Extraction workflow determines marker-specific recovery and reproducibility in leaf-litter eDNA metabarcoding

Forest-floor leaf litter is a dynamic and structurally complex ecological transition zone and thus a promising substrate for terrestrial eDNA metabarcoding. Yet, extraction workflows for this heterogeneous matrix remain poorly standardized, especially in tropical systems, making it largely impossible to compare ecological functions across space, time, and taxa. To guide workflow selection across a series of selection criteria, including biological target, research question and practical considerations, we compared DNA extraction workflows for leaf-litter eDNA collected from 42 biological samples across seven different forest sites on Oʻahu, Hawaiʻi. We evaluated four DNA extraction workflows: (1) Two low-volume approaches, with DNA extracted directly from 200 mg of homogenized litter using (i) CTAB or (ii) DNeasy PowerSoil®; and (2) two high-volume approaches using PBS wash-based from 10 g of litter followed by (i) Centrifugation or (ii) Filtration. Taxonomic recovery from each workflow was evaluated with two COI primer sets targeting arthropods (ANML and shorter NoPlant), and one ITS marker targeting fungi. Results show that eDNA workflows tested here recovered site-level differences among forest-floor communities, but biodiversity recovery depended strongly on extraction workflow and marker. For low volumes, PowerSoil recovered the highest fungal richness (with ITS marker), and produced the most reproducible PCR-replicate profiles across markers, and required the least hands-on time, while CTAB was less expensive but required handling hazardous chemicals. For high volumes workflow, Centrifugation recovered higher arthropod diversity with ANML primer. Differences in community composition were nonetheless recovered by each method. At the same time, sampling sites explained more ASV-level compositional variation than extraction workflow across markers, showing that all workflows retained site-level ecological signals. Together, these results support a workflow framework in which extraction choice depends on target organism group, DNA state, reproducibility needs, and practical constraints.