Remote Sensing Applications for Geological Mapping in the Mediterranean Region: A Review

Abstract

Remote sensing has emerged as an essential method for geological mapping, especially in complex environments such as the Mediterranean region. While earlier global reviews have been focused either on multi- and hyperspectral sensors in general for geological applications or on hyperspectral sensors using machine learning for lithological mapping and mineral prospecting, this review article provides the first regionally focused synthesis dedicated to the Mediterranean region. The review examines both passive sensors such as Sentinel-2 MSI, Landsat-8 (OLI), ASTER, MODIS, Hyperion, PRISMA, EnMAP, and active sensors such as Sentinel-1, ALOS, TerraSAR-X. Furthermore, the review emphasizes the sensor functionalities, the data integration within Geographic Information System (GIS) platforms and methodological advancements such as machine learning and multi-sensor fusion. A total of 42 case studies are assessed, covering Portugal, Spain, France, Italy, the Balkans, Greece, Turkey, Cyprus, Egypt, Tunisia and Morocco. These examples highlight how remote sensing techniques have been adapted to varying lithological, tectonic and geomorphological settings across the Mediterranean. The analysis identifies key methodological trends, including the transition from spectral indices to advanced data fusion, the growing reliance on open-access available multispectral archives, and the emerging role of new-generation hyperspectral missions (PRISMA, EnMAP) in high-resolution geological mapping. The findings illustrate the non-invasive and scalable advantages of remote sensing for geological mapping in complex terrains, while also noting current challenges such as atmospheric correction, spatial resolution mismatches, and field validation requirements. By combining region-specific applications, this review demonstrates how remote sensing contributes not only to fundamental geological understanding but also to sustainable resource management and mineral exploration within one of the world’s most geologically diverse regions.

1. Introduction

The Mediterranean region, located where the African plate subducts under the Eurasiatic plate, is a geological puzzle characterized by active deformation processes, a variety of lithologies, and a complicated history of tectonic and magmatic activity. To understand the regional geodynamic framework, the distribution of natural resources, and the hazard assessment, the mapping and frequent monitoring of these geological features and processes are critical. Although conventional field-based geological surveys are still essential for ground-truth validation, remote sensing technologies have increasingly become a state-of-the-art tool for digital geological mapping. The capacity of satellite sensors to obtain repetitive and multiscale observations has revolutionized the detection, analysis, and interpretation of geological features of the complex Mediterranean geology. Many of the strategic minerals emphasized in the European Union’s Critical Raw Materials (CRM) Act (http://data.europa.eu/eli/reg/2024/1252/oj, accessed on 12 March 2025) are found in environments with the complexity of the Mediterranean region. Geological mapping using remote sensing (RS) applications plays a primary role in supporting a secure, stable, and sustainable supply chain of strategic minerals by enabling more efficient identification and monitoring in areas enriched by these minerals.

One of the most widely used techniques in satellite-based geological mapping is pixel-based methods, which categorize land cover according to the spectral response of each single pixel. Traditional methods encompass Spectral Indices (SIs) [1,2,3]—such as the Normalized Difference Vegetation Index (NDVI), the Iron Oxide Spectral Index (IOSI), and the Normalized Hydrothermal Alteration Spectral Index (NHASI)—which are usually obtained from multispectral data (e.g., Landsat, Sentinel-2, and ASTER) to identify specific surface materials representatives.

In addition, the importance of subpixel methods [4,5,6,7,8] has increased significantly, particularly in heterogeneous or geologically complex regions, such as the Mediterranean, where one single pixel could contain various types of land cover or mineral compositions. Imaging spectroscopy (IS) is another remote sensing technique applied mainly on hyperspectral sensors like PRecursore IperSpettrale della Missione Applicativa (PRISMA), Environmental Mapping and Analysis Program (EnMAP), or the future NASA Surface Biology and Geology (SBG) mission. The IS technique applied to hyperspectral data delivers high-spectral-resolution thematic maps that facilitate the detection of minerals and alteration zones at the subpixel level. A state-of-the-art subpixel technique is Spectral Unmixing (SU), which mathematically calculates the abundance of a set of end members (pure spectral signatures) within a mixed pixel. SU is able to quantify the relative abundances and produce more accurate mineralogical maps in areas like the Mediterranean region, where the variability on the chemical composition of the terrains is extremely high.

Recent advances in machine learning (ML) have significantly improved the efficiency of both pixel-based and sub-pixel approaches for geological mapping. Supervised classification methods, such as support vector machines (SVMs), random forests (RFs), and advanced deep learning models—including convolutional neural networks (CNNs) [9,10,11,12,13,14]—are widely applied to automate lithological classification, tectonic feature detection, and hydrothermal alteration mapping [15,16,17]. These algorithms are particularly well-suited for handling large volumes of satellite data across multiple spatial scales, leveraging both multispectral and hyperspectral imagery to extract detailed geological information.

Complementing passive RS methods, active RS technologies such as Synthetic Aperture Radar (SAR) and Interferometric Synthetic Aperture Radar (InSAR) techniques [18,19] can provide information on the tectonic structures and the active surface information of active tectonic environments such as the Mediterranean region. SAR is an all-weather system and is not affected by daylight conditions such as passive satellite sensors.

In the above-mentioned methods, the results are often verified through targeted field campaigns, which both validate the precision of RS results and enhance the reliability of classification models. In addition, laboratory-based mineralogical analysis of field collected samples complements in situ geological mapping. Techniques such as X-ray diffraction (XRD), electron microscopy (SEM and EPMA), and Raman spectroscopy provide important ground truth data [20] for the validation of the mineral spectral signatures collected with RS techniques.

Subsequently, the combination of RS products with current digitized geological maps, geophysical information, and field observations on Geographic Information System (GIS) platforms [21] facilitates multi-layered spatial analysis and enables more comprehensive geological interpretations for decision makers and regional authorities. This review article (a) explores the latest developments in RS applications for geological mapping in the Mediterranean region (Figure 1), (b) highlights new trends in pixel-based and sub-pixel methods on passive satellite sensors and InSAR on active satellite sensors, and (c) reveals possible data constraints on the application of ML techniques for satellite image analysis.

2. Methods

To fully understand the state-of-the-art and future developments of remote sensing applications for geological mapping, a critical assessment of the main methodologies based on satellite imagery, InSAR and GIS is reported in

Table A1. Case studies categorized by region, results/purpose and tectonics/geological setting including the bibliographic references.

No.	Case Study/Location	Region	Results/Purpose	Tectonics/Geological Setting	References
1	Freixeda, Northern Portugal	North	Mapping hydrothermal alteration zones	Varied, old sedimentary rocks	[38]
2	Beiras Group, Central Portugal	North	Lithology mapping	Sedimentary units	[92]
3	Torrential basins, Portugal	North	Mineral mapping	River basins	[106]
4	Los Frailes, Spain	North	Hydrothermal alteration patterns	Volcanic-hosted ore	[3]
5	Cantabrian Mountains, Spain	North	Land use	Variscan belt	[5]
6	Vega Media, Segura River, SE Spain	North	Alluvial deformation behaviors	Fluvial sediments	[75]
7	Iberian Pyrite Belt, SW Spain	North	Contamination monitoring	Sulfide deposits	[91]
8	Buëch area, France	North	Lithological mapping	Jurassic/Cretaceous units	[48]
9	Campi Flegrei, Southern Italy	North	Volcanic structures/unrest monitoring	Caldera volcanism	[68]
10	San Vito, Campi Flegrei, Italy	North	Hydrothermal alteration	Volcanic geothermal field	[20]
11	Lesvos Island, Greece	North	Temporal geothermal variations	Miocene volcanic field	[16]
12	Methana Volcano, Greece	North	Deformation trends	Hellenic Volcanic Arc	[67]
13	Limnos Island, Greece	North	Hydrothermal alteration mapping	Volcanic island	[34]
14	Nisyros Volcano, Greece	North	Mapping hydrothermal field	Volcanic Hellenic Arc	[4]
15	Koutala Islet, Lavreotiki, Greece	North	Identify mineralization	Igneous intrusions	[1]
16	Aegina Island, Greece	North	Ground deformation monitoring	Hellenic Volcanic Arc	[73]
17	Kythira Island, Greece	North	Ground displacement trends	Western Hellenic Arc	[74]
18	Sivas Basin, Turkey	North	Ophiolitic rock mapping	Ophiolite units	[39]
19	Central Turkey (Evaporites)	North	Evaporite minerals	Sedimentary basin	[40]
20	Eastern Taurides, W Turkey	North	Lithological units	Ophiolite/tectonic belt	[41]
21	East Anatolian Fault, Turkey	North	Tectonic activity	Fault zone/strike-slip fault	[43]
22	Kösedağ, Central-Eastern Anatolia	North	Hydrothermal alteration mapping	Volcanic metamorphic	[44]
23	Reşadiye, Tokat, Turkey	North	Bentonite mapping	Volcanic sedimentary	[65]
24	East Oltu, Erzurum, Turkey	North	Chromite ore exploration	Ultramafic rocks	[57]
25	Afyonkarahisar (Akarcay Basin)	North	Hydrothermal alteration mapping	Geothermal basin	[50]
26	Southern Italy	North	Land use	Complex volcanic/tectonic area of Southern Italy	[105]
27	East Vardar Ophiolite, North Macedonia	North	Predictive mapping	Ophiolite complex	[47]
28	Selac, Kosovo	North	Mineral prospecting	Alpine Ophiolite	[46]
29	Panagyurishte, Bulgaria	North	Hydrothermal alteration/ore deposits	Balkan Thrust	[45]
30	Troodos Ophiolite, Cyprus	North	Lithological mapping/prospecting	Ophiolite complex	[98]
31	Rich area, Central High Atlas, Morocco	South	Lithological mapping	Atlas mountains	[22]
32	Igoudrane, Jbel Saghro, Morocco	South	Lithology/mineral mapping	Anti-Atlas	[51]
33	Tifraouine M’sirda, NW Algeria	South	Hydrothermal alteration/structural mapping	Coastal/volcanic	[32,33]
34	Hamash area, Egypt	South	Mineralization zones	Precambrian basement	[37]
35	South Eastern Desert, Egypt	South	Mineral prospecting	Precambrian rocks	[42]
36	Sidi Bou Azzouz, Morocco	South	Abandoned mining site mapping	Anti-Atlas	[49]
37	Northern Tunisia (Nappe Zone)	South	Fe ore characterization	Atlas mountains	[94]
38	Nisyros Volcano (thermal correlation), Greece	North	Correlating thermal performance with hydrothermal zones	Volcanic Hellenic Arc	[15]
39	Methana, Santorini, Greece	North	Volcano monitoring/ pre/post unrest	Volcanic arc	[69]
40	Ischia Island, Italy	North	Ground displacement	Phlegraean Volcanic	[70]
41	NE Italy coastal plains	North	Subsidence zonation	Coastal plain	[72]
42	Crete, Central	North	Karst geomorphology mapping	Mediterranean karst	[85]

, as a timeline in Figure 2, and discussed below. Pixel-based and sub-pixel methods on passive satellite sensors and InSAR on active satellite sensors are also presented, including the emerging application of ML techniques for satellite image analysis (

Table 1. Comparison between before and after preprocessing in remote sensing geological mapping.

Case Study/Sensor	Before Preprocessing	After Preprocessing
Central High Atlas, Morocco (Sentinel-2 MSI)	Spectral confusion between carbonate and silicate units; reduced classification accuracy due to lack of atmospheric correction	Radiometric calibration and atmospheric correction improved spectral separability; lithological discrimination accuracy exceeded 85%.
Nisyros island, Greece (Sentinel-1 SAR)	Speckle noise and geometric distortions masked deformation patterns linked to hydrothermal activity	Co-registration, noise filtering, and coherence computation revealed coherence loss zones correlated with hydrothermal activity.
Nisyros island, Greece (ASTER, Landsat-8 OLI, Sentinel-2 MSI)	Topographic shading and noise contamination reduced reliability of spectral indices; alteration zones poorly delineated.	Radiometric calibration, atmospheric correction, and spectral subsetting improved mapping accuracy of minerals by 20–30%.
Italy (PRISMA hyperspectral)	Atmospheric distortions and spectral overlaps limited discrimination of phyllosilicate and carbonate minerals.	Radiometric calibration, atmospheric correction, and spectral subsetting improved mapping accuracy of minerals by 20–30%.

).

2.1. Satellite Data Acquisition and Preprocessing

Recent advances in remote sensing have enabled the effective use of various satellite datasets for geological and mineral exploration, particularly in lithological and hydrothermal alteration mapping. A range of multispectral sensors, including Landsat-8 Operational Land Imager (OLI), ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer), and Sentinel-2 MultiSpectral Instrument (MSI), have been widely adopted due to their diverse spectral resolutions and spatial coverages. For example, Bentahar et al. [22] compared these three datasets in the Rich area of the Central High Atlas, Morocco, highlighting the superior performance of Sentinel-2A for lithological discrimination, largely due to its higher spatial resolution (10–20 m) and broad spectral range. Similarly, Shebl et al. [23] demonstrated that the integration of Sentinel-2 with airborne gamma-ray data significantly improved the accuracy of lithological classification when applying supervised machine learning techniques, such as support vector machines (SVMs). In addition to optical and thermal sensors, radar data have become increasingly important in hydrothermal alteration studies. Sentinel-1 Synthetic Aperture Radar (SAR) data, for example, offers valuable information on surface coherence and deformation. Multitemporal interferometric coherence analysis was applied using Sentinel-1 to detect surface changes associated with hydrothermal activity on Nisyros Island [18]. The preprocessing of SAR data—such as co-registration, noise filtering, and coherence computation—is essential for meaningful analysis. Furthermore, Tompolidi et al. [15] utilized ASTER’s thermal infrared bands to correlate surface temperature anomalies with known alteration zones, reinforcing the value of multi-sensor data usage.

Hyperspectral satellite missions such as EnMAP (Environmental Mapping and Analysis Program) and PRISMA (PRecursore IperSpettrale della Missione Applicativa) represent a significant advancement in Earth observation for geological applications, offering detailed spectral resolution across hundreds of contiguous bands. EnMAP, as described by Guanter et al. [24], provides high-reliability imaging spectroscopy data designed specifically for Earth’s surface characterization, making it suitable for mineralogical and lithological studies. PRISMA, developed by the Italian Space Agency (ASI), has been increasingly used in mineral mapping and the detection of alteration zones due to its hyperspectral capabilities and moderate spatial resolution (30 m) as outlined by [25,26]. Alicandro et al. [27] demonstrated the effectiveness of PRISMA in mapping mineral assemblages in various geological settings in Italy, with preprocessing workflows that include radiometric calibration, atmospheric correction (e.g., using ATCOR or FLAASH) and spectral subsetting to optimize mineral detection. Similarly, Sorrentino et al. [28] processed PRISMA data to delineate hydrothermal alteration zones in complex ore systems in the Chilean Andes, emphasizing the value of advanced preprocessing to correct for topographic and atmospheric distortions in mountainous terrain. Furthermore, comparisons between PRISMA and multispectral sensors like Sentinel-2 underscore the advantages of hyperspectral data for geological discrimination.

A preliminary assessment between PRISMA and Sentinel-2 in archaeological applications [27] highlighted transferable insights relevant to geology, specifically, the superior capacity of PRISMA to resolve subtle spectral differences associated with mineralogical variation. These hyperspectral missions [29] require sophisticated preprocessing pipelines, often involving noise reduction, spectral smile correction, and end member extraction to ensure accurate interpretation. As these missions mature, their high spectral accuracy will increasingly support the precise mapping of alteration minerals, lithological boundaries, and compositional anomalies in diverse geological contexts. It is extremely important to preprocess data as shown by the comparison between raw and corrected datasets. For example, Sentinel-2 MSI data used to map the Central High Atlas could not correctly detect carbonate and silicate units without the application of the atmospheric corrections, while radiometric calibration and correction raised classification accuracy above 85% [22]. Similarly, unprocessed Sentinel-1 SAR images over Nisyros were dominated by speckle noise, masking subtle deformation patterns, but after co-registration and coherence computation, hydrothermal zones were better identified [18]. Another example is PRISMA in Italy: at the beginning, the imagery showed strong atmospheric distortions and spectral overlaps that hindered mineral discrimination; then after radiometric calibration and spectral subsetting was applied, the mapping accuracy of phyllosilicates and carbonates was improved by 20–30% [27]. In rugged volcanic terrains preprocessing steps such as atmospheric corrections were equally critical, enhancing the signal-to-noise ratio and enabling confident detection of alteration minerals [4]. In summary, these examples confirm the fundamental importance of robust preprocessing to generate reliable reflectance data, and maximizing mineral detection accuracy to reduce topographic effects across multispectral, radar, and hyperspectral sensors.

2.2. Pixel-Based Methods

Pixel-based methods are fundamental for multispectral and hyperspectral satellite image analysis for geological mapping and mineral exploration. These techniques operate at the level of individual pixels, utilizing their spectral signatures to classify surface materials based on their reflectance or emissivity properties. In multispectral imagery, pixel-based methods often involve the use of spectral indices and thresholding to improve and identify specific mineralogical or alteration features, such as iron oxides, clay minerals, and silicates [30]. In hyperspectral imagery, pixel-based analysis enables more detailed mineral discrimination by exploiting narrow and contiguous spectral bands to capture diagnostic absorption features. Techniques such as the Spectral Angle Mapper (SAM), Minimum Distance Classifier (MDC), and Spectral Feature Fitting (SFF) are commonly applied to match pixel spectra with known reference spectra from spectral libraries [31]. Despite their sensitivity to spectral variability and noise, pixel-based methods remain widely used because of their simplicity, computational efficiency, and compatibility with both supervised and unsupervised classification methods.

2.2.1. Spectral Indices

In the Mediterranean region, satellite remote sensing has been widely applied to detect hydrothermal alteration zones and assess mineralization potential using spectral indices derived from multispectral and hyperspectral data. Spectral indices (SIs), particularly using ASTER and Landsat-8 (OLI) data, have been proved to be essential in mapping iron oxides, clay minerals, and silicification—key indicators of hydrothermal processes. For example, studies in the Tifraouine and M’sirda regions of Northwestern Algeria used Landsat 8 OLI SIs such as (B4/B2), (B6/B5), and (B7/B5) to highlight iron oxide and hydroxyl-bearing minerals related to epithermal and sulfide deposits [32,33]. Similarly, Anifadi et al. [34] applied ASTER SIs including (B4/B2), (B5/B6), and (B7/B6) on Limnos Island, Greece, to effectively detect argillitic, phyllitic, and propylitic alteration zones. These SIs improve the spectral features related to Fe⁺³ absorption and OH⁻ and Al–OH vibrational modes, allowing the refined discrimination of alteration facies. Tompolidi et al. [4] expanded this approach by applying spectral unmixing techniques using ASTER, Landsat-8 OLI, and Sentinel-2 MSI data, demonstrating that integrated spectral analysis improves the detection of subtle alterations in volcanic environments like in Nisyros.

The Iron Oxide Index (e.g., ASTER B2/B1 or Landsat B4/B2), Clay Mineral Index (e.g., ASTER B5/B7), and Ferric Iron Index have been frequently used to map hydrothermal systems and weathering processes in the Mediterranean region. A combination of iron oxide and hydroxyl-sensitive SIs with topographic derivatives (e.g., slope and elevation) has been used to map a variety of alteration zones on Lesvos Island [35,36].

Furthermore, a novel degradation index was introduced by Chikhaoui et al. [2] by using ASTER’s shortwave infrared bands to assess land degradation in semi-arid Mediterranean catchments. In mineralized terrains of Egypt and Portugal, spectral indices such as (B4/B2) and (B5/B6), were integrated with geophysical data to enhance the delineation of gold and antimony-rich alteration halos [37,38].

ASTER data have been successfully employed also in the far east Mediterranean (Turkey) to map ophiolitic rocks in the Sivas Basin [39], evaporite minerals in central Anatolia [40], and lithological units in the Eastern Taurides [41]. Early applications of spectral analysis techniques also demonstrated the potential of satellite data to identify clay minerals [42]. Beyond lithological mapping, satellite spectral analysis has been used to evaluate tectonic activity by detecting mineral alterations along fault zones, such as the East Anatolian Fault [43], and to delineate hydrothermal alteration zones with hyperspectral EO-1 Hyperion data in Kösedağ [44]. Comparable approaches have been applied elsewhere in Southeastern Europe, such as in Bulgaria where ASTER band ratios were tested for detecting hydrothermal alterations in the Panagyurishte ore region [45], in Kosovo for mineral prospecting using integrated remote sensing methods [46], and in North Macedonia where geophysical data were combined with multispectral imagery for predictive mapping of the East Vardar Ophiolite Zone [47]. Similarly, in Western Europe, Sentinel-2 and ASTER data were combined to map lithological units in the Buëch area of Southeastern France [48].

Moreover, the recent usage of the state-of-the-art PRISMA hyperspectral imagery—for SIs applications [28,49]—enabled the precise mapping of mineralogical zones. PRISMA, because of its high spectral resolution, is able to capture narrow absorption features of specific alteration minerals that are not detectable by multispectral sensors. These examples underscore how the proper selection of SIs tailored to the mineralogical and geological context is crucial for an effective geological mapping using passive remote sensing sensors throughout the Mediterranean.

2.2.2. Spectral Similarity-Based Techniques

Spectral similarity-based classification methods, such as Spectral Angle Mapper (SAM), Spectral Feature Fitting (SFF), and Minimum Distance Classifier (MDC), are widely used in remote sensing applications for geological mapping throughout the Mediterranean region. These methods compare the spectral signature of each pixel with known reference spectra (end members) to assess compositional similarity, enabling mineral and lithological discrimination even in complex terrains. SAM, in particular, has been proven to be robust in hydrothermal alteration and lithological mapping as demonstrated in the Akarcay Basin (Turkey) using ASTER data [50] and in the Anti-Atlas Mountains (Morocco) using Landsat 8 OLI and ASTER imagery [51]. Meanwhile, SFF, which fits the absorption features of pixel spectra to reference spectral profiles, improves the detection of specific minerals with diagnostic absorption bands and has been successfully applied in areas such as Cyprus and North Africa [52,53]. MDC, another similarity-based method, calculates the Euclidean distance between the pixel and end member spectra, offering computational efficiency and ease of interpretation in multispectral data analysis [54,55].

However, these methods have shown limitations, particularly under conditions of dense vegetation cover, surface weathering, or mixed pixels. For example, lithological discrimination using airborne multispectral data in Troodos Ophiolite (Cyprus) was significantly impeded by vegetation [52], while Chen [55] highlighted the challenges in applying spectral similarity techniques in heavily vegetated or soil-covered regions. Moreover, recent studies have successfully applied these approaches in Turkey, demonstrating their effectiveness by applying SAM in predictive mineral prospectivity mapping [56] and the exploration of ultramafic-hosted chromite deposits [57]. These methods exploit the unique spectral characteristics of target minerals, providing fast, large-scale screening of potential ore zones while reducing the need for extensive field sampling.

To address these issues, recent studies emphasize preprocessing techniques such as vegetation masking, continuum removal, and end member refinement to improve classification accuracy. Moreover, integrating SFF and MDC with machine learning algorithms has improved their adaptability and precision in mineral-rich environments [53,54]. The inclusion of hyperspectral sensors further improves feature detection since detailed spectral information enables better resolution of subtle absorption features related to hydrothermal alteration minerals [58]. In general, spectral similarity-based techniques—especially when combined with advanced data fusion and classification frameworks—remain valuable tools for mineral exploration and geological mapping across the diverse and mineralized Mediterranean region.

2.2.3. Unsupervised Classification

Unsupervised classification techniques play a pivotal role in geological mapping in the Mediterranean region, especially in data-scarce areas where traditional supervised methods are limited by the availability of the training data. Algorithms such as fuzzy c-means (FCM) and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) have proven effective in identifying alteration zones by clustering pixels with similar spectral properties without prior class labels. Ghezelbash [59] demonstrated that the application of FCM and DBSCAN for unsupervised mineral prospectivity mapping effectively distinguishes hydrothermal zones from surrounding lithologies by leveraging the subtle spectral differences captured in multispectral data. These methods are especially valuable in Mediterranean volcanic and geothermal environments—such as those in Greece, Morocco, and Southern Spain—where alteration minerals such as kaolinite, alunite, and hematite exhibit diagnostic spectral features in the ASTER and Sentinel data. Tompolidi et al. [15] further validated the utility of thermal and spectral clustering approaches in identifying hydrothermal alteration at the Nisyros Volcano, correlating ASTER thermal performance with known alteration zones.

Beyond clustering, unsupervised classification is often integrated with advanced processing workflows to enhance hydrothermal feature detection. In particular, the delineation of hydrothermal zones is significantly improved, particularly when using high-dimensional hyperspectral data such as Hyperion, if unsupervised methods like k-means are combined with object-based classification and spectral similarity algorithms (e.g., SAM) [10,60].

In structurally complex and vegetated regions of the Mediterranean, these methods help isolate alteration signals from background noise and vegetation interference. Recent studies have also integrated unsupervised classification outputs into machine learning frameworks to refine alteration mapping—for instance, in predictive models of mineralization potential using random forests [13] or neural networks trained on pre-clustered lithological units [61]. These approaches demonstrate the growing importance of unsupervised classification not only as a stand-alone method but also as a foundational step in hybrid workflows for the detection of hydrothermal alteration in the tectonically and mineralogically diverse landscapes of the Mediterranean.

2.3. Sub-Pixel Methods

Subpixel-based methods are advanced remote sensing techniques designed to extract detailed information from mixed pixels in multispectral and hyperspectral satellite datasets, where a single pixel contains a combination of multiple surface materials. These methods aim to estimate the proportion or abundance of different materials—known as end-members—within each pixel, thereby improving the spatial and thematic resolution of geological mapping. A common approach is spectral unmixing, which includes linear and non-linear unmixing models. The linear model assumes that the observed spectrum is a weighted sum of end member spectra, while non-linear models account for complex interactions like multiple scattering. Techniques such as linear spectral unmixing (LSU), Multiple End-member Spectral Mixture Analysis (MESMA), and Non-linear Mixture Models (NMMs) have been effectively applied in mineral exploration, hydrothermal alteration mapping, and soil composition analysis [62]. Hyperspectral data, with their high spectral resolution, are particularly well-suited for subpixel analysis, as they capture distinct absorption features that facilitate accurate end member discrimination [63]. Subpixel-based methods offer significant advantages in heterogeneous or geologically complex terrains, where traditional pixel-based classification may fail to represent the diversity of surface materials.

Spectral Unmixing

Spectral unmixing, particularly through linear models, has been applied effectively to hydrothermal alteration mapping in complex volcanic terrains of the Mediterranean. In linear spectral unmixing (LSU), each pixel’s spectrum is modeled as a weighted sum of pure spectral signatures (end members), assuming no significant interaction between materials. This approach has proven especially useful in identifying alteration zones composed of mixed mineralogical assemblages, such as argillitic, phyllitic, and propylitic facies. Tompolidi et al. [4] applied LSU on ASTER, Sentinel-2 MSI, and Landsat-8 (OLI) data in the Nisyros volcanic system (Greece), revealing the spatial distribution of alteration minerals and highlighting the effectiveness of combining sensors with varying spectral resolutions. Linear unmixing methods have been demonstrated to be computationally efficient and interpretable, making them a widely used starting point for geological applications [64]. Furthermore, in the far Eastern Mediterranean, Canbaz et al. [65] applied machine learning in combination with sub-pixel mixture algorithms to map bentonite deposits in Reşadiye (Tokat, Turkey), highlighting how these techniques can enhance the resolution of lithological and mineralogical maps derived from multispectral data. However, linear models may oversimplify the spectral mixing process, especially in rugged or vegetated Mediterranean terrains, where non-linear interactions between surface materials can occur due to multiple scattering and topographic effects. To address the limitations of linear models, non-linear spectral unmixing approaches have gained traction, particularly in hyperspectral geological mapping.

Non-linear models account for complex mixing behaviors, such as internal mixtures or spectral variability induced by mineral grain size, illumination geometry, and surface roughness, which are common characteristics of hydrothermal environments. For instance, Chakraborty [66] demonstrated the superiority of non-linear methods in capturing subtle spectral differences for mineral mapping in complex terrains, while Borsoi et al. [7,12] introduced data-driven and multiscale models capable of handling spectral variability across spatial and temporal scales. These methods often incorporate machine learning and geometric modeling techniques, such as deep neural networks, to improve end member extraction and abundance estimation [8]. In the Mediterranean region, where hydrothermal alteration zones are often spatially heterogeneous and spectrally variable, these advanced unmixing methods provide a more realistic and detailed interpretation of sub-pixel mineralogy, supporting high-precision mapping of potential mineralization zones and geothermal activity.

For example in the case of Nisyros, the improvement in classification accuracy was quantified by comparing LSU-derived mineral abundance maps with (a) the georeferenced geological map and (b) the hydrothermal alteration distribution map [4]. Accuracy assessment involved the calculation of confusion matrices and kappa statistics, leading to the demonstration that the usage and comparison of the three multispectral sensors of ASTER, Sentinel-2, and Landsat-8 (OLI) reduced the spectral confusion among alteration facies and increased overall classification accuracy by approximately 15–20% compared to single-sensor approaches. A further confirmation of the validity of the results is aroused by the Root Mean Square Error (RMSE) between measured and modeled abundances, suggesting that preprocessing combined with multi-sensor unmixing can significantly enhance the mapping of hydrothermal alteration zones in the Nisyros volcanic system.

2.4. SAR Interferometry (InSAR) for Geological Mapping

The Mediterranean region, characterized by its dynamic geotectonic framework, active volcanism, and densely populated coastal zones, provides a compelling environment for the application of remote sensing technologies in geological mapping. Interferometric Synthetic Aperture Radar (InSAR), particularly through multitemporal techniques like SBAS (Small Baseline Subset), has become a cornerstone method for detecting and analyzing surface deformation across the region. Notable volcanic case studies include the Methana Volcano in Greece, where Sentinel-1 MT-InSAR data combined with GNSS and seismic observations revealed deformation patterns linked to tectonic and magmatic processes [67], and the Campi Flegrei caldera in Italy, where SAR and seismic interferometric techniques were used to track uplift and unrest phases between 2011 and 2013 [68]. Similar methodologies applied to La Palma and Santorini [69] and Ischia Island [70] underscore the effectiveness of InSAR in pre- and post-unrest monitoring, allowing the detection of subsurface magma movements and fault dynamics critical to early warning systems.

In addition to volcanic terrains, remote sensing has proven highly effective for subsidence mapping and geological assessments in sedimentary and coastal environments. For example, studies in the Po River Delta and Venice Lagoon [71] and in Northeastern Italy [72] used InSAR data to delineate areas undergoing anthropogenic and natural subsidence, facilitating a better understanding of soil compressibility and land stability. Furthermore, in regions such as Aegina Island [73], Kythira Island [74], and the Vega Media of the Segura River Basin [75], SBAS and other time-series InSAR techniques have captured ground deformation trends associated with tectonic activity and alluvial soil compaction, offering important inputs for the zonation of geological hazards and environmental monitoring.

Remote sensing techniques have also been expanded to detect hydrothermal and surface alterations associated with volcanic systems and have not only been used for crustal deformation mapping in the Mediterranean region. In particular, the use of multitemporal Sentinel-1 coherence analysis was demonstrated to identify hydrothermal alteration zones at Nisyros Volcano, revealing the potential of coherence imagery as a complementary dataset to traditional deformation monitoring [18]. In addition, the integration of InSAR with complementary geophysical datasets—such as GNSS, seismic networks, and geological field surveys—enhances the interpretative power of remote sensing and provides a multidisciplinary framework for geological mapping. Advanced InSAR techniques, as proposed by [76], allow the classification and characterization of types of deformation, distinguishing between processes such as fault creep, landsliding, or subsidence. Furthermore, long-term deformation records such as those at Nisyros volcano [77] exemplify the role of SAR data in deformation time-series. As demonstrated throughout the Mediterranean basin, satellite-based remote sensing, especially when integrated with ground-based observations, has become indispensable for comprehensive geological analysis, offering high-resolution, wide-area, and cost-effective monitoring solutions.

2.5. Field Campaigns and Mineralogical Analysis

Field campaigns play an essential role in validating remote sensing outputs and interpreting spectral signatures in geological and mineral exploration. They not only confirm the presence of alteration minerals detected from space but also provide critical mineralogical and geochemical details through laboratory analysis. For example, Chen et al. [78] integrated ASTER and WorldView-3 imagery with field sampling and employed X-ray diffraction (XRD) and short-wave infrared (SWIR) spectroscopy to confirm the presence of phyllitic and argillitic alteration minerals—including kaolinite, illite, and chlorite—within the Pulang porphyry copper system. Furthermore, Ferrier et al. [36] combined satellite-based compositional data with field samples analysis by X-ray fluorescence (XRF) and petrographic microscopy to characterize different types of hydrothermal alteration across multiple scales. These examples highlight how multi-sensor data acquisition must be complemented by laboratory-based mineralogical confirmation to ensure accurate spectral classification. Hewson et al. [79] further reinforced the need for multiscale geological remote sensing by advocating for ground-validated spectral libraries and laboratory-based calibration to bridge the gap between satellite-scale and hand-sample-scale observations.

In the Mediterranean region, mineralogical laboratory analysis has been crucial for understanding active and fossil hydrothermal systems. In the study of Bobos et al. [80], samples from the Furnas Volcano (Sao Miguel, Azores) were analyzed using XRD and inductively coupled plasma mass spectrometry (ICP-MS) to quantify the mineralogical and geochemical signatures of acid-sulfate alteration, including high concentrations of high-field strength elements (HFSEs) and rare-earth elements (REEs) associated with kaolinite and alunite. Similarly, Pereira et al. [81] combined field mapping with XRD, scanning electron microscopy with Scanning Electron Microscope–Energy-Dispersive X-ray Spectroscopy (SEM-EDS), and gas geothermometry to characterize mineral alteration in the Ribeira Grande geothermal field (Sao Miguel, Azores). Moreover, Quintela et al. [82] used XRD and differential thermal analysis (DTA) to evaluate hydrothermally altered clays for their industrial and therapeutic potential in Caldeiras da Ribeira Grande (Sao Miguel, Azores). In broader Mediterranean contexts, Chikhaoui [2] validated a remote sensing-based land degradation index in a semi-arid catchment using geochemical analysis of soil and rock samples.

These integrated approaches—combining hyperspectral satellite data with laboratory-confirmed mineralogy—demonstrate the critical role of fieldwork and analytical methods such as XRD, XRF, SEM-EDS, and ICP-MS in producing geologically meaningful interpretations in Mediterranean hydrothermal systems (

Table 2. Satellite, sensor, processing method, spectral library used.

Satellite	Sensor	Processing Method	Spectral Library Used
Sentinel-2	MSI	Pixel-based (spectral indices, SVM, RF)	Customized
Sentinel-1	C-band sensor	SAR Interferometry (SBAS, PSInSAR)	N/A
Landsat 8	OLI	Pixel-based (spectral indices, PCA)	Customized
Terra	ASTER (VNIR, SWIR, TIR)	Pixel-based (spectral indices, PCA)	Customized
Terra/Aqua	MODIS	Pixel-based (thermal time-series)	N/A
EO-1	Hyperion	Sub-pixel (MESMA, spectral libraries)	USGS
PRISMA	HYC + PAN	Sub-pixel (spectral unmixing, indices)	USGS, Customized
EnMAP	HIS	Sub-pixel (unmixing, RTM models)	USGS, Customized
Sentinel-2 + Ground Data	MSI + Gamma-ray	Pixel-based (SVM, RF)	Customized
Landsat + DEM	OLI + DTM	Pixel-based (spectral indices)	Customized
Sentinel-1 + Landsat	C-band sensor + OLI	Mixed (SAR + DBSCAN, fuzzy C-means)	Customized
ALOS, TerraSAR-X, Sentinel-1	PALSAR (L-band), X-band, C-band sensor	SAR Interferometry (SBAS, MTInSAR)	N/A

N/A: not applicable; USGS: United States Geological Survey.

).

2.6. Geographical Information Systems (GISs) and Multi-Layer Analysis

According to the literature, all processed datasets—including spectral indices, mineral abundance maps, unsupervised classified lithology, InSAR, and deformation layers—were integrated within a GIS environment (ArcGIS/QGIS) [4,15,83,84,85]. Overlay analysis enabled multi-source cross-validation. For instance, tectonic structures (e.g., faults and fissures) derived from InSAR results were superimposed on digitized and orthorectified lithological maps.

This approach in most of the studies considered for the Mediterranean region enables an advanced spatial mapping and hotspot analysis of CRM-related lithologies, as well as integration with a variety of geological, geophysical, and geochemical layers [9,86]. The GIS framework, by supporting dynamic querying, can also facilitate iterative field targeting and knowledge refinement.

3. Results

In the following, we present a synthesis of remote sensing methodologies and their applications in geological mapping across the Mediterranean region (Figure 3). An overview of the multispectral, hyperspectral, and SAR data applications in various Mediterranean environments, highlighting the latest state-of-the-art methods and trends, is reported (Table A1). The second subsection explores the role of spectral libraries—both standard (e.g., USGS) and custom—in improving classification accuracy and mineral identification. Then we focus on the distinction between standard and custom spectral libraries cases, including sources, validation methods, and integration with mineralogical field analyses. Finally, we compare the effectiveness of pixel-based and sub-pixel methods, with emphasis, respectively, on their advantages and limitations in different geological contexts, such as ophiolite complexes, hydrothermal systems, and structurally complex terrains.

3.1. Overview of Remote Sensing Applications in the Mediterranean

Remote sensing applications for geological mapping across the Mediterranean region have grown substantially in the past two decades (Figure 2), with a strong emphasis on multispectral and hyperspectral satellite platforms. ASTER has been the most frequently utilized sensor (Figure 3), particularly in alteration mapping due to its effective SWIR bands, with notable case studies in Cyprus, Morocco, and Southern Italy. Sentinel-2 has emerged more recently as a complementary tool for broader regional mapping, especially when high spatial resolution is needed. Specifically, Sentinel-2 has proven valuable for structural and lithological analysis when combined with spectral indices and tectonic structure mapping (Table A1). More advanced hyperspectral missions such as PRISMA and EnMAP are beginning to appear in studies, particularly for mineral mapping in complex terrains such as Tunisia and Central Greece.

The review shows that pixel-based classification methods, support vector machines (SVMs) and maximum likelihood, are widely applied in geologically simple areas with well-defined lithologies, such as ophiolitic complexes in Cyprus or sedimentary basins in Tunisia. These methods perform well in areas where surface exposures are clear and rock units are spectrally distinct. However, in more geologically heterogeneous zones—such as the Iberian Pyrite Belt in Portugal or faulted volcanic terrains in Morocco—spectral mixing within pixels limits the effectiveness of unsupervised classification. In these settings, sub-pixel methods such as linear spectral unmixing (LSU) and Multiple End member Spectral Mixture Analysis (MESMA) have provided enhanced mapping precision, especially for detecting hydrothermal alteration. Several studies in Portugal, Cyprus, and Tunisia successfully integrated field-collected spectra and laboratory measurements with spectral reference libraries such as the USGS Spectral Library to improve the accuracy of geological mapping.

In addition to optical data, Synthetic Aperture Radar (SAR) interferometry has been applied in the Mediterranean for geological and geomorphological mapping, particularly in volcanic regions (Table A1). A key example is from Nisyros Volcano in Greece, where Tompolidi et al. [18] demonstrated the potential of multitemporal coherence derived from Sentinel-1 InSAR data to detect zones of hydrothermal alteration with different response to weathering. The findings showed that coherence loss correlated well with altered zones, optical remote sensing and available geological data [4], suggesting the strong potential of InSAR as a non-invasive tool for monitoring volcanic and geothermal systems. Complementary work by Tompolidi et al. [4,15] using ASTER, Sentinel-2 MSI, and Landsat-8 (OLI) further validated the integration of thermal, optical, and radar (SAR) datasets for detailed mapping of alteration minerals, interferometric coherence mapping and thermal anomalies across Nisyros. These studies emphasize the value of combining optical and SAR data for comprehensive geological interpretation, especially in terrains where vegetation, cloud cover, or surface inaccessibility may limit passive sensor use.

3.2. Spectral Libraries: Use and Impact on Geological Mapping in the Mediterranean Region

Spectral libraries are essential tools in geological mapping with remote sensing techniques, particularly for spectral similarity techniques (e.g., SAM, SID) and spectral unmixing techniques. In the Mediterranean region, both standard global libraries and custom-built regional libraries have been used to improve classification accuracy and mineralogical identification.

Standard and Custom Spectral Libraries

The USGS Spectral Library [31] is the most commonly referenced spectral library in remote sensing studies for geological mapping of the Mediterranean region. It has been used in studies applying pixel-based methods and sub-pixel methods for geological mapping in Mediterranean regions such as Greece [4,87], Cyprus [88], Italy [89], Spain [90,91], Portugal [81,92], Tunisia [93,94] and Turkey [95]. These applications focused on identifying alteration minerals (e.g., kaolinite, hematite, and chlorite) associated with hydrothermal systems, mining tailings, and rare-earth-elements (REEs) exploration.

However, reliance on global spectral databases poses challenges, particularly in heterogeneous terrains with very specific regional lithologies. To address this point, several studies developed custom spectral libraries using field spectro-radiometry and laboratory analysis. For example, Ferrier et al. [96] used a field spectroradiometer to collect ground reference spectra of altered rocks in Southern Spain, which were then used to calibrate and validate imaging spectrometry data for mapping hydrothermal alteration zones associated with gold mineralization. Similarly, Yalcin et al. [50] used both the USGS spectral library and a custom developed spectral library at the local level, derived from field and laboratory analyzes of rock samples, to improve the identification and classification of hydrothermal alteration minerals in the Akarcay Basin (Turkey) using ASTER imagery.

Reported bibliography demonstrated the diagnostic spectral characteristics of secondary minerals using ASTER and Sentinel-2 data. In Greece, Sykioti et al. [1] and Tompolidi et al. [4,15] identified key hydrothermal alteration minerals—kaolinite, illite, and chlorite—by their typical absorption bands in the 2.2–2.35 μm region (shortwave infrared) and spectral indices derived from ASTER and Sentinel-2 MSI bands, using the USGS Spectral Library. Ferrier et al. [35,36] also applied these procedures to map advanced argillic alteration zones associated with Fe- and Al-rich minerals, using in addition customized spectral libraries obtained from field observations. In Morocco, Bentahar et al. [22] and Baid et al. [51] successfully used ASTER and Landsat-8 OLI datasets to discriminate secondary minerals like serpentine, epidote, and limonite, from their diagnostic absorptions bands in the 0.5–2.5 μm range, using USGS and region-specific spectral libraries. With the reported examples we want to focus on the effectiveness of standard and custom spectral libraries in detecting and mapping secondary minerals in different lithologies and tectonic settings across the Mediterranean.

Custom spectral libraries are often validated by using mineralogical techniques such as X-ray diffraction (XRD), thin-section petrography, or X-ray fluorescence (XRF), ensuring the reliability of selected end members. Approximately one-third of reviewed studies explicitly combined mineralogical analysis with spectral library development [29]. From these examples, there clearly appears the need for a structured framework for regional spectral library development involving (1) robust acquisition and validation protocols (e.g., spectroscopy, XRD, and XRF measurements in the field), (2) complementing global resources like the USGS library, and (3) strategies for open-access sharing to promote broader applications in geological mapping. Building on these developments, recent research highlights the need for the following:

a

The creation of validated and customized spectral signature libraries derived from in situ geological samples (XRD, XRF, etc.) complementary to the available open-access USGS library.

b

Validating mineral identification using the reflectance spectra of Raman-validated samples, which can be convolved into satellite spectral bands.

c

An accurate convolution to satellite spectra from the available multispectral and hyper-spectral sensors, to enable automatic mineral identification.

d

Training machine learning (ML) models for pixel-based and sub-pixel analysis optimized for mineral detection, scalable to critical raw material (CRM)-rich environments.

3.3. Comparative Summary of Pixel-Based and Sub-Pixel Methods Effectiveness

Pixel-based methods, including algorithms like maximum likelihood and support vector machines (SVMs), as well as unsupervised clustering such as ISODATA and k-means, have been extensively applied in geological mapping in the Mediterranean region (

Table 3. Performance comparison of pixel-based and sub-pixel methods in geological mapping applications in the Mediterranean region.

Aspect	Pixel-Based Methods	Sub-Pixel Methods
Classification accuracy	High (>80%) in homogeneous terrains (e.g., Troodos ophiolite, Cyprus); declines (<65%) in spectrally mixed terrains like Iberian Pyrite Belt (Portugal).	Consistently higher in heterogeneous terrains; improvements of +15–25% accuracy reported in Morocco, Portugal, and Cyprus.
Sensitivity to vegetation/urban cover	Strongly affected; vegetation and urban features increase misclassification.	More robust; can extract mineral signals even under partial vegetation or weathered surfaces.
Geological context suitability	Best for simple and homogeneous terrains (volcanic islands, sedimentary basins).	Best for complex, heterogeneous, or mineralized terrains (e.g., Iberian Pyrite Belt, hydrothermal zones in Morocco).
Dependence on spectral libraries	Moderate; often uses global libraries (e.g., USGS).	High; requires accurate end member spectra (USGS or custom local libraries validated with XRD/XRF).
Computational cost	Low to moderate (fast training/classification).	High (iterative unmixing, end member optimization, ML-based decomposition).
Interoperability	High (clear class assignments per pixel).	Moderate (requires abundance maps and unmixing validation).
Integration with ground truth data	Limited; often validated with surface geology maps.	Strong; commonly integrated with XRD, XRF, and petrography for validation.

). These methods are particularly effective in geologically simple or spatially homogeneous terrains, such as volcanic islands or large sedimentary basins. For example, in Cyprus [97,98], pixel-based approaches using ASTER data have been used to delineate alteration zones in Troodos ophiolite, benefiting from the clear spectral heterogeneity between ultramafic and altered rocks. However, in complex terrains such as the Iberian Pyrite Belt in Southern Portugal [99,100], where lithological units are intermixed and spectrally homogeneous, pixel-based methods often are not able to extract the spectral variability, and this can lead to misclassification, particularly in regions with significant vegetation or urban landscapes (Table 3).

Sub-pixel methods, such as linear spectral unmixing (LSU), Multiple End member Spectral Mixture Analysis (MESMA), and more recent machine learning-based spectral decomposition techniques, enhanced performance in such complex and heterogeneous settings like the Mediterranean. In Morocco sub-pixel methods combined with hyperspectral data have successfully mapped hydrothermal alteration zones and mineralized structures at sub-pixel scales [93,94], especially when supported by reference spectra from the USGS Spectral library or custom spectral libraries for the region under investigation. Similarly, in Portugal [81,92] and Cyprus [88], sub-pixel analysis has enabled the discrimination of alteration minerals (e.g., muscovite, chlorite, and kaolinite) even within spectrally mixed pixels. These approaches are particularly valuable in mountainous or mineralized zones, where surface exposures are limited or weathered. While sub-pixel methods are computationally intensive and rely heavily on accurate end member selection, their integration with ground-truth data—such as X-ray diffraction, thin section petrography, or X-ray fluorescence—has substantially improved mapping reliability in recent studies from the Mediterranean region.

4. Discussion

4.1. Implication for Mining Exploration

An exponential growth in the demand of resources is expected as a result of the increase in global population, industrialization, digitalization and demand from developing countries. With global material demand doubled from 79 billion tonnes in 2020 to 167 billion tonnes in 2060 (OECD forecast https://www.oecd.org/en/topics/sub-issues/economic-outlook.html, accessed on 12 June 2025) the global competition for resources will become fierce in the coming decades, and the dependence of critical raw materials may soon replace the current dependence on oil [101].

The EU Green Deal recognized that the access to resources was a strategic security question for the EU. Critical raw materials (CRMs) are highly demanded for key technologies and strategic sectors such as renewable energy, e-mobility, digital, space and defense [102]. The almost complete dependence of the EU on the supply of raw materials from third countries forced the EU Commission to define in the new Industrial Strategy for the EU an Action Plan for CRM and for industry-driven raw materials alliances where the diversification of supply concerns reducing dependencies in all dimensions by increasing secondary supply of raw materials through resource efficiency and circularity, and finding alternatives to scarce raw materials [103].

One of the priorities of the EU Commission is to encourage the European production of CRMs and facilitate the launch of new mining and recycling activities to ensure the competitiveness of the EU industrial value chains: “The importance of metals and minerals to sustain businesses and the economy is particularly true for the EU, where about 30 million jobs are directly reliant on access to raw materials” (EU list of critical Raw Materials final report 2020 [104]).

The supply of CRM is therefore a priority for European industry and society. Europe has limited primary mining resources that are not being used well at the moment; therefore, new exploitation and extraction strategies to produce CRM are strategic for Europe. In this framework, the CRM Act [103] sets ambitious 2030 benchmarks for EU extraction (at least 10% annual), processing (at least 40%) and decreasing external sources from a single third country. The Mediterranean region has a complex geotectonic configuration that produces ore deposits of different types, most of which have been exploited since ancient times. They span from Volcanogenic Massive Sulfide (VMS) to epithermal, porphyry, lateritic, skarn to carbonate-replacement (MVT, CRD), and sedimentary (phosphate deposits in Tunisia and Morocco). A potential CRM resource still needs to be thoroughly characterized.

Remote sensing applications in the Mediterranean are directly relevant to the objectives of the EU CRM Act, specifically for the prospection of Cu, rare earth elements (REEs), and other CRM-bearing systems. In the Iberian Pyrite Belt of Portugal and Spain, sensors like Sentinel-2 MSI and ASTER have been successfully used to map the zones altered by hydrothermal processes associated to volcanogenic massive sulfide (VMS) deposits formation, which are major sources of Cu and Zn. In Morocco, PRISMA and Hyperion hyperspectral missions could discriminate between alteration minerals related to carbonatite complexes hosting REE mineralization whose strategic importance has increased in recent years. Similarly, in Greece (e.g., Nisyros and Milos volcanic systems) and Western Turkey, the integration of Sentinel-1 SAR and optical data has been used to detect hydrothermal systems in which advanced argillic alteration is spatially associated with epithermal and porphyry-style mineralizations now relevant for Cu and critical by-products. The reported examples demonstrate that RS contributes to the EU’s strategic priorities in securing sustainable supplies of critical raw materials, as well as to a scientific understanding of Mediterranean geology (Figure 4).

In this framework, most of the primary alteration minerals occurring in ore deposit formation are present together with secondary alteration processes (i.e., weathering of mineralized rocks) and can be used as key minerals for mining exploration to produce predictive resource assessments.

4.2. Challenges and Future Directions

Despite notable advances, remote sensing applications for geological mapping in the Mediterranean still face several challenges. Addressing these challenges is essential if remote sensing is to play a central role in supporting geologic mapping, scientific understanding, and strategic mineral exploration under the EU Critical Raw Materials (CRM) framework.

The Mediterranean landscape is highly heterogeneous, with mixtures of lithologies, weathered surfaces, vegetation, and anthropogenic cover. Pixel-based methods often misclassify such areas, while sub-pixel techniques depend heavily on accurate end member selection. A clear research priority is the systematic testing of pixel-based versus sub-pixel approaches in varied Mediterranean settings to establish best practices for different geological environments.

Most studies rely on the USGS spectral library, which does not adequately capture the mineralogical diversity of Mediterranean lithologies. This limits the accuracy and reproducibility of the classification. The development of open-access, standardized, region-specific spectral libraries built through coordinated field campaigns and validated with XRD, XRF, Raman, and petrography would significantly enhance the reliability of RS applications. Such libraries should be continuously updated and made interoperable with global datasets.

Recent advances in deep learning are new and promising in overcoming some of these challenges. Convolutional neural networks (CNNs) seem to have strong potential for lithological classification and alteration mapping in spectrally complex terrains, because they can capture both spatial and spectral features beyond the capacity of traditional machine learning methods. Similarly, deep learning-based spectral unmixing methods are arising as powerful tools for addressing non-linear mixing effects in rugged Mediterranean terrains, thereby enhancing the detection of sub-pixel mineralogical variability. Focused on the Mediterranean region, recent studies have explained practical applications of hyperspectral missions and deep learning: alteration mapping and classification in Southern Italy and SE Iberia was successfully produced from PRISMA data [105,106], while mapping semi-arid Aegean environments was successfully obtained by CNN-based semantic segmentation workflows combined with multispectral imagery with LiDAR [107]. Furthermore, mission resources and case studies summarized in the EnMAP science plan and publications highlight emerging EnMAP applications for mineral and lithological mapping [24,108,109,110], strengthening the value of spaceborne imaging spectroscopy when coupled with modern machine learning and advanced unmixing methods. Multi-sensor fusion strategies that integrate optical (Sentinel-2, PRISMA) and radar (Sentinel-1, TerraSAR-X) datasets, have proven effective in reducing vegetation interference and improving mineral identification. Integrating these approaches with region-specific spectral libraries and field validation could substantially enhance the robustness and applicability of remote sensing for CRM exploration in the Mediterranean region.

In addition to optical, hyperspectral, and radar-based approaches, it is important to highlight the contribution of satellite geophysical missions such as the Gravity Recovery and Climate Experiment (GRACE) and its follow-up (GRACE-FO) [111]. Although still rarely applied in Mediterranean mineral exploration, GRACE data provide valuable large-scale information on mass distribution and temporal gravity variations related to geodynamic processes, crustal density contrasts, and groundwater storage changes [112,113]. The importance of these parameters is increasingly recognized as complementary to traditional remote sensing methods, providing potential insights into the geodynamic controls of ore deposits formation and the sustainability of mining-related water resources. Therefore, GRACE-type data have the potential to contribute to future integrated strategies for geological mapping and critical raw materials evaluation in the Mediterranean, particularly when coupled with high-resolution optical and hyperspectral observations.

Looking ahead, the growing demand for critical raw materials (CRMs) under the EU’s Critical Raw Materials Act represents both a strategic opportunity and a research imperative for remote sensing studies in the Mediterranean region. To meet this challenge, future Mediterranean RS research should prioritize (1) building and maintaining regional spectral libraries; (2) developing standardized preprocessing and validation protocols; (3) advancing multi-sensor data fusion supported by machine learning; and (4) expanding geographical coverage beyond current case-study hotspots. Together, these efforts will enable RS studies to move beyond individual case analyses toward a more integrated operational framework that supports both geoscientific research and sustainable mineral resource management across the Mediterranean.

5. Conclusions

Remote sensing strongly increased its importance as a tool for geological mapping across the Mediterranean in recent years. Satellite platforms like ASTER, Sentinel-2, and PRISMA enabling effective detection of lithological units and alteration zones (Figure 4). Pixel-based methods are extensively applied in simpler terrains, whilst sub-pixel techniques offer improved accuracy in geologically complex or spectrally mixed areas. The integration of hyperspectral data with field-based mineralogical validation, such as XRD and XRF, has significantly improved the reliability of remote sensing interpretations. Spectral libraries—particularly the USGS Spectral Library and locally calibrated datasets—play a critical role in accurate end member selection and classification. The reported case studies from Cyprus, Morocco, Portugal, Greece, and Italy demonstrate the value of combining optical, thermal, and SAR datasets for comprehensive geological analysis, supporting early-stage mineral exploration and strategic planning under the EU Critical Raw Materials Act (CRM Act). Key conclusions and recommendations from our review include the following:

• Remote sensing, coupled with mineralogical data, offers a non-invasive and scalable approach to map mineralized zones.
• Locally developed spectral libraries are essential for improving classification accuracy in regions with unique mineralogical signatures.
• The standardization of methodologies, machine learning integration, and multi-sensor dataset merging is needed to address inconsistent methods and limited ground-truth data.
• Maps of alteration minerals can be linked to prospectivity models, highlighting zones likely to host Cu, REE, and CRMs.
• Integrating alteration, geophysical, geochemical, and structural datasets into a GIS framework enables spatially explicit resource assessments for exploration strategies.
• Expanding and standardizing regional spectral libraries will improve the trust of sub-pixel and hyperspectral analyses in complex terrains.

The implementation of these methods will strengthen the scientific basis of mineral exploration in the Mediterranean and provide decision-makers with operational tools to evaluate potential mining targets, prioritize exploration campaigns, and maximize sustainable supply chains within Europe. By explicitly connecting geological mapping with the objectives of the CRM Act, remote sensing will advance from a supporting tool to a central pillar of the sustainable resource strategy in the Mediterranean region.