Probabilistic Modeling of Provenance of Trace Soil Evidence
Abstract
Trace soil evidence may have identifiable characteristics which are more commonly found in some regions over others. The observable properties may indicate human activities, the bedrock or surficial geology/soil of the area, or the taxa present in the area. Seldom are maps available which directly represent the properties observed in a forensic soil trace, but often there are reference maps which are related to the observations. Because of the indirect relationships between reference maps and observations and the poorly constrained circumstances of soil emplacement and recovery, there are substantial uncertainties to assigning a soil to a specific map category or quantify for maps of continuous properties. To mitigate these uncertainties, a probabilistic GIS-based approach was used to model the more likely source areas of the forensic soil. Relative likelihoods of observations are assigned to the mapped features in a series of relevant maps to narrow the geographic source of a trace soil sample, and may be applied to discrete/categorical maps or to quantitative raster maps by specified functions. These probability maps are merged by: applying a grid size over the area of interest; extracting the assigned relative likelihood for each input map layer for each grid cell; multiplying these relative likelihoods together at each pixel; and normalizing these products into probabilities that sum to unity across the area of interest. This end product can be exported as a heat map of more likely regions or a polygon enclosing relevant areas to provide to law enforcement search teams.
This approach is applied retrospectively to a real case in which two bodies were buried somewhere along a transcontinental route (Webb this session) using maps representing distance from travel path, surficial geology, % illite, species range maps, and proximity to coal-fired power plants. The actual grave site location was discovered 2.5 years after burial. Observations in the report of examination of soil on digging tools were used to generate a spatial probability model. This model predicts areas where the clandestine grave site is more likely to be. In a future case, this approach could aid law enforcement investigations.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.B53F2132S
- Keywords:
-
- 0486 Soils/pedology;
- BIOGEOSCIENCESDE: 1065 Major and trace element geochemistry;
- GEOCHEMISTRYDE: 1094 Instruments and techniques;
- GEOCHEMISTRYDE: 0240 Public health;
- GEOHEALTH