UHI Research Database pdf download summary Significance of climate indices to benthic conditions across the northern North Atlantic and adjacent shelf seas

13 The northern North Atlantic Ocean and its adjacent shelf seas, are influenced by several 14 large-scale physical processes which can be described by various climate indices. Although 15 the signal of these indices on the upper ocean has been investigated, the potential effects on 16 vulnerable benthic ecosystems remains unknown. In this study, we examine the relationship 17 between pertinent climate indices and bottom conditions across the northern North Atlantic 18 region for the first time. Changes are assessed using a composite approach over a 50 year 19 period. We use an objectively-analysed observational dataset to investigate changes in bottom 20 salinity and potential temperature, and output from a high-resolution ocean model to examine 21 changes in bottom kinetic energy. Statistically significant, and spatially coherent, changes in 22 bottom potential temperature and salinity are seen for the North Atlantic Oscillation (NAO), 23 Atlantic Meridional Overturning Circulation (AMOC), Atlantic Multi-decadal Oscillation 24 (AMO) and Subpolar Gyre (SPG); with statistically significant changes in bottom kinetic 25 energy seen in the subpolar boundary currents for the NAO and AMOC. As the climate 26 indices have multi-annual timescales, changes in bottom conditions may persist for several 27 years exposing sessile benthic ecosystems to sustained changes. Variations in baseline 28 conditions will also alter the likelihood of extreme events such as marine heatwaves, and will 29 modify any longer-term trends. A thorough understanding of natural variability and its effect 30 on benthic conditions is thus essential for the evaluation of future scenarios and management 31

salinity and potential temperature, and output from a high-resolution ocean model (Viking20) 45 to examine changes in bottom kinetic energy. We provide a first look at the emergent 46 patterns, and examine and describe their spatial coherency and magnitudes. We make some 47 tentative explanations for the physical basis of some of the significant signals, but are careful 48 not to ascribe causality where we investigate only correlation. The northern North Atlantic region, which we here define as that north of 30° N including its 52 adjoining continental shelves and seas (Figure 1), has several multi-annual large-scale 53 physical processes that influence upper ocean climate and have the potential to effect deep-54 sea ecosystems. These processes are often described using a number of basin-scale climate 55 'indices'; the most common being the: North Atlantic Oscillation (NAO), the strength of the a high NAO (Alekseev et al., 2001;Dickson et al., 1996). Finally, Iceland Scotland Overflow 88 Water has a lower salinity during a high NAO (Sarafanov, 2009) with less vigorous flow 89 (Boessenkool et al., 2007). No relationship between Denmark Strait Overflow Water 90 salinities and the NAO is observed (Sarafanov, 2009 93 The AMOC is a measure of the strength of the overturning circulation in the important North combines data from all types of profiling instruments, including ship-based measurements 139 and ARGO floats, and is available as monthly averages (Good et al., 2013). EN4 has a 1° 140 horizontal resolution meaning that smaller spatial features will not be resolved, and that large 141 property gradients, such as around the boundary of basins, will become smoothed. Although 142 EN4 starts in 1900, we limit our analysis to 1959 onwards (to match with Viking20) but 143 extend the analysis until December 2017 to maximise available data. variable, which ranges from approximately zero to one. A value of zero indicates the absence 155 of any observations for that data point, during that month, and that climatology is used. 156 Conversely, a weighting of approximately one indicates a high influence of observations. Viking20 has been used in many studies, including those to investigate: the North Atlantic  the same limitations apply. For example, real-world smaller-scale frictional effects, such as 180 benthic boundary layers, will not be resolved in the velocity field. As our goal was to study 181 kinetic energy, rather than velocities, ubot and vbot were first linearly interpolated from their 182 respective grids onto the grid containing θbot and Sbot. Bottom mean kinetic energy was then 183 calculated as the sum of the squares of mean ubot and vbot, with the eddy kinetic energy 184 defined as the sum of the variances of ubot and vbot. As our data are five-day averages, 185 variances represent energy of sub-inertial flows.  (Table 1), there is no site that truly represents 195 the abyssal plains of the North Atlantic south of 50 °N. This reflects the current absence of 196 VME indicator records in these areas (Morato et al., 2018). Time-series were constructed for 197 each case study site. In EN4, data within each case study polygon were simply averaged for 198 each monthly time-step. As case studies 4 and 10 did not contain any EN4 grid points; time-199 series for these sites were instead extracted from the nearest EN4 grid point. In Viking20, the 200 model grid is curvilinear. Thus, grid points were accordingly area weighted when calculating 201 averages for each five-day time-step. 203 We consider the four most commonly used climate indices for the North Atlantic Ocean: the 204 NAO, AMOC, AMO and SPG. The NAO index used is defined as the normalised pressure 205 difference between Gibraltar and southwest Iceland (Jones et al., 1997). Data were 206 downloaded from the Climatic Research Unit (https://crudata.uea.ac.uk/cru/data/nao/) and the 207 winter (DJFM) mean calculated. As Viking20 is forced with CORE2 atmospheric data, which 208 will include a signature of the NAO, we use the observational NAO time-series to investigate 209 ocean correlations in both the EN4 and Viking20 datasets. observational AMOC may not be contemporaneous, we compute two AMOC time-series: one 215 from EN4 and one from Viking20. For the observational dataset we used the method of 216 Mercier et al. (2015), but using EN4 data along the OSNAP-EAST section (black line, Figure   217 1.a). Geostrophic velocities perpendicular to the section were calculated from EN4 218 temperature and salinity data and referenced to satellite altimetry data. The Viking20 AMOC 219 time-series was calculated using model velocities perpendicular to the OSNAP-EAST 220 section.

222
We use an AMO index downloaded from the National Oceanic and Atmospheric 223 Administration (https://www.esrl.noaa.gov/psd/data/time series/AMO/) as monthly averages.

224
This time-series consists of sea surface temperatures, averaged over 0-70 °N before de-225 trending using a 10 year running mean (Enfield et al., 2001). The AMO is also an oceanic 226 index, suggesting that again a model and observational time-series is required. Although the 227 construction of an AMO index from Viking20 was considered, it was discounted for two 228 reasons. Firstly, the AMO index is calculated using sea surface temperatures over the entire  those exceeding one standard deviation above the mean, and "low years" as those less than 253 one standard deviation below the mean. Composites for the high and low climate states were 254 calculated by averaging properties from all high and low years respectively. One of the 255 limitations of the composite approach is that it only considers high and low states, and not 256 transitional processes between the two. As the Viking20 AMOC time-series shows a long-257 term trend and we are interested in multi-annual changes, we de-trended this index before 258 creating the composites. The time-series was de-trended by assuming a linear long-term trend 259 which may not be entirely appropriate for low-frequency oscillations. However, the record is 260 too short to establish any low-frequency variability more accurately, and de-trending enables 261 us to investigate multi-annual changes over a 50 year period. The number of months averaged 262 to create each composite are shown in Table 2.  2000 m (± <0.01). However, there is also spatial variability within these general descriptions.

308
For example, variability is higher in the southern North Sea compared to the northern North 309 Sea and shelf areas west of the UK and Ireland.  as an additional control we carried out statistical testing (as detailed in section 3.5). Only 363 changes between the high and low years that are statistically significant at the 95 % 364 confidence level are discussed. As pure climatology would have no significant correlations, 365 we can be sure that any statistically significant patterns are due to the presence of data.  As the timing of changes in strength between the observational and modelled AMOC 424 compare extremely well during the contemporaneous period (Figure 2.b, f), we now use the 425 de-trended Viking20 AMOC time-series to interrogate the EN4 dataset (Figure 7.c-d).

426
Applying the Viking20 AMOC time-series to the EN4 data assumes that the relationship 427 between the observed and modelled AMOC persists outside the post-1993 era. However, the 428 advantage is that it increases the amount of observational data used in the construction of the 429 composites (Table 2) therefore increasing the spatial extent of the analysis.

452
As discussed in section 3.4, it is not appropriate to apply the AMO time-series to output from 453 Viking20; thus we examine changes using the EN4 dataset only and do not discuss KEbot.    Shelf are dominated by the SPG, whereas the shelf west of the UK is dominated by the AMO.

522
In the North Sea the AMO also dominates changes in θbot, whilst the AMOC is more 523 important for changes in Sbot.
524 525 6. Results: variability at ecosystem case study sites 526 We now move on to discussing changes at fourteen case studies ( Figure 1, location. However, we caution that most case studies cover a range of depths (Table 1), 531 which may be subject to different processes and signals, as well as lag periods. As such, 532 changes at a particular depth may be different to the mean conditions for the entire case study North Sea (Figure 1). Here, the AMOC is associated with the largest changes in Sbot in 540 southern areas, whilst the largest changes in the northern part of the case study region is 541 associated with the AMO (Figure 10). However, when the Sbot changes are averaged over the 542 entire case study region, only the AMO is associated with statistically significant changes.

543
Finally, as changes are often larger at shallower depths, case study averages are likely to be 544 biased towards processes occurring higher in the water column. As such, we advise the reader 545 to consider the results in this section in conjunction with Section 5 and Figures 5-10.

547
We again use the EN4 dataset to examine changes in Sbot and θbot, and output from Viking20   x 10 -2 m 2 s -2 ).

594
Case study 7 is located in the Gulf of Cadiz and has a mean depth of 697 m. Here, only the 595 NAO shows statistically significant changes in θbot and Sbot (0.14 °C and 0.034 respectively).

596
In Viking20, the NAO is also associated with weaker KEbot (-0.22 x 10 -2 m 2 s -2 ) during a high 597 state. Case study 8 is situated around the Azores and is the deepest site with a mean depth of 598 3064 m. There is insufficient observational data to assess changes here with EN4 weightings

640
In this paper we set out to ask whether there are statistically significant changes in bottom 641 conditions across the northern North Atlantic, and its adjacent shelf seas, associated with four 642 major climate indices, and whether these changes are spatially coherent. The answer to both 643 of these questions is 'yes', but what of the physical processes responsible for these changes?

644
This is a more nuanced question. Bottom conditions in the northern North Atlantic region 645 span shallow seas to deep oceans; thus in one map we have contrasting dynamical regimes: 646 from highly seasonal shelf seas, and energetic boundary currents, to quiescent abyssal depths.

647
It is likely that different mechanisms are important at different depths, and lag-times between 648 changes in the index and bottom manifestations will also vary. In this discussion section, we 649 touch on some possible physical processes responsible for observed significant correlations, 650 but anticipate that we only scratch the surface leaving deeper analysis for future work.       Fig. 6b, Fig. 7c