Detection of hazardous zeolites in sedimentary and volcanic rocks in New Zealand using SWIR reflectance spectroscopy and XRD

Erionite is a naturally occurring zeolite mineral that predominantly forms in altered volcanic deposits, including intercalated basaltic to rhyolitic ash layers. Erionites often have fibrous crystals. They can be disseminated, infilling in cracks or vesicles in rocks and is highly carcinogenic, causing lung cancer (malignant mesothelioma) when becoming airborne. Recent studies have confirmed sporadic occurrences of natural erionite and other zeolite minerals that can be toxic to human and livestock. New Zealand’s erionite occurrence is observed to be limited to tuff layers within the Waitematā Group volcaniclastic sediments, or in vugs within the Waitakere Group and Mt Somers Volcanic Group.

This study tests a new non-destructive and fast, in-situ method for screening for erionite and other zeolite minerals using reflected light spectroscopy. The sample suite (n=100) has been studied mineralogically in details to confirm the presence of fibrous zeolites, using scanning electron microscopy and X-ray diffraction methods. We further analysed the same sample suit using an ASD FieldSpec 4 portable spectroradiometer capturing reflected light properties from the visible-near infrared (350–1000 nm, VNIR) to shortwave infrared (1000–2500 nm, SWIR) range. The scanned rocks were then used to develop a new spectral library of zeolite-bearing rocks across New Zealand. The VNIR-SWIR range can include diagnostic absorption features to detect the presence of hydrated minerals, including zeolites, using their vibration absorption features at 1920 nm and at longer wavelengths, related to related to -OH and -H₂O molecular bonds.

We trained multiple classification models (e.g., logistics regression and linear discriminant analysis) to test the applicability of multivariate statistical methods to detect fibrous and then potentially hazardous erionite and mordenite minerals in rock and powdered samples. The samples from the spectral library show a range of absorption features at 2208 nm, 2241 nm, 2295 nm and 2340 nm, which can be linked to fundamental stretching and bending vibrations (e.g., Al, Fe and Mg bonds), and carbonates, respectively. The classification yields variable overall accuracies between 0.5–0.6, using 0.75–0.25 train-test split validation. Most methods detected all known erionite-bearing rocks using VNIR-SWIR data, suggesting absorption features can successfully be linked to zeolites. Given the number of false positive samples, we suspect this method can provide a fast-screening tool, that can be employed in-situ to flag formation and lithologies prone to contain geologically-hazardous materials, such as fibrous erionite.