Quantitative modeling method for geolocating pollen samples

Pollen producing plants are present on every continent and the taxa of these plans can be distinguished based on the pollen grains. As a pollen assemblage can only come from a vegetation that produces the pollen types in that assemblage, pollen can give an indication of vegetation composition. Knowing the assemblage that produced a specific sample can be used to determine the origins of pollen on objects of interest such as footwear, drugs and poached goods. Most studies that used forensic palynology for geolocation were qualitative analyses of regions with low biodiversity. None are studies on a continental scale that involve the tropics. There is a need for objective qualitative forensic methods, which are expected to be effective and efficient with both high and low biodiversity regions. In this study a qualitative modeling method was developed using pollen samples from 705 neotropical sites and tested on pollen samples from 55 European sites. These samples came from pollen traps, soil samples, mud water interfaces of lakes and rivers, and moss samples. Ensemble species distribution models based on presence data were run for all genera present in these samples. The predicted range of all genera present in each individual sample were combined to produce a predicted range of origin of the site. A rate of 89.9% of all Neotropical sites was correctly predicted. The highest accuracy was achieved with samples from pollen traps, which were correct in 96.2% of cases. The results of European and Neotropical sites of the same sample origin were comparable. The results indicate that qualitative analysis can be accurate and efficient in the tropics at a continental scale. This file contains all used data and R scripts. With the contents the results can be recreated. Instructions on use are enclosed