Applicability of multivariate methods for evaluating and mapping potentially toxic elements spatial distribution in soils

MSc. Ndiye Michael Kebonye

Doctoral thesis

Abstract:

This thesis contains several published manuscripts first-authored and co-authored which were sequentially worked on between 2019 and 2021. The two main author manuscripts that form the backbone of the current thesis have a common theme, ‘potentially toxic elements (PTEs)’. These manuscripts study PTE spatial distributions in soils while applying various multivariate methods (i.e. regularization methods and the self-organizing map artificial neural networks) and geostatistical methods. The remaining manuscripts cover other aspects relating to potential human health risk, soil properties and nutrient assessment. The entire thesis was carried out under the supervision of the Department of Soil Science and Soil Protection at the Czech University of Life Sciences (CZU), Prague. Various financial aids, those who carried out experiments and field activities as well as collaborators in the form of co-authors to this project are acknowledged and included in the respective sections of the published manuscripts. The primary advisor/supervisor of this thesis was Prof. Dr. Ing. Luboš Borůvka. Soil contamination is detrimental as it brings about land degradation in various soil environments. The characteristics of soil enable it to naturally host and allow for the mobility of PTEs. PTEs in soils originate from lithogenic, anthropogenic, or a combination of these sources. Mainly though, anthropogenic sourced PTEs, for example from mining and smelter (Kebonye et al., 2021a) are of concern because of their accelerated deposition as well as retention in soil environments. As soil contamination resulting from PTEs continues to surge, the need to make informed decisions regarding the remediation of these potential environmental threats is a priority. Therefore, the adoption and application of multivariate methods for mapping soils have been increasing. The current work applies multivariate methods, particularly regularization techniques to predict as well as map site-specific arsenic (As) content levels in a highly contaminated floodplain area near the Litavka River in Příbram, Czech Republic (Kebonye et al., 2021a, 2021b). Moreover, the thesis explores the suitability of the self-organizing map artificial neural networks to delineate PTE hotspots in the same soils (Kebonye et al., 2021b). The value of utilizing the abovementioned methods is that they are robust and these have not been widely applied for soil contamination studies in the Czech Republic. For the context of this Ph.D. thesis, the abovementioned methods have improved the monitoring of PTEs in the study area while providing detailed preliminary evidence that would help decision-making by policy-makers regarding PTE distribution in the Příbram District (Czech Republic). Moreover, since there is a dearth of knowledge regarding the study of floodplain soils in the Czech Republic (Skála et al., 2017), the current work is a contribution within the context of assessing the spatial distribution of PTEs. This thesis is mainly made up of the following publications: Kebonye, N.M., John, K., Chakraborty, S., Agyeman, P.C., Ahado, S.K., Eze, P.N., Němeček, K., Drábek, O., Borůvka, L. (2021a). Comparison of multivariate methods for arsenic estimation and mapping in floodplain soil via portable X-ray fluorescence spectroscopy. Geoderma, 384, 114792. Kebonye, N.M., Eze, P.N., John, K., Gholizadeh, A., Dajčl, J., Drábek, O., Němeček, K., Borůvka, L. (2021b). Self-organizing map artificial neural networks and sequential Gaussian simulation technique for mapping potentially toxic element hotspots in polluted mining soils. Journal of Geochemical Exploration, 222, 106680. Kebonye, N.M., Eze, P.N., John, K., Agyeman, P.C., Němeček, K., Borůvka, L. (2021). An in-depth human health risk assessment of potentially toxic elements in highly polluted riverine soils, Příbram (Czech Republic). Environmental Geochemistry and Health, 1-17. John, K., Kebonye, N.M., Agyeman, P.C., A

Abstract:

hado, S.K. (2021). Comparison of Cubist models for soil organic carbon prediction via portable XRF measured data. Environmental Monitoring and Assessment, 193(4), 1-15.

Keywords

Floodplain soils, Potentially toxic elements, Machine learning methods

Jazyk práce: English

Datum vytvoření / odevzdání či podání práce: 21. 6. 2021

Obhajoba závěrečné práce

• Vedúci: prof. Dr. Ing. Luboš Borůvka
• Oponent: Radim Vácha, externi

Czech University of Life Sciences Prague

Faculty of Agrobiology, Food and Natural Resources

Doctoral programme / odbor:
Agricultural Specialization / Exploitation and Conservation of Natural Resources