Hyperspectral Sensing to Boost AMD Monitoring in Post-Mining Scenery

The evolving hyperspectral sensors have become a big ally for a large range of applications in remote sensing for the monitoring of a variety of earth ecosystems and processes (natural and anthropogenic). The processes occurring within the mining life cycle are not an exception. Hyperspectral sensors have been widely used in a large number of applications ranging from exploration, operation and post-mining. In this work, the authors explore particularly the use of hyperspectral methods to contribute to the monitoring of one of the most important environmental phenomena that many mining operations might face: acid mine drainage (AMD). Failure of accurately monitoring and remediation of such complex, leads to long-term impacts on ecosystems and human health, in addition to significant financial consequences and reputational damage to operators. Hyperspectral imagery represents one solution to enhance the quality of classical geochemical analyses in post mining-related contaminated scenarios, which can increase the overall accuracy of the monitoring, allowing frequent and multi-temporal observations to detect risk areas and take fast corrective actions.

Environmental monitoring of AMD

Acid mine drainage (AMD) is an environmental phenomenon that can occur either by the natural exposition of sulfate metals to weathering conditions or as a consequence of certain mining activities. Lottermoser (1) defines AMD as a process whereby low pH mine water is formed from the oxidation of sulfide minerals. These acidic and metal-enriched waters can negatively affect the natural ecosystem’s quality and aquatic life. Mainly impacted areas are rivers, lakes, estuaries, and coastal waters. Its advancement can take years or decades and can continue spatially increasing for centuries (1). Therefore, such an environmental problem needs to be carefully monitored and ideally remediated.

Several efforts have been applied in order to monitor the spatial distribution of contamination by AMD, commonly involving systematic sampling and laboratory analysis of stream sediment followed by interpolation of the results in assembled distribution maps (2, 3) however, such approaches can be time-consuming, costly, and with limited spatial coverage.

The environmental monitoring of such complex and diverse adverse effects on earth ecosystems requires frequent and multi-temporal observations. Active control can serve as an effective method for successful conservation or rehabilitation of natural systems. In this sense, remote sensing tools have been widely used in many environmental investigations since the technique enables the use of digital imaging sensors to reveal key information from a distance, typically from satellite or aircraft (4). Thus, traditional monitoring studies based only on certain ground-sampling locations can be expanded to large areas from derived aerial-image products. In general, optical spectral analysis refers to the measurement of matter-light interactions as a function of their energy. More specifically, this comprehends any radiation that is emitted, reflected or transmitted from the investigated target (5). The development of new generations of sensors made it possible to examine processes on earth, beyond the visible spectrum of the human eye. Commonly, these devices can acquire data in different wavelength ranges – from the ultraviolet to the far-infrared spectrum of electromagnetic radiation – and have evolved from spectral over multispectral to hyperspectral sensors for different kinds of earth’s surface investigations.

The trend toward progress for higher spectral resolution (hyperspectral remote sensing) has grown in the last decades. While the majority of space-borne sensors traditionally used for geologic remote sensing like ASTER or Landsat contains information in only a few wavelengths or bands, hyperspectral sensors are able to provide a continuous spectrum for each pixel of the dataset (6). Currently, hyperspectral sensors are employed in a wide range of spatial dimensions (scales) according to the platform used for data acquisition, e. g., satellite, airborne, up to lab-scale sensing for detailed-mineralogical analyses (Figure 1).

The emergent use of unmanned aerial systems (UAS), like multi-copters, and new-generation lightweight hyperspectral sensors have become a tool to collect data at a higher spatial resolution than some of their aircraft and satellite counterparts, resulting in greater precision (higher spatial resolution of a scene and enabling the investigation of up to a few centimeters sized pixels) (7). In this article, some studies will be reviewed that have took advantage of hyperspectral imaging to monitor AMD occurrence, mineralogy and related geochemistry.

From spectral to hyperspectral sensors

The main purpose of hyperspectral remote sensing – also known as imaging spectrometry or imaging spectroscopy – is to measure quantitatively the components of the Earth System from calibrated (radiance, reflectance or emissivity) spectra acquired as images in many, narrow and contiguous spectral bands (6). Hyperspectral sensors can capture data from the visible through the near-infrared wavelength ranges over a determined terrestrial surface of the earth. Collected data results in a three-dimensional data-cube composed of a set of pixels represented as vectors, containing the measurement corresponding to a specific wavelength (8). This provides the opportunity to query a plottable spectral signature for each spatial position on a surface. The accompanying amount of information results in much larger data sizes compared to polychromatic or multispectral imagery (9). The vector size is equal to the number of bands or spectral channels. In opposition to multispectral data, which usually acquire up to tens of bands, hyperspectral data channels are able to collect several hundreds of contiguous bands along the spectral axis (6).

Regardless the scale of acquisition, hyperspectral sensors bring higher spectral resolution, in comparison to multispectral sensors, offering higher accuracy to detect targets and characterize earth surface processes. In Figure 2, is possible to distinguish the differences between a common Red Green Blue (RGB) composite, a multispectral dataset and the hyperspectral.

The visualization format of any spectral dataset is similar, regardless the covered wavelength range, scanned specimen or area, and the spectral process underlying. A spectral imaging dataset is composed by three dimensions with at least one, even indistinct, value defining the measured signal intensity along at least two spatial and one spectral axis (9).

Multi-scale approach in post-mining applications

Mining wastes are quite heterogeneous compared with other industry sectors due to their quantity, mineralogical formation, and their properties. It varies depending on the mineral preparation and enrichment process applied. Waste in mines is usually stored in dumpsite and slurry ponds, while it is stored in some mineral sites in the form of post-leaching ore piles. As introduced early AMD can occur in these waste sites and if superficially dumped, when iron sulfide in coal mines or sulfur in base metal mines, can undergo into oxidation conditions (10). With the removal of ore from the ground exposure of sulfides to water and oxygen in air takes place, in turn, the oxidation processes of pyrite FeS₂ associated with iron, coal, and sulfur deposits can produce an acidic environment (1).

Particularly, the visible to shortwave infrared electromagnetic range has been widely used to monitor AMD mineralogy at mining surroundings since iron and also REE present strong and narrow absorption features in the visible to near infrared (VNIR). Mine waste dumps, pit-lakes, stockpiles and tailings generally contain high dissolved iron and sulfate content normally associated with this kind of metalliferous drainage, which makes possible to provide qualitative and (semi-)quantitative information on the composition, characteristics and spatial distribution of AMD processes. The characteristics of drainage waters may present high concentrations of metals and ions such as iron, manganese, aluminum, and sulfate. Elements like zinc, cobalt, lead, chromium and copper are commonly found in trace concentrations (11). These elements react with the surrounding environment, and in conjunction with other abundant ions, lead to the precipitation of a broad list of secondary minerals, which are not exclusive to mine tailings and AMD waters, having also been found in high saline environments regardless of pH values (12).

Satellite-airplane scale

Several studies have shown the benefits of remote sensing data for many environmental monitoring purposes. In relation to AMD, some studies have demonstrated the feasibility to use field and imaging spectroscopy for the detection of minerals containing metals as contamination proxies in mining areas (2, 13). Another attempt to map iron-bearing minerals with satellite data was performed by Swayze et al. (14) including validation studies with XRD and field spectroscopy. Montero et al. studied the characteristics of waste rock associated with acid drainage for protecting water reservoirs (15), while Sares et al. focused on indirect pH estimations of an AMD-stream by identifying iron-bearing minerals precipitated on the stream bed (16).

Most recently, hyperspectral sensors have been used in the study of mine tailings using airborne platforms (17, 18). These studies focused on the responsible minerals of acid formation in tailings and the distribution of the secondary minerals, e. g., jarosite, ferrihydrite, goethite/hematite, as indicators of the degree of environmental pollution using reflectance spectroscopy (19). Quick mineral diagnosis of short-lived thin-crusts concentrating metals by means of high spectral resolution imagery has been gathered in a spectral library for AMD minerals by Crowley et al. (20). Over river sediments and vegetation, visible-near-infrared spectroscopy has been researched by Clevers und Kooistra (21).

Each mineral provides a unique spectral signature that allows distinguishing between them. In this sense, Figure 3 shows reflectance for four of the main distinctive iron-secondary minerals related to AMD production (goethite, jarosite, hematite, and schwertmannite).

Spectral curves reveals distinctive absorption features depths of each iron mineral. Two regions are gray-shaded in Figure 3 to analyze the shapes and wavelength positions of each mineral in the charge transfer (ligand to metal charge transfer) transition and those triggered by the crystal field effects (transitions of electrons from lower to higher energy states) (22). Hematite characteristically has a narrower absorption at wavelengths surrounding 880 nm, while goethite has a broader feature with wavelengths around 920 nm or greater (22). This feature associated with crystal field absorption around 900 nm is also found in the jarosite and schwertmannite spectral curves. However, the charge transfer shoulders around 650 nm, associated with the charge transfer of change to Fe3+ and change to Fe2+ (23) allow further distinction for schwertmannite which has no known inflection point at 650  nm and spectral peak location at 738 nm (20).

The peak location at 720 nm and a small distinctive absorption feature at 2.264 nm confirms spectral identification for jarosite. Several minerals have been collected in so-called spectral libraries for validation purposes by the USGS spectral library (24) and Crowley Library for AMD minerals (20).

Large mining operation vicinities have been monitored by means of remote sensing imagery. Davies and Calvin have studied the Leviathan lake from mine tailings (25) and the spectral behavior of surface waters (26), while Swayze et al. studied the Venir pile in California/USA (14). The Iberian Pyrite Belt (IPB) in southern Spain has been also a target area for many remote sensing and compositional studies related to AMD chemistry (27, 28). The Sokolov mining district of the Czech Republic AMD phenomenon has been generally studied by Murad and Rojík (12) and by means of airborne hyperspectral data by Kopačková and Hladíková (29) for determining water surface parameters in water.

Unmanned Aerial Systems (UAS) scale

The emergent use of unmanned aerial systems (UAS), like multicopters coupled with lightweight hyperspectral sensors has become a tool to collect data at a higher spatial resolution than most of aircraft and satellite counterparts, resulting in greater precision (higher spatial resolution of a scene enabling the investigation of down to a few centimeters pixel size) (7). Most recently Jackisch et al. (30) implemented the use of UAS-hyperspectral imaging for high-resolution, multi-temporal mapping of proxy minerals for AMD in the Sokolov lignite region, Czech Republic, while Flores et al. (31) has focused mapping not only mineralogy but also hydro geochemical properties to assess the extent of AMD in Odiel and Tintillo waters, in the Iberian Pyrite Belt in southern Spain. In this study, several techniques have been combined to produce high resolution maps (Figure 4), a machine learning approach using regression has been used to fuse geochemical data from validation points at the field with the hyperspectral dataset. Also a 2.5 photo-grammetric model was constructed using Structure-from-motion (SfM) Stereophotogrammetry to compute a digital surface model (DSM).

Laboratory Scale

In addition to the airborne monitoring approach, hyperspectral sensors have been widely used on multiple laboratory scale applications for mineralogical characterization and AMD prediction. The so called geoenvironmental risks, has been used to evaluate the potential for AMD formation based on core logging, static chemical testing, bulk- and hyperspectral mineralogical techniques (32). As means of identifying AMD drivers of acidification in Schleenhain dump, the visible-near infrared (VNIR) and short wave infrared (SWIR) regions have been investigated by the TRIM4Post-Mining – a Horizon 2020 project funded by the Research Fund for Coal and Steel – for the detection of secondary iron oxides, hydroxides and sulfates in lignite waste dump material. For this purpose, two hyperspectral sensors (FX10 and FX17; from manufacturer Specim, Spectral Imaging Ltd.) has been used to acquire part of the VNIR and SWIR on the available samples.

No sample preparation is needed for the hyperspectral scanning. A portion of each sample was set on a white paper-sheet directly under the camera. The holder where the sample lay, moves in a horizontal direction and the line scan camera captures the hyperspectral image similar to a conveyor belt. This is achieved with the rail-like construction shown in Figure 5 together with a controllable motor. The stage moves directionally depending on the position with adjustable speed to the right or to the left depending on the position.

In general all hyperspectral surveys, should be accompanied by validation campaigns, in where point spectral measurements have to be done in discrete and strategic spots of the investigated area, as well as the incorporation of further geochemical/geophysical datasets to support the spectral method. Figure 6 shows a classification map created over scanned samples for validation from AMD affected site in the Sokolov lignite region(30).

Accurate compositional information of the mine waste materials is fundamental to understand the reaction schemes associated to AMD production and needed for geochemical modelling. Either by identifying primary sulfides prone to AMD or detecting secondary-iron sub products after weathering effects needs to be analyzed in order to locate risk areas, and provide adequate mitigation or prevention routines, prior to select the best post-mining plan.

Conclusions

Together with the high demand for raw materials in post-industrial societies comes the waste generation and all the task concerning their efficient management and risk assessments. In this sense, accurate and constant monitoring on terrain or vegetation cover of spoil banks is often required for two different reasons in post-mining management:

1. to monitor and prevent adverse effect of hazards; and
2. to assess restoration success.

Hyperspectral data brings several advantages as a complement to traditional environmental monitoring studies. The development towards lighter and smaller sensors, allows easier incorporation of hyperspectral technology into different stages of mine waste management. It could be used, rather during active mining to identify potential lithologies hosting minerals prone to AMD and forecast adverse effects, or in post-mining scenarios to target affected areas and continuous monitor restored areas. Traditional monitoring of soils and water quality is mainly based on the chemical analysis of samples routinely collected over the year and on the physical parameters of the groundwater measured by instruments located in the flow path. These tasks can be expensive, time-consuming and controlled by access limitations to the areas. In general, UAS mapping compared to ground surveying represents a reduction in the time employed on acquiring data. Furthermore, UAS allow reaching locations that may be difficult to access, are under protected status or that involve personal security risks for terrestrial-sampling. Regardless the scale, hyperspectral sensors allow repeatability and recurrent data-acquisition. Therefore, multi-temporal analysis is feasible and may allow constant monitoring of sensible ecosystems. Although many instruments with higher spectral resolution and wider wavelength range have been developed. This equipment is too heavy, fragile, and costly to be mounted on UAS. Several efforts have been made on satellite development to increase their spatial resolution by enhancing band acquisition efficiency and making data available in open-source systems.