Multi-Temporal Mineral Mapping in Two Torrential Basins Using PRISMA Hyperspectral Imagery

Abstract

The Sierra Minera de Cartagena-La Unión, located in southeast of the Iberian Peninsula, has been significantly impacted by historical mining activities, which resulted in environmental degradation, including acid mine drainage (AMD) and heavy metal contamination. This study evaluates the potential of PRISMA hyperspectral imagery for multi-temporal mapping of AMD-related minerals in two mining-affected drainage basins: Beal and Gorguel. Key minerals indicative of AMD—iron oxides and hydroxides (hematite, jarosite, goethite), gypsum, and aluminium-bearing clays—were identified and mapped using band ratios applied to PRISMA data acquired over five dates between 2020 and 2024. Additionally, Sentinel-2 data were incorporated in the analysis due to their higher temporal resolution to complement iron oxide and hydroxide evolution from PRISMA. Results reveal distinct temporal and spatial patterns in mineral distribution, influenced by seasonal precipitation and climatic factors. Jarosite was predominant after torrential precipitation events, reflecting recent AMD deposition, while gypsum exhibited seasonal variability linked to evaporation cycles. Goethite and hematite increased in drier conditions, indicating transitions in oxidation states. Validation using X-ray diffraction (XRD), laboratory spectral curves, and a larger time-series of Sentinel-2 imagery demonstrated strong correlations, confirming PRISMA’s effectiveness for iron oxides and hydroxides and gypsum identification and monitoring. However, challenges such as noise, striping effects, and limited image availability affected the accuracy of aluminium-bearing clay mapping and limited long-term trend analysis.

1. Introduction

The 19th and early 20th centuries marked the golden age of mining in Spain, particularly for metallic ores such as lead (Pb), zinc (Zn), copper (Cu), iron (Fe), and mercury (Hg), as well as coal [1]. This boom centred in several areas, including the southeast of the Iberian Peninsula, where lead became a the main extracted and exported mineral in Spain, driving a resurgence of private mining ventures [2].

By the 20th century, however, many of Spain’s lead deposits became less profitable due to depletion, global competition, and the 1929 recession. Despite these challenges, some mining districts, notably the Cartagena-La Unión area, experienced a resurgence in the second half of the 20th century [1]. This revival was largely due to the advancement in hydrometallurgical techniques and open-pit mining operations, which revitalised the industry [3]. Yet, the metal crisis of the 1970s, the low prices of raw materials, and the depletion of reserves led to abrupt mine closures, leaving little time or resources for environmental protection and corrective measures.

The Sierra Minera Cartagena-La Unión, located in the southeast of the Iberian Peninsula, is a good representative of this scenario. This area has been mined for silver (Ag), lead (Pb), zinc (Zn), copper (Cu), and iron (Fe) since Phoenician and Carthaginian times (3rd century BC) until its closure in 1991 [4]. As a consequence, the region has faced significant environmental impacts, particularly in the decades before the closure date due to large-scale mining and open-pit methods targeting Pb, Zn, and Fe [2,3]. These metals naturally occur as poly-metallic sulphides such as galena (PbS), sphalerite (ZnS), and pyrite (FeS₂), which were the main ores exploited [5].

In the Sierra Minera de Cartagena-La Unión, the sulphide minerals were extracted from low-grade deposits with scarce ore distribution within the host rocks [1]. This resulted in the generation of substantial volumes of waste dumps. This issue is particularly concerning given the historical use of less efficient extraction and processing techniques in this region, coupled with limited environmental awareness and controls [2]. The overall factors have contributed to the accumulation of large volumes of sulphur-rich deposits throughout the Sierra Minera de Cartagena-La Unión.

Sulphide minerals in ore deposits are typically formed under reducing conditions without the presence of oxygen, anoxic conditions in which these minerals are stable. However, when exposed to atmospheric oxygen or oxygenated waters due to mining and material-moving processes, these minerals become unstable and oxidise upon contact with air and rainwater, potentially causing acid mine drainage (AMD) [6,7]. The exposure of waste rocks to oxygenated environments creates chemically unstable conditions that trigger complex weathering reactions. This occurs due to the chemical disequilibrium between the mine waste rock and the surrounding environment. While this process occurs naturally, mining activities significantly accelerate it by increasing the exposure of sulphide minerals to weathering processes involving air, water, and microorganisms.

Pyrite (FeS₂), one of the most common sulphide minerals, is the main contributor to AMD. Its oxidation results in the release of sulphate, acid, and metals into the environment. Pyrite oxidation can proceed through several pathways, involving surface interactions with dissolved oxygen (O₂), ferric iron (Fe³⁺), and other catalysts such as certain minerals (e.g., MnO₂) or bacteria. This process can be summarised in three major steps [7,8]:

(I)

First, pyrite oxidation occurs through the interaction of iron sulphide with oxygen and water, producing ferrous iron (Fe²⁺), sulphate (SO₄²⁻), and hydrogen ions (H⁺).

(II)

Second, the ferrous iron (Fe²⁺) is oxidised to ferric iron (Fe³⁺) by dissolved oxygen, generating additional hydrogen ions and increasing solution acidity. Ferric iron (Fe³⁺) further oxidises sulphide minerals, increasing sulphate and acid production.

(III)

Third, ferric iron (Fe³⁺) hydrolyses in water, forming ferric hydroxide (Fe(OH)₃) and releasing more hydrogen ions. This hydrolysis helps remove other metals from the solution, leading to precipitation of secondary mineral [9].

These reactions can be summarised as follows (Equation (1)):

4FeS₂ + 15O₂ + 14H₂O → 4Fe(OH)₃ + 16H⁺ + 8SO₄²⁻

(1)

Secondary mineral precipitation occurs through two main mechanisms [7,9]: (1) precipitation as low-solubility oxides, hydroxides, or hydroxysulphates, such as schwertmannite (Fe₈O₈(OH)₆(SO₄)) and jarosite (KFe₃(SO₄)₂(OH)₆), which remove trace metal(oid)s from solution. Over time, these mineral phases will transform into more stable forms like goethite (FeO(OH)) or hematite (Fe₂O₃) [8]. And (2) the formation of soluble evaporitic salts, which are mainly composed of a mixture of minerals of Ca (gypsum, CaSO₄·2H₂O), Fe (copiapite, Fe²⁺Fe³⁺₄(SO₄)₆(OH)₂·20H₂O, and melanterite, Fe²⁺SO₄·7H₂O), or Mg (hexahydrite, MgSO₄·6H₂O). Climate significantly influences evaporitic sulphate salt formation; while these minerals are often ephemeral in wet climates, semiarid climates favour their persistence during long dry periods [10].

Numerous studies have assessed AMD and heavy metal distribution in Sierra Minera de Cartagena-La Unión [11,12,13,14,15]. These studies involved systematic sampling and laboratory analysis of sediment samples, followed by interpolation to generate distribution maps. While valuable, these methods are destructive, labour-intensive, and lack temporal dynamics, which does not allow for large-scale, rapid, spatial, and dynamic monitoring.

Remote sensing—particularly imaging spectroscopy—has emerged as an efficient alternative to conventional field-based methods for lithological and mineral mapping, especially in arid regions [16,17,18,19]. In this context, hyperspectral satellites offer unique opportunities for monitoring Earth’s surface by capturing reflected sunlight across a continuous spectral coverage from visible and near-infrared (VNIR) to shortwave infrared (SWIR) wavelengths (approximately 400–2500 nm).

In recent years, several spaceborne hyperspectral missions have enhanced Earth observation. These includes: DLR Earth Sensing Imaging Spectrometer (DESIS) by the German Aerospace Center (DLR) and Teledyne Brown engineering (TBE) launched in 2018 [20], PRecursore IperSpettrale della Missione Applicativa (PRISMA) launched by the Italian Space Agency (ASI) in 2019 [21]; and Earth Surface Mineral Dust Source Investigation (EMIT) by NASA [22], Hyperspectral Imager Suite (HISUI) by the Japan Aerospace Exploration Agency (JAXA) [23], and Environmental Mapping and Analysis Program (EnMAP) by the DLR, all launched in 2022 [24].

These missions have been successfully employed in various mineralogical studies, including the mapping of alteration zones in carbonate-hosted Zn-Pb deposits in Yemen [25], porphyry Cu-Au systems in Pakistan [26], carbonatite-related alteration in Namibia [27], and detecting minerals like alunite, calcite, and kaolinite in Cuprite, Nevada, United States [28], characterising and identifying mineral and rock type in India [29], and detecting copper deposits in Iran [30]. Hyperspectral data has proven effective in identifying AMD-impacted regions [16,17,18,19], as secondary iron oxides, hydroxides, and evaporitic salts show distinct spectral signatures in the VNIR and SWIR regions [31]. These spectral characteristics allow the detection and mapping of AMD-affected areas [16,17,18,19]. In the VNIR range, electronic transitions of metallic ions such as ferrous and ferric iron produce diagnostic spectral features [32,33]. This enables the identification of iron oxides and hydroxides, common AMD-related minerals. In the SWIR region, vibrational processes involving anions like hydroxyl (OH⁻) generate characteristic absorption features, allowing the detection of evaporitic minerals like gypsum [34,35]. The SWIR region can also be used to identify clays, which play a significant role in mining-contaminated environments due to their ability to adsorb heavy metals. Among the various types of clays, aluminium-bearing clays are of particular interest in this region, as they dominate the clay fraction in the local mineralogy [36]. These clays, such as kaolinite, illite/muscovite, or montmorillonite, exhibit distinct vibrational features in the SWIR region that allow for their identification and quantification. These spectral characteristics allow remote sensing techniques to successfully analyse mineral distribution from tailings dumps to the sedimentation areas where they are deposited by fluvial dynamics.

Despite these advances, most studies rely on single-date hyperspectral imagery [25,26,27,28,29,30], limiting the analysis to static mineralogical conditions. Very few studies have explored the potential of multi-temporal hyperspectral data for tracking mineralogical evolution, particularly in torrential mining basins, where surface mineralogy can change rapidly due to episodic rainfall, erosion, and sediment redistribution.

Among the current hyperspectral sensors, PRISMA and EnMAP are best suited for AMD mapping and monitoring [37]. Both are sun-synchronous satellites offering near-global coverage, 30 m spatial resolution, and consistent revisit intervals—potentially acquiring imagery every 30 days. They provide full spectral coverage from approximately 400 to 2500 nm. In contrast, DESIS, EMIT, and HISUI are mounted on the International Space Station (ISS), meaning acquisitions depend on the irregular and unpredictable station passes over the area of interest. Moreover, their spatial coverage is restricted to the ISS orbit range—approximately between 51.6° N and 51.6° S. In addition, EMIT’s coarser spatial resolution (60 m) may reduce its effectiveness for detailed mapping, and DESIS covers only the 400–1000 nm range, excluding SWIR wavelengths, which are key for identifying many alteration minerals [31].

Between PRISMA and EnMAP, PRISMA currently offers the highest number of scenes over the Sierra Minera de Cartagena-La Unión, with five scenes between 2020 and 2024, making it the most suited for multi-temporal monitoring. However, performing temporal monitoring using only hyperspectral data remains challenging due to their limited acquisition frequency. To overcome this limitation and increase the number of observations, this study incorporates Sentinel-2 data, a multi-spectral satellite mission. Although Sentinel-2 lacks the spectral resolution of hyperspectral sensors—with only 13 discrete bands across VNIR and NIR—it provides a higher temporal resolution (5–6 days) and higher spatial resolution (10 m for bands used in this study) (

Table 1. Technical specifications of PRISMA and Sentinel-2 satellites.

Specification	PRISMA	Sentinel-2
Orbit	Sun-synchronous, ~615 km altitude	Sun-synchronous, ~786 km altitude
Swath Width	30 km	290 km
Revisit Time (Equator)	~29 days	~5 days (S2A + S2B)
Spatial Resolution	30 m (hyperspectral) 5 m (panchromatic)	10 m (bands: 2, 3, 4, 8) 20 m (bands: 5, 6, 7, 8A, 11, 12) 60 m (bands: 1, 9, 10)
Spectral Range	VNIR: 400–1010 nm SWIR: 920–2500 Panchromatic: 400–700 nm	443–2190 nm
Spectral Width	≤12 nm	15–180 nm (depending on band)
Number of Bands	VNIR: 66 bands SWIR 173 bands Panchromatic: 1 band	13 bands
Temporal Coverage	2019–present	2015 (S2A) and 2017 (S2B)–present
Scene Size	~30 km × 30 km	~290 km × ~290 km

). Although PRISMA’s 5 m panchromatic band could improve spatial detail, preserving spectral integrity is key for mineral identification, so pansharpening is avoided due to potential distortions [38,39]. These attributes make Sentinel-2 ideal for generating dense time series, enabling temporal validation and contextual interpretation of PRISMA-derived mineralogical maps. The combination of both datasets, which are independent data sources, thus supports a more reliable assessment of AMD evolution in the mining district.

Similar approaches combining multi-spectral and hyperspectral data can be found in the literature [40,41]. For example, ref. [40] used Sentinel-2 for its frequent (5-day) and high-resolution images alongside DESIS, which provided superior spectral resolution but fewer scenes, to map the 2021 Cumbre Vieja eruption in La Palma (Spain). In another case, ref. [41] combined EO-1 Hyperion hyperspectral data with long-term Landsat time series to study salt crust dynamics in Namibia. However, to our knowledge, no previous study has implemented a combined multi-temporal approach using PRISMA and Sentinel-2 to monitor AMD-related surface mineral changes in mining-affected torrential basins.

This study assesses whether infrequent PRISMA hyperspectral imagery, combined with the frequent Sentinel-2 multi-spectral data, can effectively monitor spatial and temporal patterns of AMD-related mineral groups across two mining-affected catchments. The aim is to (i) detect spatial distributions and temporal variations in key mineralogical indicators of AMD, (ii) evaluate the integration of hyperspectral and multi-spectral data for operational monitoring, and (iii) evaluate the added value of hyperspectral data in identifying diagnostic absorptions not captured by multi-spectral sensors.

2. Study Area

The Sierra Minera de Cartagena-La Unión, located in the southeast of the Iberian Peninsula, is a coastal mountain range that extends 26 km from east to west, from Cabo de Palos to the city of Cartagena (Figure 1). Bounded by the Mediterranean Sea to the south and the east and by the Campo de Cartagena basin and the Mar Menor to the north, this area covers approximately 100 km². The highest elevation is the “Peña del Águila” at 431 m above sea level.

Historically, the Sierra Minera de Cartagena-La Unión has been extensively impacted by mining activities. Twelve open-pit mines were excavated, along with the drilling of more than 3000 wells and the construction of several kilometres of mining galleries [42]. All this mining activity has resulted in hundreds of piles of mining wastes located across the landscape, significantly altering its natural state.

The climate in the Sierra Minera de Cartagena-La Unión is semiarid Mediterranean, with an average annual precipitation of 300 mm. Precipitations are sporadic and concentrated in short, intense events. The average temperature ranges from 12 °C in January to 28 °C in August, with an average annual temperature of 17 °C. The potential evapotranspiration is approximately 900 mm/year [43].

As in other parts of the SE of the Iberian Peninsula, rainfall dynamics are closely associated with the intrusion of cold air masses (−23 °C) into the upper layers of the atmosphere, generating the so-called Depresión Aislada en Niveles Altos (DANA) or Isolated High-Altitude Depression, mainly during the months of October and November [44]. The rapid evaporation rates after the summer months trigger a localised and rapid vertical development of convective clouds, often giving rise to torrential rainfall [45].

These climate conditions, combined with the region’s geology and topography, have resulted in the formation of ephemeral watercourses known as “ramblas” (local name for dry watercourses due to the aridity of the climate). These dry riverbeds can experience sudden, significant flows during intense precipitation events, contributing to transport of sediment and potentially dispersing heavy metals downstream from waste deposits. The hydrographical network consists of ten watercourses. Five of them drain into the Mediterranean Sea and the other five into the Mar Menor (Figure 1), a coastal saltwater lagoon protected at national and international levels. The northern ramblas (Mar Menor ramblas) have gentler slopes compared to the southern ramblas (Mediterranean Sea ramblas), which descend sharply towards the Mediterranean coast [42].

Soil development in the Sierra Minera de Cartagena-La Unión is limited due to the arid climate, with predominant types including Gypsic Regosol, Haplic Cambisol, Haplic Calcisol, and Petrocalcic Calcisol [46]. Vegetation primarily consists of shrubs and small areas which are reforested with Pinus halepensis.

2.1. Geological Context

The Sierra Minera de Cartagena-La Unión belongs to the Internal Zones of the Betic Cordillera, an arc-shaped mountain belt formed by the convergence of the Iberian and African plates during the Mesozoic and Cenozoic periods [47] (Figure 2).

The geomorphological configuration is related to the tectonic activity of sinistral strike–slip NE-SW-oriented faults. An example is the ENE-WSE Cartagena-La Unión Fault or Southern Fault of the Campo de Cartagena, which condition the distribution of the relief in this study area, with the Sierra Minera de Cartagena-La Unión in the south as the main high relief and the Campo de Cartagena as the main sedimentary basin in the north [48,49].

The presence of significant differences in height enables the genesis of large quantities of debris on the relief slopes, which are remobilised during episodes of torrential rains. According to the orographic, lithological, and climatic characteristics and, additionally, the limited fixation power of the debris by the scarce existing vegetation cover, erosion and transportation processes are normally very effective [50]. In the mountain (mainly metamorphic) reliefs, most of the materials are remobilised by gravitational phenomena to the bottom of ravines and ramblas. These materials are then transported during the rare, but intense, heavy rainfall episodes characteristic of the area. In this way, the torrential processes and landforms (alluvial fans) are dominant in the Campo de Cartagena [47], and also in the ramblas draining to the Mediterranean Sea, as small fan-deltas in the south. Anthropogenic landforms are widely represented throughout the area as open pits, mine waste dumps, and mining ponds.

2.2. Studied Basins

From the ten ramblas in the study area, two were studied: in the Mar Menor, the Rambla del Beal and in the Mediterranean Sea the Rambla del Gorguel. Rambla del Beal was chosen because it has the most natural and least anthropogenically altered alluvial fan. Other ramblas in the Mar Menor region are located near agricultural or urban areas, which limits their suitability for analysing natural sediment dynamics; in the Mediterranean, the Gorguel drainage was selected. Other basins were excluded for the following reasons: the Rambla de Escombreras is heavily modified by urban development at its mouth; Rambla de Cobaticas is minimally impacted by mining activity; and both Rambla de Portmán and Barranco del Moro are influenced by other mining problematics that fall outside the scope of this study. This study, therefore, focuses on the alluvial fans developed at the mouths of the Beal and Gorguel drainage basins, where well-developed alluvial fans make them suitable for assessing the spatial distribution of AMD-related minerals (Figure 3).

The Beal and Gorguel basins share a similar geomorphological structure, with mining residues concentrated in their headwaters, which are transported downstream during precipitation events, where deposition occurs within alluvial fans [50]. The Beal basin drains into the Mar Menor, while the Gorguel basin discharges into the Mediterranean Sea.

However, the distinct morphologies of each basin lead to differing sedimentary dynamics [50]. The Beal basin, covering 7.60 km² with a gentle average slope of 1.90%, is more extensive which allows for a more gradual dispersal of sediments. Mine wastes, covering nearly 10% of the basin’s area, are mainly located in the headwaters section near the villages of Llano del Beal and El Beal [42]. These residues are then transported downstream, accumulating in a broad alluvial fan that intersects the Lo Poyo saltmarsh. In contrast, the Gorguel basin, covering 3.4 km² with a steeper average slope of 4.61%, has a more abrupt sediment transfer. With over 20% of its surface affected by mining residues, the basin transports contaminated materials through a narrow riverbed before deposition occurs in a smaller alluvial fan at Gorguel Bay [42], where a small settlement is inhabited mainly during the summer months.

3. Materials and Method

This section describes (a) the PRISMA and Sentinel-2 imagery; (b) the spectral analysis focusing on the application of band ratios to identify iron oxides and hydroxides, gypsum, and aluminium-bearing clays; and (c) the validation with X-ray diffraction (XRD) analysis, laboratory spectral curves, and the comparison of PRISMA and Sentinel-2 data (Figure 4).

3.1. Material

3.1.1. Image Datasets

Freely available data from two sensors PRISMA and Sentinel-2 were used in this study. PRISMA Level 2D images were selected among currently available hyperspectral sensors because they provided the highest number of hyperspectral scenes (five) over the study area, enabling a multi-temporal analysis of various AMD-related minerals due to their high spectral resolution. Sentinel-2 Level 2A images were used to complement PRISMA data due to their higher spatial resolution (10 m for the bands used in this study) and frequent temporal coverage (5–6 days revisit time), making them suitable for enlarging AMD temporal data and helping with contextual interpretation of PRISMA-derived mineralogical maps.

PRISMA, a sun-synchronous hyperspectral sensor developed by OHB Italy and launched by the Italian Space Agency (ASI) in March 2019, provides 239 spectral channels across a wavelength range of 400–2500 nm [21,52]. It captures spectral information in the VNIR (66 channels from 400–1100 nm) and SWIR (173 channels from 920–2500 nm) regions, with a spectral width of ≤14 nm and a calibration accuracy of ±0.1 nm. It has a spatial resolution of 30 m for hyperspectral bands and 5 m for the panchromatic band. Each PRISMA scene covers an area of 30 × 30 km with a 29-day revisit time.

Five PRISMA scenes were acquired from the ASI portal (https://prisma.asi.it/, accessed on 1 November 2024) (Table 1). Four scenes were cloud-free, while one had 20% cloud cover over the Mediterranean Sea. From the overlapping wavelength region between the VNIR and the SWIR spectrometer channels, only the bands from the VNIR cube were used, leaving 230 bands in total. The priority of the VNIR cube is based on the objective of the study, which is the identification of key mineral indicators of acid mine drainage, specifically iron oxides, which exhibit spectral absorption features in the VNIR region.

The Sentinel-2 A/B satellites, equipped with the MultiSpectral Instrument (MSI) sensor, capture spectral data between 440 nm and 2200 nm, distributed in 13 bands [53]. Sentinel-2 provides a spatial resolution of 10 m (VIS, NIR), 20 m (NIR, SWIR), and 60 m (auxiliary bands) with a 5- to 6-day revisit time. In this study, only the 10 m resolution bands B3 and B4 were used for band ratioing to detect the presence of iron oxides.

Two different Sentinel-2 datasets were employed. The first set included five images closely matching the PRISMA acquisition dates (

Table 2. PRISMA and Sentinel-2 acquisition dates and cloud cover over the Sierra Minera de Cartagena-La Unión.

PRISMA Date	Cloud Cover (%)	Sentinel-2 Date	Cloud Cover (%)
17/02/2020	20.24	15/02/2020	5.8
17/03/2021	0.12	11/03/2021	1.00
24/06/2023	0.00	24/06/2023	0.01
13/10/2023	0.00	02/10/2023	0.00
15/04/2024	0.08	14/04/2024	0.91

), which were obtained from the Copernicus Open Access Hub. The second set involved a broader multi-temporal analysis using Google Earth Engine (GEE, version released on 18 November 2024) with a time series from 1 January 2019 to 1 May 2024, from which only images with less than 10% cloud cover were selected, resulting in a total of 200 images.

3.1.2. Sample Collection

For validation and correct interpretation in terms of mineral composition of the image datasets, a total of 52 sediment samples (Figure 5) were collected during two field campaigns in October 2021 and October 2023, the latter (42 sediment samples) coinciding with a PRISMA image acquisition. The samples belong to a larger sample collection and for this reason the sample number of these (52) is not correlative (see Appendix A).

3.2. Analysis

3.2.1. Spectral Analysis

To minimise vegetation impact on the iron absorption feature and eliminate noise from water bodies, the normalised difference vegetation index (NDVI) (Equation (2)) and normalised difference water index (NDWI) (Equation (3)) were calculated for each image from both sensors. Pixels above 0.3 DN were masked for vegetation and pixels below 0.0 DN were masked for water (Figure 6).

Figure 6. (A): Vegetation mask, (B): water mask, and (C): combined mask of vegetation and water from PRISMA and Sentinel-2 masks. Solid blue lines indicate the fluvial channels within the two study basins, while dashed blue lines represent channels in the neighboring basins. Red lines indicate the boundaries of the two study basins.

Figure 6. (A): Vegetation mask, (B): water mask, and (C): combined mask of vegetation and water from PRISMA and Sentinel-2 masks. Solid blue lines indicate the fluvial channels within the two study basins, while dashed blue lines represent channels in the neighboring basins. Red lines indicate the boundaries of the two study basins.

$$NDVI={\displaystyle \frac{NIR-Red}{NIR+Red}}$$

(2)

$$NDWI={\displaystyle \frac{Green-NIR}{Green+NIR}}$$

(3)

where NIR corresponds to the near-infrared band,

Red corresponds to the red band,

Green corresponds to the green band.

For Sentinel-2, the NIR band corresponds to Band 8 (842 nm), the Red band to Band 4 (665 nm), and the Green band to Band 3 (560 nm). To ensure direct comparability between both sensors, PRISMA hyperspectral data were spectrally resampled to match Sentinel-2’s spectral resolution. Specifically, PRISMA Bands 22–26 (542–572 nm) were averaged to represent the Green band, Bands 46–51 (655–685 nm) for the Red band, and Bands 48–59 (785–902 nm) for the NIR band. Once the vegetation and water masks were generated for both sensors, the Sentinel-2 mask was resampled to PRISMA’s 30 × 30 m grid. This step was necessary to ensure consistency, as the coarser spatial resolution of PRISMA led to a higher number of vegetated pixels being excluded compared to Sentinel-2. The two masks were then combined to ensure that both sensors excluded the same pixels (Figure 6).

After the preprocessing, spectral analysis was conducted using the Relative Band Depth ratio algorithm [54], which highlights the relative abundance of a mineral (or mineral group) based on the depth of an absorption feature with respect to the shoulder pads that frame it. This technique enhances spectral differences between bands and minimises topographic influences [55]. The result is a new image that enables the visualisation of wavelength position (chemical composition) and absorption depth (quantification).

For this study, three band ratios—based on the Relative Band Depth algorithm—were calculated for the three studied mineral groups: iron oxides and hydroxides, gypsum, and aluminium-bearing clays. These ratios were applied to PRISMA data, while only iron oxides and hydroxides could be applied to Sentinel-2, as its bands do not cover the absorption features of gypsum and aluminium-bearing clays. Therefore, the iron oxide and hydroxide band ratio was used to compare the temporal trends observed in PRISMA and Sentinel-2 and to assess PRISMA’s sensitivity in detecting these minerals.

Iron Oxides and Hydroxides

According to ref. [56,57], iron oxides show high reflectivity in the red (600–700 nm) and low reflectivity in the green (500–570 nm) intervals of the electromagnetic spectrum. Based on these spectral characteristics, ref. [57,58,59] proposed a ratio (Equation (4)) which provides a proxy to classify iron oxides and hydroxides (hematite, goethite, and jarosite) by their colour response in the visible spectrum. As described in ref. [59], high ratio values suggest dominance of iron phases with stronger red reflectance (e.g., hematite-rich areas), intermediate values indicate phases with prominent green reflectance (e.g., jarosite-rich areas), and low values are typically associated to iron phases with more yellowish reflectance (e.g., goethite-rich areas). For this ratio, ref. [59] proposed threshold values to guide interpretation: a lower limit of 0.5 as the minimum indicative value for the presence of goethite and an upper limit of 3.3 as the maximum value indicative of the presence of hematite. This classification should be considered as a soft classification [57,58,59]. Unlike hard classification methods—which assign each pixel to a single, discrete class—soft classification provides continuous values that reflect the likelihood, or dominance, of a particular feature or material. In this context, band ratio values are indicative of dominant spectral responses but should not be interpreted as exact mineral identifications due to potential spectral overlap and mineral mixing [59].

$$Iron oxides and hydroxides Ratio={\displaystyle \frac{Green}{Red}}$$

(4)

For Sentinel-2, Band 3 (543–578 nm) was used as the green wavelength and Band 4 (650–680 nm) as the red wavelength (Figure 7). To ensure a direct comparison between both sensors, PRISMA bands were spectrally resampled to match Sentinel-2’s spectral resolution as was performed with vegetation and water masks. In this case, Bands 22 to 26 (542–572 nm) were used for green and Bands 46 to 51 (655–685 nm) for red (Figure 7). This band ratio was calculated for both Sentinel-2 datasets—the five Sentinel-2 images matching PRISMA acquisition dates and the entire series of cloud-free Sentinel-2 images from January 2019 to May 2024—as well as the five PRISMA images. In this way, the capability to monitor iron oxides and hydroxides combining both sensors was validated, as well as the reliability of PRISMA results for temporal evolution using a simple and replicable metric. In this case, Sentinel-2 did not need its applicability to be validated due to its known stability and proven capability to map iron-bearing minerals [60,61]. The resulting maps were stretched using the same linear stretch parameters and consistent threshold values. This approach was applied to maintain visual consistency and minimise interpretation bias across different dates and the two sensors.

Gypsum

Gypsum displays several distinct absorption features in the SWIR spectral region, notably around 1500 nm, as well as at 1750 nm, 1900 nm, and 2200 nm (Figure 7). The most notable feature is around 1500 nm; however, this overlaps with strong atmospheric water vapor absorption, making it unsuitable for remote sensing applications. Similarly, the 1900 nm feature is also affected by water vapor. The 2200 nm feature can be confused with clay absorption, particularly in soil mixtures [63,64]. Therefore, the 1750 nm absorption feature is chosen as the most effective discriminator for gypsum.

An empirically proposed ratio for gypsum identification based on PRISMA’s band configuration is given by Equation (5):

$$Gypsum Ratio={\displaystyle \frac{R1667}{R1746}}={\displaystyle \frac{B72s}{B80s}}$$

(5)

where R is the reflectance value in nm and B is the corresponding PRISMA spectral band, and s indicates that the band corresponds to the SWIR cube of PRISMA.

Aluminium-Bearing Clays

Aluminium-bearing clays exhibit distinct absorption features at approximately 1400 nm, 1900 nm, and 2200 nm. However, the first two overlap with atmospheric water vapor, limiting their use in remote sensing analysis. The 2200 nm feature, attributed to OH stretching and Al-OH bending modes, is thus the main spectral indicator for these clays (Figure 7).

As previously mentioned, the 2200 nm spectral feature overlaps for both aluminium-bearing clays and gypsum. The presence of one mineral significantly influences the spectra of the other, particularly affecting the 1400 nm and 1900 nm features. However, the 2200 nm feature is more stable, with the clay-related absorption centred around 2208 nm, which shows limited interaction with the nearby gypsum absorption at 2217 nm. This stability makes the 2200 nm feature the best indicator for aluminium-bearing clays, despite its overlap with gypsum [63,64].

An empirically proposed band ratio for detecting aluminium-bearing clays using PRISMA’s band configuration is provided by Equation (6):

$$Al-bearing clays={\displaystyle \frac{R2135}{R2206}}={\displaystyle \frac{B124s}{B133s}}$$

(6)

where R is the reflectance value in nm and B is the corresponding PRISMA spectral band and s indicates that the band corresponds to the SWIR cube of PRISMA.

3.2.2. Validation

The methodology validation involved three approaches. First, PRISMA scenes were compared with Sentinel-2, which has proven its capability to effectively map secondary minerals resulting from pyrite oxidation, such as iron oxides and hydroxides [60,61], including within the study area [8]. To ensure temporal consistency, the five PRISMA scenes were compared with corresponding cloud-free Sentinel-2 images acquired within a maximum time frame of ±10 days from the PRISMA acquisitions (Table 1). This comparison aimed to validate the reliability of PRISMA mineral mapping and to assess whether this sensor noise or artifacts compromised consistently detecting iron oxides. Additionally, these images were contextualised within the broader multi-temporal dataset from Sentinel-2 to further evaluate the compatibility of both sensors and results in terms of consistency across different dates. For the pixel-wise comparison with PRISMA, the corresponding Sentinel-2 images were resampled to match PRISMA’s 30 m spatial resolution grid. In contrast, the extended multi-temporal Sentinel-2 dataset used to contextualise these comparisons retained its original 10 m resolution.

Second, PRISMA data from October 2023 were validated with X-ray diffraction (XRD). XRD is a non-destructive technique that uses X-rays to identify minerals in powdered rock or sediment samples [65]. The X-rays interact with the mineral’s crystal structure, creating a diffraction pattern unique to each mineral [65,66]. In this study, a Bruker D2 PHASER X-ray Diffractometer (Bruker AXS, Berlin, Germany) was used to analyse the 52 powdered sediment samples to identify and quantify mineral phases with the DIFFRACT.EVA software (version 6.0; Bruker AXS, Berlin, Germany).

Last, for the 52 sediment samples, their spectral curve was measured in the laboratory using an ASD FieldSpec-4 Standard-Res Spectroradiometer (Analytical Spectral Devices, Inc. Boulder, CO, USA). The samples were air-dried before measurements. Spectral measurements were taken in the VNIR-SWIR range (350–2500 nm) with a spectral resolution of 3 nm for 350–1000 nm and 5 nm for 1000–2500 nm. A Spectralon panel was used as white reference and was measured before each sample. All samples were measured at three different points using a contact probe, and the average of the three measurements was used to generate the final spectral curve. These laboratory-derived spectral curves were then compared to the corresponding pixels in the PRISMA October 2023 image.

4. Results

This section will first show the comparison of band ratios between PRISMA and the two Sentinel-2 images sets, including the five matching PRISMA acquisition dates. Then, the spatial and temporal distribution through image band ratioing of iron oxides and hydroxides, gypsum, and aluminium-bearing clays with PRISMA will follow. The latter is focused on the lower parts of both the Beal and Gorguel basins, as the lower areas are where sedimentation processes are dominant.

4.1. Cross-Sensor Comparison

The comparative analysis of iron oxide and hydroxide ratios derived from Sentinel-2 and PRISMA is shown in Figure 8. The top graph shows the average values of the iron oxide and hydroxide ratio (as described in Section 3.2) for each satellite, plotted alongside precipitation data from January 2019 to May 2024, obtained from the Spanish Weather Agency (Agencia Española de Meteorología, AEMET). As an example, the bottom images display maps of the iron oxide and hydroxide ratio for 17 February 2020, for both Sentinel-2 and PRISMA.

In the top graph of Figure 8, PRISMA values fall within the range of the Sentinel-2 time-series values, with PRISMA generally displaying slightly higher values. When compared with the precipitation data, Sentinel-2 shows a clear correlation between increased mean ratio values and precipitation events. While less evident in PRISMA due to the limited number of images, the same pattern can also be observed. This holds particularly for the October 2023 image, in which the mean value increased after a precipitation event on 2 September 2023.

A visual comparison of the Sentinel-2 and PRISMA maps (bottom of Figure 8) shows that both the two sensors have largely similar mineral distribution patterns, with only minor differences. In both maps, high ratio values associated with hematite are located in agricultural areas, while lower values associated with goethite and jarosite are mainly found near the watercourses, marshes, and in the Sierra Minera de Cartagena-La Unión. The PRISMA map appears slightly lighter than Sentinel-2, as also observed in the graph (top of Figure 8), which results in an overrepresentation of jarosite and an underrepresentation of goethite when comparing PRISMA with Sentinel-2.

A regression analysis was performed for the iron oxide and hydroxide band ratio values in each pixel within the study area, comparing the five PRISMA images with the closest cloud-free Sentinel-2 images (Table 1). The results show a strong correlation between PRISMA and Sentinel-2 data (Figure 9), with R² values of 0.81 for February 2020, 0.82 for March 2021, 0.83 for both June and October 2023, and 0.85 for April 2024.

4.2. Iron Oxides and Hydroxides

The spatial and temporal distribution of iron oxides and hydroxides in the lower parts of the Beal and the Gorguel basins is shown in Figure 10 and Figure 11, respectively.

In the lower part of the Beal basin (Figure 10), areas with intermediate ratio values—associated with jarosite-dominated spectral responses—predominate in all the images, mainly located along the alluvial fan associated with this rambla near its main active channel. Higher ratio values—suggesting hematite-rich areas—are found in two areas of the basin: a small region next to agriculture soils in the southwest of the image and in the northwest area in a small area in the distal part of the alluvial fan. This area is particularly noticeable in the 17 February 2020 (Figure 10A) and 17 March 2021 (Figure 10B) images, suggesting that rainfall events may have influenced hematite precipitation. In contrast, lower ratio values—indicating a goethite dominance—are initially scarce but become more abundant in the final three images (Figure 10C–E), especially in the distal part of the alluvial fan in the 15 April 2024 scene (Figure 10E).

In the Gorguel basin (Figure 11), intermediate ratio values—which suggest the presence of jarosite-rich areas—dominate most of the scenes across the alluvial fan. However, in the 17 March 2021 scene (Figure 11B) areas with higher ratio values—suggesting hematite-rich surfaces—become more predominant. This distribution correlates with the deposit observed in the northwest area of the Beal basin (Figure 11B), also suggesting a precipitation event that caused mineral transport and deposition. This deposit is also visible in the 24 June 2023 image (Figure 11C). Lower ratio values, possibly associated with goethite-rich surfaces, are found in the distal parts of the alluvial fan, following the coastline, with the highest concentrations observed in the last two scenes (Figure 11D,E), similar to the pattern of the Beal basin (Figure 11).

4.3. Gypsum

The results associated with the spatial and temporal gypsum distribution in the lower parts of both the Beal and Gorguel basins obtained from the gypsum image ratio (Section 3.2) are shown in Figure 12 and Figure 13, respectively.

In the Beal basin, gypsum is present in all the scenes (Figure 12), with higher concentrations during spring (Figure 12A–C,E), coinciding with a major torrential event (Figure 12A) and two dry periods (Figure 12C,E). On the other hand, lower concentrations are observed in autumn (Figure 12D), after moderate precipitations, as well as in Figure 12B, following another event of moderate precipitation. This pattern suggests a solubility-driven behaviour of gypsum, where its lowest concentrations occur after moderate rainfall, while higher levels are observed during dry periods or after torrential events, likely due to sediment transport. Initially, gypsum is evenly distributed within the fan (Figure 12A), but over time, concentration increases towards the basin mouth, while gypsum content in the upper part of the fan decreases, disappearing by the final image (15 April 2024). The last three scenes (Figure 12C–E) show a gap in gypsum distribution at the basin mouth, corresponding to a pedestrian path and human activity.

In the Gorguel basin (Figure 13), gypsum is present in all the scenes, with higher concentrations observed during spring (Figure 13A–C,E) and lower concentrations in autumn (Figure 13D), similar to the pattern observed in the Beal basin (Figure 12). Gypsum is distributed across most of the alluvial fan but is less abundant in the most distal areas.

4.4. Aluminium-Bearing Clays

The spatial and temporal distributions of aluminium-bearing clays within the lower parts of the Beal and the Gorguel basins are shown in Figure 14 and Figure 15, respectively. Compared to the previous studied minerals, aluminium-bearing clay content mapping shows a more diffuse spatial distribution pattern with a noticeable “salt and pepper” effect. Nevertheless, some general trends can be observed in both basins. In the Beal basin, aluminium-bearing clays are present in all the scenes (Figure 14), with the highest content in the last two scenes (Figure 14D,E), especially in the autumn scene (Figure 14D). Aluminium-bearing clays are mainly located at the distal area of the alluvial fan.

In the Gorguel basin (Figure 15), aluminium-bearing clay content is low in the first two scenes (Figure 15A,B). However, in the last three scenes (Figure 15C–E), a moderate content is observed across the alluvial fan.

4.5. Data Validation

The mineral phases identified through XRD analysis and the laboratory spectra of the 52 sediment samples collected in the study area, along with the corresponding minerals identified using the PRISMA satellite image from 13/10/2023, are shown in

Table A1. Location and characteristics of sediment samples measured by XRD and PRISMA in October 2023.

Sample ID	Latitude (UTM)	Longitude (UTM)	XRD A	PRISMA B October 2023	Laboratory Spectra C
SM-001	692622	4170952	Qrtz, Ha, Kln, Gpsm, Jrs, (Gth)	Jrs, Gpsm, Al-clays	Jrs, Gpsm, Al-clays
SM-002	692653	4170791	Qrtz, Gpsm, Jrs, (Gth)	Gth, Gpsm	Gth, Gpsm
SM-003	692770	4170765	Qrtz, Ha, Ilt, Kln, Jrs, (Gth)	Jrs, Gpsm, Al-clays	Jrs, Gpsm, Al-clays
SM-004	692063	4170197	Qrtz, Clct, Kln, (Ilt) (Hem)	Hem, Al-clays	Hem, Al-clays
SM-005	691311	4169486	Qrtz, Clct, Kln, (Ilt) (Hem)	Hem, Al-clays	Hem, Al-clays
SM-006	691749	4169460	Qrtz, Gpsm, Jrs, Gth, (Ilt)	Jrs, Gpsm	Jrs, Gpsm, Al-clays
SM-007	692499	4170225	Qrtz, Gpsm, Jrs, (Ilt)	Jrs, Gpsm	Jrs, Gpsm, Al-clays
SM-008	690642	4167014	Kln, Qrtz, Jrs, Gpsm, (Ilt)	Jrs, Gpsm	Jrs, Gpsm, Al-clays
SM-009	690671	4166309	Qrtz, Dlmt, Cal, Gpsm, Jrs, (Ilt)	Jrs, Gpsm	Jrs, Gpsm, Al-clays
SM-012	691280	4169469	Qrtz, Clct, Kln, (Ilt) (Hem)	Hem, Al-clays	Hem, Al-clays
SM-016	686034	4165850	Qrtz, Clct, Kln, Ilt (Al)	Al-clays	Al-clays
SM-017	686821	4173970	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-018	687022	4173314	Qrtz, Clct, Kln, Ilt		Al-clays
SM-019	687234	4172652	Qrtz, Clct, Ilt, (Kln), (Hem)	Hem, Al-clays	Hem, Al-clays
SM-020	688393	4173901	Qrtz, Clct, Ilt, (Kln)	Al-clays	Al-clays
SM-021	688208	4172949	Qrtz, Clct, Ilt, (Kln), (Hem)	Hem, Al-clays	Hem, Al-clays
SM-022	688144	4172943	Qrtz, Clct, Ilt, (Kln)	Al-clays	Al-clays
SM-023	690300	4172194	Qrtz, Clct, Ilt, (Kln), (Hem)	Hem, Al-clays	Hem, Al-clays
SM-024	690578	4173576	Qrtz, Clct, Ilt, (Kln)	Al-clays	Al-clays
SM-025	691136	4171422	Qrtz, Clct, Ilt, (Kln)	Al-clays	Al-clays
SM-026	691279	4171054	Qrtz, Clct, Ilt, (Kln)	Al-clays	Al-clays
SM-027	690547	4168433	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-028	690801	4166932	Qrtz, Kln, Ilt (Gth)	Al-clays	Al-clays
SM-032	689157	4165983	Qrtz, Kln, Ilt (Gth)	Gth, Al-clays	Gth, Al-clays
SM-033	689100	4162862	Qrtz, Gpsm, Jrs, Gth, (Ilt)	Jrs, Al-clays	Jrs, Al-clays
SM-034	688985	4163013	Qrtz, Jrs, Gth, (Ilt)	Jrs, Al-clays	Jrs, Al-clays
SM-037	693387	4163543	Qrtz, Clct, Kln, Ilt (Hem)	Al-clays	Al-clays
SM-038	694656	4163533	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-040	693171	4164411	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-041	695693	4165457	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-042	695601	4165532	Qrtz, Clct, Kln, Ilt		Al-clays
SM-043	696901	4166148	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-044	698423	4165822	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-048	696988	4166618	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-049	696847	4166584	Qrtz, Clct, Kln, (Ilt)		Al-clays
SM-051	696369	4169218	Qrtz, Clct, Kln, Gpsm (Ilt)	Gpsm, Al-clays	Gpsm, Al-clays
SM-052	694262	4169509	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-053	694072	4167346	Qrtz, Clct, Kln, (Ilt)	Al-clays	Al-clays
SM-054	694005	4167439	Qrtz, Clct, Kln, (Ilt) (Hem)	Hem, Al-clays	Hem, Al-clays
SM-055	694030	4167398	Qrtz, Clct, Ilt, (Kln)	Al-clays	Al-clays
SM-056	693324	4166865	Qrtz, Clct, Kln, Ilt	Al-clays	Al-clays
SM-060	687554	4161144	Gpsm, Qrtz, Gth, (Ilt)	Gth, Gpsm, Al-clays	Gth, Gpsm, Al-clays
SM-061	687287	4161083	Gpsm, Qrtz, Jrs, (Ilt)	Jar, Gpsm, Al-clays	Jar, Gpsm, Al-clays
SM-065	684979	4160908	Qrtz, Clct, Kln, (Ilt)	Al-clays	Al-clays
SM-066	683410	4161714	Qrtz, Clct, Kln, (Ilt)	Al-clays	Al-clays
SM-067	683238	4161640	Qrtz, Clct, Kln, (Ilt)	Al-clays	Al-clays
SM-068	683593	4161880	Qrtz, Clct, Kln, (Ilt)		Al-clays
SM-069	684342	4164688	Qrtz, Clct, Kln, (Ilt)	Al-clays	Al-clays
SM-070	681045	4164535	Qrtz, Clct, Kln, (Ilt)	Al-clays	Al-clays
SM-071	680868	4164654	Qrtz, Clct, Kln, (Ilt)		Al-clays
SM-072	682822	4167063	Qrtz, Clct, Kln, (Ilt)	Al-clays	Al-clays
SM-073	683370	4166746	Qrtz, Clct, Kln, (Ilt)	Al-clays	Al-clays

Al Alunite, Al-clays Aluminium-bearing clays, Qrtz Quartz, Ha Halite, Kln Kaolinite, Gpsm Gypsum, Jrs Jarosite, Gth Goethite, Hem Hematite, Clct Calcite. ^A XDR data: major phase (>20% by weight) in bold, minor phase (5–20% by weight) as plain text; () indicates trace phase (<5% by weight). ^B PRISMA data: plain text indicates moderate to high content for Al-clays and Gpsm. For Jrs, He, and Gth plain text only indicates presence of the mineral. ^C Laboratory spectral data: Dominant iron oxide and hydroxide phases were identified based on their diagnostic minimum wavelength features: Hem at 880 nm, Gth at 960 nm, and Jrs at 930 nm. Gpsm was determined by the absorption feature at 1750 nm, while Al-clays were identified by the absorption feature around 2200 nm.

. The chosen PRISMA scene is the closest in time to the second field campaign, during which the majority of the sediment samples (42) were collected.

The comparison between XRD results and PRISMA data (Table A1) for iron oxides and hydroxides shows consistent agreement. In all samples where XRD identified one of the three iron-bearing studied minerals (hematite, goethite, or jarosite), PRISMA has also detected the same mineral. This correlation is also present when mixtures of two of these minerals occur: where XRD identified two phases of iron-bearing minerals within a single sample (SM-01, SM-02, SM-03, SM-33, SM-34), PRISMA was limited to classifying the pixel as containing only one mineral. In these cases, PRISMA was able to identify the phase with the higher abundance.

For gypsum, its identification was consistent across all samples. Whenever XRD analysis detected gypsum, PRISMA also identified this mineral in the corresponding pixel.

The validation for the clay minerals, specifically illite and kaolinite—both indicative of aluminium-bearing clays—shows that PRISMA results are generally aligned with XRD identifications. In most samples where XRD identified kaolinite and/or illite, PRISMA also detected these minerals. However, there were some exceptions. PRISMA was not capable of mapping their presence in some samples where kaolinite and illite were present as trace minerals (SM-06, SM-07, SM-08) or where clays were not a major phase (SM-18, SM-049, SM-68, SM-73).

Laboratory spectral curves further support these findings (Table A1). The spectral curves obtained in the laboratory revealed similar mineralogy to those identified by PRISMA (Table A1). Gypsum was identified by its characteristic absorption features at 1750 nm (Figure 16A), while clay was identified by its diagnostic absorption near 2200 nm (Figure 16C), associated with the presence of illite and kaolinite. However, as with the XRD validation, some samples showed clay features in the laboratory spectra that were not detected in the PRISMA spectral curves (Figure 16D) (SM-06, SM-07, SM-08, SM-18, SM-049, SM-68, SM-73) (Table A1). For iron oxides and hydroxides, the laboratory spectra showed broad absorption features between 800 nm and 1000 nm, which align with the characteristic feature of these minerals. PRISMA spectra showed similar features, although with notable noise (Figure 16B). The dominant iron oxide and hydroxide phases were identified based on their diagnostic minimum wavelength positions of a fitted parabola, and these matched the phases detected in the PRISMA data.

5. Discussion

The results confirm PRISMA’s capability to identify and map key AMD indicators, specifically iron oxides and hydroxides and gypsum. Band ratios applied to PRISMA imagery provided accurate mineral distribution maps, validated by X-ray diffraction (XRD) analysis and laboratory spectral curves (Table A1) and a cross-comparison with Sentinel-2 data (Figure 8). Both the iron oxide and hydroxide and gypsum band ratios (Figure 9, Figure 10, Figure 11 and Figure 12) showed consistent agreement with the XRD and laboratory spectral curves results across all sediment samples (Table A1). A strong correlation (R² > 0.81) was also observed between PRISMA and Sentinel-2 data for the iron oxide and hydroxide band ratio. However, PRISMA images consistently show higher values compared to Sentinel-2 (Figure 8), which may indicate differences in sensor sensitivity or atmospheric correction algorithms. These discrepancies were more pronounced in earlier scenes, suggesting improvements in PRISMA’s correction algorithms over time [21,52].

PRISMA-derived maps of the Beal and the Gorguel basins show the impact of sedimentary processes. Distribution of minerals such as hematite, goethite, jarosite, aluminium-bearing clays, and gypsum exhibits spatial and temporal variability influenced by hydrological and climatic factors such as seasonal rainfall and flood events.

Iron oxides and hydroxides associated with AMD—specifically hematite, goethite, and jarosite—show distinct spatial and temporal patterns, reflecting environmental factors. Jarosite, a meta-stable iron mineral that forms under acidic, sulphate-rich conditions with high redox potential, can be used as an indicator of recent AMD deposition [67,68]. In this study, jarosite was identified as the dominant iron-bearing mineral in all PRISMA scenes (Figure 9 and Figure 10), with higher concentrations following the torrential precipitation events of September and December 2019 (Figure 9A and Figure 10A). These heavy rainfall events likely facilitated the oxidation of sulphide minerals, leading to the formation of jarosite. Subsequent scenes (Figure 9B–E and Figure 10B–E) show an increase in the presence of more stable oxides, such as goethite or hematite. This increase can be attributed to the absence of intense rainfall events after December 2019 and dried conditions. Moderate precipitation events likely remobilised previous jarosite sediment and transported and precipitated more oxidated iron phases such as hematite and goethite, iron minerals favoured by the prevalence of oxidising environments during this period [69].

As noted previously, PRISMA imagery may overrepresent jarosite compared to Sentinel-2 (Figure 8), potentially misclassifying goethite-rich zones as jarosite-dominant. Previous research [8] supports this, showing lower jarosite concentrations with more confined distributions in the Beal basin’s alluvial fan.

Goethite-rich areas increased in the later scenes (Figure 9D,E and Figure 10D,E), which were captured after moderate precipitation events. This pattern suggests that goethite forms under moderately oxidising environments, where periodic moderate precipitations facilitate partial oxidation processes [67,69]. Additionally, some of the observed increase in goethite content in later scenes may be related to the improvements in atmospheric correction processes applied to PRISMA data [52].

Hematite, a stable iron oxide, predominates in well-drained, oxidised environments such as floodplains. Although direct observation of floodplain areas in this study is limited due to agricultural vegetation cover, previous studies [8] and XRD analyses from samples SM-4, SM-5, SM-12, SM-19, SM-21, and SM-23 confirm the dominance of hematite in these zones. PRISMA images further support this pattern, as seen in two hematite deposits observed in February 2020 and March 2021 (Figure 9A,B and Figure 10B). These periods followed the torrential rainfalls in September and December 2019, which transported great volumes of sediment downstream. This was followed by a dry and hot period after June 2020, which enhanced oxidation processes, particularly evident in March 2021. This sequence of heavy rainfall followed by elevated temperatures likely accelerated the oxidation of iron-bearing minerals, facilitating the formation of hematite.

In contrast to classical AMD models that show a downstream progression from jarosite to goethite and hematite [18,19], in this study hematite-rich zones can be observed upstream and jarosite/goethite concentrations downstream (Figure 9). This atypical pattern reflects the basins’ geomorphology: both ramblas are short channels flowing over mining-contaminated Quaternary deposits with limited neutralising capacity. Hematite was mostly found in elevated, well-drained agricultural areas with prolonged oxidative conditions, while jarosite and goethite dominated flood-prone lower basins, where reduced redox conditions favour their precipitation.

Gypsum shows a seasonal pattern driven by solubility, with lower concentrations in autumn (Figure 12D and Figure 13D), following moderate torrential events, and higher concentration in spring (Figure 12B,E and Figure 13B,E) after dry periods or major torrential events. This fluctuation correlates with evaporation cycles, as warmer temperatures and low precipitation in spring promote evaporative mineral formation [7,10]. In contrast, during autumn, moderate precipitation dissolves gypsum, facilitating its removal through water transport due to its high solubility [70].

Although gypsum itself is not a contaminant, its presence can indicate areas where more hazardous evaporite sulphates—such as copiapite, melanterite, or hexahydrite—may also form, potentially contributing to acid generation [7,9].

Aluminium-bearing clays, known for their capacity to adsorb heavy metals, including Ni, Co, Cu, As, or Fe [71], showed higher concentrations in October 2023 (Figure 14D and Figure 15D) and lower levels in March 2021 (Figure 14B and Figure 15B). However, due to the “salt and pepper” effect in PRISMA images and the limited temporal dataset, establishing detailed temporal trends for these clays is challenging. Despite employing band ratios to reduce noise effects, the longer wavelengths of PRISMA data remain affected by random and fixed noise patterns [72,73]. This susceptibility does not allow establishment of a temporal trend for clays, as evidenced by the “salt and pepper” effect observed in PRISMA images. Moreover, some sediment samples identified by XRD and the laboratory spectral curve as containing aluminium-bearing clay minerals (Table A1) were not accurately represented in the PRISMA-derived mapping (Figure 14 and Figure 15). Nevertheless, no false positives were observed, as PRISMA did not identify aluminium clays (e.g., kaolinite, illite) in any samples where these minerals were absent according to XRD analysis, indicating a low likelihood of false detection.

Overall, the spatial and temporal distribution AMD-related minerals in the Beal and Gorguel basins reflects the interplay of climatic, geochemical, and geomorphological factors. Jarosite and gypsum commonly co-occur in the lower basin areas, forming under similar conditions—acidic, sulphate-rich environments typically following intense rainfall (Figure 17B,C,E and Figure 18B,C,E). Jarosite, a meta-stable iron sulphate, indicates recent AMD activity, while gypsum, a calcium sulphate, precipitates when evaporation concentrates dissolved sulphates. In contrast, clays tend to dominate in more neutral periods, when pH and sulphate concentrations are comparatively lower (Figure 17A,D and Figure 18A,D). Areas with higher clay content often show reduced gypsum presence, as their formation is favoured by different geochemical conditions. Geomorphology further shapes these patterns: hematite-rich deposits form in the elevated, well-drained upstream areas where oxidising conditions persist (Figure 17), whereas the lower basins accumulate mixtures of jarosite, gypsum, and clays due to episodic flooding, sediment deposition, and limited drainage (Figure 17).

Despite PRISMA’s proven capability, several challenges remain, with the most significant ones identified in this study being data availability and noise-related limitations. The limited number of available PRISMA scenes in the study area restricted the possibility to analyse long-term trends. This study utilised only five PRISMA scenes captured between 2020 and 2024, which have allowed to us observe the previously exposed mineral distribution patterns related to seasonal changes and hydrological processes. However, the restricted temporal resolution of PRISMA, with images acquired only on demand and that are weather-dependent, has limited the possibilities to study inter-annual effects or provide detailed before-and-after images of precipitation events. Although EnMAP, a hyperspectral sensor with comparable characteristics to PRISMA, is currently operational, no EnMAP images were taken for this study area before 2024.

Another major limitation is the significant noise present in PRISMA imagery, which affects spectral quality and classification accuracy. PRISMA hyperspectral data offers a high spectral resolution with 239 bands, theoretically allowing for the identification of multiple AMD-related minerals. While hyperspectral imaging enables the selection of the most relevant bands for specific minerals, the excessive number of bands also leads to data redundancy, exacerbating the impact of noise and striping effects [21,52]. Despite the application band ratioing to normalise noise and extract relevant information, the results remain inconsistent, such as the Al-clay band ratio.

Furthermore, due to the high noise levels, vegetation was masked using NDVI instead of more refined methods such as spectral unmixing, which would have allowed for a more precise separation of vegetation and mineral pixels.

To address these challenges, band ratioing was employed, as it allows for the selection of specific band combinations while being less sensitive to noise and striping effects [48]. This makes it a practical approach for routine or real-time AMD monitoring. Furthermore, it offers a simpler alternative to more complex methods like supervised and unsupervised classification (e.g., SAM, PCA, MNF), which often require extensive preprocessing to correct noise and artifacts.

Looking ahead, current and upcoming satellite hyperspectral missions are expected to overcome these limitations and enhance capabilities for detailed surface composition analysis and environmental monitoring. The NASA EMIT sensor currently provides 60 m resolution pixels with spectral data with minimal noise, ensuring clean spectral curves [74]. Also, the systematic acquisition of EnMAP images in the study area [75] will improve temporal resolution. Future missions, such as the NASA Surface Biology and Geology (SBG) observing system [76] and the Copernicus Hyperspectral Imaging Mission for the Environment (CHIME) [77], are expected to provide higher temporal coverage, enabling a more comprehensive analysis of AMD dynamics in the Sierra Minera de Cartagena-La Unión.

Overall, this study has demonstrated that PRISMA hyperspectral data, combined with band ratio methods, is effective for mapping and monitoring minerals in the AMD-affected area of Sierra Minera de Cartagena-La Unión. The successful application of this approach in this region suggests its potential adaptability for other mining-contaminated areas with similar environmental conditions.

6. Conclusions

This study demonstrates the effectiveness of PRISMA hyperspectral satellite data for mapping and monitoring minerals associated with acid mine drainage (AMD) in the Beal and Gorguel basins, areas impacted by historical mining activities. By applying band ratios, it was possible to identify variations in AMD-related minerals, such as iron oxides and hydroxides and gypsum, across five different scenes from 2020 to 2024. These results demonstrate the capability of hyperspectral data in capturing mineral variations, essential to understand the spatial extent and seasonal behaviour of contaminants in mining-affected regions.

The temporal analysis of PRISMA data revealed distinct seasonal patterns in mineral distribution, influenced by hydrological processes and climatic factors. Areas with spectral responses indicative of hematite, jarosite, and goethite showed distribution patterns associated with the basins’ geomorphological characteristics and redox conditions. Gypsum, although not directly a contaminant, served as an indicator of evaporative formation conditions, often linked to the presence of other secondary contaminant minerals.

While PRISMA hyperspectral data offers significant advantages for AMD mapping, it also has limitations. Noise and striping effects, particularly evident in the aluminium-bearing clay band ratio, contributed to a “salt and pepper” effect, not allowing a clear distribution mapping of these minerals. Future hyperspectral missions, with improved sensor atmospheric corrections, are expected to address these challenges and provide more accurate and detailed information on AMD-affected areas.

Overall, this study highlights the potential of satellite-based hyperspectral sensors for multi-temporal monitoring of the environmental impacts of mining activities in regions affected by AMD. By integrating PRISMA data with ground truth such as X-ray diffraction, laboratory spectral curves, and cross-comparison with Sentinel-2 time series, it was possible to obtain the spatial and temporal distribution of minerals associated with AMD. These findings could contribute to future environmental management and remediation strategies.