The Clarion-Clipperton Zone (CCZ) is an area of approximately 5.1 x 10⁶ km² in the eastern equatorial Pacific Ocean, which has been designated by the ISA as a potential region for DSM due to its high abundance of high-grade (i.e., enriched in copper, nickel and cobalt) polymetallic nodules (Hein et al., 2013; Petersen et al., 2016). The exploration contract area of the Belgian company GSR is located in the north-eastern part of the CCZ, and the 72.728-km² large area is subdivided into three zones (Figure 1). The Trial site of the pre-prototype collector vehicle PATII was located in the central zone B4 (Figure 1), at the south-eastern flank of a 140-m high seamount in water depths of 4390-4530 m, exhibiting a nodule abundance of 20-24 kg/m² wet weight with a size of more than 4 cm, During the SO268 sampling campaign in 2019, baseline samples were also obtained within the Trial site (Haeckel and Linke, 2021).

Prior to the PATII test (14-17 April 2021), an extensive sensor array for monitoring the sediment plume was installed on the seabed in the Trial area (Vink et al., 2022). It consisted of 20 sensor platforms with a total of 50 optical and acoustic sensors to monitor and quantify the spatiotemporal dynamics of the sediment plume created by PATII, together with two additional Sedimentation Level Indicator (SLIC) boxes to collect and measure the amount of resettled sediment. These devices were placed on the seabed by means of a Remotely Operated Vehicle (ROV) around the collector impact area (in a circle on a northwest-southeast axis) and at various distances (200 to 1000 m) along the south-southeast axis (Vink et al., 2022). This positioning corresponded with the expected sediment plume dispersal orientation, downslope of the dominant bathymetry gradient and downstream of the prevailing bottom currents, which was based on a priori predictive modelling as described by Purkiani et al. (2021).

The nodule collection trial with PATII (dive number PAT009) in the GSR exploration contract area started on the 18^th of April 2021. The vehicle was built at sub-industrial scale (1:4, 25 x 10³ kg in water, L8 m x W4.8 m x H4.2 m) and is composed of a tracked propulsion (i.e., caterpillar tracks) and a hydraulic nodule collector system (https://deme-gsr.com/). The nodule collector vehicle was equipped with several sensors and cameras to monitor environmental conditions such as turbidity, current velocity and sedimentation. Nodules were collected in three strips along parallel 50-m long lanes following a continuous “fishtail” trajectory (Figure 1). Due to local seabed topography, the southernmost strip 1 and middle strip 2 could not be fully completed, resulting in a total of 55 and 31 lanes, respectively. The northernmost strip 3 had a more favorable topography and a total of 85 lanes were completed. Since no riser system was present to bring the nodules to the surface, collected nodules were stored in a container at the back of the PATII vehicle and deposited in the turns of the tracks. Discharge waste, composed of sediments and fragmented nodules, was returned to the seabed by a diffuser at the rear end of the vehicle. The mining trial ended on the 20^th of April 2021 after a deployment on the seabed of approximately 40 hours in which approximately 6.6 x 10⁵ kg of nodules were removed, from an area larger than 30.000 m² (Gazis et al., submitted)¹.

A 20-core multicorer (MUC, Oktopus GmbH), equipped with 60 cm-long cores (inner diameter of 96 mm), was used to obtain soft sediment samples. Pre-impact or “no-impact” samples were obtained within the GSR Trial site throughout two consecutive sampling campaigns. During the SO268 campaign in 2019, four MUC deployments were performed along the northwest-southeast axis (Haeckel and Linke, 2021, Figure 1, Table 1). During the IP21 campaign (2021), one MUC deployment was performed prior to the trial (Figure 1; 05MUC). In addition, two deployments were taken after the trial, located at 500 m northwest of the impact site. This area was also considered as “undisturbed” due to the south-southeast dominated dispersal of the produced sediment plume, based on preliminary data from a turbidity sensor mounted on the AUV and a stationary sensor (NIOZ_PFM_05) placed at 200 m distance in between the MUC locations and the PATII trial site (Gazis et al., submitted)¹. As such, both samples were categorized and further referred to as “Pre-impact” within this paper. This correct categorization could be confirmed after analyzing the AUV-based photo mosaics for plume deposition on the seafloor (Gazis et al., submitted)¹ and numerical modelling of the plume dispersion matching the sensor data (Purkiani and Haeckel, unpublished).

In addition, three MUC deployments were performed after the PATII trial during the IP21 campaign within the tracks of the PATII vehicle (Collector Impact, 1-3 days post-impact, Figure 1, Table 1). Samples were also obtained in the area adjacent to the collector test site (Plume Impact), subject to sedimentation (3-5 days post-impact, Figure 1, Table 1). The post-trial impact categorization (Collector Impact and Plume Impact) and positioning of the MUC deployments were chosen on board based on sensor readings (e.g., AUV and CTD-mounted sensors) and inspection of AUV/ROV derived imagery (Vink et al., 2022). From each deployment, 1-2 cores and 2-3 cores were allocated for meiofaunal morphological and metabarcoding analysis, respectively. Sediment was also collected from 2-3 cores during each deployment for environmental analyses. The exact amount of cores allocated for each type of analysis are provided in Appendix Table 1.

Processing of Pre-impact samples from the SO268 campaign are described in Lefaible et al. (2023) and are comparable to the methods used for the samples collected in IP21. The IP21 cores for meiofaunal analyses were photographed and stored on board at 4°C. The upper 0-5 cm layer were used for further morphological and metabarcoding analyses. Sediments were collected in plastic containers and fixed with 4% formaldehyde-seawater solution and 98% denatured ethanol for morphological and metabarcoding samples, respectively. Cores obtained for environmental analyses were sliced at different intervals depending on the type of variable under study. For the pigment measurements, cores were sliced at a 1 cm depth resolution (down to 20 cm depth), whereas cores were sliced at 0-1 cm, 1-5 cm and 5-10 cm sediment depth for the grain size, total organic carbon (TOC) and total nitrogen (TN) measurements. Sub-samples of 5 ml were taken for pigment analyses from each depth layer using a cut-off syringe, which was then sealed with a cap and wrapped in aluminum foil. Morphological samples were stored at room temperature, whilst metabarcoding and environmental samples were preserved at -20°C until transported to the Max Planck Institute for Marine Microbiology for pigment analysis (MPI-MM Bremen) and Ghent University for morphological, metabarcoding and environmental (grain size, TOC, TN) analyses.

Samples from one deployment (40MUC) were removed from the final biological IP21 dataset because sediment cores appeared highly disturbed after sampling. Final environmental datasets for the all the sampling categories were based on the mean of the pooled 0-5 cm sediment layers. Comparisons between the Pre-impact samples obtained during the two campaigns (SO268 vs. IP21) in terms of univariate environmental and meiofaunal properties were performed by means of independent t-tests. Since pigment measurements were done with different methods during each campaign, these variables were not included for the Pre-impact analysis.

For the immediate-disturbance study, univariate environmental and biological differences between the sampling categories from the IP21 campaign (i.e., Pre-impact, Collector Impact and Plume Impact) were assessed through a one-way analysis of variance (ANOVA). Data normality and homoscedasticity were assessed by means of a Shapiro and Levene’s test, respectively. If the ANOVA analysis was significant, Tukey post-hoc tests were performed to compare all possible pairs of groups/categories. In addition, power analyses were performed for the univariate community descriptors under study for the PATII impact assessment (IP21). Omega squared (ω²) was chosen as a measure of effect size to quantify the magnitude of an effect, as it is considered to be a reliable estimate when sample sizes are small (Field, 2013). Firstly, ω² values were calculated based on the one-way ANOVA models of each variable, through the omega_squared() function and were classified as very small (ω² < 0.01), small (0.01 ≤ ω² < 0.06), medium (0.06 ≤ ω² < 0.14) or large (ω² ≥ 0.14) according to Field (2013).

A one-way permutational analysis of variance (PERMANOVA, distance metric: unweighted UniFrac (Lozupone and Knight, 2005) was also applied to test for differences in terms of nematode ASV genus composition between Pre-impact sampling campaigns (SO268 vs. IP21) and impact categories during the IP21 campaign. Permdisp tests were performed to assess the homogeneity of multivariate dispersions. All statistical analyses and graphs were produced with R Statistical Software (v4.1.2; R Core Team 2021) using the following R packages: car (Fox and Weisberg, 2019), tidyverse (Wickham et al., 2019), vegan (Oksanen et al., 2019) and pwr (Champely, 2020).

Abyssal sediments in the CCZ area are characterized as fine-grained, mainly consisting of clay and silt, rich in manganese-oxide (MnO₂) (Pape et al., 2017; Hauquier et al., 2019; Volz et al., 2020). Primary productivity in the surface waters of the CCZ is oligotrophic, resulting in low particulate organic carbon (POC) fluxes to the seafloor (< 2 mg C_orgm^-2d^-1) (Lutz et al., 2007; Volz et al., 2018a). Pre-impact upper sediments (0-5 cm) within our study site exhibited environmental properties (e.g., silt > 70%, clay < 20%, TN < 0.2% and TOC < 0.6%), comparable with previously described baseline conditions in the GSR exploration contract area (De Smet et al., 2017; Pape et al., 2017; Volz et al., 2018; Hauquier et al., 2019). Deep-sea nematode assemblages tend to be characterized by a few prevailing and many rare taxa (Lambshead and Boucher, 2003; Pape et al., 2017; Hauquier et al., 2019; Lefaible et al., 2023). This pattern was also observed in our study, with highest relative read abundance for the genera Acantholaimus and Halalaimus, which are commonly reported as dominant in nodule-bearing abyssal sediments (Pape et al., 2017; Hauquier et al., 2019; Macheriotou et al., 2020; Lefaible et al., 2023). Previous metabarcoding studies further corroborated that Desmoscolex, Chromadorita and Phanodermopsis are also important genera in terms of relative read abundance (> 5%) in the CCZ (Macheriotou et al., 2020; Lefaible et al., 2023).

Nevertheless, significant geochemical and biological differences were found between baseline sediments taken at the Trial site during two consecutive campaigns (i.e., SO268 and IP21). Stations sampled immediately prior to the PATII trial (Pre-impact IP21) were characterized by significantly higher TOC contents and labile organic matter fractions (C/N ratios) then those sampled during SO268. Differences in food availability may be a consequence of temporal variation in seafloor POC flux but lateral heterogeneity in TOC has also been attributed to interactions between bottom-water currents, topographic relief (e.g., basins, ridges) and the degree of organic matter degradation (Lutz et al., 2007; Volz et al., 2018a; Vonnahme et al., 2020). Since benthic fauna in abyssal sediments is driven by sediment composition and food availability, it can be expected that spatial differences in these variables will also result in spatially patchy meiofaunal communities (Lambshead and Boucher, 2003; Pape et al., 2017; Hauquier et al., 2019; Kuhn et al., 2020; Washburn et al., 2019). Next to this natural heterogeneity on a larger spatial scale (i.e., between sampling stations), abyssal sediments and especially meiofaunal communities are known to exhibit small-scale heterogeneity even between cores of the same deployment (Rosli et al., 2018; Uhlenkott et al., 2021; Lefaible et al., 2023), which was also observed within this study.

For example, average nematode abundances of the Pre-impact IP21 samples varied considerably, ranging from 7 up to 145ind./10 cm². This high degree of variability regarding nematode abundance is probably also linked to the presence of polymetallic nodules, as they represent the only natural hard substrate in these areas and will therefore influence local habitat complexity, which is considered to be an additional driver of benthic faunal distribution (Vanreusel et al., 2010; Zeppilli et al., 2016; Uhlenkott et al., 2021). A very low number of nematodes (7 ind./10 cm²) occurred in one core sample that contained a large, superficial nodule, whereas higher (≥ 140 ind./10 cm²) to moderate (60-70 ind./10 cm²) abundances were observed in cores containing smaller nodules, located in deeper sediment layers (Appendix Figure 5). Reduced nematode abundance has consistently been found in nodule-rich areas compared to nodule-free areas (e.g. Miljutina et al., 2010; Singh and Ingole, 2016; Pape et al., 2017; Hauquier et al., 2019; Pape et al., 2021; Uhlenkott et al., 2021) and it has been proposed that a negative relationship between nodule presence and faunal abundance exists due to reduced sediment volume for nematodes (Pape et al., 2017; Hauquier et al., 2019). While the current increase in deep-sea research will provide better understanding of baseline conditions in proposed mining areas, we would like to emphasize that the high level of natural heterogeneity observed in our study has strong implications for effective impact assessments and distinguishing anthropogenic disturbances from natural variability (see section 4.3).

Studies such as ours, in which a sub-industrial scale vehicle was used, will be crucial to perform detailed DSM impact assessments and quantify the spatio-temporal extent of the effects that are induced during nodule collection. Therefore, the most pronounced DSM effects in the form of sediment removal, seabed alterations and enhanced sediment deposition are discussed separately throughout the following sections (4.2.1, 4.2.2) in an attempt to identify the mechanisms behind the different stressors and the associated meiofaunal responses. Based on our findings, we also provide methodological recommendations for future DSM research on benthic habitats (4.3).

The comparison of environmental conditions within the Pre-impact and post-impact categories of our study confirm the hypothesis that a considerable fraction of the upper sediment layer within the collector impact site was removed (a minimum of 5 cm, Gazis et al., submitted)¹ during nodule collection and that surrounding areas were subject to sediment blanketing (± 2-3 cm). Although average values (i.e., reduced levels in Collector Impact sites, enhanced in Plume Impact sites) for the Nematoda community indices support that environmental changes will also have impacts on the associated biota, differences between categories were not statistically significant. Post-hoc power analyses did, however, reveal that nematode abundance and ASV Shannon diversity tests were characterized by moderate to strong effect sizes, indicating the biological relevance of changes in these variables (Field, 2013). Moreover, univariate biological tests had low statistical power and a substantial increase in samples size (n = 30 per category) would be required to achieve appropriate statistical sensitivity (power level 80%). As such we propose that the lack of significant biological results might be attributed to the insufficient replication combined with noticeable within-category variability. This conclusion further illustrates what can be considered as the main challenge of DSM research, namely: quantifying multiple sources of variability associated with i) the intricate collector-induced seabed disturbances and ii) the natural heterogeneity of abyssal sediments (Clark et al., 2020).

Nodule collection by the PATII vehicle induced a sequence of pressures (i.e., sediment removal, intricate seabed modification by collector vehicle and sediment blanketing), resulting in complex seabed modifications with different disturbance types such as i) central lanes of sediment removal, ii) alternating bands of depressions/ripples in caterpillar tracks, ii) microscale variations in topography due to unstable sediment “clumps” and varying levels of sediment deposition in the Collector Impact site. It will therefore be necessary to distinguish and integrate these disturbances into the data collection of future benthic DSM studies. A proposed solution would be to apply a nested sampling design within the “Collector Impact” category, where sufficient replication is performed for the major subgroups (e.g., central lanes, depressions and ripples of caterpillar profile). It could also be opted to combine MUC samples with ROV push cores, which are taken in a more targeted manner owing to the live camera feed. As shown in this study, a priori methods like the visual inspection of high-resolution images from the collected samples can help to determine specific seabed alterations (e.g., estimation of position in PATII tracks, exact sedimentation level). Whereas our results indicate that both post-impact categories were affected by enhanced sedimentation, many uncertainties remain regarding the effects of sediment blanketing on overall seabed integrity. Regarding meiofaunal responses, one very important aspect will be to find methods (e.g., staining, DNA/RNA techniques) to determine whether the assemblages were alive or not at time of sampling. Moreover, continued monitoring will be crucial to identify the long term viability of the present populations.

Environmental and biological properties of the Pre-impact samples in our study fell within the range of previously described baseline conditions in the same area. Nevertheless, heterogeneity across small and large spatial scales has been a consistent pattern in the CCZ. As such, an important implication is that environmental variables should preferably be obtained from the same samples used for biological analyses. Furthermore, the sediment depth resolution (i.e., entire 0-5 cm layer) used in this study was not appropriate to distinguish redeposited from natural sediments in adjacent areas (i.e., Plume Impact category) where sediment blanketing was up to 3 cm. Therefore, it should be explored if it is feasible to adjust the slicing depth resolution of the core samples, depending on the expected sediment deposition thickness. Furthermore, we want to emphasize that findings in our study reflect immediate (i.e., one week after trial) benthic responses, and results from the follow-up SO295 campaign will provide additional insights on how the benthic environment has evolved over a period of 1.5 year (MI2, 2018). Lastly, given the observed inter-annual variability in terms of environmental and faunal baseline conditions in our study. Understanding natural temporal variability based on adequate time series and a sufficiently high number of replicates of abyssal benthic communities is therefore crucial for establishing and predicting the impacts of mining activities in deep-sea ecosystems.