University of Twente Student Theses
From multidimensional soil properties to reduced summary scoring curves for soil health
Simioniuc, Ruxandra (2024) From multidimensional soil properties to reduced summary scoring curves for soil health.
PDF
8MB |
Abstract: | The study of soil is a multidisciplinary field, spanning domains such as ecology, biology and chemistry. Most approaches for evaluating soil health use basic statistical techniques, rely heavily on field knowledge, are very specific or do not provide a start-to-end pipeline. We propose a framework with clear steps and new methods, with the potential to be generally applicable. The most frequent soil health indicators for chemical, physical, and biological pillars are selected from relevant literature and aligned with the dataset. The samples are split on texture type and processed separately. Redundant features are discarded using correlation analysis. The remaining subsets are processed using sparse principal component analysis, a DRT unused in related research, attaining a 90% reduction in the number of indicators. Scoring curves for individual indicators are obtained from literature, adapted to the data, and combined into summary indicators, using the SPCA weights and components. Summary scoring curves are computed, reducing the evaluation of each pillar to only 2 dimensions, while retaining over 75% of the explained variability for biological and physical pillars. The final step is providing a simple and interpretable visualisation of the results, so stakeholders can interpret the health of a sample. |
Item Type: | Essay (Master) |
Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
Subject: | 54 computer science |
Programme: | Computer Science MSc (60300) |
Link to this item: | https://purl.utwente.nl/essays/104438 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page