Variability of surface currents in the area of study
Before investigating the particle trajectories and patterns of connections between different source and destination locations (Table 1 and Fig. 1), it is useful to analyze the variability of surface currents in the domain of study. The mean ocean and sea ice circulation in the Arctic Ocean and the North Atlantic is well known19,22; the data used as input in this study, averaged over the entire period 1993–2020, reproduce that circulation (Fig. 2a). At the large scale, the domain can be divided into two subregions, the northern North Atlantic with the European shelf seas (North Sea, Barents Sea), which remain mostly ice-free throughout the year, and the Arctic Ocean, where the presence of perennial or seasonal sea ice cover has a strong influence on the surface ocean dynamics. The dominant circulation features in the first region are as follows: the strong (southerly) flow along the east coast of Greenland and Newfoundland/Labrador (mean velocities up to 0.3 m s\(^{-1}\)), a cyclonic recirculation region in the area to the southwest of Iceland/east of southern Greenland (south of the Denmark Strait), and several branches of relatively stable north-easterly flow in the region west and north of the British Isles, further along the west and northwest coast of Norway and up to the west coast of Svalbard (mean velocities locally exceeding 0.1 m s\(^{-1}\)). Importantly, those main flow branches are largely independent of the local, instantaneous wind forcing, as they are part of the large- and regional-scale ocean circulation. In the remaining areas the vector-averaged currents are weak, at the level of a few centimetres per second. Crucially, however, that does not mean that the amplitudes of currents are similarly low in these areas (Fig. 2c); rather, their high short-term (synoptic scale) directional variability is reflected, associated with passing mesoscale weather systems. Over wide areas far from shore, the standard deviation of current directions (Fig. 2b) is close to 80\(^\circ\), i.e., the value corresponding to the directionally uniform distribution (see “Methods”). Not surprisingly, that lack of directional stability is further enhanced by the Stokes drift (Fig. 2d,e). Apart from a few isolated locations at the coast of Greenland, the standard deviation of Stokes-drift directions is higher than 55\(^\circ\) in all areas, and over most areas, it is higher than 70\(^\circ\). Therefore, the vector-averaged Stokes velocities tend to be very low (Fig. 2d). However, their average amplitude over large parts of the North Atlantic is comparable or even larger than that of the wind-generated currents (Fig. 2c,f,i and Supplementary Fig. 1). Consequently, as mentioned in the introduction and as demonstrated in the next section, the contribution of the Stokes drift to the net drift of surface-floating material is substantial. Moreover, in near-shore and shallow areas waves tend to propagate towards the coast due to refraction and to increase their steepness due to shoaling, thus increasing the amplitude of the Stokes drift (note that those effects are not visible at the scale of maps in Fig. 2 and Supplementary Fig. 1, but they are present in the data). Therefore, in coastal areas Stokes drift facilitates the stranding of floating material, or at least its transport towards the inner coastal zone (see Supplementary Fig. 13c for an example).
Due to strong wave energy attenuation in sea ice (especially in the high-frequency range of the spectrum, strongly contributing to the net Stokes drift), the role of wave-induced flow in the ice-covered regions of the Arctic Ocean is limited, so that the total currents are close to those associated with sea ice drift within the Beaufort Gyre and the Transpolar Drift, with average amplitudes of 0.05–0.15 m s\(^{-1}\). Still, the contribution of waves is noticeable in the seasonally ice-free areas of the southern Beaufort and Chukchi Seas, i.e., around coastal locations relevant for the subject of this study (Supplementary Fig. 13a,b).
Probability distributions of drift duration (a,c) and path sinuosity (b, d) for trajectories originating at GBR (a, b) and ALA (c, d). The bin widths equal 1 month in (a, c), 0.5 in (b) and 0.25 in (d). Only those destination regions are shown, which are reached by at least 0.01% of all paths originating at a given source region (nine for GBR and six for ALA; values in the legend entries in panels a, c).
(a–d): Time series of total monthly release (blue) and landing (red) events from GBR to NOR (a), SHE (b), FAR (c) and ICE (d). (e,f): Time series (normalized with maximum values) of the atmospheric indices \(i_{\mathrm {GBR}\rightarrow \mathrm {FAR}}(t)\) and \(i_{\mathrm {GBR}\rightarrow \mathrm {ICE}}(t)\) defined in Eqs. (1)–(3). In all plots the period 1996–2017 is shown, for which a robust statistics of trajectories with duration 1 day–3 years can be computed. The labels at the x-axes are placed at the years’ begin. The insets in (a, b) show the mean seasonal cycles of release (blue) and landing (red) events.
Patterns and variability of material transport
The large-scale, aggregate view of the surface material transport in the domain of study can be depicted in the form of a connection matrix, with rows/columns listing the source/destination regions, and cell values corresponding to the probability of particles travelling from source A to destination B (Fig. 3a). As for all source regions, a substantial (and, in some cases, a dominating) part of trajectories end on land within that same region after a few days of drift (Fig. 3b); hence, it is useful to exclude all A\(\rightarrow\)A-type trajectories from the connection matrix. In other words, for each source region we analyse the fate of those drifters that manage to escape that region and either reach one of the remaining 15 destinations or move out of the domain of study/stay adrift throughout the maximum time span considered (the first, unlabelled column in Fig. 3a). Notably, the fraction of drifters that escape their source region, normalized with the respective \(N_r\) (proportional to the length of that region’s coastline), is roughly constant for all regions with an exception of small islands or groups of islands: BEI, FAR, JAM and SHE (red line in Fig. 3b). Not surprisingly, escaping is particularly difficult from elongated regions, especially those with relatively strong Stokes drift directed towards the shore (e.g., NOR, GLS or the west coast of GBR), as well as from regions with concave coastlines, trapping drifters within their semi-enclosed basins (e.g., RUE or SBW).
Considering the mean surface currents described earlier, the dominating transport directions are not surprising. For example, the south-east coast of Greenland (GLS) is the main destination for trajectories originating at GLN, JAM, ICE and SVA, i.e., source regions to the north, north-east and east of GLS, in the (average) upstream direction. Similarly, the Norwegian coast (NOR) is the main destination of floaters from GBR, SHE and FAR. Notably, due to the already mentioned relatively high speeds and directional stability of the major currents, these dominating pathways tend to be relatively fast (see the median and minimum travel times in Fig. 3c,d) and to have small sinuosity (defined as a ratio of the total path length to the shortest possible path, i.e., the straight-line distance between the start and end point).
As said, the connections described above are related to the main, long-term-average circulation. Arguably, the less frequent, episodic or anomalous paths are more interesting, as they provide means to shorten the usual, multi-stage trajectories and create connections between regions that otherwise are isolated from each other. The character of those anomalous connections is hard to decipher from the aggregate statistics in Fig. 3, but becomes clear in the in-depth analysis below, performed for two example source regions, one (GBR) representing the North Atlantic part of the domain of study, the other (ALA) the Arctic Ocean.
Undoubtedly, the British Isles (GBR) are a main source of anthropogenic drifters among the 16 regions analyzed (as our simulations suggest, it is also one of the main receivers, see Fig. 3a). On average, as this region is located close to continental Europe and to the external boundary of the domain of study (Fig. 1), the great majority (>80%) of trajectories originating in the southeasternmost and southwesternmost release locations around GBR are ‘lost’ to nearby destinations at the southern North Sea coast and through the boundary, respectively. Of the remaining ones, the majority reach Norway (NOR; \(\sim\)23%), the Shetland Islands (SHE; \(\sim\)7%) and the Faroes (FAR; \(\sim\)0.7%). Overall, there are nine destination regions that receive at least 0.01% of all particles leaving GBR (Figs. 3a and 4a,b), enough for computing robust statistical measures of those paths. The remaining regions receive no (ALA, CAA, SBE) or only very few isolated particles (FJL, GLN, SBW).
As might be expected due to their spatial proximity, the typical drift duration from GBR to NOR, SHE and FAR is short (Fig. 3c,d), peaking at 1–2 months and, especially for SHE, rarely longer than 5–6 months (Fig. 4a). The corresponding pdfs of sinuosity are nearly exponential (Fig. 4b). However, the character of each of these three drift connections is very different, as can be seen from the time series of monthly particle departure and arrival events (Fig. 5a–c). In the case of NOR, the main receiver of material from GBR, transport takes place with comparable intensity every year, with high-amplitude, repeatable seasonal fluctuations, which are particularly strong in the case of landing times: all high peaks of landing visible in the time series in Fig. 5a occur between October and March, with the largest values in November and December. Although seasonal variability is also present in the GBR\(\rightarrow\)SHE case (Fig. 5b), due to a short distance (note that particles reaching SHE originate mostly in the northern part of GBR, i.e., in Scotland and Northern Ireland) the seasonal variability is overlapped by synoptic-scale weather patterns. The short distance and thus drift time makes successful GBR\(\rightarrow\)SHE transport possible even if duration of favourable conditions is relatively short. This is not the case for GBR\(\rightarrow\)FAR and, even more so, GBR\(\rightarrow\)ICE transport. The straight-line distance GBR\(\rightarrow\)FAR, depending on the exact release and landing location, averages 550 km, with a minimum of 330 km; for GBR\(\rightarrow\)ICE, the values are \(\sim\)1100 km and 800 km, respectively. Thus, with usual drift speeds of, say, 0.1–0.2 m s\(^{-1}\), 20–40 days (respectively 45–90 days) are necessary to cover the smallest distance from GBR to FAR (respectively from GBR to ICE). In both cases, the net drift direction (relative to the ocean floor) has to be roughly perpendicular to the mean northeasterly currents in that region (Fig. 2), i.e., apart from a velocity component perpendicular to the mean flow, a component opposing that flow is necessary to prevent the floaters from moving too far east before they drift sufficiently far north. To make that possible, anomalous wave/current forcing, and thus anomalous atmospheric circulation patterns, are necessary, either present over extended periods of time or causing exceptionally high drift speeds in north-westerly directions in the area in question. Accordingly, the GBR\(\rightarrow\)FAR paths tend to have low sinuosity, often close to the minimum value of 1, corresponding to a straight line, i.e, the drift is fast and unidirectional. In the case of GBR\(\rightarrow\)ICE, the distribution of path sinuosity has a broad peak in the range 2.5–5.0 and a wide tail, but, remarkably, the paths from the two largest GBR\(\rightarrow\)ICE transport events, clearly seen in Fig. 5d, are characterized by well-below-average sinuosity values.
Over the whole 28-year-long period of study, more than one particle leaving GBR in a thousand (0.13%) landed on ICE. This is a substantial number, especially if one considers that almost all of those particles were transported within just two events (Fig. 5d), one in 2000–2001 (with release peak in August 2000 and landing peak in February 2001) and one in 2013–2015 (with the release peak in December 2013 and landing distributed over several sub-events in late 2014 and throughout 2015). The scale and exceptional character of those two events indicates highly anomalous wind forcing. Indeed, their occurrence coincides with a rare combination of those atmospheric circulation indices that have their associated spatial patterns active over the relevant part of the North Atlantic: the North Atlantic Oscillation (NAO), East Atlantic (EA), East Atlantic/Western Russia (EA/WR) and Scandinavian (SCAND) patterns (see “Methods” and Supplementary Note S3). For an anomalous wind forcing in the northwesterly direction, positive phases of EA and SCAND and a negative phase of EA/WR are favourable, together with a small amplitude of NAO, i.e., weak overall northeasterly flow. Let us denote the time series of those indices, smoothed with a moving-average window of 9 months, with \(i_\mathrm {NAO}(t)\), \(i_\mathrm {EA}(t)\), \(i_\mathrm {SCAND}(t)\) and \(i_\mathrm {EAWR}(t)\), respectively, and define a new index, \(i_{\mathrm {GBR}\rightarrow \mathrm {ICE}}(t)\), as a product of four terms:
$$\begin{aligned} i_{\mathrm {GBR}\rightarrow \mathrm {ICE}}(t) = \max \{i_\mathrm {EA}(t),0\} \max \{i_\mathrm {SCAND}(t),0\} |\min \{i_\mathrm {EAWR}(t),0\}|\alpha _\mathrm {NAO}(t), \end{aligned}$$
(1)
where:
$$\begin{aligned} \alpha _\mathrm {NAO}(t) = \left\{ \begin{array}{ccc} 1 &{} \mathrm {if} &{} |i_\mathrm {NAO}(t)|<0.25, \\ 0 &{} \mathrm {if} &{} |i_\mathrm {NAO}(t)|\ge 0.25, \end{array}\right. \end{aligned}$$
(2)
The two periods of non-zero values of \(i_{\mathrm {GBR}\rightarrow \mathrm {ICE}}\) correspond well to the maxima of GBR\(\rightarrow\)ICE transport (Fig. 5d,f). (Although it is impossible to verify that prediction based on the available data, the values of \(i_{\mathrm {GBR}\rightarrow \mathrm {ICE}}\) computed for the period 1950–2020 suggest very strong GBR\(\rightarrow\)ICE transport events in 1952 and 1988, and weaker ones in 1961 and 1972.) For GBR\(\rightarrow\)FAR, the best combination of the basic indices is found as:
$$\begin{aligned} i_{\mathrm {GBR}\rightarrow \mathrm {FAR}}(t) = \max \{i_\mathrm {EA}(t),0\} \max \{i_\mathrm {SCAND}(t),0\} |\min \{\nabla i_\mathrm {EAWR}(t),0\}|, \end{aligned}$$
(3)
i.e., it is independent of NAO and includes the gradient of EA/WR instead of its values (that is, conducive for GBR\(\rightarrow\)ICE transport are periods of decreasing EA/WR). As can be seen in Fig. 5c–f, the agreement is not perfect—and hard to verify statistically due to the small number of events. Nevertheless, the indices (1)–(3) seem to provide a good rough guess of the likelihood of the GBR\(\rightarrow\)FAR and GBR\(\rightarrow\)ICE transports to occur. It is important to stress that in both cases it is the combination of the basic indices that provides a good predictor of transport and not any of these indices alone: none of EA, EA/WR, SCAND had exceptionally high amplitude during the events in question; and omitting any of the terms in Eqs. (1) and (3) spoils the performance of \(i_{\mathrm {GBR}\rightarrow \mathrm {ICE}}\) and \(i_{\mathrm {GBR}\rightarrow \mathrm {FAR}}\). In general, as shown in Supplementary Note S3, the correlation between the individual atmospheric indices and time series of monthly release and landing events (like those in Fig. 5a–d) is low and rarely exceeds ±0.2. Non-negligible values tend to occur for the ‘typical’ routes, e.g., to NOR from GBR, SHE and FAR (positive correlation with NAO and negative with SCAND). Otherwise, as in the example of GBR\(\rightarrow\)ICE discussed above, a combination of several factors is necessary.
The hitherto analysis concentrated on locations in the North Atlantic sector of the domain of study. The trajectories originating at ALA provide a good example of transport patterns in the second sub-domain, the Arctic Ocean. There are six destinations that receive floating material from ALA in significant quantities (Figs. 3 and 4c,d): two neighbouring ones, CAA to the east (i.e., upwind/upstream) and SBE to the west (i.e., downwind/downstream), and four remote ones, GLN, GLS, ICE and JAM. As on average the drift velocities in the Arctic are very low (Fig. 2), all features of trajectories leading to those two groups of destinations are markedly different. Connections ALA\(\rightarrow\)CAA and, especially, ALA\(\rightarrow\)SBE are active every year and exhibit a typical seasonal variability, although a different one than that observed in lower-latitude, ice-free regions: here, the maxima of landing frequency occur between August and December, with the highest values in August–October (Fig. 6a), that is, during the Arctic summer/autumn, when the nearshore areas of the Beaufort and Chukchi Seas are free of ice or covered with low-concentration, drifting ice pack. Usually, the drift duration to SBE and CAA is a few months and the paths often have relatively high sinuosity (Fig. 4c,d), reflecting the varying wind conditions (see Supplementary Fig 9 for selected path examples).
As in Fig. 5, but for transport ALA\(\rightarrow\)SBE (a), ALA\(\rightarrow\)GLS (b), GLN\(\rightarrow\)CAA (c) and CAA\(\rightarrow\)GLN (d).
Maps of the spatial density of final positions of particles released from ICE in the period 1993–2017 (in number of particles per grid cell, i.e., \(\sim\)156 km\(^2\)) in simulations with (a) and without (b) Stokes drift. The color scale is logarithmic, grid cells with no particles are white. (Maps created with Matlab version 2016b, https://www.mathworks.com/).
In order for the drifters to reach Greenland or Iceland, however, they have to not only cover large distances, first within the Beaufort Gyre and then the Transpolar Drift, but also encounter conditions favorable for crossing the Transpolar Drift in the easterly direction; otherwise, they return towards the North American coast within the eastern branch of the gyre. As can be seen in Fig. 6b with the example of the ALA\(\rightarrow\)GLS connection (time series for GLN, ICE and JAM are similar), there was only one ‘mega-event’ in the whole study period when the drift successfully completed within the time limit of 3 years: a very large number of drifters that left ALA between early 2006 and late 2007 reached GLS from throughout 2009 until early 2010, i.e., during the very strong negative phase of the Arctic Oscillation (AO) index (Supplementary Figs. 17 and 19e). The timing of that event corresponds to anomalous intensification of the Beaufort Gyre in the period 2007–2010, manifested in its increased doming23 and amplification in ice-drift curl24, combined with anomalies in the speed and orientation of the Transpolar Drift (TPD)20,25. In particular, the year 2007 was characterized by the strongest cross-TPD sea ice transport in the whole period 1979–201525. Remarkably, the period 2007–2009 seems also to mark a period of reversal in material transport between the two neighbouring regions of GLN and CAA (Fig. 6c,d), with transport GLN\(\rightarrow\)CAA dominating in the first and CAA\(\rightarrow\)GLN in the second part of the analysis period. Again, climate-change driven changes of the sea ice cover are likely responsible for that rapid change of drift pattern in that area.
Overall, the results of this study demonstrate the very clear separation between the two sub-domains of the area of study. The maps showing spatial density of the final positions of all trajectories originating from a given source region (Supplementary Note S4) can be broadly divided into two groups. Drifters originating in the western part of GLN, as well as in CAA, ALA and SBE stay within the Arctic Ocean or leave it through the Bering Strait and, especially, in a wide southerly stream along the east coast of Greenland. Crucially, none of these drifters reach any of the European destinations between GBR, NOR and RUE, and only a few (from SBE) reach SBW. Analogously, drifters originating in the northern North Atlantic only sporadically enter the Arctic Ocean. Notably, this is true not only for sources in Iceland, British Isles and continental Europe, but also for the European Arctic islands: SVA, BEI, FJL and islands belonging to RUE. The only exception is SBW: although its main receivers are located in the European Arctic, trajectories from this source regularly enter the Arctic Ocean (however, they only sporadically reach the coast there; rather, they leave again with the East Greenland current).
Support Lumiserver & Cynesys on Tipeee
Visit
our sponsors
Wise (formerly TransferWise) is the cheaper, easier way to send money abroad. It helps people move money quickly and easily between bank accounts in different countries. Convert 60+ currencies with ridiculously low fees - on average 7x cheaper than a bank. No hidden fees, no markup on the exchange rate, ever.
Now you can get a free first transfer up to 500£ with your ESNcard. You can access this offer here.
Source link