Swath-bathymetry covering the whole continental shelf was acquired in 2004 as part of the ESPACE Project “Characteristics of the systematic study of the Spanish continental shelf and upper slope” by the Spanish Institute of Oceanography, using a Kongsberg EM-3000D multibeam echosounder ( Figure 2 ). The EM-3000D is a system with two sonar heads, each of them with a swath width of 130°. This system uses 254 beams and was operated at a frequency of 300 kHz to assure a narrow beam width (1.5°) and a maximum ping rate of 40 Hz, resulting in a depth accuracy of 5 cm root mean square. Sound velocity profiles (SVP) were collected to correct the acoustic data for water-column sound speed variations. An Octans Inertial Measurement Unit was used for motion corrections and a DGPS provided vessel position during the cruise. A new swath bathymetry was obtained in 2017 by the company GEOMYTSA using a RESON SeaBat 8101, as part of the project “Extension of the marine geophysical survey to a depth of 100 m between the port of Barcelona and Portbou (Girona)” commissioned by the General Subdirectorate for Coastal Protection of the Spanish Ministry for Ecological Transition and the Demographic Challenge. The SeaBat 8101 operates at a frequency of 240 kHz, using 101 beams (beam spacing 1.5°). The surveyed area in this cruise covered the middle continental shelf from 45 m to 115 m water depth ( Figure 2 ). The 2004 survey was used to describe the morphology and backscatter characteristics of the study area whereas the 2017 survey was used to compare both datasets and assess morphological changes.

Post-processing of multibeam data (including correction for heading, depth, pitch, heave and roll) collected during the 2004 survey was conducted using the Teledyne CARIS HIPS and SIPS TM 11.7 Hydrographic Data Processing System. Tidal and sound velocity corrections were applied and the sounding data were cleaned to remove erroneous soundings. Filtered soundings were gridded into 4 m resolution bathymetric surface, providing 100% coverage of the swathed seabed. Backscatter data was processed with QPS Fledermaus software using the Geocoder algorithm for radiometric and geometric corrections, and mosaicking. The original time series registered by the multibeam sonar were corrected for beam pattern and angle varying gains (AVG) using a window size of 300, with a range of normalization between 30° and 60°. The intensity was filtered between 0 and -70 dB. No range filters were applied during processing. Backscatter strength, often measured in decibels (dB), is a measure of the amount of acoustic energy scattered from the seafloor back to the transducer. It depends not only on the environmental and system properties, but also on several seabed characteristics (Collier and Brown, 2005). Main seabed factors include seabed topography (slope), micro-topography, heterogeneity within the near-surface sediments, biotic elements, and sediment characteristics (grain size and sorting or shell content), among others. Therefore, its correlation to grain size is not straightforward (Gaida et al., 2019). In this work, backscatter imagery was used to qualitatively interpret sediment type, based on previous works in the study area (ITGE, 1994; García-García et al., 2012; Lo Iacono et al., 2012; Durán et al., 2014; Durán et al., 2017) and bottom sediment samples. According to this information, in seabed areas with homogeneous relief (no significant variations in slope), fine sediments typically result in weaker backscatter signals (dark tones in the backscatter imagery), while coarser sediments tend to produce stronger backscatter responses (light tones in the backscatter imagery).

Bathymetric and backscatter data were integrated into a geographic information system (QGIS) for the morphological analysis. Bathymetric data from the 2017 survey were processed using HYPACK HYSWEEP^® software and provided as a 5 m bathymetric grid by GEOMYTSA. Final maps were projected into Universal Transverse Mercator projection zone 31 N (UTM 31N) in the European Terrestrial Reference System 1989 (ETRS89).

Side scan sonar (SSS) data were collected in the Roses continental shelf in September 2017 aboard the R/V Sarmiento de Gamboa. The side scan sonar used was the deep-towed large-scale Edge-Tech 2400-DSS/DT-1 system, emitting at a frequency of 120 kHz ( Figure 2 ). Side scan sonar records were visualized and analyzed using the SonarWiz software from Cheasepeake Technology. Trawl marks identified and mapped in the side-scan sonar data using SonarWiz were incorporated into the geographic information system to determine density and direction.

Sediment cores were also collected during the ABIDES-ROV cruise in 2017, using a HAPS bottom corer ( Figure 2 ). Sampling locations were selected based on the spatial distribution of the areas of high backscatter observed in the 2004 bathymetric survey. A total of 18 sediment cores were retrieved in the Roses Bay (12 cores) and off Llançà (6 cores). The upper 10 cm of each sediment core were subsampled at 1 cm intervals. Grain size fractions were determined using a Horiba Partica LA-950V2 particle-size analyzer, with an accuracy of 0.6% and 0.1% precision. Prior to this analysis, 1–4 g of each sample were oxidized using 20% H₂O₂ for a week and then left overnight with a solution of P₂O₇- to disaggregate the sediment particles.

The bottom trawling activity in the Llançà continental shelf during the study period (2005-2017) was obtained using data from the satellite-based tracking Vessel Monitoring Systems (VMS) provided by the Fishing Monitoring Centre of the Spanish Secretariat of Maritime Fishing (SEGEMAR). Since VMS data is broadcasted and stored in 2-hours intervals, vessel tracks were interpolated to 10-minute intervals using the R package VMSbase (Russo et al., 2014). With this higher-resolution data, fishing hauls were identified following the method explained in Paradis et al. (2021). This method first identifies trawling speeds, which for this dataset ranged between 2 and 4.3 knots. Hauls were then defined as consecutive entries that met the trawling speed criteria for at least 60 minutes, taking into account possible false-positives and false-negatives. Fishing intensity was then computed as swept area ratio (SAR) by counting the number of hauls within a 50 m radius, assuming a gear width of bottom trawlers operating in this area being ~100 m wide. SARs were then represented in 25 m x 25 m grid cells to report fishing intensity as hauls per hectare.

The quantification of the morphological impact of bottom trawling on the seafloor morphology was particularly challenging due to the low relief of the morphologies produced by trawling in relation to the general seafloor gradient. In this work, we used different terrain attributes derived from bathymetric data using QGIS. These terrain attributes include: terrain slope using an analysis sale of 3 x 3 pixels’ neighborhood; terrain aspect (3 x 3 pixels) and Bathymetric Position Index (BPI), as modified from the topographic position index defined by Weiss (2001), with an inner radius of 4, and an outer radius of 40. The BPI compares the elevation of a focal point with the mean elevation of its neighboring cells that fall within a user-defined annulus. Positive BPI values represent locations that are higher than the average of their surroundings (positive reliefs), whereas negative BPI values indicate locations that are lower than their surroundings (negative reliefs).

BPI was used to identify low-relief features developed on a slope and trace their boundaries. Since bottom trawling causes erosion of the seabed, the morphologies generated from recurrent trawling can be defined by areas of negative BPI values relative to the surroundings. The joint analysis of bathymetry, backscatter and BPI data allowed us to establish the limits of these erosional forms, defined by BPI values of 0.1-0.2. The boundaries were automatically extracted and manually verified by means of cross profiles spaced 50 m apart for each dataset (2004 and 2017). No significant changes were detected between the two boundaries extracted in each dataset, with the exception of newly identified depressions. Therefore, a single boundary resulting from the intersection of the two limits was used to calculate volume changes. For each bathymetry, the top surface of the depression was created from the elevation of the feature boundary using natural neighbor interpolation. The depth (or incision) and volume of the features were calculated by subtracting the elevation of the feature from the top surface using raster differencing. The volume change between the two surveys was derived from the difference obtained by comparing the volumes calculated independently for each bathymetry.

Satellite-based navigation tracks from bottom trawlers operating in a study area over a 13-year period (2005-2017) provided valuable insights into long-term trends in the location and fishing pressure of the most recurrently visited fishing grounds ( Figures 3A , S2 ). Several highly-impacted fishing grounds were observed on the continental shelf over the study period, particularly off Llançà, off the Cap de Creus Peninsula and in the Roses Bay. Off Llançà, a large (13 km long) fishing ground extends primarily along the 52-62 m depth range from Portbou to Cap de Creus, following the isobaths. Total fishing intensity computed for the period 2005-2017 displays the maximum values in this area (400-750 hauls/ha), particularly to the north of Llançà ( Figure 3A ). Additionally, other fishing grounds can be observed in the middle and outer shelf showing different orientations (N-S, NE-SW and NW-SE) but with lower fishing intensity (40-160 hauls/ha). Off Cap de Creus Peninsula, the most important fishing ground (160-325 hauls/ha) is located off and southward of the Cap de Creus headland, at 75-100 m water depth. It displays a curved path between the coastline and an elongated rocky outcrop that extends along 6 km at 95-115 m water depth. In Roses Bay, trawlers typically follow two main N-S paths that are parallel to the isobaths at 52-56 m and 67-69 m water depth, with fishing intensities of 240-290 hauls/ha ( Figure 3A ).

The backscatter map reveals marked across- and along-shelf variations in reflectivity ( Figure 3B ). The inner shelf shows high backscatter values, particularly to the north and off the Cap de Creus Peninsula, with elongated patches of low backscatter that extend across-shelf from the shoreface down to 25-35 m off small coastal bays. Only in Roses Bay, the inner shelf is characterized by homogeneous medium backscatter. The middle shelf is dominated by low backscatter, particularly to the south of the Cap de Creus Peninsula, where a N–S oriented, 18-km wide low backscatter area extends southwards along the mid shelf, as well as to the north of the Cap de Creus Peninsula. Backscatter is high over the entire continental shelf between the Cap de Creus Peninsula and the Cap de Creus Canyon as well as along the outer shelf. The backscatter imagery also displays large, elongated areas of high backscatter off Llançà and Cap de Creus Peninsula, as well as in Roses Bay, which shows a similar pattern to that observed in the trawling activity ( Figure 3 ).

Side scan sonar data acquired in an area of high backscatter of the Roses continental shelf ( Figure 2 ) shows numerous trawl marks displayed as long, parallel to subparallel features with high reflectivity and different levels of visibility ( Figure 8 ).

The spatial quantification of trawl marks density, length and orientation reveals significant differences between the area of high backscatter and the adjacent shelf. The density of trawl marks (TM) in the area of high backscatter is much higher (320 TM/km²; Figures 8B, C ) than in the surroundings (40-100 TM/km²; Figures 8B, D ). TM in the area of high backscatter are longer (more than 8 km) and mostly show the same curvilinear morphology and N-S orientation than the high backscatter area identified in the backscatter imagery ( Figure 8C ). On the contrary, TM outside this area are more widely scattered and show different orientations, predominantly NE-SE ( Figure 8D ).

In the area of high backscatter, pairs of quasi- parallel tracks separated 40-70 m between them can be distinguished, attributed to the two doors of a given haul ( Figure 8C ). However, in the nearby low backscatter area, some trawl marks are hardly traceable and cannot be clearly distinguishable from the background, having a rather blurry appearance ( Figure 8D ). Detailed observations of the trawl scars show 1 m-wide marks corresponding to the size of the otter doors ( Figure 8E ). Shadows in these trawl marks indicate sediment removal at both sides of the scars produced by lateral push from the doors.

A total of 6 sediment cores were collected off Llançà, inside and outside the area of high backscatter ( Figure 9 ). The sediment cores retrieved in the area of high backscatter (between -29 and -24 dB) corresponding to the trawl-generated erosive feature display sediment coarsening in the upper layers (0-4 cm) (cores L3 and L4, Figure 9C ). The sediment grain size of the upper layers consists mainly of silt and sand (28-62% of silt, 22-67% of sand and 5-13% of clay), coarsening upwards, whereas the lower layers are composed of silt and clay (60-79% of silt and 17-21% of clay), with a minor sand fraction (16-21% of sand). In contrast, the grain size of the sediment cores collected seaward of this area, in an area characterized by lower density of trawl marks and lower backscatter (-38 dB), is more homogeneous (i.e., cores L1 and L2, Figure 9 ). They are mostly composed of silt and clay (55-69% of silt and 9-13% of clay) with a low content of sand (18-36%) and with no differences along the sediment core. Sediment core L5, collected landwards of the area of high backscatter ( Figure 9A ), shows intermediate characteristics between the cores inside and outside the high backscatter region. It is mostly composed of silt and sand (51-70% silt and 16-40% of sand) coarsening upwards ( Figure 9C ). Sediment is finer (62-70% of silt and 16-26% of sand) in the lower layers (7-10 cm) of the sediment core compared to the upper layers (0-4 cm), where the sediment grain size consists of silt and sand (50-51% of silt and 37-40% of sand). Sediment core L6 was collected in the inner shelf, in an elongated area of moderate backscatter (-35 dB) that extends from the shoreline down to 40 m water depth, off the mouth of an ephemeral stream ( Figure 9A ). It displays a different textural pattern dominated by well-sorted fine sand (83-91% of sand and 0.23 mm mean grain size) with a very low content of finer fractions (7-15% of silt and 1.4-2.5% of clay).

A similar pattern is observed in the sediment cores collected along two sampled transects crossing the high reflectivity area in the Roses Bay ( Figures 10 , S4 ). Sediment cores R2 and R8, located in an area of moderate trawl marks density ( Figure 10A ) show slight coarsening trend upwards, more noticeable in the core R8 ( Figure 10C ). The sediment grain size of the upper layers of core R8 consists mainly of silt and sand (63-69% of silt and 18-26% of sand) with a low clay content (11-12%), whereas the lower layers are composed of silt and clay (67-73% of silt and 12-20% of clay), with a minor sand fraction (6-20% of sand). The sediment coarsening in the upper layers is more evident in the sediment cores retrieved in the area of high backscatter, corresponding to the highest density of trawl marks (i.e., cores R4 and R9, Figure 10 ). The sediment grain size of the upper layers (0-5 cm) consists mainly of silt and sand (61-62% of silt and 27-30% of sand) with low clay contents (9-12%), whereas the lower layers are composed of silt and clay (60-77% of silt and 11-22% of clay), with a minor sand fraction (1-28% of sand). Conversely, the sediment cores located in the area of low density (or even absence) of trawl marks (cores R5 and R11, Figure 10 ) are composed of homogeneous fine sediments. Sediment mostly consists of silt and sand (63-72% of silt and 17-29% of sand) with a low clay fraction (9-12%).

The physical impact of bottom trawling on the continental shelf seabed has been widely recognized by the identification and quantification of trawl marks generated by the otter boards (Friedlander et al., 1999; Smith et al., 2007; Buhl-Mortensen et al., 2016; Depestele et al., 2016; Bruns et al., 2020). In the northern Catalan margin, trawl marks appear widely distributed over the whole continental shelf, but concentrated in specific areas, as observed off Llançà and Cap de Creus, and the Roses Bay ( Figures 8 , S3 ). They are better recognized in muddy sediments, but they have also been observed in coarser sediments, even showing a decreased visibility, particularly around the Cap de Creus Peninsula.

Large-scale erosive morphological features have been also identified on the continental shelf as a result of recurrent trawling activities. They are characterized by regions with higher densities of trawl marks relative to the surrounding area ( Figures 3B , S3 ) and exhibit a similar pattern to trawling activities, being localized in areas of increased fishing intensity ( Figure 3 ). These morphologies are displayed on the backscatter data as narrow (less than 300 m wide) and elongated (up to 8 km long) areas of high backscatter relative to the surroundings. The increase of backscatter values in these areas can be mainly attributed to changes in the seabed topography caused by the high density of trawl marks and the reworking of the sediment due to the trawling gear. In the trawled areas, an upward-coarsening trend in the sediment is recorded in the sediment cores, produced by a reduction in the silt content of 14-50% in the first 4-5 cm of the core. Such trend is not observed in the sediment cores of the untrawled area, which are composed of fine sediments and show homogeneous vertical grain size distribution ( Figures 9 , 10 ). The coarsening of the uppermost layers suggests that part of the finer fraction resuspended by trawling gears is winnowed, increasing the sand content of the surface sediment. The fine sediment winnowed could be subsequently deposited on the bottom around the trawl marks, as observed in the Llobregat River prodelta (Palanques et al., 2001), or transported away from the trawled areas by dominant SW along-shelf currents. Slight increases in the grain size of surface sediments and upward coarsening trends caused by recurrent trawling have been observed in other fishing grounds: in the Llobregat River prodelta (Palanques et al., 2001), the Gulf of Lions (Ferré et al., 2008), the Ebro shelf (Palanques et al., 2014) or the Bay of Biscay (Mengual et al., 2016). The coarsening of the surficial sediments in the trawl-generated seafloor features in the Llançà and Roses continental shelf suggests ongoing trawling activities in the fishing grounds, also evidenced by VMS data ( Figure S2 ).

The erosive features observed in the fishing grounds with low trawling intensity (140-160 hauls/ha over the study period), such as the fishing grounds of the middle and outer shelf off Llançà ( Figure 3A ), show no discernible relief in the bathymetric data but they are well displayed on backscatter data ( Figures 3B , S3 ). However, the erosive features located in the fishing grounds with higher trawling intensity (240-750 hauls/ha during the study period) exhibits reliefs ranging from 0.2 to 1.2 m ( Figures 4D, E ). The shallower incision is in accordance with previous works that reported lithological and structural changes on the sediment column induced by recurrent bottom trawling down to 20 cm on the Ebro continental shelf (Palanques et al., 2014) and to roughly 35 cm on the NW Iberian shelf (Oberle et al., 2016b). The areas off Llançà with greatest incision (up to 1.2 m) correspond to areas of highest fishing intensity, up to 750 hauls/ha over the study period. These observations suggest that the repeated passage of the otter boards during successive fishing hauls of the trawling fleet operating in the same fishing ground can result in deeper excavations of the seafloor, compared to the individual tracks caused by the otter boards, inducing larger bottom trawling impacts than previously observed.

The volume of sediment eroded by recurrent trawling in the most intensively trawled areas varies from 0.11-0.18 m³/m² in the fishing ground of the Roses Bay to 0.41 – 0.44 m³/m² in the fishing grounds off Cap de Creus Peninsula and Llançà, respectively, based on the erosion observed in the seafloor features ( Table 1 ). To assess the importance of fishing activities on the sediment dynamics of the continental shelf, these values were compared with the volume of sediment eroded by natural processes in other erosive features, such as lineations, small depressions and obstacle marks identified on the continental shelf between the Cap de Creus peninsula and the submarine canyon ( Figures 1 , S5 ). Lineations have lengths ranging from 0.7 to 2 km, with a negative relief between 0.2 and 1 m. Oval depressions are 120-150 m wide and up to 2 m deep. Obstacle marks are variable in length (100-400 m) and width (50-150 m) with reliefs of up to 2 m. The anthropogenically-derived trawl-generated features and the aforementioned natural features are products of distinct processes occurring in different types of sediments (muddy sediments in the fishing grounds and coarse sand and gravels in the bottom current-generated features). However, the results reveal comparable volumes of eroded sediment per unit area. It is estimated that the volume of eroded sediment ranges between 0.11 and 0.44 m³/m² in the trawling-generated large-scale features, and between 0.10 and 1.22 m³/m² (0.45 m³/m² on average) in the bottom current-generated features. However, the area affected by trawling is significantly larger than the area occupied by individual bottom current-generated features ( Figure S5 ).

Previous studies reported that suspended sediment fluxes induced by trawling erosion in muddy areas are comparable to natural-resuspension events by waves and currents (Durrieu de Madron et al., 2005; Ferré et al., 2008). These trawling-induced processes can become a dominant process controlling water turbidity during periods of high trawling intensity (Palanques et al., 2001; Palanques et al., 2014; Mengual et al., 2016). In the Bay of Biscay, in the French Atlantic continental shelf, trawling contribution to annual resuspension is in the order of 1% of the annual resuspension load, winter storms being the dominant processes. However, during the high fishing season, trawl resuspension can indeed be several times greater than natural resuspension in calm conditions (Mengual et al., 2016). In the Ebro delta, trawling introduces additional resuspended sediment that more than doubles the suspended sediment load of the bottom nepheloid layer on the middle shelf, suggesting that recurrent trawling can be able to change the modern sediment record in continental shelf environments (Palanques et al., 2014). The identification of large erosive features generated by recurrent bottom trawling on the continental shelf shows that repeated trawling over the same fishing ground can also lead to persistent changes on the seafloor morphology comparable to those caused by natural processes.

The characteristics and distribution of the erosive features generated by trawling activities can provide insights into trawling behavior and its potential cumulative impacts on the seafloor. In the northern Catalan continental shelf, the distribution of trawl marks and large-scale erosive features shows a good correspondence with the fishing grounds identified in the trawler’s tracking data during the period 2005-2017 ( Figure 3 ). Off Llançà, most trawl marks and large-scale erosive features are concentrated along the 52-56 m isobaths, indicating that part of the trawling activities are conducted parallel to the isobaths ( Figures 4A , S3 ). Large trawl marks also extend along the whole middle shelf converging towards the narrower shelf between the Cap de Creus Peninsula and the Cap de Creus submarine canyon head, further supporting trawlers’ preferential haul path parallel to the isobaths ( Figure S3 ). Other trawl marks, however, appear crossing the middle shelf showing an orientation independent from water depth, probably related to the fishing practices (e.g. transits between the harbor and deeper areas).

The comparison of the 2004 and 2017 bathymetric datasets shows the bifurcation of two incisions ( Figure 5C ), as a result of the existence of two main fishing grounds located at different depths and presumably exploited in different periods. The NNW-SSE oriented incision off Llançà, mapped in 2004, turns continued northward along the 43-52 m depth range ( Figure 4E ), while the incision mapped in 2017 was located at deeper water depths (52-56 m) and shown a N-S orientation, running parallel to the 50 m isobath ( Figure 5 ). This change in the fishing grounds could be attributed to the implementation of fishing regulations in the mid-1970s (Order of 30 July 1975, BOE-A-1975-17128), when bottom trawling in the Spanish Mediterranean continental shelves at depths less than 50 m was banned, and more precisely to the implementation of the VMS for fisheries monitoring in 2005. Prior to the fishing regulation, trawlers operated as close to shore as possible without risking the fishing gear due to the presence of shallow coastal rocky outcrops. These continued practices over decades explain the formation of a large trawl-generated erosive feature parallel to the coast along the edge of the coastal rocky outcrops, fishing at depths of less than 50 meters. The relief and visibility of this morphology in the backscatter data, coupled with the presence of trawl marks in the 2004 dataset, suggest that despite the implementation of regulations in the mid-1970s, trawlers continued operating in this coastal fishing ground.

The implementation of the VMS in 2005 resulted in the offshore displacement of the fishing activities, which led to a new fishing ground and area of erosion, as observed in the 2017 bathymetric dataset ( Figure 5B ). Although fishing intensity indicates activity in the new fishing ground as early as 2005, it was not until the year 2010 that fishing along this new ground intensified ( Figures 11A , S2 ). The continuous erosion caused by fishing activities conducted over more than a decade produced an erosive feature with a sufficiently significant relief (0.2 m) to be measured with high-resolution multibeam systems ( Figure 11C ).

To the north of Cap de Creus Peninsula, there is also a shift in location of the fishing grounds and the erosive morphologies caused by trawling. The ancient fishing ground that extended along the 52-56 isobaths following the northern limit of the MPA ( Figure 11A ), turned slightly northwards away from the coast in the more recent years ( Figure 11B ). This change in the fishing ground could also be related to the fishing regulations associated with the Cap de Creus Natural Park and its inclusion as MPA within the framework of the Natura 2000 network. Despite trawling was banned in 1998, when the MPA was established, previous works reported occasionally incursions of trawlers inside the MPA (Lloret and Riera, 2008). De Juan et al. (2013) and Demestre et al. (2015) also noted the presence of trawl marks on side scan sonar data collected in 2009, thus evidencing that periodic fishing activities were still occurring in 2009. These observations are supported by VMS data indicating fishing activity in this area up to 2010. In 2011, trawling activities shifted offshore the boundary of the MPA of the Cap de Creus Natural Park ( Figure S2 ). As observed to the north of Llançà, this change in the trawling pattern is evidenced in the bathymetric data by the formation of another emerging incision in the new fishing ground ( Figure 11D ).

Off the Cap de Creus Peninsula, trawl marks show the same orientation as the trawlers track, following a direction parallel to the coast up to the Cap de Creus headland, where they then turn southwards, to continue parallel to the mid-shelf rocky outcrop ( Figures 3 , 6 ). Trawl marks observed in the multibeam echosounder data appear concentrated in the area of higher fishing activities, where the repeated passage of trawlers resulted in the formation of a slightly depressed area that became incised over time, reaching an incision of up to 0.8 m in 2007 ( Figure 6E ).

In the Bay of Roses, seafloor features and vessel tracking data show a similar pattern but covering a wider area than in Llançà and Cap de Creus. Overall, the fishing grounds extend along the bay, running parallel to the isobaths. Only in the central part of the bay, the trawling pattern draws a curve away from the 50 m isobath ( Figure 3A ). Bottom trawlers draw this curve to avoid two obstacles that can be challenging and risk potential gear loss or damage. These include a known shipwreck, the Saint Prosper shipwreck, a 106 m long oil tanker sunk in 1939, and a foul ground of unknown origin ( Figure 7A ). Erosion caused by recurrent trawling in this area (maximum incision of 0.6 m, Figure 7D ) is not as pronounced as that observed off Llançà and Cap de Creus. This is likely associated to the fishing practices, as trawling activities in the Roses shelf are not concentrated in a single location, and instead trawlers spread out over a larger area, reducing the localized impact. Besides, natural sedimentation processes also play a crucial role in the recovery of the impacts by trawling, and the fishing grounds of the Roses Bay are likely directly affected by modern fluvial inputs from the Muga and Fluvià Rivers. Fine sediments delivered by these rivers and deposited in the middle shelf can help to fill any large-scale depressions and trawl marks caused by trawling, reducing their preservation.

The long-term morphological changes observed in the fishing grounds provide information on the persistence of the trawling impact on the seafloor morphology, and its potential recovery after cessation of activities. In the northern Catalan continental shelf, the three studied areas have shown very different morphological evolution over the study period.

The continental shelf off Llançà is the area most affected by trawling, with maximum incisions (up to 1.2 m) and volumes of sediment (> 500 000 m³) removed. In addition, the morphological evolution of this area over the study period shows increased erosion due to continued trawling in active fishing grounds. Volume variations in the shallow fishing ground (located at 43-52 m depth range) was not feasible due to the limited coverage of the 2017 dataset. However, the morphological changes observed in the small area mapped in both years suggest a moderate infilling of the depression ( Figure 5C ). North of Cap de Creus, however, the variations in the eroded sediment in the fishing ground that was abandoned in 2011 still do not show significant changes in 2017, seven years after its abandonment ( Figures 5C , 11D ), evidencing the slow recovery of the fishing ground several years after the cessation of the fishing activity.

The persistence of seafloor features caused by bottom trawling on the continental shelf has been intensively explored through morphological evolution of individual trawl marks. Their preservation is strongly influenced by multiple factors like water depth, sediment type, hydrodynamics (e.g., tidal currents, waves, storms) and trawling patterns. A minimum preservation time of 5–7 days was observed in coastal areas (Depestele et al., 2016; Bruns et al., 2020), whereas they can last for months or even a few years in rather offshore waters (Palanques et al., 2001; Gilkinson et al., 2015; Bruns et al., 2020). The penetration of the trawling gear into the seabed and the durability of the trawl marks was shown to be higher in muddy sediments compared to sandy sediments. Whereas scars caused by gears on sandy uncohesive sediments of energetic areas are covered by ripples in a few hours (BEON, 1990), trawling on muddy sediment can have longer-term effects on the seabed. Palanques et al. (2001) observed that trawl marks in muddy sediments on the Ebro Shelf (20–70 m water depth) did not show any changes after a few days but after a year they appeared with lower backscatter values. They related the relative longevity to the cohesive properties of the muddy sediment (mud content >60%). Gilkinson et al. (2015) also examined the effects of experimental clam dredging on the seabed at greater depths (65–75 m water depth) on the Scotian shelf, Canada, observing trawl marks were still visible in side scan sonar data after three years, although showing a blurry appearance. As observed in the Ebro shelf, storms were the main factor in reworking the sediments and, therefore, in flattening the trawl marks (Palanques et al., 2001; Gilkinson et al., 2015).

The long-lasting effects of trawling in the seafloor morphology in the northern Catalan continental shelf could be mainly explained by the intensity and frequency of trawling activities, sediment inputs and local hydrodynamics. The reduced impact of bottom trawling in Roses Bay compared to Llançà and Cap de Creus could be explained by the local fishing practices and natural riverine sedimentary inputs. As previously mentioned, the annual evolution of fishing activity shows a variable distribution of the vessel tracks in Roses Bay over time, minimizing the impact on a specific area ( Figure S2 ). Conversely, the higher frequency and intensity of fishing activities in Llançà and Cap de Creus fishing grounds hinder the capacity of the seafloor to recover between trawling events. The recovery of trawl-generated features in these areas could be favored by stronger local hydrodynamics. However, despite the strong hydrodynamics of the north Catalan continental shelf, as evidenced by large-scale current-erosive features (Lo Iacono et al., 2012; Durán et al., 2014), the main fishing grounds are located in the more sheltered areas, dominated by weak currents and limited sedimentation (Durán et al., 2014), thus contributing to the persistence of trawling impacts. Only off the Cap de Creus, stronger hydrodynamics could compensate the impact of trawling, favoring its recovery. This would explain that despite being a long fishing ground, the trawling impact is concentrated only in the area of maximum fishing intensity. However, as long as fishing activities continue in these areas, the affected seafloor area will increase and become deeper, resulting in larger and long-lasting effects on the seafloor, with their subsequent impact on benthic habitats.

The penetration depth of otter trawl doors has been associated to adverse impacts on benthic habitats. Bottom trawling impacts have been shown to change benthic species composition when disturbed by gear having a significant larger impact with increasing penetration depth (Hiddink et al., 2017; Sciberras et al., 2018). In terms of recovery, benthic community abundance does not recover within 3 years when impacted by gears with penetration depth of ≥16 cm (Sciberras et al., 2018). The seabed excavation produced by recurrent bottom trawling on the north Catalan continental shelf (up to 1.2 m), deeper than the penetration depth of most benthic trawling equipment, suggests longer-lasting effects in such benthic habitats.