For several years, mining of manganese (Mn) nodules from the deep-sea floor is again considered a worthwhile option to meet future resource demands. Several nations have secured exploration contracts in areas beyond any national jurisdiction as they seek economic benefits and/or aim for technological leadership in terms of deep-sea mining. Current plans for a future mining scenario involve collectors that will move on the sea floor gathering Mn nodules from the top 10 to 30 cm of the sediment, most likely using a hydraulic collection mechanism (Kuhn et al., 2011; Oebius et al., 2001). This principle implies considerable consequences for the benthic environment in the mined area. Besides the removal of the Mn nodules as an important hard-substrate habitat on the abyssal plains (Purser et al., 2016; Vanreusel et al., 2016; Thiel et al., 1993), the mining activities will completely rework the top sediment layers and resuspend large amounts of sediment into the water column. Depending on the plume properties such as particle size, flocculation behaviour, sediment mass per litre and the prevailing current conditions, these sediment particles might be transported outside the mined area. The deposition of this material will cause a secondary impact on the environment by clogging filter feeders and burying the sessile fauna, which are both adapted to the low sedimentation rates in the deep sea (Thiel and Schriever, 1989). Resedimentation of this material can also lead to differences in local geochemical gradients and consequently might influence the recolonization processes of the primary and secondary disturbed areas. To evaluate these effects on the environment, several benthic impact experiments (BIEs) and one Recolonization Experiment, the German Research Project “Disturbance and Recolonization Experiment – DISCOL” (http://www.discol.de, last access: 5 March 2020), have been conducted in the past within different large Mn-nodule areas, including the Peru Basin (Thiel and Schriever, 1989), the central equatorial Pacific (e.g. Burns, 1980; Fukushima, 1995) or the Indian Ocean Basin (Desa, 1997). Information about the sediment plume dispersal during the different large-scale disturbances is compiled in Sect. 1.2. A review of the biological responses to such BIEs was recently presented by Jones et al. (2017), and studies by Simon-Lledó et al. (2019) in the DISCOL Experimental Area (DEA) show that colonization pattern differences still exist between the disturbed and undisturbed areas even after 26 years.

In the late 1970s, the first so-called “mining test” operations were conducted in the central North Pacific as part of the DOMES project (Ozturgut et al., 1978, 1980) that used a suction dredge towed on skis to create a disturbance for illustrating potential mining impacts. Here, the experimental area was surveyed before, during and after the experiment, with each disturbance lasting for several hours (see Table A1 in Appendix A for details on location, duration, monitoring techniques and impacted area). For the first three tests in spring 1978, operated by Ocean Mining Inc. (OMI), detailed data about the induced sediment plume were derived from different sampling methods including sediment coring and sediment traps (see Burns, 1980, and details in Table A1), and results indicate a plume dispersal of up to 16 km downstream of the created disturbance (Table A1). Model results based on the OMI experiment indicate a sediment blanketing thickness of ≤1 mm beyond a 400 m distance to single disturbance tracks. An extrapolation of these results for a potential mining scenario was performed and predicted a distribution of resuspended sediment particles of up to 160 km distance (Lavelle et al., 1981). Another mining test phase in November 1978 focused on the distribution of a surface discharge plume (Ozturgut et al., 1980). During an 18 h operation by Deepsea Ventures Inc. and Ocean Mining Associates (OMA), a second sea floor mining test was conducted in November 1978 by the Ocean Minerals Company (OMCO) using a remote-controlled self-propelled miner (RCM) (Chung, 2009). This vehicle removed approximately 4 cm of the upper sediment layer (Khripounoff et al., 2006), creating a track of 1.5 m width (Miljutin et al., 2011). The aim of this experiment was mainly to test the mining technology and not to monitor the benthic impact of the plume. Hence, detailed information regarding the sediment plume dispersal right after the impact is missing. In 2004, the disturbed area was revisited and investigated for its ecological recovery (Mahatma, 2009; Miljutin et al., 2011), indicating only a near-track influence of redeposited sediment.

Chronologically the next and largest ever created disturbance was conducted in the DISCOL Experimental Area in the Peru Basin. For creating the disturbance, a plough harrow (8 m width) was towed 78 times crisscrossing through a circular area of 2 nmi in diameter (Thiel and Schriever, 1989). Due to technical problems the deployed nephelometers at that time did not detect the sediment plume and the amount of suspended material remains largely uncertain. Nevertheless, the presence of a plume in the water column about 6 h after the last plough deployment was confirmed by visual observations (Thiel and Schriever, 1989). Numerical modelling predicted a dispersal of the suspended sediment for several kilometres with coverages of resettled material of >100 g m-2 up to a distance of 2 km (Jankowski et al., 1996; Table A1). The effects of the disturbance were investigated just after the experiment (RV Sonne cruise SO61; Thiel and Schriever, 1989) as well as 0.5 (cruise SO64; Schriever, 1990), 3 (cruise SO77; Schriever and Thiel, 1992), 7 (cruise SO106; Schriever et al., 1996) and finally 26 years later (cruise SO 242; Boetius, 2015; Greinert, 2015) to document the environmental impact, the recolonization and the sediment geochemical equilibration of the disturbed sites in comparison to a number of undisturbed reference sites in the vicinity.

Again north of the Equator, the first large-scale benthic disturbance experiment in the eastern Clarion-Clipperton Fracture Zone (CCFZ) conducted by the United States was the Benthic Impact Experiment II (BIE-II) in 1993, using the “Deep Sea Sediment Resuspension System” (DSSRS) (Brockett and Richards, 1994; Tsurusaki, 1997) as the disturbance tool (Trueblood and Ozturgut, 1997). The initiated sediment plume was monitored with camera systems, sediment traps and transmissiometers, which were moored at different distances from the tow zone in order to estimate the distribution areas of resettled sediment and the plume dispersal in the water column. The studies revealed an area of strong sediment blanketing within the first 50 m downstream of the disturbance and a decreasing blanketing thickness with increasing distance. Moorings located 400 m away still detected suspended material passing by and also deployed sediment trap samples indicating a maximum “blanketing” thickness of 1 mm. In contrast to these data, camera observations suggested a sediment blanketing thickness of 1–2 cm close to the disturbance zone (Jones, 2000) already indicating that the sediment traps might have missed the additional sediment transport of initiated gravity flows just above the sea floor.

A total of 1 year after the American experiment, the Metal Mining Agency of Japan (MMAJ) carried out another disturbance study within the CCFZ, the “Japan Deep sea Impact Experiment” (JET) in 1994 (Fukushima, 1995). The disturbance was again created with the DSSRS (Tsurusaki, 1997). The distribution of the initiated sediment plume was analysed using two different approaches. One approach measured the thickness of the blanketing sediment layer using sediment traps and spatially interpolated the results using kriging. A dispersal of 2.5 km in length and approximately 1 km in width was calculated and a maximum blanketing thickness of 2.6 mm was determined (Barnett and Suzuki, 1997). The second approach used visual data from deep-towed camera surveys to estimate the extent of the sediment blanketing that covered the Mn nodules. Respective results show that the “heavy” resedimentation area, defined by a thickness >0.26 mm, did not extend for more than 100 m away from the disturbance track. Thinner blanketing <0.26 mm was observed over an area of ∼3 km length and ∼2.5 km width around the disturbance (Yamazaki and Kajitani, 1999), covering a much wider area compared to the kriging approach.

In 1995, the Interoceanmetal Joint Organization (IOM) conducted a benthic disturbance experiment (IOM-BIE) over an area of 2000 m × 1500 m also in the eastern CCFZ, once more using the DSSRS (Kotlinski and Stoyanova, 1999; Radziejewska, 2002). Studies focused on the physical and chemical properties of the resuspended and resettled sediments rather than on the spatial distribution of the material; this leads to only limited information on the amount of resuspended material. Radziejewska (2002) estimated the volume of resuspended material to be approximately 1800 m3 over the entire duration of the experiment, but the actual volume is not known.

In 1997, the “Indian Deep sea Environment Experiment” (INDEX) was carried out in the Central Indian Ocean Basin. For the fourth time, the DSSRS was used to create the disturbance during 9 d of operation (Desa, 1997; Sharma and Nath, 1997). Results from sediment traps distributed up to 800 m away from the track show an increase in average particle fluxes from 48 to 150 mg m-2 d-1 during the disturbance phase. The flux decreased to 95 mg m-2 d-1 within the first 6 d after the disturbance stopped (Sharma, 2001). Based on visual observations, most of the sediment particles resettled already within 150 m from the edge of the disturbance area (Sharma et al., 2001), with the major part of material settling within an approximately 100 m distance (Sharma, 2000).

The last large-scale BIE was conducted in 1997 by MMAJ within the area of the Marcus-Wake Seamounts in the North Pacific Ocean (Yamada and Yamazaki, 1998). The induced sediment plume was visually monitored (Yamazaki et al., 1999) and data revealed a sediment blanketing thickness on top of Mn nodules of up to 0.2 mm (Yamazaki et al., 2001). Due to the different geological setting (seamount in 2200 m water depth) and different sediment properties (calcareous sediments, coarser sediment particles, stronger currents), these results are not directly comparable to the results from most of the other BIEs mentioned above.

Reviewing the different large-scale BIEs and pilot mining tests conducted between the late 1970s and late 1990s, it becomes obvious that the different experimental setups and the missing uniform definition of “a” plume (grain size distribution, flocculation behaviour, total mass per litre, settling velocity, etc.) make it impossible to use the presented information for a meaningful prediction of the behaviour of a sediment plume created during a real deep-sea mining operation (Peukert et al., 2018). Thus reconstructing the initial disturbance of 1989 in the DISCOL area, which is considered to be the most extensively sampled and monitored BIE site, might help to gain new and more conclusive insights in terms of the distribution of resuspended and redeposited sediment during and shortly after conducting the disturbance and consequently be used as a basis for sample interpretation from this area.

This study presents new data from the DEA, which were acquired in 2015 during RV Sonne cruise SO242-1 with state-of-the-art AUV multibeam and side-scan sonar systems, cameras and underwater navigation technology (Greinert, 2015).

Since 1989, major technological advancements improved deep-sea investigations with regard to data acquisition technologies and positioning accuracy. In 1989 GPS for example was not as sophisticated and high-resolution acoustic sea floor mapping with multibeam echo sounder systems (MBES) was not as developed as it is today (e.g. 59 beams compared to 432 beams; single swath compared to dual swath; Lurton, 2017). AUV-based technologies did not exist.

To acquire the most accurate data of the old plough tracks, the entire DISCOL area was re-mapped using ship- and AUV-based hydro-acoustic MBESs with different resolution (Boetius, 2015; Greinert, 2015). This provided new information for reconstructing the extent and impact of the initial disturbance experiment, the different geological settings within and next to the DEA, and related varying habitats. The results presented in this study mainly focus on the data collected by GEOMAR's AUV Abyss (Linke and Lakschewitz, 2016, 10.17815/jlsrf-2-149). The AUV was deployed in three different modes running MBES, side-scan sonar (SSS) or a photo camera system enabling autonomous mapping with a resolution of 2 m for bathymetric data, 0.5 m for SSS data and a few millimetre per pixel for photo surveys. All systems show clear evidence of the disturbance tracks created by the plough harrow 26 years before.

This study presents the best geo-referenced data set of the study area through a combined processing of the available ship- and AUV-obtained acoustic and optical data. In addition to this mapping exercise, the succession of the disturbance tracks as well as their correct location are reconstructed, as this could not accurately be documented in 1989. Although the 78 plough tracks were created over a period of only 4 weeks (Thiel and Schriever, 1989), a more detailed understanding of their sequence is relevant regarding faunal differences from within or close to plough tracks in strongly or more weakly disturbed parts of the DEA. Furthermore for the understanding of varying down-core geochemical gradients, the spatial thickness change of the resettled sediment, the “blanketing”, needs to be understood. This thickness distinctly differs between the plough tracks depending on if they were created in an earlier or later stage of the disturbance, which highlights the difference between high plume sedimentation rates and natural deep-sea low sedimentation rates. Next to this an unbiased and correct comparison between areas that have not been impacted by any resettled sediment with areas that have been impacted to various amounts should be performed. Interpreting biological or geochemical results correctly requires a very precise knowledge of the exact and absolute sample or footage location on the sea floor and their spatial relation to the tracks which are only a few metres wide and apart from each other. Thus a correct geo-referencing of all different data layers was a significant task of this study, and, although highly developed positioning systems were used in 2015, uncertainties and deviations of tens to a few hundreds of metres occurred. This task became even more important for geo-referencing legacy data from 1989 for conclusively defining changes between 1989 and 2015 and spatial sediment resettling differences established already during the plough experiment.

Until 2015, the location and path of the disturbance tracks as well as the position of video and photo material of the past OFOS (Ocean Floor Observation System) surveys only existed as a vast collection of analogue (i.e. cruise reports, printed large navigational charts (Fig. 1), video cassettes and slide films) and some digital records (i.e. OFOS annotation files, sample analysis as text or EXCEL files; e.g. Bluhm, 1994; Bluhm and Thiel, 1996; Thiel and Schriever, 1989; Schriever, 1990; Schriever and Thiel, 1992; Schriever et al., 1996). In preparation for the 2015 cruise, these records were digitized and compiled in a database, also including all other available sampling stations (i.e. box corer (BC), multi-corer (MUC), moorings, baited traps; see Drazen et al., 2019) that were a part of the first four expeditions to the DEA. This database was used for station planning prior to the SO242 cruises and allows comparison of past and present disturbance levels and sea floor and ecosystem conditions at their best possible correct location.

AUV operation in great water depth suffers from inaccurate positioning of acquired data sets. Underwater positioning is typically determined using hydro-acoustic techniques as ultra-short baseline (USBL, measurement between the ship and the AUV) or long-baseline systems (LBL, triangulation of the AUV using sea-floor-deployed transponders). The AUV Abyss navigates autonomously using a combination of different navigational methods (Linke and Lakschewitz, 2016). During our studies LBL navigation was only used to set an accurate starting position of the AUV at the beginning of each survey after arriving at the sea floor. No additional LBL fixes were considered as this often results in abrupt track corrections that cause unwanted artefacts, particularly in SSS data. Instead, navigation after the initial LBL fix relied on Doppler velocity log (DVL) data, inertial navigation sensors and dead reckoning data fusion as supplied by the AUV system (Linke and Lakschewitz, 2016). Typically, such navigation is prone to slow drifts, which over the course of an entire mission (up to 20 h operation time) can add up to offsets of several tens or hundreds of metres. These navigational shifts need to be derived and corrected during processing when comparing or combining several different data sets such as MBES, SSS, imagery of the AUV, imagery of OFOS and ship-based bathymetry.

To achieve the best possible alignment and absolute geo-referencing, the ship-based EM122 bathymetric data with a spatial resolution of 38 m were taken as the absolute reference layer (Fig. C1 in Appendix C). The AUV bathymetric data with a spatial resolution of 2 m were resampled to match the 38 m resolution enabling a direct grid comparison (e.g. grid subtraction, Fig. E1 in Appendix E) and correction of vertical and lateral offsets of the AUV bathymetric grid relative to the ships' data layer. Using 5 m contour lines to visualize morphological features in the area, the 38 m AUV bathymetry was shifted/stretched manually onto the EM122 data (Fig. 4). Subsequently, the high-resolution AUV bathymetric digital terrain model was shifted in the same way using the ArcGIS 10.2 Georeferencing Toolbox for geographic corrections (contour lines were derived with the Spatial Analyst Toolbox and grids were subtracted to see z offsets using the Raster Calculator function). Figure 4 shows the high-resolution AUV bathymetry with the contour lines derived from the ship-based bathymetric grid to visualize the accordance of both data sets after the alignment.

The SSS map was geo-referenced relative to the AUV MBES data using a number of disturbance tracks, visible in the bathymetry (Fig. 5a, b and c) and the SSS data (Fig. 5d), and three prominent Mn-nodule-free depressions, which appear distinctly dark in the SSS map (Fig. 5c), as anchor features.

Based on the same structures, the photomosaics of the DEA could be aligned to the SSS map (Fig. 5e). Finally, visually detectable sampling locations of BC, gravity corer (GC) or MUC impacts were used to validate the accuracy of the geo-referencing by comparison with their actual USBL positions (see Appendix D for details).

Disturbance tracks visible in SSS data (Fig. 5d) were manually digitized using functionalities of ArcGIS. Each track was given a unique identifier and was assigned to one of four classes reflecting the general orientation of the respective track: H for E–W-orientated tracks, D for NW–SE- and NE–SW-orientated tracks, V for N–S-orientated tracks, and P for non-continuous tracks and track segments (H: horizontal; D: diagonal; V: vertical; P: parts of tracks; Fig. 6; Table 1). The track IDs were arbitrarily given during the digitizing and were not renamed after the sequencing; thus the numbers do not reflect the age sequencing. During the digitizing it occurred that some of the tracks labelled as “P” for “parts” are as long as others assigned to one of the “entire track groups (V, H, D)”, due to subsequent extension after further investigations. Because of this there is no clear definition when tracks are labelled P; however, P tracks are typically shorter than 1200 m.

Generally during the ploughing in 1989 several tracks were undertaken during one deployment of the plough (station name PFEG-1 to PFEG-11; PFEG1 was a gear handling test a few nautical miles south of the DEA). After the first two groups PFEG2 and PFEG3, OFOS dives OFOS009 and OFOS010 were conducted during SO61 and the photo and video material collected during these two OFOS dives could be examined for track occurrences. The track orientation was determined from each sea floor image and matched to the track orientation on the SSS map considering the course over ground (COG) and heading of the OFOS (Fig. 7) to distinguish the correct plough track. This way and considering the log files from cruise SO61, which gave an idea about rough course and location of the ploughs, the tracks corresponding to the first two plough groups (15 tracks) could be identified. Considering photo and video material from remotely operated vehicle (ROV) (GEOMAR Helmholtz-Zentrum für Ozeanforschung, 2017), AUV and OFOS surveys and the SSS data, intersections were visually examined to establish a relative age succession between the investigated tracks (Fig. 8).

For some intersections the sequence could not directly be established and was inferred considering the relative track age information of other intersecting tracks (Fig. 8). Unfortunately, this workflow could not be applied for the later groups of tracks (PFEG4 to PFEG11) since they were not directly followed by an OFOS survey. Thus the track density until the next OFOS observation became too high, and, considering the navigation uncertainties, an unambiguous assignment of the tracks is not possible. The reconstruction of the age succession of all tracks was finally carried out using an 84×84 matrix (including 60 identified tracks and 24 track segments, Table F1 in Appendix F) where all observed crossings were included. Logical process of elimination and cross-referencing of individual tracks relative to all other tracks in combination with their position and the reconstructed ships' navigation during the time of the experiment (Fig. 1a) were performed. Based on this, the tracks were assigned to their respective PFEG and to track IDs (Table 1).

The initial impact of the plough tracks is given through the mixing (ploughing) of the top 20 to 30 cm of the sediment and the related suspension of sediment into the bottom water (Foell et al., 1990). Nodules were not removed from the sea floor but ploughed under (Thiel and Schriever, 1989). The resedimentation of the initiated sediment plume is considered the secondary impact. For reconstructing the initial impact and the proximal (visible in images) sediment blanketing, the course of the plough tracks was used in combination with bottom current information recorded during the time of the experiment to establish a disturbance intensity map (including initial and secondary impact). Also considering the plume deposition information from other BIEs (Table A1) and the recent study by Peukert et al. (2018), the qualitative sediment blanketing thickness within the DEA was determined based on the following assumptions and set parameters. Each track was assumed to have a width of 8 m, not considering the possible handling problems with the plough harrow (e.g. being towed only on the side, short loss of bottom contact; Thiel and Schriever, 1989). The intensity of the disturbance was assumed to be the highest within and close to the tracks, and the sediment blanketing thickness was assumed to decrease with increasing distance off the track. Studies from other BIEs showed visual sediment blanketing distances between 70 and 150 m in current direction away from the track. It is assumed that the majority of the resuspended sediment (about 90 %) resettled over this distance (Lavelle et al., 1981; Peukert et al., 2018).

The main factors controlling the redeposition are current speed and direction and particle settling velocity with the last being describable as a function of the particle size according to Stoke's law and the method described by McCave (1984; Jankowski et al., 1996). The sediments within the DEA are composed of layered clayey silts or silty clays, with a sand fraction of ca. 5 % consisting of foraminiferous residues and shell fragments (Grupe et al., 2001). According to Lavelle et al. (1981), Schriever et al. (1996) and Becker et al. (2001) the stirred-up sediment mixture induced flocculation and aggregation of particles, causing a very rapid resedimentation (≥1 cm s-1) of the plume within the first 20 m away from the track. The latest research on mining-induced sediment plume behaviour also indicated a near-track heavy sediment blanketing (Gillard et al., 2019). The sediment blanketing decreases as a function of reduced particle settling velocities as finer particles dominate the plume composition and stay longer in suspension (Lavelle et al., 1981).

Bottom current direction and velocity determine the direction of the resedimentation area and sediment spreading (Lavelle et al., 1981; Jankowski et al., 1996; Greinert, 2015). Bottom currents in the central Pacific are reported to be distinctly different even at locations only a few kilometres apart (Robinson and Kupferman, 1985). Several measurements in the DISCOL area revealed a predominantly northern to northwestern direction, with maximum current speeds of 17 cm s-1 (Thiel and Schriever, 1989; Schriever and Thiel, 1992) indicating a transport of the resuspended particles primarily in this direction. The undertaken measurements showed that the currents in the DEA alternate between strong (>5 cm s-1) and quasi-unidirectional currents towards the NNW and weaker currents (< 1–3 cm s-1) with greater directional variability (Klein, 1993, 1996). This variability has also been observed during the first cruise SO61 to the DISCOL area (Thiel and Schriever, 1989), with the “strong” current regime occurring during the first leg (February 1989) and the creation of PFEG1 to 7 and the weaker currents towards the end of the second leg (March 1989) and the creation of PFEG8 to 11, where the currents showed semidiurnal change of current direction from predominantly NNE to predominantly SSE. This certainly affected the sediment plume dispersal.

Since no information about the amount of resuspended material is available, the impact is reconstructed qualitatively using values resembling disturbance intensity between 1 within the disturbance tracks and 0.1, representing the deposition of 90 % of the resuspended material at the maximum distance of the proximal disturbance. With regards to other impact monitoring results from large-scale disturbances (e.g. Lavelle et al., 1981, Table A1) and the results of small-scale disturbance experiments conducted during SO239 (Martinez-Arbizu and Haeckel, 2015; Peukert et al., 2018), SO242/1 (Greinert, 2015) and SO242/2 (Boetius, 2015), the maximum distance affected by sediment blanketing was assumed to be 120 m with and 20 m against the current direction for the strong current regime. These distribution values and a distribution direction of 334∘ were set for PFEG 2 to 7 and all recognized parts of plough tracks which could not be assigned to a distinct PFEG. To account for the changing conditions during weaker bottom currents (PFEG 8 to 11), the distances were set to 100 m with and 30 m against the current direction. Based on the statistics of the current directions (Thiel and Schriever, 1989) during the creation of 31 recognized tracks of that period, the plough tracks were divided in two groups, one considering a NNE current (towards 18∘, 19 tracks) and the other group considered a SSE current (towards 143∘; 12 tracks). Considering the semidiurnal current direction change, the assignment of the tracks to one of those groups was based on the estimation of the track creation time and duration in consideration of the length of one track and the speed of the ship while ploughing and the determined relative sequence of the single tracks.

For calculating the sediment plume deposition down-current and up-current (due to turbulence), the following simple function was used: 1y=e-xR, with y representing the relative sediment thickness at distance x from the disturbance track.

An exponential function was chosen to account for the effects of flocculation and aggregation of the resuspended sediment closer to the track. The factor R was introduced to meet the assumption that 10 % of the resuspended material remains in the water column and is redeposited at greater distances (Lavelle et al., 1981; Jankowski et al., 1996). This factor was considered for the particle transport with the prevailing bottom currents. Against the bottom currents, the resuspended material was assumed to completely resettle within either the first 20 m for strong currents or 30 m for weak currents. The relative sediment thickness was calculated in 0.8 m steps away from each disturbance track considering the above-mentioned current directions. The final blanketing map was produced by adding all relative sediment thicknesses within each square metre of the DEA area using the blockmean command in Generic Mapping Models (GMT, argument -Ss to get the sum; Wessel et al., 2013) and producing an interpolated grid using the nearneighbor command. It is assumed that the plough intensity and sediment resuspension did not change as each plough track was created.

Navigational offsets were detected between the different AUV missions with lateral offset between AUV and ship data of 30 to 80 m (Fig. 4). As AUV data sets from four different MBES surveys are used, a good geo-referencing of the completely compiled AUV data set on the ships' bathymetry was not possible. Therefore a focus for the best possible alignment was set to the DEA region with only three AUV data sets. To check for the improvement of the geo-referencing, the AUV bathymetry was subtracted from the ship-obtained data set at identical resolution (Fig. E1). Prior to shifting and stretching of the AUV grid, the depth differences showed a mean offset of -9 m. Thus 9 m was added to the entire AUV bathymetric grid to account for this absolute z offset, and after shifting/stretching the difference between the AUV and ships' bathymetry showed the mean to be at 0 m depth difference with only a ±0.5 m median range (Fig. E1).

As for the MBES data, the lateral offset between different SSS data surveys was not constant but varied between 40 and 50 m. Geo-referencing the combined SSS map onto the AUV bathymetry showed offsets between 30 and 80 m that were corrected (Fig. G1 in Appendix G). The photomosaics, which could be aligned to the SSS map very accurately, show sampling locations that we compared to the USBL position during the time of sampling for validating the geo-referencing results (Fig. 9; Table G1 in Appendix G). The mean difference between the geo-referenced photomosaic sampling locations and those from the USBL navigation is 14 m (Table G1), whereas BC positions on average differ by 11 m and MUC positions differ by 19 m. These values indicate the overall absolute accuracy of the navigation and geo-referencing that could be achieved. The accuracy decreases in the outer regions of the SSS data, as no additional information such as sea floor sampling locations or characteristic features is available.

Within the SSS map a total of 60 continuous tracks were identified and assigned to three different classes: V (11), H (28) and D (21) (Fig. 6). In addition, 24 track segments were found and represent the fourth class P (Fig. 6; Table 1). Some of these partial tracks were assigned to other track identifiers (Table 1), based on the same course, but this could not be accomplished for all of the segments.

The quality of geo-referencing different data layers towards each other highly depends on prominent morphological features that are detectable in all available data layers. The depth differences of more than 5 m between the ship- and AUV-based multibeam data after geo-referencing (marked red in Fig. E1) are related to two different AUV surveys, which seem to be inconsistent. However, the range of vertical depth deviation is still within the given depth resolution of the EM122 ship system (Kongsberg, 2007) of maximum 0.2 % of the water depth (=∼8 m).

The alignment of the SSS and optical data onto the AUV MBES data was reasonably easy due to the presence of the plough marks visible in all of the high-resolution data sets; this provided a sufficient number of anchor points over the entire area. A considerable shift of 80 m encountered during the geo-referencing highlights the importance of accurate navigation and being aware of these kinds of offsets and deviations between the different data sets. When detailed sampling is requested, especially in terms of small-scaled (tens of metres) habitat mapping, wrong navigation and geo-referencing will make analyses and correct interpretations impossible.

The remaining shift of 14 m between the USBL recorded sample positions and the position visible in the photomosaics (Fig. 9) is, for the given purpose, within an acceptable range. Regarding the punctual position of sampling compared to the lateral grid resolution of 38 m of the ship-obtained MBES data as the base of the geo-referencing, such an offset might be unavoidable without considerably more technical effort of sea-floor-based navigation systems that build a temporary reference system for all deployed gear. In our case the USBL transponder for the sampling gear was mounted 100 m (MUC) or 50 m (BC) above the gear on the wire, which might be already one source of error, if the rope most likely is not vertical above the sampling gear while this touches the bottom. Small, not completely correct evaluated static offsets between GPS and the USBL antenna as well as misalignments between the motion reference unit and USBL antenna might cause another problem that is difficult to evaluate without dedicated tests. Therefore the remaining range of 10 to 20 m deviation should be considered “normal” for sample interpretation and navigational accuracy between differently derived map and location data sets.

About 77 % of the reported disturbance tracks (60 out of 78) could be identified, most of them based on the SSS data (Sect. 2.4). The 24 track segments of class P might account for the missing 18 tracks (e.g. P04 has been assigned to D21, Table 1).

The high-resolution MBES data did not fully capture the disturbance tracks due to the small morphological differences between plough tracks and the surrounding sea floor (circa 15–30 cm; Boetius, 2015) and the internal structure of the plough marks. The reconstruction of the initial disturbance was mainly based on the SSS mapping because of the higher along-swath resolution of the SSS compared to the MBES data. The penetration depth of the plough harrow in combination with its very characteristic pattern facilitates the detection of the disturbance tracks. Morphological changes that are ensonified perpendicularly (tracks parallel to the AUV flight path) cause higher reflections of the emitted signal compared to perpendicular tracks to which the small ridges and valleys of a plough track are ensonified parallel (Lurton, 2017). Thus some tracks can be seen more clearly in the SSS data than others, which also causes the sequence at some crossings to not be finally determined. The very first disturbance tracks are clearly visible within the SSS data, again indicating that the amplitude of the signal reflectance cannot be used as an indicator for their relative age. This becomes even more evident when comparing acoustic and optical data of the AUV. Some tracks that were barely visible in the SSS image (resolution: 0.5 m) could be clearly detected in sea floor photographs. Following this, the most reliable data source to establish the relative age sequence is the image and video material recorded by the various devices (AUV, ROV, OFOS) deployed during SO242 and the OFOS data from the previous cruises. The different survey altitudes and operation plans influence the area that was covered by each instrument and the quality of the images (Greinert, 2015). The AUV photomosaics turned out to show the best results in resolving the age relation of multiple tracks even in highly disturbed areas within the DEA. There were a total of 9 AUV, 18 ROV and 57 OFOS surveys conducted within the DEA between 1989 and 2015. However, since the DEA was not entirely covered by visual investigations, it is possible that some tracks which were not detected by the SSS were also not seen with the optical devices.

In general, the age reconstruction was successful, where more than one data set was available. The plough tracks could be reconstructed with the highest amount of certainty for the very first and second sets of disturbance tracks (PFEG 2 and PFEG 3). The uncertainties within the sequence regarding the absolute ages, especially with later sets of tracks (PFEG 4–11), increase since they are mainly based on statistical information and the logical method of elimination (see Sect. 2.4).

Mesoscale numerical sediment distribution modelling by Jankowski et al. (1996) and Jankowski and Zielke (2001) considering all plough tracks of the DEA experiment predicted blanketing of resettled material of >100 g m-2 up to a distance of 2 km. Due to the lack of data that measured the amount of resuspended sediment, actual mass values of the blanketing cannot be given for our near-field estimate approach. The settling velocity is also highly dependent on the sediment plume concentration (Gillard et al., 2019). As stated, only 90 % of the sediment is assumed to settle immediately due to flocculation and aggregation, causing resettling of particles within proximal distances (Becker et al., 2001; Gillard et al., 2019). The changing current conditions over the course of the plough experiment, especially in the later phases of the disturbance with a clear semi-diurnal signal (Thiel and Schriever, 1989), combined with the residence time of the resuspended particles in the water column for more than 10 h (Thiel and Schriever, 1989; Greinert, 2015) indicate that these remaining 10 % were most likely spread across the entire DEA and beyond. The sediment blanketing map should thus be considered as the minimum impact, with the SE sector being least impacted as already suggested by Thiel and Schriever (1989).

The sectors with the highest sediment blanketing are central southeast (CSE) and central west (CW) (Fig. 11a), where a high density of disturbance tracks also occurs (Fig. 11b). X-ray studies aiming at measuring the deposition thickness were performed on selected MUC samples during SO077 (Fig. 11); results imply that sectors CS, central north (CN), peripheral southeast (PSE) and central northeast (CNE) are most heavily influenced with thicknesses between 5 and 30 mm (Schriever and Thiel, 1992, Table H1). In the disturbance map, for example within sector PSE, only low disturbance is indicated, due to the very low density of tracks. A sample taken in sector CW (SO077_110MC_358, Fig. 11, Table H1) only shows a thin resedimented layer (1–2 mm), despite it being located in one of the most heavily disturbed areas (Fig. 11b) with high blanketing (Fig. 11a). This discrepancy could be explained by the more inaccurate positioning during SO77 (positions represent the ship position at the bottom contact time of the sampling gear) which hinders a punctual comparison on such a small scale. As implied in the disturbance map in the close vicinity of sampling station MC_358, there is a ∼100 m × 70 m wide patch where only thin sediment blanketing has been calculated (Fig. 11a). In this respect differences between the ships' logged position and the devices' position on ground in 2015 varied sometimes by more than 100 m (Greinert, 2015). In 1992, this distance might have even been greater as the old RV Sonne did not have dynamic positioning systems. Considering the high blanketing thicknesses proximal to the tracks, a sample location offset of several tens of metres could considerably change the result. Samples during SO077 could have been taken within or next to a track or from one punctual location within a disturbed area, where not much sediment has been deposited (Fig. 11a). This again highlights the importance of being aware of the exact sampling positions and thus the need for detailed geo-referencing for the interpretation of the data.

However the generation of the disturbance intensity map is based on simplifications, not considering the specific sediment settling parameters as particle sizes, density of particles and water turbulence. It also did not include the local morphology, which has been proven to influence the sediment plume distribution (Peukert et al., 2018). Furthermore the microrelief of ripple crests and furrows within the track will also have an influence on the sediment blanketing thickness results from sediment cores, which again require detailed position knowledge for accurate sample interpretation. These factors could also be a reason for the deviating results from the SO77 X-ray studies compared to the disturbance map of this study. For more detailed investigations, this should be considered and implemented into calculations and further sampling methods to allow an appropriate comparison of the results.

The numerous optical data acquired by OFOS, AUV and ROV during all expeditions to the DEA facilitate a comparison of the impact and its evolution over the 26 years that passed between the first and the most recent visit to the DISCOL area. Due to the explained navigation uncertainties, especially during the early visits to the DEA (SO061, SO064), a direct comparison of exactly the same square metre of the sea floor is difficult, but the comparison of different locations within an about 150 m long section of one track seems more reasonable (Fig. 12). Generally, the fine morphology of the disturbance tracks appears to be smoothed out over time by currents and natural sedimentation, although the characteristic sequence of alternating crests and valleys is still clearly visible after 26 years (Boetius, 2015; Greinert, 2015).

Track V02 is one of the first tracks that have been created during PFEG 2 (Table 1) and could still be detected in different OFOS surveys in 1989 (SO61_OFOS10), 1990 (SO64_OFOS19) and 2015 (SO242_1_OFOS05; Fig. 12a). Figure 12b shows the track only a few hours after it has been created. The characteristic plough structures are very prominent and the freshly broken-up sediment lumps appear brighter than the surrounding sediments. Half a year later, the track appears distinctly smoothed and covered by sediment (Fig. 12c) due to the resettled sediment from plough deployments PFEG 4 to 11. Using the assumed sediment plume distribution of 100–120 m down-current and 20–30 m up-current of the location where OFOS 19 crossed track V02, the observations are within the proximal deposition areas of later PFEGs (Fig. 12a). The high sedimentation visible within this track is in accordance with the predictions of the disturbance map (Fig. 11a). Over the following 25 years until 2015 (Fig. 12d) the track structures continued to be smoothed out, but the differences between 1990 and 2015 are less distinct. This illustrates the immense impact of the evolving sediment plume and the proximal resettling of the sediment compared to the natural sedimentation and current induced shaping of the sea floor in the deep sea.

Track H15 was one of the last tracks, created during PFEG 11 (Table F1 in Appendix F). This track could be captured in OFOS dives from different deployments in the centre of the DEA; H15 was further covered by the AUV photomosaic of SO242 (Fig. 13a). Figure 13b of SO61 (OFOS17 from 1989) shows the freshly ploughed sediment within the disturbance track comparable to Fig. 12b. During SO106 in 1996 the track morphology is smoothed but broken-up sediment lumps are still visible (Fig. 13c). The sediment cover within this track appears less than for track V02. The smoothing continued until 2015 but the track ripple structures are still apparent. At one location captured in the photomosaic the H15 track crosses the V02 track (Fig. 13a), which allows a direct comparison of two tracks from different PFEGs. In 1992 V02 already appears much less distinct (Fig. 13e) than track H15 4 years later (Fig. 13c), again pointing at strong resedimentation initiated by the plough activities after PFEG2. In 2015 the track ripples appear even weaker for V02 (Fig. 13f). This illustrates that the still observable levels of the secondary disturbance through the sediment plume need to be interpreted with respect to their sequential age and respective PFEG deployments. It underlines the importance of careful interpretations of the disturbance state of samples inside and near tracks. Furthermore, the track orientation with regard to the bottom current direction plays an essential role in terms of estimating the ecological impact coming along with the sediment plume; V02 runs parallel to the prevailing current direction, causing higher sedimentation within and very close to the track, whereas H15 runs perpendicular to the north- or south-directed bottom currents, which transported the sediment plume away from the track.