We split the Paleocene to lower Eocene succession in the Froan Basin into six units (Figure 4): three in the Selandian to Thanetian (Units T1-T3) and three in the Ypresian (Units Y1-Y3). The depositional setting and evolution of these units are described and interpreted in the following section using thickness and seismic attribute (RMS) maps (Figure 5) together with well log data (Figure 3, Figure 6). The seismic geomorphology is described from frequency decomposition (FDC) maps in Figure 7.

The Late Cretaceous and earliest Paleocene coastline was located landward of the Froan Basin and seismic data show that the basin had a ramp-like geometry with gradual deepening toward the Bremstein Fault Complex in the west (Figure 2). Initial hinterland uplift in the earliest Danian is documented by input of clastic material to the margin south of the study area (Gjelberg et al., 2001; Sømme et al., 2013). The first phase of sand delivery to the Froan Basin was in the early Thanetian (UT2) (Figure 8), but the shallow-marine system feeding these channels and lobes is not preserved. When the coastline reached the Froan Basin in the late Thanetian (easternmost UT3 clinoforms) (Figure 8), progradation was characterized by an up-stepping shelf-break trajectory, back-stepping landward terminations and fore-stepping seaward terminations (UBF, sensu Aali et al., 2021) (Figure 9), indicating moderately high δA/δS conditions during normal regression. Moderately high δA/δS is also expressed by sigmoidal clinoforms showing rapid basinward thinning and downlap onto the T2 surface (Figure 4). Continued progradation is characterized by a static shelf-break trajectory and fore-stepping landward and basinward terminations (SFF, sensu Aali et al., 2021), indicating gradual decrease in accommodation. This SFF unit represents a transitional phase between normal and forced regression. Backstripping and decompaction of the dip-transect in Figure 9 and the individual clinothems provide an indication of the water depth in front of the advancing system, suggesting ∼350 m near the shelf break and ∼800 m at a reference location 20 km along the dip transect (near well 6407/11-1) during this time (Figure 10).

Progradation at the base of UY1 is dominated by a down-stepping shelf-break trajectory and fore-stepping landward and seaward terminations (DFF, sensu Aali et al., 2021), indicating negative δA/δS conditions and forced regression. The unit is also marked by a basinward increase in thickness of the lower foresets and bottomsets pointing to loss of accommodation at the topset and increasing sediment accumulation on the slope and toe-of-slope region (Figure 9). The lower part of this unit coincides with the PETM (Figure 8). The FDC map in Figure 7C shows that the longest run-out of channel complexes and development of basin-floor fans occurred during the PETM, at the early phase of forced regression. This forced regressive unit is succeeded by a thin package (above T4) showing up-stepping and backstepping terminations (UBB, sensu Aali et al., 2021), high δA/δS conditions, and base-level rise. Backstripping and decompaction suggests water depths of 150–250 m near the shelf break and ∼750 m at the reference location at the PETM (Figure 10).

Subsequent progradation of UY1 above T4 and after the PETM is characterized by continued forced regression (DFF). The clinothem thicknesses rapidly decreased as relative sea-level fell and as the system prograded over older slope and toe-of-slope deposits (Figure 4), indicative of negative accommodation (Posamentier and Morris, 2000). Backstripping and decompaction of clinothems (Y1) suggests that the water depth was less than 50 m at the shelf edge and ∼345 m at the reference location during maximum regression of UY1 in the earliest Ypresian (Figure 10). At the reference location near 6407/11-1, this relative sea-level fall of ∼460 m reflects ∼260 m of deposition post PETM combined with ∼200 m of sea-level fall due to tectonic uplift.

Valley incision (up to 130 m deep) and erosional features on the outer shelf developed as a part of the subaerial unconformity (Y1) and the expression of this surface is shown in Figure 7E. At the time of maximum regression and subaerial exposure, ∼0.5 Myr after the PETM, deposition was largely constrained to toe-of-slope aprons within ∼15 km of the shelf break. The aprons were fed directly by unconfined shelf-edge deltas in the south and incised valleys and erosional features in the north (Figure 9). Valley incision ∼130 m below the shelf break at maximum regression suggests a total relative sea-level fall of more than 300 m between ∼57 and ∼55.5 Ma (between the earliest phase of UT3 progradation and maximum regression of UY1).

Y1 is overlain by another UBB (transgressive) unit indicating high δA/δS conditions and flooding of the shelf (Figure 9). Inside the canyons, the upper Y2 maximum flooding surface is marked by a trough in seismic reflection data and can be traced from the outer shelf to the distal toe-of-slope (Figure 4). This flooding resulted in a landward shift of the shoreline of at least ∼5 km as suggested by the most proximal clinoform above Y2.

Progradation of UY2 is characterized by a stationary shelf-break trajectory and fore-stepping landward and seaward terminations (SFF, sensu Aali et al., 2021), indicating negative to balanced δA/δS conditions and normal to forced regression (Figure 9). In contrast to UY1, UY2 deposits are preserved over a larger part of the shelf indicating higher accommodation during this phase of deposition. The decompacted clinothems are up to ∼85 m thick near the shelf break, showing that significant accommodation existed on the shelf after flooding of UY1. UY2 progradation stopped as the shoreline reached the previous shelf break. Similar to UY1, forced regression resulted in the formation of a subaerial unconformity; however, no erosional networks, valleys or large shelf-indented canyons formed at this time. The lack of shelf erosion and significant valley incision at the end of UY2 deposition is interpreted to reflect higher accommodation conditions compared to UY1. The FDC maps in Figures 7F,G show that the main area of deposition during maximum regression was within ∼15 km of the shelf break, similar to UY1, and there is no evidence of sediment bypass to the more distal Gimsan Basin. Backstripping suggests that the water depth at the reference location was ∼340 m during the end of UY2.

UY2 is locally overlain by another thin UBB package indicating high δA/δS conditions and transgression. The upper Y4 surface represents another maximum flooding surface, which can be mapped in canyons and locally on the shelf (Figure 7F, section A). This transgression shifted the shoreline at least 6 km landward. Above the Y4 surface, UY3 comprises three clinoforms sets (bounded by downlap surfaces Y5 and Y6) showing up-stepping shelf-break trajectories, backstepping landward terminations and fore-stepping seaward terminations (UBF) indicating high δA/δS conditions and normal regression (Figure 7F). UY3 is deeply eroded at the seafloor in the southern part of the study area and the stacking pattern cannot be determined. Decompacted clinothem thicknesses are 30–70 m for the three sets, giving an indication of the water depth at the outer shelf during each regressive phase. Despite higher shelf accommodation during UY3 deposition, each regressive phase was able to extend to the previous shelf-break before transgressing and backstepping. No deep valleys or canyons are observed within UY3 and the outer shelf is only cut by gullies that did not develop into larger canyons (Figure 7I). Backstripping and decompaction suggests that the water depth was 360 m at the reference location at the upper Y7 flooding surface.

Flooding and condensation commenced at ∼54 Ma in the distal part of the system as UY3 topsets started to aggrade on the shelf (Figure 9). This flooding is marked by increasing GR values, more diverse microfossils indicating open marine circulation and the absence of terrestrial derived pollen and organic matter. On the toe-of-slope and on the shelf, final flooding took place around 51 Ma (Figure 8), marked by the last input of sand to the basin (Figure 3). This transgression is expressed by backstepping shorelines and local wave-cut terraces during step-wise transgression, similar to what has been documented in Pleistocene systems (Lebrec et al., 2022).

In summary, the UT1-UY3 succession documents a long-lived (∼8 Myr) phase of relative sea-level fall of at least ∼300 m followed by aggradation, backstepping and abandonment (Figure 9). Within this long-term trend, at least four short-lived (0.5–3 Myrs) phases of regression and transgression occurred, reflecting relative sea-level fluctuation of 30–80 m, suggesting higher-frequency fluctuations in sediment supply and accommodation along the margin.

Several dendritic incisions, valleys, canyons, and gullies formed during development of UY1 and UY2 (Figures 7E,G; Figure 11). A seismic strike-line through UY1 shows that the northern and southern dendritic incisions did not form at the same time (Figure 11, section A). The southern incision formed during early forced regression, but a short base-level rise and a shift in progradation direction caused the network to be abandoned and overlain by shallow-marine deposits. This shift was driven by littoral sediment transport, evident from the observed change in progradation direction along the outer shelf. As the system approached maximum UY1 regression, the northern dendritic incision formed. This development is evident from the different incision depths in the seismic cross section (Figure 11F). Both incisions occurred landward of the shelf break at maximum regression, and both are associated with low-sinuous, up to 150 m wide channels bypassing erosional products to terminal lobes at the toe-of-slope (Figure 11). These observations suggest that incision commenced during formation of the subaerial unconformity. Remnants of other dendritic incisions also occur farther south, pointing to a common development in areas away from main sediment entry points. Similar incisions have also been documented in Quaternary systems along-strike of large sediment entry points (Nordfjord et al., 2006).

Farther south, the FDC map shows a ∼12 km long and up to 700 m wide sinuous valley incising up to 130 m into the underlying shelf and slope deposits. Remnants of fluvial channels on the subaerial unconformity and toe-of-slope channels within UY1 suggest that the area occupied by the sinuous valley also received sediment during forced regression. A seismic dip-line through the valley (Figure 11, section B) shows that the high amplitude reflections at the base of the valley terminate near the area of maximum regression, where a clinoform is observed. This clinoform is interpreted to represent the coastline at maximum forced regression. Basinward of this clinoform, the erosive valley changes into a channel attached to a terminal lobe in its distal end. The valley fill is expressed by a low-amplitude trough in seismic data, suggesting backfilling by fine-grained material during subsequent transgression. Despite being primarily formed towards the end of forced regression, it is possible that headward erosion extended the valley landward during subsequent transgression.

Toward the area of direct sediment input to the south, the UY1 valleys have a more funneled morphology (Figure 11). The central valley is up to ∼80 m deep and ∼1.5 km wide at the shelf break where it terminates. The valley head is located ∼5 km landward of the shelf break. Basinward, the FDC maps show a network of distributary channels (tens of meters wide) where the valley fed sediments to the toe-of-slope aprons. A cross section along the valley (Figure 11, section C) shows a clinoform at the position of maximum regression, indicating that valley incision primarily occurred during the last phase of forced regression. However, the canyon shape may also here have changed during subsequent transgression, resulting in the funneled morphology typical of transgressive estuarine systems (Dalrymple et al., 1992).

The southern valley has no preserved clinoforms at maximum regression but shows similar characteristics as the central valley. Incision is highest landward of the maximum regressive shoreline, reaching as deep as ∼70 m. The valley widens to ∼2.4 km near the shelf break, where it changes into submarine channels feeding sediment to toe-of-slope aprons. Up-dip, the valley head is located ∼5 km from the shelf break.

In the southern part of the shelf, the maximum regressive shoreline is eroded at the seafloor and the shelf break is not preserved. However, the FDC map shows that the area is dominated by nested channels up to 150 m wide instead of valleys, suggesting unconfined sediment supply from shelf-edge deltas to the slope during maximum forced regression.

Seismic reflection data show that the UY1 valley-fills are expressed by dim seismic facies (Figure 7G, Figure 10E), suggesting that subsequent transgression resulted in passive onlap and back-filling of valleys by fine-grained material, similar to many Pleistocene valleys (Blum et al., 2013; Wang et al., 2020).

As regression of UY2 commenced, previous incisions on the subaerial unconformity were filled to the extent that shorelines, slope channels and later incised canyons occupied new positions along the outer shelf. In contrast to UY1 valleys, UY2 canyons did not erode as far back across the shelf (maximum 1.5 km). Because these canyons eroded into the youngest clinoforms deposited at maximum regression, it is clear that UY2 canyon formation must have occurred during transgression. The FDC maps in Figures 7F,G show toe-of-slope channels linking up with the two canyons at the same time as there is no evidence of fluvial channels at the canyon heads. The absence of channels at the canyon heads, together with the northward prograding clinoforms, suggest that longshore transport from littoral drift was the main source of sediment to these northern canyons (Figure 8).

Collectively, the observations from UY1 and UY2 suggest that sediment was supplied by rivers directly to the shelf break along a wide section of the shelf. Incised valleys were not the main conveyers of sediment to the slope during forced regression. The different canyons and incision morphologies observed along the shelf are interpreted to reflect the proximity to the area of high sediment input. In areas of direct sediment delivery, the slope gradient was lower, and the system was graded, preventing deep canyon incision (Figure 12). Toward the north, where shelf-slope progradation mainly was maintained by littoral transport, steeper, out-of-grade slopes developed. These slopes were prone to mass-wasting, gully and canyon incision (Micallef et al., 2014; Prélat et al., 2015).

The North Atlantic margins experienced regional uplift in the Paleocene and early Eocene (e.g., Parker, 1975). Although controversy still exists regarding the driving mechanisms behind the uplift, recent studies have documented several short- and long-lived phases of uplift along the Faroe-Shetland Basin (FSB) and East Shetland Basin (ESB) during the Paleocene and earliest Eocene (Nadin and Kusznir, 1995; Underhill, 2001; Mudge and Jones, 2004; Mackay et al., 2005; Champion et al., 2008; Hartley et al., 2011a; Stucky de Quay et al., 2017; Soutter et al., 2018; Conway-Jones and White, 2022). Comparison of surfaces and events between the FSB, ESB and the Froan Basin (Figure 13) is relevant because the regions are located with variable distance from the outer rifted margin, areas of active volcanism and the center of the Iceland mantle plume. Similarities and differences between the areas may help to understand the nature of allogenic controls and their stratigraphic expressions in different basin settings.

The Selandian to Ypresian shelf-slope wedge in the Froan Basin represents ∼8 Myr of progradation in response to hinterland uplift. The base Paleogene unconformity below the unit documents a ∼30 Myr hiatus in the proximal part of the basin where the entire Upper Cretaceous succession is missing. The hiatus is marked by an angular unconformity, similar to what has been documented in the British Isles (Gale and Lovell, 2018) and East Greenland (Hovikoski et al., 2021). Along the Norwegian margin, this initial phase of uplift resulted in hinterland erosion and increased sediment supply during the Danian (Martinsen et al., 1999; Gjelberg et al., 2001; Sømme et al., 2019).

Although regional uplift started in the Danian, the first arrival of coarse sediment to the Froan Basin occurred in the early Thanetian (59–58 Ma), expressed by remobilized submarine channel complexes and lobes within UT2 (Figure 7A, Figure 9). This phase of initial coarse clastic input was followed by erosion of the Thanetian T2 unconformity (∼57–58 Ma). The T2 unconformity is time equivalent with the Upper Paleocene Unconformity in the Faroe-Shetland Basin (Jolley et al., 2021; Walker et al., 2022), where Jolley et al. (2021) suggested a causal link to a phase of increased volcanism.

Rapid uplift and transition from normal to forced regression occurred just before the PETM. Although this phase coincides with a period of eustatic sea-level fall of ∼40 m (Figure 13), the observed relative sea-level fall of at least ∼300 m in the Froan Basin is an order of magnitude larger, suggesting a tectonic control. Continued uplift culminated in the formation of subaerial unconformities (Y1 and Y3) in the Froan Basin between 55.5–55 Ma. These are coeval with the “R” and the “Bressay” landscapes observed along the ESB (Stucky de Quay et al., 2017; Conway-Jones and White, 2022), and the Flett unconformity in the FSB (Hartley et al., 2011b; Jolley et al., 2021; Walker et al., 2022), suggesting a common mechanism for their formation across the North Atlantic region.

Another similarity between the Froan Basin and FSB is the stratigraphic position of the PETM within the prograding shelf-slope wedge. Jolley et al. (2021) demonstrated that the PETM predated subaerial exposure and the formation of the Flett unconformity in the FSB in the same way as the PETM predates the Y1 subaerial unconformity in the Froan Basin. Since the PETM is considered a regional datum (Denison, 2021), this comparison shows that there was no delayed transient uplift in the Froan Basin compared to the FSB. Within the resolution of biostratigraphic data, it appears that maximum uplift and subaerial exposure was broadly time equivalent across the region.

The Y2 flooding event (between UY1 and UY2) formed between 55.5–55 Ma at the time of peak transient uplift in the Froan Basin. A similarly aged flooding above the Forties sandstone member is also present in the Moray Firth area (Thomas and Hartley, 2014). If dynamic topography resulted in vertical motion of several hundred meters between 55.5 and 55 Ma, how can an ∼85 m relative sea-level rise between maximum regression of UY1 and UY2 be explained? One possibility is a combination of eustasy, compaction, and tectonic- and load-induced subsidence. Isostatic considerations suggest that the total load-induced subsidence of the UT3-UY1 wedge amounts to ∼200 m (which equals to an average rate of 0.1 mm/yr for the ∼2 Myr unit). Such a subsidence rate superimposed on a ∼40 m eustatic sea level rise between 56 and 55 Ma (Figure 13) might be sufficient to explain the rapid transgression and flooding observed between the two units during long-lived regional uplift. Increasing load-induced subsidence in combination with post Cretaceous tectonic subsidence could also explain the gradual shift from overall forced regression during UY1 towards higher accommodation during UY2, to aggradation and final backstepping at the end of UY3.

Flooding of UY1 and higher accommodation during deposition of UY2 demonstrates that regional subsidence commenced between 55.5 and 54.5 Ma. This subsidence coincides with transgression and backstepping of Ypresian shorelines in the FSB, which commenced at 54.8 Ma (Hardman et al., 2018), suggesting that increasing accommodation likely was related to regional subsidence.

After deposition of UY2, accommodation increased significantly on the shelf, preventing large volumes of sediment reaching the toe-of-slope. The data show three phases of regression within the Dracodinium simile zone (between 54.5 Ma and prior to final flooding at 51 Ma), suggesting that UY3 is largely time equivalent with the Halibut Delta in the Moray Firth area (Zimmer et al., 2019). The eustatic curve shows only one sea-level fall during this period (between 53–52 Ma) but Conway-Jones and White (2022) also recognize two phases of uplift in the ESB (“S” and “T” landscapes), suggesting a possible common control across this area.

Mudge et al. (2014) showed that the Ypresian transgression occurred at ∼54 Ma along the East Shetland Basin, which coincides with abandonment and condensation in the distal part of the system in the Gimsan and Froan Basin (Figure 13). In the more proximal area represented by well 66407/12-2, however, sedimentation continued until ∼51 Ma, when the last sand was deposited, and the shoreline finally stepped back (onlap of Y7) (Figure 3). Final flooding in the proximal part of the Froan basin at ∼51 Ma correlates with transgression of the Halibut Delta, which drowned around 50.4 Ma (Zimmer et al., 2019), suggesting that flooding of shelf-slope wedges occurred concurrently across larger areas of the North Atlantic.

Comparison between surfaces and events along the North Atlantic margin suggests a common forcing for several of the stratigraphic events. Yet, the areas also show significant discrepancies in the age of unconformities, stratigraphic stacking patterns and phases of sand delivery to the shelf margins. Lovell (2010) and Conway-Jones and White (2022) suggested that repeated uplift and subsidence along the ESB reflect fluctuation in mantle plume activity causing regressions and transgressions on the scale of hundreds of meters. Although several transgressive-regressive phases of similar age are present in the Froan Basin, the presence of local flooding surfaces indicates that the preservation of tectonic signals like the 2–3 Myrs fluctuations suggested here may be modulated by local controls. These discrepancies could reflect variable distance to the area of dynamic uplift or basin configuration (Friedrich et al., 2018), but could also reflect sampling and data bias, or local controls such as variations in subsidence and sediment supply. Thus, even if the margins were exposed to the same tectonic forcing, a non-unique stratigraphic response may have resulted across the region (Burgess and Prince, 2015).

The short-lived PETM climate event lasting 100–200 kyr is known to have been associated with significant global warming (Sluijs et al., 2007), changes in the hydrological cycle and increasing sediment supply to North Atlantic basins (e.g., Dypvik et al., 2011; Kender et al., 2012; Pogge von Strandmann et al., 2021; Rush et al., 2021; Stokke et al., 2021; Jin et al., 2022).

In the Froan Basin the PETM is constrained by the presence of A. augustum in a ∼54 m thick succession in 6407/12–2 (Figure 3), a ∼20 m thick succession in 6407/11-1 (Figure 6), decreasing to 10–20 m in more distal wells (Figure 6). The lithology of the PETM succession in the Froan Basin is typically non-calcareous, grey, green to brown claystone, locally grading to siltstone with streaks of sandstone (Supplementary Table S1). In log data, the PETM is expressed by an abrupt increase in GR values followed by a gradual decline (Figure 6), similar to what has been described elsewhere in the North Sea (Vieira and Jolley, 2020).

Figure 14 shows that the ∼20 m thick PETM section in 6407/11–1 correlates with the two low-sinuous channel complexes at the lower part of UY1, NW and SE of the well location. The FDC map demonstrates that these channel complexes are connected to the basin-floor fan in the Gimsan Basin and that they can be traced landward across the toe-of-slope aprons towards the shelf break where they terminate. The map thus documents the shallow to deep-marine depositional expression of the PETM in the Froan Basin.

Although the PETM often is linked to a period of eustatic sea-level rise (Sluijs et al., 2008; Harding et al., 2011), the climate event coincides with a period of forced regression in the Froan Basin and the only indication of base-level rise is at the top of the shallow-marine PETM succession (Figure 7C, section B). One possible explanation for this apparent discrepancy from other global PETM locations is that the amplitude of eustatic sea-level rise was too low to be expressed by onlap in seismic data (i.e., below seismic resolution). Another, partly complementary explanation, is that the rate of dynamic uplift balanced the rate of eustatic sea-level rise during the main phase of the PETM. We prefer a model where rapid dynamic uplift combined with high sediment supply outpaced eustatic sea-level rise during the main phase of the PETM, resulting in net forced regression. However, it is possible that the indication of base-level rise observed in Figure 7C, section B marks the waning phase of the PETM when the sediment supply rate decreased. Interestingly, a maximum flooding surface has also been interpreted above a sand-rich PETM succession in the deep-marine part of the North Sea (Vieira and Jolley, 2020; Jolley et al., 2022).

The seismic data show that the shoreline was at the shelf break and that the fluvial system delivered sediment directly to the upper slope during the PETM. The sediments delivered to the shelf-edge delta were primarily transported down-slope in up to 400 m wide and erosive channel complexes (Figure 7D, section A). When reaching the toe-of-slope, the erosive PETM channel complexes started to deposit sediment locally, but parts of the sediment load bypassed the toe-of-slope and continued toward the Gimsan Basin in two depositional channel complexes. The development of such wide and long channel complexes is in stark contrast to the pre- and post-PETM situation when the slope was dominated by narrow (up to 100 m wide) channels terminating near the toe-of-slope.

Cross-sections through these channel complexes show that they change morphology basinward. As they exit the toe-of-slope aprons (Figure 14, section A) they are typically 0.7 km wide, up to 80 m thick and show only minor erosion at the base. They have no pronounced levees and show considerable differential compaction, suggesting that they are sand-prone (Jennette et al., 2000; Bryn and Ackers, 2014; Edwards et al., 2018). Internal reflectivity and lack of deep incision at the bases show that they are overall aggradational. Farther basinward (Figure 14, section B), the channel complexes are wider (up to 1 km) and thinner (up to ∼50 m). They show less internal geometries and less differential compaction compared to the more proximal location, indicating that they are less sand prone. Toward the Bremsein Fault Complex (Figure 14, section C), the channel complexes are up to ∼1.8 km wide, ∼30 m thick and have no internal seismic character. They show subtle differential compaction, but significantly less than in the more landward parts, indicating lower sand content. The basin gradient decreases basinward of the Bremstein Fault Complex and this transition marks the channel-lobe transition zone where the channel complexes become distributary as they deliver sediment to the basin-floor fan in the Gimsan Basin. The basin floor fan is expressed by a single seismic reflection and has no measurable thickness and show no indications of differential compaction. Wells penetrating the distal part of the basin-floor fan (e.g., 6407/8-1, Figure 6) show that the PETM succession typically is <20 m thick and dominated by claystone and minor siltstone (Supplementary Table S1).

Based on these observations it seems that most of the coarse material supplied to the deep-water during the PETM was deposited in the toe-of-slope aprons and in the proximal part of channel complexes landward of the Bremstein Fault Complex. Only the fine-grained fraction was transported >50 km into the Gimsan Basin. However, there was not sufficient time and/or volume of sediment to develop thick levees along the channel complexes, pointing to an abrupt and short-lived pulse of sediment delivery to the distal basin during the PETM.

A transition from narrow (<100 m wide) channels terminating at the toe-of-slope before the PETM, to wide and erosive slope channel complexes feeding sediment more than 50 km into the basin during the PETM support a significant change in flow properties during the climate excursion. With time, deep-water channels tend to develop towards a state of equilibrium characterized by sinuous channel morphology and deceasing width and depth down-stream (Kneller, 2003; Pirmez and Imran, 2003; Shumaker et al., 2018). Wider channel complexes during the PETM suggest that gravity flows were larger and more erosive (Konsoer et al., 2013; Ferguson et al., 2020), and it is likely that they were denser and had more mud entrained in the flow in order to travel longer distances (Baas et al., 2004; Traer et al., 2018; Pope et al., 2022). A fundamental shift in depositional style to larger channels and lobes, combined with increased transport distance of clastic material observed in the Froan and Gimsan basins, is thus in line with recent work from other locations in the North Atlantic where the PETM was associated with higher onshore weathering rates, increased erosion and delivery of sand and clay to the marine realm (Pogge von Strandmann et al., 2021; Rush et al., 2021; Jin et al., 2022). As the fluvial system delivered sediment directly to the shelf edge during the PETM, more seasonal variations in discharge (Eldrett et al., 2014; Jolley et al., 2022) could have resulted in rapid deposition and slope failure on the upper slope together with larger and more frequent hyperpycnal flows (Petter and Steel, 2006; Henriksen et al., 2011), triggering the inferred change in gravity-flow size and composition inferred during the PETM. However, in contrast to Jin et al. (2022) who demonstrated increased sand delivery to the deep-water North Sea basin, the Froan example documents increased mud-delivery to the distal deep-water Gimsan basin during the PETM. A similar transition toward more mud-rich deposits during the PETM has also been described from deep-water settings in the Pyrenees (Pujalte et al., 2015) and Gulf of Mexico (Sharman et al., 2022).

Jin et al. (2022) also documented an order of magnitude increase in sedimentation rates during the PETM in the North Sea from well data. When calculating sedimentation rates for wells 6407/12-2 and 6407/11-1 (Supplementary Table S2), both wells similarly show an increase in sedimentation rate from 8–75 m/Myr before the PETM, to 270–350 m/Myr during the PETM, and back to <140 m/Myr after the event. This variability supports the idea that the climate excursion was associated with an order-of-magnitude increase in sedimentation rates when measured in one-dimensional well data. However, when comparing volumetrics from seismic units it becomes clear that the change in sediment supply across the PETM was less dramatic (Supplementary Table S3). When we use the PETM thickness maps in Figure 5 and assume a PETM thickness of 10–20 m in the distal basin (where the interval is below seismic resolution), and a duration of 100–200 kyr, we can calculate high and low sediment supply scenarios before, during and after the PETM (Supplementary Table S3). The results show that the transition from UT3 to the PETM was associated with up to an order of magnitude increase in sediment supply, from 46 km³/Myr during deposition of UT3 to 185–470 km³/Myr during the PETM. After the PETM, sediment supply decreased to 111 and 153 km³/Myr during UY1 and UY2, respectively. There are significant uncertainties associated with these numbers since the units have internal unconformities and as parts of the sediment budget may not be accounted for, but the overall trend is that although the transition into the PETM was associated with up to an order of magnitude increase in sediment supply, the sediment supply immediately after the hyperthermal was only deduced by 50%–75%. One possibility is that the order-of-magnitude increase in sedimentation rate observed in well data might be biased towards local areas of active deposition and/or a result from averaging over short time intervals (Schumer and Jerolmack, 2009). Another key factor is that the PETM in the Froan Basin occurred during forced regression, when shelf accommodation was lower compared to the preceding UT3 unit. This transition from normal to forced regression shifted the depocenter basinward and the full volumetric extent of the unit is thus better preserved in the rock record. We suspect that the real increase in sediment delivery to the Froan Basin during the PETM was more moderate (2–4 times the pre- and post PETM supply), and that the order-of-magnitude values identified in individual wells may not be indicative of changes in sediment supply for an entire source-to-sink system during such rapid climate events.

From the discussion above it is clear the PETM response in well data is highly dependent on the sample location along the sediment routing system and the local interplay between accommodation and sediment supply. In the North Atlantic, PETM eustatic sea-level rise appears to have been balanced by rapid dynamic uplift in many locations, resulting in overall regression and deposition of a sand-rich unit in areas that were already dominated by basin-floor fan deposition (Vieira and Jolley, 2020; Jin et al., 2022). The expression of the PETM climate signal is further dependent on the sampling location along the routing system. The sand-dominated PETM succession sampled by Jin et al. (2022) was in a medial to distal position of an older sand-rich basin floor fan system and it is possible that a well in a more distal position will capture a more mud-rich PETM signal as observed in well 6407/8-1 in the Froan Basin (Figure 14). Similarly, a more proximal sampling location along the channel complexes and toe-of-slope deposits in the Froan Basin would likely have recorded a thicker and more sand-rich PETM succession. The proximal, but off-axis position of well 6407/12-2 captures a relatively thick, silty, and mud-rich PETM succession. The integration of seismic geomorphology is therefore key to understand the system-wide context of one-dimensional well data and the response of an entire source-to-sink system to rapid climate change.

Sediment partitioning, stratigraphic architecture and the spatial-temporal link between shallow and deep-water deposits reflect the interplay between external controls on accommodation and sediment supply, autogenic processes, and local environmental processes (Gawthorpe et al., 1994; Madof et al., 2016; Straub, 2019; Paumard et al., 2020). In supply-dominated settings, sediment dispersal to the basin floor can occur throughout the sea-level cycle (e.g., Carvajal and Steel, 2006), especially in greenhouse systems where shelves may have been relatively narrow compared to icehouse systems (Burgess et al., 2022).

Our data analysis from the Froan Basin demonstrate that long-term transient uplift and subsidence drove the large-scale partitioning of sediments along the margin, whereas short-lived eustatic fluctuations, isostasy and environmental processes controlled spatial changes in sedimentation. During the initial phase of normal regression (Figure 4), sediments were largely trapped locally in topsets and proximal foresets (Figure 15), expressed by high proximal sedimentation rates (up to 350 m/Myr) and rapid basinward thinning of foresets during UT3. Sediment partitioning changed at the onset of forced regression when the shelf-break trajectory fell and when there was no accommodation landward of the advancing shoreline (UY1). Lack of accommodation resulted in thick foresets and bottomsets, aggradation and progradation of toe-of-slope aprons across a larger area and with high sedimentation rates (up to 200 m/Myr). The observed shift in sediment partitioning is similar to previous studies (Porębski and Steel, 2003; Johannessen and Steel, 2005), showing that basin floor fans (or toe-of-slope aprons) can develop in front of supply-dominated shelves even with rising shelf-break trajectories (Helland-Hansen and Martinsen, 1995), but that those fans tend to be thinner and have shorter run-out lengths compared to the ones developed when the trajectory is flat or falling (Carvajal and Steel, 2006; Koo et al., 2016). As accommodation increased during UY2 and UY3, sediments were trapped on the shelf. UY3 covers around 3–4 Myrs of deposition and the resulting sedimentation rates on the shelf are thus low (up to 50 m/Myr), suggesting that sediment accumulated landward of the shelf, possibly combined by an overall lower sediment supply.

Within such a framework of tectonically controlled sediment partitioning during normal regression, forced regression and aggradation, local autogenic processes also controlled sediment distribution along the margin. Figure 7I, section A shows that each successive phase of clinothem progradation stopped near the shelf break established during maximum regression of UY1. The concept of autoretreat suggests that increasing space in front of the advancing delta eventually will cause backstepping (Muto and Steel, 2002), but the seismic geomorphology also shows that the process regime changed as the delta of UY2 approached the shelf break. Northward sediment transport by littoral currents removed parts of the sediment from the advancing shoreline, preventing further progradation. The littoral drift became stronger as the delta approached the shelf break (Dixon et al., 2012), evident from the change in clinothem progradation direction near the shelf break (Figures 7E,F). Although the point of autoretreat can be determined by a suite of intrinsic variables (Burgess et al., 2008), we suggest the process of sediment removal by longshore drift may have been just as important as increasing accommodation in partitioning sediment across and along the shelf during each phase of clinothem progradation.

Figure 12 shows how the slope gradient changes along strike at the top of UY2 from 1 to 3 degrees in the area of direct fluvial supply in the south to as much as 7 degrees northward where the margin was wave-dominated and where sediment was supplied by longshore drift. Here the slope and toe-of-slope is dominated by mass wasting and local development of small lobes in front of canyons fed by longshore drift (Figure 7D). The toe-of-slope in the northernmost part of the study area shows few signs of gravity deposits (Figure 7G), indicating that most of the sediment transported by longshore drift was captured by canyons and gullies before reaching the northernmost part of the shelf. As documented in several other areas, the change in process regime along strike shows that the fluvial-dominated shelf-edge environment in Froan was more efficient in bypassing sediment to the toe-of-slope than the wave dominated part of the systems (Dixon et al., 2012; Cosgrove et al., 2018; Paumard et al., 2020). This along-strike change in gradient and mechanism of sediment delivery from the shelf to the deep-water is similar to the two modes of shelf margin development described in Fisher et al. (2021): out-of-grade margins are typically associated with sediment delivery through shelf-incised canyons resulting in slope onlap and basin floor aggradation, whereas direct-fed margins often show more unconfined sediment delivery to the slope and basin floor causing progradation along the entire shelf-slope system (Hadler-Jacobsen et al., 2005). In the Froan Basin these two modes of system development occurred along strike within the same system. This along-strike variability was facilitated by a change in process regime in which longshore transport drove shelf progradation northward, resulting in steepening of the slope and initiation of canyon and gully incision capturing parts of the littoral drift and supplying it down-slope. In the case of the Froan Basin, along-strike variation in slope gradient is thus an intrinsic control and not related to large-scale basin or margin configuration.

Submarine canyons fed by littoral drift in modern systems show that they tend to be most active during sea-level rise or highstand, when vigorous longshore currents develop, and that they often are disconnected from sediment supply during lowstands (Covault et al., 2007; Bernhardt and Schwanghart, 2021). Supply to canyon heads during transgression and highstand conditions may also have occurred in the Froan Basin, but in addition, the change in clinoform progradation during deposition of UY1 and UY2 shows that longshore transport must have been prominent also during forced regression. Such an interpretation is backed by a simple mass balance comparison showing that the volume of sediment stored in the lobe in front of the northern incision in Figure 11 is ∼3.8 km³, whereas the volume missing within the incision is only 0.008 km³ (Figure 11F). This calculation shows that the dendritic incision itself cannot have supplied the material to the lobe and that longshore drift must have been the main source of sediment for the toe-of-slope lobe deposits.

The preferred model is one where sediment supplied from longshore currents was trapped and conveyed down slope during forced regression when dendritic incisions and incipient valleys formed along the shelf (Figure 11). Since canyons did not extend far across the shelf, the amount of sediment delivered by longshore drift during transgression was likely relatively low. The key mechanism for highstand feeding of canyons in modern systems is the presence of deep, shelf-indented canyons that can bypass sediment from the highstand shoreline. In the Froan Basin UY1 canyons were largely backfilled during transgression and UY2 canyons did not extend far across the shelf, suggesting that these were not important sediment conveyers during highstands.