1Diverse and remote deep-sea communities are critically under-sampled and increasingly 2 threatened by anthropogenic impacts. Environmental DNA (eDNA) metabarcoding could 3 facilitate rapid and comprehensive biotic surveys in the deep ocean, yet many aspects of the 4 sources and distribution of eDNA in the deep sea are still poorly understood. In order to examine 5 the influence of the water column on benthic eDNA surveys in regions targeted for deep-sea 6 polymetallic nodule mining, we investigated the occurrence of pelagic eDNA across: (1) two 7 different deep-sea habitat types, abyssal plains and seamounts, (2) benthic sample types, 8including nodules, sediment, and seawater within the benthic boundary layer (BBL), and (3) 9 sediment depth horizons (0-2 cm, 3-5 cm). Little difference was observed between seamounts and 10 the adjacent abyssal plains in the proportion of legacy pelagic eDNA sampled in the benthos, 11 despite an > 1000 m depth difference for these habitats. In terms of both reads and amplicon 12 sequence variants (ASVs), pelagic eDNA was minimal within sediment and nodule samples (< 13 2%), and is unlikely to affect benthic surveys that monitor resident organisms at the deep 14 seafloor. However, pelagic eDNA was substantial within the BBL (up to 13 % ASVs, 86% 15 reads), deriving both from the high biomass upper ocean as well as deep pelagic residents. While 16 most pelagic eDNA found in sediments and on nodules could be sourced from the epipelagic for 17 metazoans, protist legacy eDNA sampled on these substrates appeared to originate across a range 18 of depths in the water column. Some evidence of eDNA degradation across a vertical sediment 19 profile was observed for protists, with higher diversity in the 0-2 cm layer and a significantly 20 lower proportion of legacy pelagic eDNA in deeper sediments (3-5 cm). Study-wide, our 21 contracts granted by the International Seabed Authority (ISA). Benthic communities in the CCZ 43 are typically characterized by high biodiversity and low biomass (Ramirez-Llodra et al., 2010, 44 Kaiser et al, 2017, Smith et al. 2019, due to severe food limitation at abyssal depths (Smith et al. 45 2008a). A high proportion of species are rare, rendering adequate sampling coverage particularly 46 difficult to achieve (Smith et al., 2008a; Simon-Lledó et al., 2019;De Smet et al., 2017). 47 Thomsen et al. 2012), larger volumes are required to offset expected declines in eDNA 132 concentration and adequately sample communities at depth (e.g., as in Shulse et al. 2017). 133 Therefore, filtered seawater volumes varied across depth in this study, with 5 L per replicate 134 sampled at 5 mab, 50 mab and bathypelagic depths, 4 L filtered in the deep mesopelagic 135 (1000m), 2 L in the mesopelagic (500m), and 1 L filtered per replicate at the DCM and in the 136 near surface. Four to six replicates were taken from each CTD cast and depth. To assess and 137 eliminate cross-contamination, negative controls (double-distilled water; ddH 2 O) were collected 138 for each CTD c...