Mineral exploration employs various techniques such as geological, geochemical, geophysical, and drilling methods. In contemporary exploration, hyperspectral remote sensing has proven to be particularly effective. The objective of this study is to enrich the TUNL (TUNisian Library) spectral library of minerals within the Tunisian mining context, initiated by Mezned [1] and expanded upon by Alayet [2]. The focus of this work lies in the analysis of spectral responses for the characterization of iron mineral surfaces and clays, establishing VNIR/SWIR spectral signatures at different particle size scales. As a result of our analysis, hematite, goethite, kaolinite, and montmorillonite were successfully identified spectrally. Additionally, sulfur was detected based on a distinct spectral absorption feature in the VNIR region. The purity of various clay minerals and iron oxides/hydroxides was also considered. The mineralogical compositions confirmed by X-ray diffraction (XRD) analysis align with the presented results. A comparative analysis with public spectral libraries, USGS and JPL, was conducted, with future plans to further enrich the TUNL library. The establishment of a geoportal, showcasing and providing data from the TUNL spectral library, holds significant importance. These advancements will be crucial for a more profound understanding of the mineralogical composition of the studied sites and geological mapping.

Mine tailings of Hammam Bent Jedidi (F-Ba-Zn-Pb), which are part of the mining residues, may contain a large amount of valuable minerals such as by heavy metals (Pb, Zn, etc,) and minerals with high pollution potential (Ba and F). However, these tailings are continually exposed to erosion and thus destroys the normal function of soils and endangers human health. The characterization, detection, and prediction of the abundance of minerals with high pollution potential is the first step towards mitigating their impact. In this context, X-ray diffraction (XRD) method was used to characterize of mineral types and their associated contents on each collected soil samples. The partial least squares regression (PLSR) method was processed for the prediction of mineral contents using: i) VNIR-SWIR field Hyperspectral reflectance data and ii) SENTINEL-2 Multispectral data.

In the abandoned Hammam Zriba mine site, mine tailings were caused a real desertification phenomenon under the prevailing wind action leading to the invasion of olive trees. Therefore, timely and accurate detection of stressed olive trees may be a critical problem for food safety. Thus, this study aims at detecting the stress of olive trees induced by heavy metal pollution around a particular abandoned mine site through the spectral feature investigation and the using of spectral indices. Results have shown a convergence between olive leaves spectra and the measured chlorophyll fluorescence, which reveals the stress of olive trees. Olive leaves spectra indicate progressive decrease in Near-IR reflectance and the FV/FM ratio (Variable Fluorescence/Maximal Fluorescence ratio) was about 0.773 and 0.3 for stressed olive trees compared to 0.8 for healthy olive tree. Moreover, the correlation between all used vegetation indices and chlorophyll fluorescence was calculated.

Machine learning methods particularly, partial least squares regressions (PLSR) and Support Vector Machine (SVM), were conducted in this study for the abundance prediction of phosphate minerals. Abundances of fluorapatite and dolomite minerals were predicted using VNIR-SWIR hyperspectral reflectance and abundance results issued from X-Ray Diffraction analysis. The performance statistics of the generated PLSR models, for both fluorapatite and dolomite minerals, revealed by the X-Ray Diffraction analysis, were calculated. According to the results, the short-wave infrared Short Wave InfraRed SWIR region has been shown to be the most important for the prediction of dolomite and fluorapatite contents. The PLSR model, developed for the fluorapatite content prediction, has shown an interesting performance compared to the once based on the SVM method. For the dolomite, the SVM model shoed the better results.

Heavy metal contamination is a worldwide environmental safety issue, directly threatening ecosystem security and human health, especially near abandoned mines. Reflectance spectroscopy has proven its capabilities for the identification and the prediction of the toxic metals. The current study evaluates the performance of field hyperspectral spectroscopy for lead (Pb), Zinc (Zn) and cadmium (Cd) heavy metal contents in soils around the Jebel Ressas mine site in the north of Tunisia. A total of fifty-two soil samples were collected and surface reflectance was measured in situ. Continuum removal (CR) method was used as a normalization function to determinate spectral characteristics. The linear regression method was used to correlate the spectral absorption depth of both bands at 2202 nm and 2338 nm, related respectively to clays and carbonates, with heavy metal contents, which were determined by geochemical analysis. Results indicated that the prediction was of greater accuracy and appear sensitive to spectral characteristics of clay and carbonate absorption features.

Mine tailings, charged with toxic minerals, of the abandoned mining of Hammam Zriba in the north of Tunisia are continually exposed to erosion and thus destroys the normal function of soils and endangers human health. Characterization, estimate and efficient monitoring of the abundance of minerals with high pollution potential in a quick and timely fashion is crucial for protecting the natural environment. Particularly, detecting changes in abundance of mineral concentrations measured in the same collection points at different time series is of special interest in this application domain. In this context, field Hyperspectral spectroscopy and X-ray diffraction (XRD) method were used to characterize of both several minerals. The partial least squares regression (PLSR) method was processed for the prediction of mineral contents.

Mine tailings, left after the mine activities without any control, are charged of heavy metals and contaminating minerals, qualified as risks minerals. In this study, we explore the potential of hyperspectral field spectroscopy in predicting the abundance of risk minerals in agricultural soils around the Hammam Zriba mine site in Northern Tunisia. Sixty-nine samples were gathered from the top surface of soils and dikes (tailing deposits) and measured using the ASD Field Spec HiRes operated in VNIR/SWIR regions (350-2500 nm). All samples were analyzed by X-ray diffraction method XRD for the estimation of both several minerals, including Barite and Fluorite contents. The partial least squares regression (PLSR) was used to relate the measured VNIR/SWIR spectra to XRD analysis results for the prediction of risk mineral contents. The prediction performances of Fluorite and Barite contents were evaluated using particularly, the coefficient of determination R2 and the ratio of performance to deviation RPD. Results have shown an accurate prediction of both risk minerals using VNIR/SWIR hyperspectral field spectroscopy.

Hammam Zriba district mine (36°20'N and 10°13.5'E) which includes the Tunisian fluoric province (F-Ba-Zn-Pb) is among the most important one in the Maghreb. The important waste quantities which were left over after smelting have a wide spread distribution without any control. Particularly, the residues of lead zinc and above all fluorine and barium cause the environment degradation and threaten the human health. Reflectance spectroscopy provides information about minerals and heavy minerals present within wastes, between visible (VIS) and shortwave infrared (SWIR). In this study, we propose an approach based on VNIR and SWIR reflectance spectroscopy with respect to XRD laboratory analysis. Particularly, we are interested on the content estimation of risk minerals through linear modeling using proposed hyperspectral parameters. The selection of all parameters is based on the continuum removal method. Regression results show that our proposed approach for the prediction of risk minerals is efficient.

Remote sensing is widely used for alteration mineral recognition and mapping. Since it covers spectral features of hydroxyl-bearing minerals as well as C-O bearing minerals like phylosilicates, sulfates, and carbonates, ASTER SWIR region allows detailed spectral characterization of alteration minerals. This study focuses on investigating the ability of ASTER SWIR data for mapping alteration minerals in western Nefza farmer in the northwest of Tunisia, an area with hight vegetation. Our contribution consists on a cascade classification based on partial sub-pixel mapping of alteration minerals. The classification method was implemented using the mixture tuned matched filtering (MTMF) method, on two steps away, with respect to ASD data field truth and XRD laboratory analyses.

Hyperspectral measurements were carried out at tree canopy as well as leaf levels in an olive field around the Jebel Ressas mine site using the ASD (350 - 2500 nm) instrument. The main objective of this study was to explore hyperspectral reflectance spectroscopy to detect heavy metal vegetation stress as an indicator of soil contamination. Moreover, the used families of Vegetation Indices were optimized and tested. The results indicated that vegetation indices based on the near infrared shortwave infrared (NIR - SWIR) domain can be obtained at tree canopy and leaf levels. At leaf level, a good detection of heavy metal vegetation stress was showed using optimized indices.

Abandoned mines cause critical problems which are among the most challenging environmental ones. This is the case in Tunisia, were effects of mining activity are well illustrated particularly in Mejerda and Mellag river watersheds. Mapping is the first step for the inventory, characterization and survey of mine wastes elsewhere. Particularly, detecting changes in images of the same scene taken at different time series is of special interest in this application domain. In this context, we applied a based MAD (Multivariate Alteration Detection) transformation method for multi temporal classification and change detection of mine wastes. For these purposes we used Landsat ETM+ time-series data.

Mine tailings constitute an alarming source of pollution that threatens soils, vegetation and human health around Jebel Hallouf-Bouaouane mine site. Because of their widespread geographical distribution; their location and extent, the characterization of mine tailings using traditional field work alone is both costly and inefficient. In this study, we explore remote sensing techniques based on multispectral and multisensory data fusion. Our contribution consists in enhanced mine tailing map generation using both SPOT 5 panchromatic and Landsat ETM+ multispectral data. Results of the linear spectral unmixing of the resulting hybrid image tend to field truth. The inter-comparison of unmixing results indicates that this methodology can be applied successfully to multispectral data for multi-temporal monitoring of mine tailings.

Field spectral data, measured on the ground by a spectroradiometer are very important and needed for tailing classification. The Spectral Modeling Approach SMA based on library reflectance's fusion is an optimal solution in absence of field data. In this study, we give a generalization of the approach we developed in [5]; [6] for other types of mine in Tunisia. The main contribution is to estimate the tailing components' spectral reflectance by direct means. This allows us to replace the lack of the field data needed for mine tailing mapping. Thus, we can generate tailing maps from multispectral data particularly those acquired in the previous date. The results indicate that SMA can be applied successfully.

Mine tailings may have a widespread geographical distribution; their location and extent may also vary along time, due to reprocessing and disposal activities. Remote sensing techniques have been proven extremely valuable in the inventory, characterization, and remediation of mine tailings else where. In this study we focus on mine tailing mapping around Jebel Hallouf-Bouaouane mine using Landsat multispectral data. Field spectral data, measured on the ground by a spectroradiometer are very important and needed for mine tailing classification. These informations, more faithful to the natural conditions, are unavailable in some dates. Our methodology is based on a Spectral Modelling Approach SMA using the JPL reflectance library. The main contribution is to evaluate the tailing modelled reflectance proposed to replace the lack of the field data needed for mine tailing mapping. Linear spectral unmixing is applied for mine tailing map generation using the tailing modelled spectrum, the ASD measured spectrum and the ETM+ derived spectrum. The comparison of results indicate that the SMA approach can be applied successfully to multispectral data analysis, particularly those acquired during previous periods. The SMA can be an optimal solution to replace the lack of measured field data (by the spectroradiometer).

The spatial cartography is very effective for the study of the mine tailing which often show a not uniform and wide spatial distribution. Field spectral data, measured on the ground by a spectroradiometer are very important and needed for tailing classification. In this study, we explore for the first time the mine tailing cartography with remote sensing techniques for Jebel Hallouf-Bouaouane mine in Tunisia. Our methodology is based on a Spectral Modelling Approach SMA using the library reflectance fusion. The main contribution is to estimate the tailing components' spectral reflectance by direct means. This allows us to replace the lack of the field data needed for mine tailing cartography. The results indicate that SMA can be applied successfully to multispectral data.

The mine of Bouaouane-Djbel (Hill) Hallouf which is exploited for the lead and zinc ores is among several types of mines in the Medjerda river watershed. We propose a multispectra inter-images fusion using a simplified version of multisensor multiresolution technique (MMT) for mine tailing cartography refinement. We use Landsat MS/Pan fused image and ASTER SWIR image acquired in the same period to conserve mineral state. Classification of the resulting Hybrid multispectral image based on constrained and unconstrained linear spectral unmixing is performing using end member library spectra. Unmixing results coincide with ASTER TIR interpretation as well as laboratory analysis. Moreover, the given results show that Hybrid multispectral image is more precise for certain mineral detection than ASTER fused image.

In the north of Tunisia, the mine tailings constitute a chronic problem causing multiple environmental consequences in the basin of Medjerda river. In this paper, we are interested on the mine tailings cartography around "Djebel" (hill) Hallouf- Bouaouane mine. We propose a new method for tracking the terrain contamination evolution based on the relationship between the degree of contamination and the terrain elevation (using digital elevation model, DEM). We used the unmixing multispectral technique on the landsat ETM+ imagery for generation the land contamination map. We consider the library mineral spectra for replacing the lack of the field spectra data needed in the unmixing process. This process use as endmembers the three first principal component resulting from the principal component analysis (PCA). The given results show that the contamination degree distribution is correlated with the elevation terrain.