Digital soil mapping methods of soil nutrients and soil properties

Kingsley John

Doctoral thesis

Abstract:

The selected publications presented in this thesis were compiled as part of my research activities between 2019 and 2022. All the studies are interconnected to the subject matter, digital soil mapping of soil properties and nutrients. The studies were conducted in Nigeria, the Czech Republic, and Morocco, characterized by different climatic conditions, land use, and geological forms. All the research was performed at the Department of Soil Science, University of Calabar, Nigeria, National Institute of Agricultural Research (INRA), Morocco, and the Department of Soil Science and Soil Protection at Czech University of Life Sciences in Prague, as part of various grant obligations, independent research, or part of the thesis work. Grant providers and co-authors are acknowledged within the respective publications. During the doctoral study, a central team evolved. Members were responsible for significant contributions to the conceptualization, consultation, and development of the research and the practical aspects such as sampling and analysis. Consistent primary contributors to the listed research papers are Prof. Chengzhi Qin, Ndiye Michael Kebonye, Ph.D., Ing. Prince Chapman Agyeman, and Isong Abraham Isong, Ph.D., with a project overview and thesis supervision by doc. Ing. Vít Penížek and RNDr. Tereza Zadorova, Ph.D. Digital soil mapping introduction into Soil Science is still significantly evolving from the research phase into the global development and creation of soil maps. Approaches and methods to improve digital soil mapping of soil properties to ensure the accurate representation of the reality of the soil situation still borders around financial budget, optimum sample size, available soil data, and expert knowledge. Therefore, this current research work reveals pragmatic models in predicting soil properties with few samples and ways to improve modelling by simulating sample ratios with different predictive models to cover individual needs where necessary. The present doctoral thesis applied a few sample sizes obtained from existing databases to model some selected soil nutrients in low relief using a more pragmatic kriging model (John et al., 2020). We compared kriging and inverse weighting distance interpolations (John et al., 2020) and also tested different machine learning models on the variability of soil organic carbon (John et al., 2020). Besides that, we tested the combination of cokriging and Gaussian process regression in modeling soil sulphur (John et al., 2021). Furthermore, we evaluated the appropriate sample ratio for predicting soil organic matter (John et al., 2021) and tested different external drift for soil modelling soil properties via kriging with external drift (John et al., 2022). And finally we attempted different linear functions in modelling soil organic matter. The research initiation and development approach considers the workability of digital soil mapping models with few samples, improvement by different machine learning models, variability of sample ratios and environmental covariates for modelling, and the influence of different linear functions on soil modelling. And while focusing on these challenges allowed for the synergizing of a compilation of holistic study during the Ph.D. research.

Keywords

Predictive soil mapping, Soil sampling, Soil nutrients, Environmental covariates

Language used: English

Date on which the thesis was submitted / produced: 20. 10. 2022

Thesis defence

• Date of defence: 11. 4. 2022
• Supervisor: doc. Ing. Vít Penížek, Ph.D.
• Reader: Bořivoj Šarapatka, Bc., externi

Czech University of Life Sciences Prague

Faculty of Agrobiology, Food and Natural Resources

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