This research seeks to characterise the GC amenable lipid component of soils as a means of discriminating between them based on subtle differences in the determined whole distributions. Furthermore, the use of models based on the results obtained, was explored as a means of potentially identifying soils of unknown origin, i.e. those that may constitute forensic evidence obtained as part of a criminal investigation. The work establishes a lipidomic workflow for characterising soils, three scenarios are considered: (1) soils from different geographical locations, (2) soils from the same location but under different land-use, and (3) peat samples dominated by two species of vegetation. The effect of seasonal change was also investigated. An analytical workflow was developed and comprised ultrasonic extraction and saponification yielding a total lipid extract. Accurate mass GC-MS was used to analyse extracts and the acquired data processed using Mass Hunter Profinder B.08 and Mass Profiler Professional. The results were assessed using principal component analysis (PCA) and hierarchical clustering analysis. An un-paired t-test and Kruskal-Wallis test were performed on each data set to select discriminant entities responsible for the observed clustering. Partial least square discriminant (PLSD) models were built to classify unknown samples. For (1) results revealed a distinct separation, based on land-use and vegetation, for the eight sites studied. Established biomarkers, e,g, dehydroabietic acid (DHAA) for pine (highest levels at the coniferous sites), were among the discriminant entities selected. For (2) this approach separated four close-proximity plots from Broadbalk long-term experiment based on the overlying vegetation arising from differing land-use (wooded, grazed, stubbed and winter wheat). Another long-term experiment, Park Grass, showed separation based on soil pH and species diversity where four clusters were observed from an extremely acidic plot (3.7), to a slightly acidic plot (5.7), control plots (5.9 and 6.9) and higher pH plots (6.5 to 6.9). Finally, (3) enabled discrimination between two peats dominated by sedge and Sphagnum. Artificial peat mixtures of sedge and Sphagnum were separated in accordance with origin along PC1 in the PCA. PLSD models gave 100 % correct prediction for both Broadbalk and peat samples showing the potential of this method for characterising soils of unknown provenance. Overall, this work demonstrates that lipidomics offers a rapid, means for profiling soils applicable to a diverse range of forensic (and wider) contexts.
|Date of Award||25 Jun 2019|
- The University of Bristol
|Supervisor||Ian D Bull (Supervisor) & Paul J Gates (Supervisor)|