Identifying Unique Soil VNIR Spectra for Application to Forensic Soil Science.

Identifying Unique Soil VNIR Spectra for Application to Forensic Soil Science

 Catherine Kobylinski, Cristine LS. Morgan, Y. Ge, C. Tom Hallmark 

Forensic soil characterization is an under-explored field in the forensic sciences.  One aspect of forensic sciences is Locard’s Exchange Principle, which states that every contact leaves a trace.  As soil characterization technology improves, applications of soil characterization to forensic soil science can be made to accurately identify whether a soil sample collected as evidence corresponds to samples collected at a crime scene. This research focuses on the use of visible near and infrared, diffuse reflectance spectroscopy (VNIR DRS) as a means to spectrally “fingerprint” soils. Our hypothesis is that VNIR spectra of soils from a crime scene are unique from other soils, even soils of the same soil series or mapping unit. If soil spectra from a crime scene are unique, this data can be used to accurately assess Locard’s Exchange Principle. To test this hypothesis, soil samples were collected within in a thirty-mile radius of a designated “crime scene” in the Brazos River floodplain near Texas A&M University. The crime scene is mapped as a Weswood silt loam (Udifluventic Haplustepts). Three other similar soil series were identified an collected, including Yahola fine sandy loam (Udic Ustifluvents), Ships clay, (Chromic Hapluderts) and Silawa fine sandy loam (Thermicultic Haplustalfs). Sixteen soil samples were collected from randomly located soil mapped Weswood, and eight soil samples were collected from Ships, twelve from Yahola and ten from Silawa (n=48). At the crime scene, an X-shaped sampling geometry was constructed, and along the X, surface soil samples were collected at 5-m intervals with a single sample being collected at the center of the X (n=17). The soil samples were air dried, ground to pass through a 2 mm sieve, and scanned with a VNIR spectroradiometer (350-2500 nm). Principle components analysis and linear discriminate analysis were used to deduce the uniqueness of the soil spectra between the four soil series and among the crime scene samples. PCA results, using the first three principal components, suggest the spectral properties of the crime scene soils were mostly unique with three other soil samples from Weswood and Yahola.  Of nine crime scene soils removed from the training set to develop an LDA model, eight were correctly classified as belonging to the crime scene. VNIR fingerprinting of soils from a particular location is a promising technique for forensic soil science.