Mineral Identification and Abundance Mapping through the hyperspectral PRISMA images on the Dallol Planetary Analog

Introduction

Studying planetary field analog environments is a key point in order to define the physical and chemical parameters that favor life on Earth and other planets. Terrestrial hydrothermal springs have long been considered among the most significant planetary analogs searching for traces of life on Mars [1].

Hyperspectral data have been recognised to be more suitable for the detailed mapping and identification of rocks and minerals identification of land surface, especially for minerals [2].

Notwithstanding the technological advances, hyperspectral satellites are still poorly represented in spaceborne missions for Earth Exploration compared to multispectral ones. In this context, the Italian Space Agency (ASI) EO mission named PRISMA (PRecursore IperSpettrale della Missione Applicativa, [3]) offers a great opportunity to improve the knowledge about the scientific applications of spaceborne hyperspectral data.

PRISMA, launched in March 2019, includes a panchromatic and a hyperspectral camera with 239 spectral bands. Specifically, the PRISMA satellite comprises a high-spectral resolution Visible Near InfraRed (VNIR) and Short-Wave InfraRed (SWIR) imaging spectrometer, ranging 400-2500 nm, with 30 m ground sampling distance (GSD) and 5 m GSD for the panchromatic camera [4].

Our analysis with PRISMA images was mainly performed on an arid environment in a remote region of NE Ethiopia (Dallol; Long: 40.299351, Lat: 14.244367), representing an exceptional Mars analog due to diffuse hydrothermal alteration and the sulfate deposits evocative of past hydrothermal activity on Mars. This work aimed to obtain an identification map of minerals and their relative abundance using hyperspectral imaging to understand the potential of PRISMA as analog probe of Mars orbital instruments to detect and study possible analogs on Earth.

Study Area

Dallol is situated in the Danakil Depression, which is part of the East African Rift System. Principal geothermal features of the central crater area of Dallol are salt pillars, circular manifestations and acidic ponds. The northern and southern part is dominated by a salt dome structure and Salt pinnacles in the SW salt canyon area. The Black Mountain and the super-saline Black Lagoon, just south-southwest of Dallol, is an area of salt extrusions, geothermal manifestations and brine upflows.

One advantage of this area is that the nebulosity is generally low, in fact the image selected during the dry season has a cloud coverage percentage of less than 1%. A salt suite was deposited and re-worked by hydrothermalism in the selected site. The characteristic minerals of the area are: carbonate, halite, carnallite and bischofite, anhydrite, gypsum, native sulfur of hydrothermal origin [5; 6].

Flooding episodes from the Lake Assale to the north due to intense winds acting over the flat topography of the depression. The PRISMA SWIR Land/Water band combinations on the image selected, helped us to choose the region of interest around the Dallol area.

Operational Hyperspectral Processing

PRISMA images have three different levels of processing. Level 2C and 2D geolocated and atmospherically corrected images were used in this work and dated 21 August 2021. it is worth noticing that the images acquired on Dallol prior to the image selected for analysis had several preprocessing problems, particularly for stripe removal.

The operational hyperspectral processing is composed of three main processing steps: (1) dimensionality reduction; (2) endmember identification and (3) mineral map distribution and abundance estimation.

An unexpected result was obtained by applying an additional atmospheric correction, the Internal Average Relative Reflectance with Dark Subtraction, on the L2C image already corrected during the principal component analysis (PCA). The corrected atmospheric PCA allows better highlighting of geomorphological features.

As for step (1), since hyperspectral images are composed of hundreds of extremely correlated bands, it is possible, and indeed beneficial, to reduce the effective dimension of the input data by removing bad bands.

Step (2) was performed using the THOR Hyperspectral Material Identification (in ENVI 5.6) to identify unknown spectral signatures by comparing them with spectral libraries. This tool considers background statistics and image endmembers and can therefore provide accurate responses and spectra plots for rare or sub-pixel targets.

Finally, the Spectral Angle Mapper (SAM) and the Linear Spectral Unmixing (LSU) tools were adopted for step (3). SAM determines the spectral similarity between two spectra by calculating the angle between the spectra and treating them as vectors in a space with dimensionality equal to the number of bands. LSU is a standard technique for spectral mixture analysis that infers a set of endmembers and fractions of these, called abundances. The mineral distribution and the abundance maps are shown respectively in Fig.1 and Fig.2.

Conclusion

Six minerals have been recognised with the SAM classification from ENVI spectral library, in excellent agreement with the previous studies: carnallite, jarosite, kainite, polyhalite, gypsum and nontronite. The results confirm the mineralogical variability typical of the Dallol; in Fig.2, high mineral abundance values are shown in blue. The error calculated with the RMS is very low over the entire area of interest, except for the central zone where there are sulphur pools and therefore the presence of water does not favour this type of analysis. 

To better constrain the mineralogical mapping, future work will be conducted by a field exploration campaign to collect spectral signatures to be added to the ENVI library used, which so far could not be organised due to the ongoing civil war in Dankalia.

To sum up, the study of terrestrial analogs can provide insights into the probable presence and nature of spring deposits on Mars, as well as help develop methods for classifying them from remote sensing data. PRISMA represents a valuable satellite for distinguishing not only the geometric characteristics of observed objects, but also the chemical-physical composition of the surface of the Earth.

References: [1] Walter, M.R. and Des Marais, D.J., 1993. Icarus 101:129–143 [2] Chang, C.I., 2007. John Wiley & Sons. 10.1002/0470124628 [3] Candela, L., et al. 2016. IEEE international geoscience and remote sensing symposium (IGARSS), 253-256. 10.1109/IGARSS.2016.7729057 [4] Loizzo, R., et al. 2019. IEEE (IGARSS), 4503-4506. 10.1109/IGARSS.2019.8899272 [5] Cavalazzi, B., et al. 2019. Astrobiology, 19(4), 553-578. 10.1089/AST.2018.1926 [6] López-García, J.M., et al. 2020. Frontiers in Earth Science, 7, 351. 10.3389/FEART.2019.00351