Fig. 3: Distance to predicted origin (km) for different tracing methods and their combinations.

Depicted as % of the total number of test trees per distance bin per method. The lowest bar represents correct site origin assignment, which is the identification accuracy, also represented by a dot and error bar that indicates the variation (st. dev.) in identification accuracy, estimated using different test and training datasets of the Random Forest models. Methods are faceted by analyses based on a single method (1), a combination of two methods (2) or all three methods (3). SNP = pSNPs, ISO = stable isotope ratios, EL = multi-element analysis, ALL = all three methods, RDM = the occurrence of pairwise distances in the reference dataset, representing the expected result of fully random assignments. The dotted line indicates 90% accuracy. A version of this figure including standard errors for all distance bins is provided as Supplementary Fig. S5.