Network-based geoforensics: Connecting pollen and plants to place

Despite the potential for pollen to provide key geographic information, law enforcement agencies have histor-ically underutilized forensic palynology. Pollen samples are rarely collected at crime scenes or on objects of interest, and experts needed to manually identify pollen and its morphological characteristics are often in short supply. Fortunately, the advent of DNA barcoding and metabarcoding is poised to change this, enabling law enforcement agencies to easily and quickly identify individual plant species from mixed pollen samples. How-ever, determining the location of pollen deposition based on identified plant species remains challenging, requiring comprehensive plant geodatabases and complex analytical approaches. This type of analysis is also complicated because pollen samples from forensic objects are not comprehensive. Furthermore, while detecting common plant pollen in a sample is likely, common plants provide less explanatory power due to their ubiquity. Thus, we propose a network-based approach to construct a categorization of 'forensically-useful' plants - those species that characterize geographically distinct plant mixes quickly for forensic applications and thus provide significant explanatory power to a geolocation model. The developed tool can better inform more detailed forensic search models, significantly reducing their overall computational burden and enabling high-fidelity forensic computational tools to operate faster, at a higher resolution, and over a more extensive study area.