Soil eDNA reflects regionally dominant species rather than local composition of tropical tree communities

Environmental DNA (eDNA) is increasingly used for biodiversity monitoring, but validation of the spatial scale(s) at which eDNA reflects extant communities is scarce, particularly in tropical forests: the terrestrial biome with the most concentrated diversity on earth. We leveraged spatially explicit tree inventory data from the 16-ha Luquillo Forest Dynamics Plot (LFDP) in Puerto Rico to validate soil eDNA as a spatially explicit indicator of tree diversity/composition. Using a comprehensive local chloroplast trnL-P6 reference library, we analyzed soil samples at multiple scales through eDNA trnL-P6 metabarcoding. We compared eDNA taxonomic diversity/composition, considering several bioinformatic thresholds, with inventory data across a range of spatial scales, as well as random points to compare observed correlations with random expectations. Despite considerable fine-scale heterogeneity in soil plant eDNA composition, we detected 53 tree Operational Taxonomic Units (OTUs) across the LFDP, corresponding to 68% of tree OTUs from the census data. Encouragingly, this equated to 98% of the total basal area (and 98% of the total stems). An initial confusion matrix evaluation suggested a highly localized eDNA signal (within 5 m of the sampled locations). However, comparison with random expectations revealed a lack of support for a fine-scale spatial signal due to misclassification (i.e., eDNA false presence or/and false absence) of relatively common taxa. Our study shows that “universal” PCR primer metabarcoding of tree eDNA in tropical soils may be useful for assessing dominant taxa at landscape scales, but not for spatially explicit characterization of rare species and community composition at local scales.