation and associated acidification. Keywords: Spectral Mixture Analysis, Soil Contamination, Acid Mine Drainage, Mining Accident.
1 INTRODUCTION The collapse of the tailings pond at the Aznalcóllar mine (Seville, Spain) in April 1998 had left several thousands of hectares in the Guadiamar floodplain contaminated with pyritic sludge with high concentrations of trace metals [1]. Pyrite oxidation is one of the most acid producing natural weathering processes, in which trace metals are mobilized and released into the river system and the groundwater. It is a complex process that proceeds rapidly, when pyrite is exposed to air. It produces in a first step a solution of ferrous sulphate and sulphuric acid. The dissolved ferrous iron continues to oxidise and hydrolyse producing additional acidity. During the oxidation process the pyrite transforms first to copiapite, then to jarosite, schwertmannite, ferryhydrite and eventually to hematite or goethite [2]. The wide-spread accidental contamination made a rapid remediation necessary, because the finegrained sludge started to oxidise rapidly. Few days after the accident an intense clean-up started and the sludge and the contaminated topsoils had been removed within 6 months after the accident using heavy machinery. However, the mechanical clean-up left considerable amounts of pyritic sludge in the ground, which was buffered with limerich material. This study aimed to develop and demonstrate methods for providing quantitative measurements of residual contamination using hyperspectral data in a form that can be understood by environmental officers and used directly in the remediation process. The conventional method for the assessment of the spatial distribution of contamination is by a raster sampling and a time-consuming laboratory analysis followed by geostatistical interpolation. However, the quality of the results is strongly depending on raster width and the spatial distribution of the contamination. Imaging spectrometers can be used as a rapid screening method to identify area-wide residual Presented at the 3rd EARSeL Workshop on Imaging Spectroscopy, Herrsching, 13-16 May 2003
333
surface contamination, even in very heterogeneous areas. This information is achieved by combination of abundances from spectral mixture analysis with results of a laboratory mixture experiment. Furthermore, hyperspectral data can be used for monitoring of oxidation/acidification processes by identification of the above mentioned iron-bearing secondary minerals, which are iron-rich and hydroxyl- and/or water bearing, making it possible to identify them on the basis of their diagnostic spectral reflectance features.
2 MATERIALS AND METHODS During two extensive field campaigns in May/June 1999 and June/July 2000 seven representative sites along the affected river catchments were selected. Spectral measurements using GER S-IRIS and ASD Fieldspec II resulted in a VIS to SWIR high-resolution library hierarchically organised of rocks, soils and vegetation. Furthermore, in both years a detailed soil sampling was carried out, collecting 214 samples in 1999 and 132 in 2000 from different depths. The soil samples were geochemically analysed in the laboratory for their element composition using standard methods (XRF, AAS) and spectrally measured. For a better understanding of the influence of the sludge on reflectance an artificial mixture series was set up, in order to assess under controlled conditions the influence of the sludge on the overall reflectance of the soils. Noncontaminated samples (i.e. not affected by the accident) of the three most important soil types in the area were used for this experiment to add sludge in increasing weight percentage from 0 to 100 percent. The imaging spectrometry data were acquired with the HyMap sensor on behalf of the JRC performed by DLR. The data takes covered the entire contaminated area and were accompanied by a radiometric calibration field campaign. The HyMap system provides 128 wavebands over the range 403 – 2480 nm with a spectral resolution of 13 – 17 nm. The data were delivered atmospherically and geometrically corrected by DLR according to the methods described by Schlaepfer & Richter [3] and Richter & Schlaepfer [4]. Spectral Mixture Analysis (SMA) is a widely used method to determine the sub-pixel abundance of vegetation, soils and other spectrally distinct materials that fundamentally contribute to the spectral signal of mixed pixels [5]. This is of particular importance to obtain quantitative estimates of distinct materials, which is a typical application of hyperspectral data. SMA aims to decompose the measured reflectance spectrum of each pixel into the proportional spectral contribution of so-called endmembers (EM). These EMs are known reflectance spectra considered to represent the spectral characteristics of the relevant surface components constituting the pixel surface cover proportional to their spatial occurrence (i.e. the area covered) within the pixel. The most widely used method for SMA consists in employing the same EMs to the whole image, and using all available EMs at the same time. In recent years, many authors have proposed and used a more complex model where both the number and the set of EMs vary dynamically on a per-pixel basis [6], which has become known as multiple endmember spectral mixture analysis (MESMA). The idea consists in restricting the large set of possible EMs to a small set of more appropriate EMs, which can be different for each pixel, thereby allowing an accurate decomposition using a virtually unlimited number of EMs. In this work we have used an improved strategy called Variable MESMA (VMESMA), which allows a segmentation of the image to increase the flexibility and accuracy [7]. Instead of reflectance spectra, both library and image spectra were standardized prior to the analysis. Standardised SMA is based on the same principles as conventional SMA, simply solving the problem using standardised coordinates, which introduces some differences in the equations. The formulation of standardised SMA is expressed as follows [7]: c
r = ∑ Ei f i + ε
(1)
i =1
where
r is the standardised pixel vector, E i represents the i-th standardised endmember,
f
i
is the proportion
ε
is the of such an endmember in the standardised coordinates, c is the number of components in the pixel and residual vector (expressed in standardised units). Standardised SMA consists of unmixing the pixel vector r using standardised EMs E i (i =1,…c) in order to obtain the proportions i . A constrained analytical estimator is used
f
according to which the sum-to-one condition in standardised SMA is expressed as follows [8]: c
fi
∑σ i =1
=
Ei
334
1 σr
(2)
σ E i and σ r represent the standard deviation of the vectors E i and r , respectively. Finally, the proportions in the original system, f i , are obtained using the expression: where
c
σr ) fi i =1σ E i
fi = ∑
(3)
This transformation ensures the sum-to-one condition and also preserves the positive proportions.
3 RESULTS AND DISCUSSION 3.1 Geochemical Analysis The results of the geochemical analysis revealed that after the end of the first clean up operation in summer 1999, the contamination is still well above background concentrations (Table 1). Moreover, there is a wide range of concentration levels, indicating a very heterogeneous situation. In the year 2000, the situation improved due to a repeated clean-up, but for some elements, particularly arsenic the situation remained critical. Table 1. Heavy metal concentration (average, range) of contaminated soils compared to background soils in June 1999 and the ranges of concentration for normal soils
Background
Contaminated
Normal*
[ppm]
Mean
Min
Max
Mean
Min
Max
Min
Max
As
16.75
8
27
61.26
7
442
1
20
Cd
0.43
0.05
1.88
1.26
0.05
14.8
0.1
0.6
Cu
50.65
18
178
120.44
17.5
521
2
40
Hg
0.06
0.01
0.29
0.45
0.01
13.9
0.02
0.2
Pb
52.65
18
221
202.21
17.5
3331.5
2
80
Sb
275.25
196
382
438.91
196
3362
0.2
10
Zn
191.69
98
748
380.44
94
3887
10
80
* according to Alloway [9]
3.2 Artificial Mixture Series The visual analysis of the measured soil spectra showed a large difference between contaminated and uncontaminated soils. Figure 1 shows the reflectance change for an uncontaminated soil with increasing sludge content. The most evident change was a strong decrease of overall albedo from a maximum reflectance of 53 percent to 14 percent. In particular, at longer wavelengths the spectrum levelled off strongly. Besides the general decrease in albedo, some changes in the absorption features are apparent. With the building up of a very wide absorption feature between 700 and 1400 nm the water and hydroxyl absorption features at 1400, 1900 and 2200 nm diminished and were almost extinguished at sludge concentrations above 50 weight-percent. Smaller absorption features, like the weak calcite feature at 2350 nm, disappeared even earlier.
335
0.80
0.60 Sludge Concentration [weight-%]
Reflectance
0% 10 % 30 %
0.40
50 % 75 % 100 %
0.20
0.00 500
1000
1500
2000
Wavelength [nm]
2500
Figure 1. Spectra of mixture series of soil sample AZ086 (10 nm)
3.3 Spectral Mixture Analysis The selection of appropriate EM is crucial for a successful application of SMA. It has to consider the changing spectral significance of EM as a function of the variability of the occurring surface materials, the spatial and spectral resolution of data and the thematic purpose of the study. For the extraction of the residual sludge signal it was necessary to separate sludge related spectral information from other ‘background’ information. Thus, typical spectra of green and dry vegetation and two different soils were selected from the spectral database as background information (Figure 2). The selection of sludge EM had to consider the potential secondary minerals of pyrite, like jarosite, copiapite, ferrihydrite and goethite. However, for the images of 1999, the oxidation was limited to areas with a constant water supply and for the major part of the contaminated area no considerable oxidation took place. Thus, a single spectrum of pure sludge was sufficient to represent the sludge fraction, which was also confirmed by the automated EM selection, which also identified one sludge spectrum besides soil and vegetation spectra. In the year 2000 other sludge related spectra had to be taken into consideration, because field and image assessment confirmed that the oxidation processes had produced various crusts on the soil surface. During the second field campaign, spectral measurements of the most abundant crusts were recorded from which the EM were selected (Figure 3). 2.00
standardised Reflectance
absolute Reflectance
0.60
0.40
0.20
0.00
0.00
-2.00
-4.00
0.40
0.80
1.20
1.60
Wavelength [µm]
2.00
2.40
0.40
0.80
1.20
1.60
Wavelength [µm]
2.00
2.40
Figure 2. Background EM: green vegetation (), dry vegetation («), soil 1 (U), soil 2 (²). Absolute reflectance (left), standardized reflectance (right)
336
2.00
standardized Reflectance
absolute Reflectance
0.60
0.40
0.20
0.00
0.00
-2.00
-4.00
0.40
0.80
1.20
1.60
2.00
2.40
0.40
Wavelength [µm]
0.80
1.20
1.60
Wavelength [µm]
2.00
2.40
Figure 3. Sludge and secondary mineral EM: pyritic sludge (), yellowish jarosite dominated crust («), bright gypsum crust (U). Absolute reflectance (left), standardized reflectance (right)
In the first unmixing step, the four background spectra were used to unmix the entire scene. The root mean squared error (RMSE) after the first unmixing clearly separated the sludge-affected areas with a high RMSE from the non-contaminated areas. The next unmixing step was performed on the segmented image. Areas outside the contaminated area were neglected, and in the affected area, a RMSE threshold was used to separate areas that were already sufficiently modelled from areas, which were sludge affected. The sludge abundance map for June 1999 (Figure 4) shows that the sludge abundances were still very high with an average abundance of 0.51 (calculated for areas with sludge abundances > 0). The results were obtained with only one sludge EM. In fact, tests with additional sludge EMs were not successful. This is a clear indicator that at this point in time, 13 months after the accident, oxidation of pyrite did not yet reach a high intensity and consequently it was possible to map the sludge using only the pyrite sludge EM. The retrieved sludge distribution corresponds well with the field observations and the geochemical analysis and reflects the discontinuous distribution pattern caused by the mechanical clean up. The same unmixing strategy was applied for the data collected in 2000. However, the situation had changed significantly. The remediation activities proceeded with a second cleaning phase in combination with fixation of the trace metals. This was achieved by augmentation of the pH-level of the soil through addition of lime-rich material. At the time of the second field and flight campaign the work was still ongoing. In many areas, particularly in the northern part, efflorescent crusting could be observed. According to Nordstrom [10], these crusts are most commonly formed during dry periods when evaporation promotes the rise of subsurface water to the uppermost soil surfaces by capillary action. As the water reaches the surface it becomes progressively more concentrated and finally precipitates various salts in efflorescence. The formation of these iron sulphate salts is an intermediate step, which precedes the precipitation of more common insoluble iron minerals such as goethite and jarosite. In order to account for this change in surface composition two new EMs representing these efflorescent crusts were included in the SMA modelling. The abundance map for July 2000 shows a considerable reduction of areas with sludge EM abundances compared to the abundances obtained for June 2000 (Figure 5). This reduction was achieved by the second remediation campaign. However, in areas where the remediation was not finished, higher abundances of secondary minerals were found. In these areas, jarosite was found to be abundant. The presence of gypsum is restricted to shallow depressions, in which more water gathered after rainfalls and the humidity was sufficient for the formation of gypsum when the water evaporates. However, these few millimetres thin gypsum crusts are only on the surface, below them secondary minerals are found. Thus, the presence of gypsum is on the one hand an indicator for buffering of acidity by the distributed material Ca-rich material; on the other hand it shows that there is still residual sludge in the soil, which produces acidity.
337
Sludge Abundance 0%
100 %
Figure 4. Sludge abundance map 1999. The affected area (black) superimposed on the HyMap false colour image for better orientation. Sludge abundance within the affected area is scaled from zero (black) to one hundred percent (white).
338
Jarosite
Pyrite
0
25
50
75
100
Gypsum
Figure 5. Mineral abundance map 2000. The mixtures of the different EMs can be derived from the colour coded ternary diagram. Red is the pyrite EM, green is the jarosite EM, blue the gypsum EM.
3.4 Quantification of Residual Contamination and Assessment of Acidification Risk The artificial mixture series of soils and sludge was initially prepared to gain insight in the influence of sludge on soil reflectance. However, it is also possible to regard these samples as a spectral mixture and apply SMA as described previously. By comparison of sludge abundance and known sludge concentration of the artificial mixture series a direct link can be established, which can be exploited further for the estimation of heavy metal concentrations from the image abundances. This is possible, because after the clean-up the residual sludge was worked into the soil by ploughing forming an intimate mixture very similar to the one of the mixture series. The artificial mixture series was prepared with three different soils and pure sludge. The spectra of the noncontaminated soils and the pure sludge were used as EM in the unmixing. The unmixing was performed with standardised variables, thus a shade EM was neglected. Figure 6 shows the obtained abundances with respect to the weight percentage of the sludge in the mixture. The clearly non-linear behaviour can be explained by the fact that the soil – sludge mixture is an intimate mixture, where the darker pyrite dominates because photons are absorbed when they encounter a dark grain [11].
339
100
Sludge [weight-%]
80
60
40
20
0 0
20
40
60
Sludge Abundance [%]
80
100
Figure 6. Comparison of sludge abundance derived from SMA of the mixture series and sludge weight percentage. A second order polynomial is fitted to the data (R2 = 0.973).
Knowing the relationship between sludge abundance and weight percentage offers the possibility for a quantitative assessment of the residual sludge and the related heavy metal contamination, because it allows deriving element concentrations and other parameters directly from the sludge abundance. The element concentrations are obtained by regression analysis of sludge abundance and respective elements (see also Figure 6). For lead (Pb), arsenic (As) and sulphur (S) the best fits were obtained with following relations: Pb = 2541.1 × (Abundance)2 + 148.24 × (Abundance) + 100.66
(4)
2
As = 328.18 × (Abundance) + 18.345 × (Abundance) + 24.809
(5)
2
Sulphur = 13.98 × (Abundance) + 0.8465 × (Abundance) + 0.3751
(6)
Based on the sulphur content also other information can be derived. Sulphur plays a key role in the assessment, because it is one of the two elements forming pyrite (FeS2), which is the main component of the sludge (75 – 80 weight-%), and it is the main source for acidity when pyrite oxidizes. The sulphur content is also used in conventional mining waste analysis for prediction of the geochemical behaviour of mining waste in order to identify wastes that are likely to be acid generating or susceptible to heavy metal leaching [12]. The total amount of pyrite in the uppermost soil layer was derived from the amount of sulphur, based on the molecular composition of pyrite, which consists of 53.45 % sulphur and 46.55 % iron. Finally, the amount of residual sludge after the remediation was calculated. The overall quantity in the first 20 centimetres totalled 148976 to 186220 tons. Compared to the amount of tailings that was spilled (between 1.3 and 1.9 million tons [13]), this means that after the first remediation campaign between 9.4 and 14.3 % of the released tailings remained in the area. The amount of residual sludge is an indicator for the quality of the first clean up. This information is very valuable for remediation planning, because it allows the estimation of the maximum potential acidity and the neutralisation requirements. The calculations of the maximum potential acidity are based on the overall reaction for pyrite oxidation [14]: FeS2 + 3.75 O2 + 3.5 H2O → Fe(OH)3 + 2 SO4 2- + 4 H+
(7) +
This means that each gram of sulphur in pyrite produces 3.06 grams of acidity (in the form of H and SO42-). The two additional cations H+ were derived from the hydrolysis of ferric iron with water to form solid ferric hydroxide. The calculations of the neutralisation requirements are based on the following assumed stoichiometry [14]: FeS2 + 2CaCO3 + 3.75O2 + 1.5H2O → 2SO42- + Fe(OH)3 + 2Ca2+ + 2CO2
(8)
For each mole of pyrite that is oxidised, two moles of calcite are required for acid neutralisation, or on a mass ratio basis, for each gram of sulphur present, 3.125 grams of calcite are required for acid neutralisation. These functions
340
were applied directly to the sludge abundance image derived from unmixing of the HyMap images from 1999 (Figure 7).
Arsenic Contamination 0
As [ppm]
370
Lead Contamination 0
Pb [ppm]
2500
Residual Sludge 0
Weight [g/kg]
140
Potential Acidification Risk 0
H2SO4 [g/kg]
120
Calcite-Buffering Need 0
CaCO3 [g/kg]
90
Figure 7. Contamination parameters derived from the sludge abundance image for June 1999: arsenic and lead contamination in ppm, residual sludge in g/kg, maximum potential acidification after full oxidation of pyrite in g/kg, and neutralisation requirements for full oxidation of pyrite in g/kg
3.5 Validation The transformation of abundance information into heavy metal concentrations allows validation of the obtained results with the results of the geochemical analysis of soil samples collected in the field. It is based on the 56 surface soil samples collected during the fieldwork, and results of the analysis of 180 samples taken by the Consejeria de Mediomabiente in parallel to our fieldwork. Lead (Pb) and Arsenic (As) were selected for validation, because they were still present in high concentrations and represent heavy metals with a different mobility. For lead, a total of 236 sampling points was available for validation, but 128 samples were excluded, either because no sludge could be detected in the vicinity or the concentrations measured in the laboratory were below 100.66 ppm. This value is the detection limit in the remote sensing maps, which was imposed by the offset in the regression equation. Figure 8 shows the comparison between laboratory and image derived lead concentration. After excluding 5 outliers, a coefficient of determination of 0.729 was obtained. However, the data are strongly controlled by the skewed distribution of the data. Generally, the model works better for medium and high concentrations; e.g., when considering only concentrations above 300 ppm, the coefficient of determination improves (R² = 0.852).
341
500
6000
Y = 0.367394 * X + 69.9218 R² = 0.732 400
Y = 0.526836 * X + 439.639 R² = 0.729 Image Pb [ppm]
Image As [ppm]
4000
300
200
2000
100
0
0 0
2000
Lab Pb [ppm]
4000
0
6000
200
400
600
800
1000
1200
1400
Lab. As [ppm]
Figure 8. Comparison of Pb (left) and As (right) concentrations for samples analysed in laboratory with concentrations derived from HyMap data. Outliers (U) were not included into the regression. Note that the scales of the axes for arsenic are chosen differently for illustration purposes.
For As, 122 samples could be used after removing samples without sludge abundances in the vicinity or with concentrations measured in the laboratory below the detection limit of 24.8 ppm for As. The range of arsenic concentrations in the laboratory is much wider than in the image-derived data (Figure 8). The concentration maximum reaches 282 ppm for the image data and 2649 ppm for the laboratory data. Therefore, the match of high laboratory data with image data is very poor. In the low and medium range the match is much better; if laboratory concentrations above 400 ppm are left out, a strong coefficient of determination (R² = 0.732) can be obtained. Both examples show that there is a strong correlation between laboratory samples and field data particularly for the medium concentration ranges. There is generally a tendency to underestimate the contamination level due to the fact that the image data are averaged over the area of one pixel; locally high concentrations are reduced due to the averaging with less high concentrations within one pixel. However, the spatial information derived here is not meant to substitute conventional geochemical analysis methods (in fact it is based on such information), but to provide area-wide spatial information on a per-pixel basis, which is one of the major advantages of remote sensing data. Imaging spectroscopy can provide information on heterogeneous environments, where an in situ sampling would need an enormous number of samples to cover the variability of the entire area.
4. CONCLUSION This study demonstrated successfully the application of combined field and imaging spectroscopy for quantification of residual heavy metal contamination and for monitoring of the oxidation process. The mapping of residual sludge and sludge derivatives was achieved with a variable multiple endmember SMA system (VMESMA), which offered greater flexibility and new possibilities to get improved performances and more accurate interpretation of the unmixing results. The iterative approach of VMESMA starting with a simple endmember set of background material worked very successful; the delineation of the affected area using the RMS error after the first unmixing was almost congruent with the GIS layers of the affected area. The standardisation of spectra was very helpful for the detection of sludge, which is a very dark and almost spectral featureless material. In 1999 the sludge could be characterised using a pure sludge EM, because the oxidation did not start yet. The obtained sludge abundances were in good agreement with the field observations. In the year 2000, the remediation efforts of the CMA resulted in a drastic reduction of the surface contamination, which is clearly reflected in the unmixing results. However in areas, where the remediation was not concluded yet the oxidation of the residual sludge had started. The complex weathering lead to an efflorescence of easily dissolved salt crusts and more stable secondary minerals. Using spectra of oxidation indicators like jarosite as EM, the spatial distribution of the oxidation became evident highlighting the usefulness of imaging spectroscopy for monitoring purposes. The abundances are already useful information for the remediation process. However, the transformation of relativ abundance information into quantitative information through the artificial mixture series converts them into
342
quantitative information. It is the same type of information, which is obtained by conventional analysis, and as such easily understandable by environmental officers. However, the spatial resolution is much more detailed and offers faster and more exact remediation planning.
ACKNOWLEDGMENTS We would like to express our gratitude to the head of the IES-Land Management Unit (IES-LM) Mr. Meyer-Roux and to the head of the IES Soil & Waste Unit (IES-SW) Mr. Bidoglio for supporting this study as part of the JRC institutional work program. M. D’Alessandro, Fabrizio Sena and the members of the IES-SW inorganic laboratory team provided invaluable support during the field campaign and through the laboratory analysis of the soil samples. Furthermore, we want to thank the Consejeria de Medioambiente of the Junta de Andalucia in Seville for their support, particularly for providing the geochemical analyses for the validation of our results.
REFERENCES [1] GRIMALT, J.O., FERRER, M., AND MACPHERSON, E., 1999: The mine tailing accident in Aznalcóllar. Science of the Total Environment, 242(1-3), pp. 3-11 [2] SWAYZE, G.A., CLARK, R.N., PEARSON, R.M., AND LIVO, K.E., 1996: Mapping Acid-Generating Minerals at the California Gulch Superfund Site in Leadville, Colorado Using Imaging Spectroscopy. Proc. of the 6th AVIRIS Airborne Geoscience Workshop. http://popo.jpl.nasa.gov/docs/workshops/96_docs/34.PDF [3] SCHLAEPFER, D. AND RICHTER, R., 2001: Geo-atmospheric processing of airborne imaging spectrometry data. Part 1: parametric orthorectification. Accepted by International Journal of Remote Sensing, ftp://ftp.geo.unizh.ch/pub/rsl2/paper/2001/IJRS_2000_parge.pdf [4] RICHTER, R. AND SCHLAEPFER, D., 2001: Geo-atmospheric processing of airborne imaging spectrometry data. Part 2: atmospheric/topographic correction. Accepted by International Journal of Remote Sensing, ftp://ftp.geo.unizh.ch/pub/rsl2/paper/2001/ijrs01_atcor.pdf [5] ADAMS, J.B., SMITH, M.O., AND JOHNSON, P.E., 1986: Spectral mixture modeling: A new analysis of rock and soil types at the Viking Lander 1 Site. Journal of Geophysical Research, 91, B8, pp. 8098-8112 [6] ROBERTS, D.A., SMITH, M.O., ADAMS, J.B., AND GILLESPIE, A.R., 1991: Leaf Spectral Types, Residuals, and Canopy Shade in an AVIRIS Image. Proc. of the 3rd AVIRIS Airborne Geoscience Workshop, http://popo.jpl.nasa.gov/docs/workshops/91_docs/7.PDF [7] GARCÍA-HARO, F.J. AND SOMMER, S., 2001: Un entorno integrado, operativo y eficiente para el análisis y estimación de parámetros en teledetección. Teledetección, Medio Ambiente y Cambio Global. Proc. IX Congreso Nacional de Teledetección, Universidad de Lleida, Editorial Milenio, pp. 495-499 [8] GARCÍA-HARO, F. J., GILABERT, M. A., AND MELIÁ, J., 1999: Estimation of endmembers from spectral mixtures. Remote Sensing of Environment, 68, pp. 237-253 [9] ALLOWAY, B.J., 1999: Schwermetalle in Böden, Analytik, Konzentrationen, Wechselwirkungen. Springer, Berlin, Heidelberg, New York [10] NORDSTROM, D.K., 1982: Aqueous pyrite oxidation and the consequent formation of secondary iron minerals. In: Kittrick J A, Fanning D S, and Hossner L R (eds.): Acid Sulfate Weathering, Soil Science Society of America Special Publication No. 10, Madison, Wisconsin, pp. 37-56 [11] CLARK, R.N., 1999: Spectroscopy of rocks, minerals, and principles of spectroscopy. In: Rencz, A.N. (ed.): Remote sensing for the earth sciences: Manual of remote sensing, 3rd ed., vol. 3, Wiley & Sons, New York, pp. 3-58 [12] TREMBLAY, G.A. AND Hogan, C.M., 2001: MEND manual: volume 1, summary. Ottawa, Ontario, Canada [13] ERIKSSON, N. AND ADAMEK, P., 2000: The tailings pond failure at the Aznalcóllar mine, Spain. Proc. of the 6th International Symposium in Environmental Issues and Waste Management in Energy and Mineral Production, Calgary, Alberta, Canada, 30 May – 2 June 2000, http://www.mineralresourcesforum.org/workshops/regulators/2000/docs/LosFrailes.pdf [14] CRAVOTTA, C.A., BRADY, K.B.C., SMITH, M.W., AND BEAM, R.L., 1990: Effectiveness of alkaline addition at surface mines in preventing or abating acid mine drainage: part1, geochemical considerations. Proc. 1990 Mining and Reclamation Conference and Exhibition, West Virginia University, Charleston, West Virginia, pp. 221-226
343
