Research Articles

Hg-Contents in Tailings of Gold Mine Area

Estimation of Mercury Content in Tailings of the Gold Mine Area of Pocont, Mato Grosso, Brazil Wolf von Tiimpling jr. 1, Peter Zeilhofer2, Ulrich Ammera, J/irgen Einax 3 and Rolf-D. Wilken t tlnstitute for Physical and Chemical Analytics, D-21502 Geesthacht 2Chair of Landuse Planning and Nature Conservation, University of Munich, D-85354 Freising 3Institute for Anorganic and Analytical Chemistry, (Friedrich Schiller University) Lessingstrafle 08, D-07743 Jena

Abstract As in many other parts of the world, gold is produced in the surface mining region of Pocont, Mato Grosso, Brazil, using mercury. The goal of this investigation was to estimate the amount of mercury in certain tailings and to determine the area of the land that has been contaminated by the gold mining operations. Mercury concentrations from 2 to 495 ng/g (dw) were determined in the tailing materials. It was observed that only isolated sites were acting as central points of contamination. Using digital Landsat satellite data (Thematic Mapper) and aerial photos, the sites degraded by the mining were classified, and their total area was estimated to be 12.3 km 2 in the region of Pocon& It was estimated, that 4.9 km 2 were occupied by the contaminated tailings. The mean height of the pile slags was determined to be 4.5 m. From the experimentally calculated average density of the material in the tailings, 2.01 g/cm 3, the total mercury content in the piles of tailings was estimated to be 1600 • 350 kg.

Key words: Gold mine; mercury content in tailings; contamination through gold mining operations, Landsat satellite data

Fig. 1: Investigated area

1

Introduction

On a global scale, the process of amalgamation is commonly used for gold extraction. In Brazil, excessive quantities of mercury are used in the gold mining regions (about 200 t alone in 1992). One of these areas is located around Pocon6 in the south west of Mato Grosso in central South America near 56040 , western longitude and 16010 , southern latitude (--* Fig. 1). On the northern border of the Pantanal wetland, primary deposits in folded Precambrian rocks have been mined since the 18th century. In the 1980s, about 5000 people were employed in about 130 gold mines called "Garimpos". It is estimated, that between 10 and 15 t of mercury [2, 3, 6] were used during this time to amalgamate the preconcentrated gold particles for the separation of the gold from the slag according to specific weight. The waste contaminated with mercury from the separation process, and combined with slag from the preconcentration process, is collected as tailings. During heavy tropical rainfalls, the permeability of the soils is insufficient to store the rain water. The heavy surface runoff results in intensive erosion processes. Contaminated material from tailings thereby reaches the surface water bodies and river sediments. Temperatures up to 60 ~ in the upper centimetres of the surface tailings support the gaseous emission of mercury. PreESPR-Environ. Sci. & Pollut. Res. 2 (4) 225-228 (1995) 9 ecomed publishers, D-86899 Landsberg, Germany

dominant northern winds in the region can cause an additional pathway for mercury contamination in the Pantanal. To evaluate the actual potential of mercury contamination in the tailings of the Pocon~ region, mercury analyses of the material in the slag piles as well as an evaluation of aerial photos and data from the Landsat Thematic Mapper have been performed. In addition, the distribution and degree of mercury contamination has been described and the maximum volatile mercury content in the railings has been estimated.

2 2.1

Experimentals Estimation of the total mercury content

The base area of the tailings, the average tailing height, the average mercury concentration above the natural background and the mean density of the sediments were used for the estimation of the total mercury content. In reality, the forms of the tailings are similar to frustums or pyramid stupid and posses with a mean gradient of 39 ~ Neither field work nor image analysis allow the counting of 1 From cooperation between Max-Planck-Institut Rir Limnologie, AG Tropen6kologie, Pl6n, and Universidade Federal de Mato Grosso (UFMT), Cuiabfi, MT, under the Govermental Agreement on Cooperation in the field of Scientific Research and Technological Development between Germany and Brazil.

225

H g - C o n t e n t s in Tailings of Gold Mine Area

tailing numbers or the definition of one typical degree of "mean" largeness. Therefore, calculating the volume of tailings by approximating their form as a cuboid was the most "accurate" manner with data. This includes a methodical overestimation of the tailing volume. Working on an assumption of different realistic base forms and largesse of the slag piles as well as different numbers of tailings in the region, the overestimation varies between 10 and 18 %. This was taken into account in the estimation term (1) as an average relative error: rnHg = C--He " h " d " A

mHg CHg

" 0.86

dX

estimated mercury content in the tailings average mercury concentration mean height of the tailings mean density of the heaps material mean total surface area of the tailings

2.2

Sampling and analyses of total mercury

(1)

Ten "Garimpos" of the region were selected for sampling. Ten samples were taken with the "Pfirkhauer" corer out of three different layers (0 - 5 cm, 50 - 55 cm, 100 - 105 cm) from the tailings of these "Garimpos" in each case. The samples were stored in polyethylene bags at a temperature of 4 ~ Approximately 1 g of the sample was used for analyses. In addition, aliquots were used for the determination of the dry weight and the relative humidity of the sediments. Because the total mercury content should be determined in the waste, and the sediment is already homogenised by the gold extraction process, the undivided total fraction was used. To determine the total mercury concentration in the sediments, a digestion in quartz tubes was performed with concentrated nitric acid. According to the method of BLOOM and CRECELIUS [1], all samples were analysed with the Atomic Fluorescence Spectrometer (AFS) (Brooks Rand Ldt. company) for total mercury concentration. To confirm the accuracy of digestion and measurement, an inter calibration test was performed German laboratories. The average total mercury concentration above the natural background was calculated using results of the single measurements. To confirm the representativeness of the average value, the mean calculation was repeated ten times whereby 50 samples were randomly selected and dropped from the calculation in each case.

2.3

Determination of the mean density of the tailings

For the estimation of the mean density d of tailing material, 60 samples, each with a volume of 30 cm 3 (twO samples of each layer per "Garimpo'), were collected and dried at 105 ~

226

Research Articles 2.4

Investigation of the tailing heights

To calculate the volume of the waste material, 60 tailings were levelled out by trigonometric measurement using the "Sunto" Hypsometer PM-5/1520P and the average height was thereby determined.

2.5

Estimation of the tailing surface

For mapping the total area covered by "Garimpos", a subset of 1100"1300 pixels of the Landsat T M scene 227/71 (acquisition date 7 / 1 6 / 1 9 9 3 ) was used. (Processing of digital Landsat Thematic Mapper (TM) data was performed at the institute of Landuse Planning and Nature Conservation of the Ludwig-Maximilians-University, Munich, using an ERDAS PC system.) The scanned area of about 1287 km 2 covers the central part of the mining region. The evaluated T M bands #1 - #5 and #7 were transformed to reflectance and corrected for atmospheric effects by the regression intersection method according to CRIPPEN [4]. To establish the "Garimpo" areas, the image data were classified by a supervised iterative algorithm. Streets and settlements were interactively digitised on the monitor and combined with the results of the classification. One thousand randomly selected reference pixels of known land use were used to verify classification results. The geometric resolution of 30 m of the T M sensor and the similar reflection characteristics of the mine and the tailing material, however does not allow the identification of only the contaminated tailings in the gold mine areas. Therefore, during a flight in February 93, about 40 colour off-nadir aerial photographs at a scale of about 1:1000 were taken with a conventional camera. Analysing the photographs, permits an estimate of the ratio beetween the regular-formed elevated, contaminated slag piles and the non-contaminated, abnormally-shaped, sometimes waterfilled mining of the "Garimpos". With this information, it was possible to estimate the total surface of the tailings as a part of the total "Garimpo" area.

3 3.1

Results and Discussion Concentrations of mercury in the tailings

Analyses of the rocks used for the gold extraction showed a natural mercury concentration between 2 and 6 ng/g (dw). The determined total mercury concentrations of the tailing material ranged from 2 ng/g (dw) to 495 ng/g (dw). A comparison of the frequency distribution of observed mercury concentrations with the natural background level shows that large areas are not or are only slightly contaminated by mercury (--" F i g . 2 ) . Strongly contaminated areas are distributed locally and diffusely. A distribution trend could not be observed. Those results described above are performable by the gold mining process. The waste from the preconcentration process is deposited in tailings. The mercury contaminated waste is deposited in the slag heaps only at irregular intervals. Hot spots in the

ESPR-Environ. Sci. & Pollut. Res. 2 (4) 1995

Research Articles

Hg-Contents in Tailings of Gold Mine Area

Gatim~ w~l

130

80 ~

Ganmpo d~

120 - ~

110

,60 }-

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

300 sarnples

~

--ili i ii i L!I

20

01,,,~ 0

....

40

,,,

80

.....

120

,

.....

160

200

TI

24(}

. . . . . . . .

28[)

320

360

400

440

480

520

-

Scab OefOre~ed

lqo

- - - > - , - Pastum

~I

-- x

- Burned scrub

-.~-

waler body

80

. .-'" .- " .

"

~ 70 ~ 60 N 50

40 +

3 0 42 0 410+ 0 400

I

J

I

I

500

600

700

800

I

I

I

I

I

I

I

t

I

-I"-i--V-r-"P

900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300

wavelength

M e r c u r y c o n c e n t r a u o n in ng/g ( d w )

[nm]

Fig. 3: Grey level at different wavelength for different landscapes Fig. 2: Mercury concentration in the tailings

tailings with up to 100 times higher mercury contents compared to the background levels are the result. The amount of the pore water in sediment samples collected at the end of the dry season in August 1993 revealed a significant increase in the depth. During the dry season, a complete drying up of the surface horizons takes place. The layer deeper than one metre with relative water contents of more than 25 % restrict the mercury and associated species from becoming volatile. Kinds of diffusion processes in lower horizons were therefore disregarded by the following estimations of the volatile mercury potential. The total average mercury concentration above the natural background was calculated to be 41 ng/g (dw). To confirm the representativeness of the value, the mean calculation was repeated ten times whereby 50 samples were randomly selected and dropped from the calculation each time. A mean relative error of 5 % for the calculated average was determined. The t-test showed that there is no significant difference between the calculated averages in a 99 % interval. This is proof for the hypothesis of a statistically sufficient representative sampling.

of water content in the soil surface. Furthermore, the spectral signature of some "Garimpos" is similar to dry pastures and deforested areas. This is a potential source of misclassification. The separability of these classes was improved by using the 'Normalized Difference Vegetation Index' (NDVI) which is calculated by the T M bands #3 and #4 as follows

(2): ( ( T M 4 - T M 3) N D V I = "(TM 4 + T M 3) +1) 9 127

(2)

According to COHEN [5], this index is closely correlated with the amount of green vegetation cover and is less influenced by soil humidity than the infrared T M bands #4, #5 and #7. Figure 4 shows the spectral mean and two standard deviations of some selected training sites in the NDVI.

NDVI

l

]

!

!

[

I

I

-

-

l

l

l

l

Gadmpo 1 Gadrnpo 2

~

I

Gadmpo 3

3.2

Density of the material and the mean tailing height

Similar densities are observed in all samples. That is, the relative error was 2 % from the calculated mean value of 2.01 g/cm 3. The comparatively high density of this material (cf. SCHEFFER and SCHACHTSCHABEL[7]) is caused by the gold mining process whereas a preconcentration of the ore in the fine crushed waste as an suspension is also deposited in the railings. Designations of the heap heights gave an average value of 4.5 m, with a mean relative determination error of 10 %.

Pasture 1

Deforested Scrub

Mapping of the "Garimpo" area with satellite data

To determine the suitable T M bands for the classification, the statistics of the spectral signatures were analysed. Figure 3 shows the mean values for some selected training sites in the original TM bands. Field studies showed that the large spectral difference of the "Garimpo" signatures can be explained with the variation

ESPR-Environ. Sci. & Pollut. Res. 2 (4) 1995

.,

!

I--i-~--I

Deforested Forest

Burned scrub Waterbody

3.3

,

Pasture 2

Scrub Flooded Savanna

l--II~

Ill

l--I--I

I

Z

!

Fig. 4: Spectral mean and two standard deviations of some selected training sites in the N D V I

227

H g - C o n t e n t s in Tailings of Gold Mine Area

Research Articles

The separation of "Garimpo" areas from pastures, in particular, could be improved. In some cases, the classification errors between deforested scrub or forest areas and "Garimpos" could be reduced. In a first classification step, all areas covered with vegetation were separated by thresholding using the NDVI.

error of 10 % was assumed. (Large scaled ortho-photos, which were not available, would facilitate a more exact determination of the area covered by tailings.)

Recently burned areas which reveal similarly low values in the NDVI as compared to those in the "Garimpos", were separated with the bands of the VIS ( ~ Fig. 3). Sparsely vegetated water bodies reflect very low levels in the middle infrared range: Thus, bands #5 and #7 could be used for thresholding ( ~ Fig. 3). In a second step, all unclassified pixels were divided in the classes "pasture", "flooded grassland" (ft. grassl.), "savanna", "forest" and "water body" using a maximum likelihood classifter. Considering the spectral similarities to other Landuse classes, the classification accuracy of the "Garimpo" areas can be judged as being satisfactory (--' Table 1). The classification of some inundated "Garimpo" areas as "water bodies" is unproblematic because these parts are normally uncontaminated mining holes filled with ground water. During the evaluation of the aerial photos, this methodical classification error was taken into account. Accidental classifications of roads or the interior of settlements were reduced by overlaying the classification results with the generated polygons of the interactive digitising. Table 1: Classification accuracy of randomly selected test sites (1000 reference pixels). Overall classification performance is 82.7 % Class

Reference pixel

Correctly classified

Producer's accuracy

74

65

87.8 %

312

259

83.0 %

26

17

65.3 %

Savanna

311

234

75.2 %

Forest

150

136

90.7 %

Burned area

24

17

70.1%

Water body

11

7

63.6 %

Settlement

76

76

100.0 o/0

Roads

16

16

100.0 %

Garimpo

Pasture Flooded grassland

Figure 6 shows the classification result for the "Garimpo" areas of the central part of the mining region. A "Garimpo" area of 12.3 + 1.5 km 2 was calculated with a relative error of 12.2 %. Especially with regard to coarse scale monitoring in regions with a poor map base, the evaluation of Landsat T M data for mining activity seems to be an appropriate tool. 3.4

Evaluation of aerial photos

The analysis of aerial photos demonstrated a proportion of the tailings to the total "Garimpo" area of 40 %. A relative

228

3.5

Estimation of the total mercury content

Using the above described method for estimation, the total mercury content above the natural background in the tailings of the gold mining region of Pocon6 was calculated to be 1600 _+ 350.kg. An estimation of the maximum volatile mercury content in the upper horizons of the tailings in this region were shown to be 400 + 100 kg and as determined in the atmosphere of the Pocon6 region, could be one of the reason for the higher mercury concentrations, compared to the natural background.

4

Conclusions

From the 10 - 15 t mercury used in the last ten years in the region of Pocon6, only about 1.6 t have recently been stored in the tailings. This result shows that other pathways of mercury than the deposition in the tailings also have to be taken into account. Because of the diffuse distribution of mercury hot spots in the tailings, an efficient decontamination seems to be impossible in certain sites. Assuming the continued use of mercury for amalgamation, only the separated deposition of uncontaminated and contaminated material can be the base of a possible decontamination. The mercury potential in the upper layers of the tailings is one reason for the significantly increased mercury concentration in the atmosphere of the study area.

5

References

[1] N. BLOOM;E. A. CRECELIUS:Determination of Mercury in Seawater at Sub-Nanogram per Liter Levels. Mar. Chem 14 (1983) 49 - 59 [2] CENTRO DE TECNOLOGIA MINERAL(CETEM): Relat6rio anual. Rio de Janeiro (1989) 3 8 - 51 [3] CENTRO DE TECNOLOGIA MINERAL(CETEM): Pocon6, um Garimpo de Estudos do Impacto Ambiental do Garimpo, S6rie tecnologla ambiental, Rio de Janeiro (1991) 9 - 21 [4] R. E. CII'PEN: The regression intersection method of adjusting image data for band rationing. International Journal of Remote Sensing 8(2)(1987) 1 3 7 - 1 5 5 [5] W. B. COHEN: Response of Vegetation Indices to Changes in Three Measures of Leaf Water Stress. Photogrammetric Engineering & Remote Sensing 57(2) (1991) 1 9 5 - 2 0 2 [6] L. D. LACERDA;W. C. PFEIFFER;R. V. MARINS;S. RODRIGUES;C. M. SOUZA;W. BASTOS:Mercury Dispersal In Water, Sediments and Aquatic Biota of Gold Mining Tailings Drainage in Pocon6, Brazil. Water, Air, Soil Poll. 56 (1991) 7 8 5 - 7 9 6 [7] F. SCHEFFER;P. SCHACHTSCHABEL:Lehrbuch der Bodenkunde.

Enke, Stuttgart, (1989) 180-190 Received: June 12, 1995 Accepted: September 6, 1995

ESPR-Environ. Sci. & Pollut. Res. 2 (4) 1995