Edinburgh Research Explorer Historic scale and persistence of drill cuttings impacts on North

13 Despite its long history of hydrocarbon exploitation, the United Kingdom lacks 14 scientific protocols to monitor ecological impacts of drill cuttings (mixtures between rocky 15 material excavated from wells and drilling mud). The present study used the UK Benthos 16 industry database to apply standardised variance partitioning and measure the scale and 17 persistence of these effects at 19 sites across the UK sector of the North Sea. Generally, 18 effects were limited to within 1 km from the platform, but two platforms historically drilled 19 with oil-based mud were impacted up to 1.2 km away. Impacts persisted for at least 6-8 20 years in the northern and central North Sea, but were undetectable in the south where 21 cuttings piles do not accumulate. This study underpins new recommendations to 22 implement regional, phase-based approaches to drill cuttings monitoring, and to apply a 23 precautionary approach in considering decommissioning options that will minimise 24 disturbance to cuttings piles.


Operational landscape and environmental impacts of the North Sea oil and gas 31 industry 32
The North Sea is a mature hydrocarbon province in the northeast Atlantic that 33 continues to be explored for new reserves that are being extracted in new ways via 34 enhanced oil recovery methods (Muggeridge et al. 2014). North Sea oil and gas reserves 35 have been exploited for over five decades, with over 770 subsea installations currently in 36 place in the waters around the United Kingdom alone (Fig. 1a). The proliferation of artificial 37 structures in the sea, or "ocean sprawl", has necessitated an evolving but complex policy 38 landscape with the dual purpose to manage licensed activities and foster economic growth 39 Partial variance partitioning was applied to selected datasets to test whether benthos in 163 the near-field (<1 km) and the far-field (>1 km) were affected by contaminant gradients 164 associated with drilling, and how the strength and statistical significance of these gradients 165 changed with time. The results were then compared across structures to derive a regional 166 appraisal of drill cuttings impacts on benthos and recovery. Recommendations are also 167 proposed on methodological changes to guide further improvements in BEP. The data associated with each structure was assessed to ensure it met several criteria for 176 statistical standards: (i) contains both macrobenthos and environmental data; (ii) include 177 near-field (<1km) and far-field (≥ 1km) stations; (iii) includes pre-and post-drilling survey 178 data; (iv) consistently used the same biological sampling gear including grabs and sieve 179 sizes. This strict selection narrowed the number of usable platform or well datasets from 180 237 to 19 (Fig. 1b, Table 1). Each station was categorised as being near-field (<1km M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT away) or far-field (>1km away) from the structure, with distance from platform, well or 182 manifold ranging between 0 to 14.4 km away. 183

Measurements of benthic contaminant gradients 184
All statistical analyses were conducted using the software Canoco v5.02 (ter Braak 185 and Šmilauer 2012). Species data were square-root transformed when abundance data 186 were available, or presence-absence transformed when colonial benthos had been 187 present. Species data were Hellinger-transformed to suppress spurious statistical effects 188 of rare taxa (Legendre and Gallagher 2001). Each dataset was then split into individual 189 survey years (Table 1). 190 In some survey years, the number of environmental variables was greater than the 191 number of stations, which posed a statistical challenge: using all variables for direct 192 gradient analysis would spuriously attribute changes in benthos to these variables when in 193 fact the relationship was an artefact. Thus, a forward-selection procedure was applied that 194 would identify a smaller suite of variables to explain most of the variation in benthos 195 (Supplementary Material Table I used to check the forward-selection procedure, and also to identify any relationships 204 between gradients and species diversity. 205 The strength and statistical significance of distance (near-or far-field) and 206 environmental gradients on benthic communities were analysed using canonical variance M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT partitioning for each survey year. Two methods of partitioning were used, depending on 208 how species responded to environmental gradients: canonical correspondence analysis 209 (CCA, in the case where species responded to gradients in a non-linear way) and 210 redundancy analysis (RDA, when species responded linearly, e.g., the gradient is not very 211 large relative to the species' niche). Partial variance partitioning was used to separate the 212 amount of variation in benthos explained by purely (i) environmental variables, (ii) the 213 distance category (near-or far-field), and (iii) the interaction between (i) and (ii). Changes    Table II). Gradients relating to normal alkanes, organic content, bioavailable barium, 254 silt/clay fraction, sediment grain size and oil content occurred most frequently, with other 255 significant contaminants including lead, zinc, 4-and 6-ringed PAHs, mercury, and total 256 barium ( Fig. 2). Three signs of ecological impacts related to these gradients could be 257 The first sign of ecological impact was a statistically significant post-drilling 259 environmental gradient established in the first post-drilling survey. This sign occurred at all 260 structures except at the Brae B and Murchison platforms (Supplementary Material Table  261 II). 262 A second sign was that this contamination gradient differed from pre-drilling 263 environmental variability by its directionality and/or magnitude. Unlike pre-drilling gradients 264 that were orthogonal to distance or of small magnitude, inspection of the direct gradient 265 ordinations showed that post-drilling contamination gradients were closely aligned to the 266 distance category (Supplementary Material Table III Table III), but a strong significant effect of distance 279 was discerned in the post-drilling statistical analyses and ordinations but not in the pre-280 drilling data (Supplementary Material Table II). 281 A third sign was that post-drilling contamination gradients were always associated 282 with opportunistic and/or pollution-tolerant indicator taxa (Fig. 3). In one instance, at the 283 showed that no taxa actually characterised the contamination gradient, but rather the 285 impacted post-drilling community was distinguished by extremely low species diversity 286 (Table 2). Taxa most frequently associated with post-drilling contaminant and/or organic 287 enrichment gradients included the polychaetes Capitella spp., Ophryothrocha spp., and 288 the bivalve Thyasira spp., all of which have been recorded associated with cuttings piles in 289 the North Sea and elsewhere (Ugland et al. 2008). Notably, this community now 290 constitutes its own habitat on the UK continental shelf. It closely corresponds to what is 291 described by the pan-European habitat classification system EUNIS known to characterise 292 organically enriched offshore circalittoral sandy mud associated with deep offshore sandy 293 mud adjacent to oil or gas platforms and organic enrichment from the cuttings piles, 294 specifically, EUNIS habitat type SS.SMu.OMu.CapThy.Odub. 295

Maximum ecological footprint of discharged drill cuttings 296
Maximum ecological footprint was determined by the distance from the structure of 297 the furthest station where the biota remained in alignment with the post-drilling 298 contaminant axis. Out of the 12 datasets where significant ecological impacts were 299 detected, 10 had a maximum ecological footprint limited to less than 1000 m from the 300 structure (Table 2). In 11 out of the 12 time series of data the extent of the ecological 301 footprint was at a maximum at the time of the first post-drilling survey. 302 Increasing strength of the interactive effect was apparent in some cases, but not all 303 (Supplementary Material Table II). Impacts were occasionally limited to a more restricted 304 footprint, in which case the interactive effect was nearest the structure. For example, the 305 Only two datasets exhibited ecological impacts on benthos beyond 1000 m: Forties E 310 and Murchison platforms (Table 2). In both cases, inspection of the direct gradient 311 ordinations showed that stations lined up along a contamination gradient extending out to 312 1200 m. At Forties E, effects of distance were never strong or significant (Supplementary 313 Material Table II) Table III). After 10 years post-318 drilling, effects were no longer strong or significant, but just three years later in 1993, 319 strong significant effects of normal alkanes and distance appeared to re-establish, and 320 extended out to 1200 m. 321

Recovery periods 322
Recovery time varied across the North Sea, with only 6 out of 19 sites showing no 323 impacts of drill cuttings i.e., minimum persistence time was zero at the 44_12 well, and the 324 Amethyst (A1, B, C), Buchan A, and Audrey A platforms (Fig. 4). When strong significant 325 impacts were detected, more than 50% of these sites showed effects of cuttings piles 326 persisting for at least 6 years post-drilling (Fig. 4), with most slow-recovery sites being 327 located in the northern North Sea. Benthos in the southern North Sea were not altered by 328 drill cuttings except at the Caister platform, where benthos were profoundly affected by a 329 mercury gradient up to 895 m away ( Fig. 4; Supplementary Material Table II). 330   Table II). 349

Discussion 350
Despite major advances in the regulatory framework over the operational lifespan of 351 many platforms, the lack of standardised methods in field sampling, statistical analysis, 352 and confounding factors such as natural variability and industry effects have made it 353 difficult for the UK to assess and monitor the wider regional-scale and persistence of drill 354 cuttings impacts on its seafloor communities. This knowledge gap introduces uncertainty 355 about the potential spatial and temporal extent of cuttings piles disturbance during future 356 decommissioning activities. 357 The empirical variance partitioning approach adopted in the present study is 358 emerging as a technique that allows monitoring results from many offshore extractive 359 industries (oil and gas, renewables, mining) to be extrapolated from a small focal area to

Duration of monitoring: a regional and phase-based approach 373
The results of this study provided the first scientific evidence base for establishing a 374 Gray 1995). However, as more data from WBM drilled sites from the northern and centraol 415 regions are collated into the UK Benthos database, it will be possible for future analyses to 416 evaluate whether a shorter minimum monitoring programme is justified for these regions or 417 whether this should just be the case for sites drilled with WBMs. 418 After 8 years of post-drill monitoring, operators could relax the intensity or frequency 419 of benthos monitoring and take less costly approaches during late-life production phases.

Considerations for decommissioning 427
Findings from the present study supports recommendations made by others for new 428 evidence to inform future policies regarding decommissioning (e.g., Jørgensen 2012). An 429 abrupt re-disturbance event seems to have occurred at the Murchison platform between 430 1990 and 1993, re-establishing a significant contaminant gradient that spread up to 1200 431 m away. Murchison's planned platform decommissioning programme will leave the drill 432 cuttings in situ on the seabed to degrade naturally. The present study demonstrates that at 433 least for the Murchison platform, activities that could significantly disturb cuttings e.g., 434 fishing or cable-laying, must be restricted in the vicinity of the platform for at least another 435 decade. Drill cutting monitoring during the decommissioning phase should therefore revert 436 to a more intensive approach similar to that recommended for the earlier stages of the 437 development. This will allow operators to detect and mitigate activities during this phase, 438 and minimise conflicts with other sectors such as fisheries and telecommunications.

Standards of industry data collection and analysis 440
Only 19 out of 351 installations from the UK Benthos v4.06 database had data 441 standardised in such a way that permitted the present study. Moving the industry beyond 442 basic compliance monitoring to ecologically robust standards would permit analyses to be 443 used more effectively and across greater spatial and temporal scales (Hawkins et al. environmental contractor is used for species identification. It is also recommended that an 453 integrated approach to statistical analyses is used. The present study used a combination 454 of direct gradient analyses at the benthic community level but also supplemented these 455 with ordinations, species diversity, and indicator species to detect impacts and persistence 456 time. 457

Flexible adaptive approaches 458
Activities and usage at wells and platforms change with time. Accidental spills, 459 discharges or disturbance events such as those caused by storms, maintenance, 460 additional drilling, or pipeline work could reset recovery. In these cases, it is recommended 461 that operators revert to a monitoring programme similar to the earlier phase. The present 462 study suggested that the cuttings pile around the Murchison platform was disturbed in the 463 early 1990s, and entirely re-set the recovery trajectory. The implication of this evidence is 464 that there must be a recommendation for drill cutting monitoring to evolve in line with

Conclusion 492
This study provided new trend-based assessments to strengthen the UK's capacity to 493 chart progress towards policy targets. Analysis revealed that the spatial scale and 494 temporal persistence of drill cuttings impacts can be measured using existing industry 495 data. The standardised approach across 19 different sites in the North Sea allowed new 496 recommendations to be made. Based on the findings of this study, it is recommended that 497 industry adopt: regional and phase-based approaches to drill cuttings monitoring; drill 498 cuttings monitoring for decommissioned sites that are as intense and frequent as those 499 occurring in earlier phases; more rigorous benthic sampling design; a flexible adaptive 500 approach to monitoring that can account for changes in management regimes and policies 501 and that can also help disentangle natural variability from man-made impacts. 502

Acknowledgements 503
The analyses reported in this study were supported by funding from Oil & Gas UK for 504 the project "A spatial and temporal assessment of North Sea benthos using UK Benthos v.   Table 1: Selected UK Benthos industry surveys, including the maximum sampled distance away from structure, and accompanying total oil content, aromatic hydrocarbon, trace metal and sedimentological data following categories defined in the UK Benthos v6 database ('X' indicates these data were available) with benthos sampling method (Van Veen or Day grab).