Geoforensics with Pollen Quantification: A Spatial Perspective

• Published in: Annals of the American Association of Geographers 2023-10, Vol.ahead-of-print (ahead-of-print), p.1-17
• Main Authors: Mu, Wangshu, Tong, Daoqin, Grubesic, Tony H., Liu, Hung-Chi, Helderop, Edward, Miller, Jennifer A., Bienenstock, Elisa Jayne
• Format: Article
• Language: English
• Subjects: Biological markers; Biology; Biomarkers; Computers; Counterfeit; Criminology; Ecology; Explosives; GEOFIND; geoforense; Geographical distribution; Geography; Geology; GOFIND; Locations (working); Measurement; Model accuracy; optimización espacial; Optimization; polen; Pollen; Probability; Probability theory; Provenance; Simulation; Statistical analysis; 地理法医学; 空间优化
• ISSN: 2469-4452; 2469-4460
• DOI: 10.1080/24694452.2023.2211155

Geoforensic science investigates the location and time of criminal occurrences by integrating multiple fields, including geography, criminology, ecology, biology, and geology. The ubiquity, durability, and spatial-temporal predictability make pollen a frequently used biomarker in geoforensic investigations to help determine the provenance of hard-to-trace items, including computers, counterfeit products, digging equipment, clothing, and undetonated explosives. The recently developed Geoforensic Interdiction (GOFIND) model links the pollen combination collected from a sample object with the probability of locations traversed by the object. Although the GOFIND model improves over the traditional single-site joint probability approach and can be used to identify multiple locations simultaneously, substantial limitations remain. In particular, GOFIND requires specifying the number of locations traversed by an object in advance-a priori knowledge that is almost impossible to obtain in real-world applications. This article aims to introduce the GOFIND + model that leverages detected and undetected pollen to establish a probabilistic relation between pollen and the corresponding species distribution in the environment. Our simulation tests using the USDA CropScape data for the state of Texas show that the GOFIND + model outperforms the GOFIND model in predictive accuracy. Further, GOFIND + does not require that users specify the number of geographical stops and sites a priori. Key Words: geoforensics, GOFIND+, pollen, spatial optimization.