Detailed characterization of iron-rich tailings after the Fundão dam failure, Brazil, with inclusion of proximal sensors data, as a secure basis for environmental and agricultural restoration
Rafaella T Silva de Sá,Marcelo Tesser Antunes Prianti,Renata Andrade,Aline Oliveira Silva,Éder Rodrigues Batista,Jessé Valentim Dos Santos,Fernanda Magno Silva,Marco Aurélio Carbone Carneiro,Luiz Roberto Guimarães Guilherme,Somsubhra Chakraborty,David C Weindorf,Nilton Curi,Sérgio Henrique Godinho Silva,Bruno Teixeira Ribeiro
DOI: https://doi.org/10.1016/j.envres.2023.115858
2023-07-01
Abstract:Following the Fundão dam failure in Brazil, 60 million m3 of iron-rich tailings were released impacting an extensive area. After this catastrophe, a detailed characterization and monitoring of iron-rich tailings is required for agronomic and environmental purposes. This can be facilitated by using proximal sensors which have been an efficient, fast, and cost-effective tool for eco-friendly analysis of soils and sediments. This work hypothesized that portable X-ray fluorescence (pXRF) spectrometry combined with a pocket-sized (Nix™ Pro) color sensor and benchtop magnetic susceptibilimeter can produce substantial data for fast and clean characterization of iron-rich tailings. The objectives were to differentiate impacted and non-impacted areas (soils and sediments) based on proximal sensors data, and to predict attributes of agronomic and environmental importance. A total of 148 composite samples were collected on totally impacted, partially impacted, and non-impacted areas (natural soils). The samples were analyzed via pXRF to obtain the total elemental composition; via Nix™ Pro color sensor to obtain the red (R), green (G), and blue (B) parameters; and assessed for magnetic susceptibility (MS). The same samples used for analyses via the aforementioned sensors were wet-digested (USEPA 3051a method) followed by ICP-OES quantification of potentially toxic elements. Principal component analysis was performed to differentiate impacted and non-impacted areas. The pXRF data alone or combined with other sensors were used to predict soil agronomic properties and semi-total concentration of potentially toxic elements via random forest regression. For that, samples were randomly separated into modeling (70%) and validation (30%) datasets. The pXRF proved to be an efficient method for rapid and eco-friendly characterization of iron-rich tailings, allowing a clear differentiation of impacted and non-impacted areas. Also, important soil agronomic properties (clay, cation exchange capacity, soil organic carbon, pH and macronutrients availability) and semi-total concentrations of Ba, Pb, Cr, V, Cu, Co, Ni, Mn, Ti, and Li were accurately predicted (based upon the lowest RMSE and highest R2 and RPD values). Sensor data fusion (pXRF + Nix Pro + MS) slightly improved the accuracy of predictions. This work highlights iron-rich tailings from the Fundão dam failure can be in detail characterized via pXRF ex situ, providing a secure basis for complementary studies in situ aiming at identify contaminated hot spots, digital mapping of soil and properties variability, and embasing pedological, agricultural and environmental purposes.