Latitudinal patterns in the concentrations of biologically utilised elements in the surface ocean

Understanding of controls on the spatial distributions of chemical elements in the surface ocean has improved over time. Macronutrients were understood first, followed by dissolved inorganic carbon and alkalinity. Utilising data collected in the Atlantic by the ongoing GEOTRACES programme, controls can now start to be investigated for other elements. Here we investigate the generality of the rule that, in surface waters, higher concentrations occur at higher latitudes. Our analyses of Atlantic GEOTRACES data show that, after salinity normalisation, all biologically utilised elements except iron follow this rule (ρ ≥ 0.45). Most elements 10 (nitrate, phosphate, cadmium, barium, and nickel) are even more strongly correlated (ρ > 0.6) with latitude. We attribute this pattern to upwelling and/or entrainment of deep water at high latitudes. Although only Atlantic data was analysed here, we predict that this rule will be found to hold true for all oceans in which surface and deep waters exchange more readily at high latitudes. The rule does not hold in the central western Arctic Ocean, where a year-round strong halocline prevents exchange of surface and deep waters. 15

Alkalinity: Because of interest in the ocean's role in climate through its uptake of carbon dioxide from the atmosphere, large datasets of two other biogeochemical variables, DIC and alkalinity, have also been accumulated over the last 20 years. The first large global data sets containing alkalinity data (GLODAP and GLODAPv2;Olsen et al., 2016Olsen et al., , 2019Key et al., 2015) have shown a more complicated picture for alkalinity than for macronutrients. It is apparent that raw (untransformed) alkalinity data does not 40 show a strong correlation with latitude. Instead it exhibits a more complex distribution in the surface ocean, with highest values in the subtropical gyres, lower values close to the Equator and intermediate values towards the poles (Lee et al., 2006;Millero et al., 1998;Takahashi et al., 2014;Carter et al., 2014). Data analysis reveals a strong correlation with salinity (Millero et al., 1998;Friis et al., 2003;Jiang et al., 2014;Fry et al., 2015); high alkalinity values co-occur with high salinity values in the subtropical gyres because both are produced by an excess of evaporation over precipitation. Evaporation (removal of fresh water containing no 45 dissolved ions) raises the concentrations of all the dissolved elemental constituents left behind in seawater and therefore also raises total alkalinity (TA), because TA is a weighted sum of ionic concentrations.
Although not evident in unprocessed alkalinity data, a hidden high-latitude elevation is revealed when evaporation/precipitation effects are removed (Fry et al., 2015). Moreover, when the effects of other processes known to influence alkalinity are also removed, 50 to leave behind a tracer controlled only by calcium carbonate production and upwelling of dissolution-affected deep waters, it is seen to be upwelling that drives the high latitude elevation in salinity normalised TA (nTA) (Fry et al., 2015). This effect is masked in maps of TA but is apparent in maps of nTA and of related tracers such as potential alkalinity (Takahashi et al., 2014) and Alk* (Fry et al., 2015).

55
Dissolved inorganic carbon (DIC): Understanding of the distribution of DIC in surface waters has advanced at the same time as that of TA. A low-latitude to high-latitude gradient is apparent in non-normalised DIC data (Lee et al., 2000;Key et al., 2004;Takahashi et al., 2014) and even more strikingly apparent in salinity-normalised data (nDIC) (Wu et al., 2019). This latitudinal pattern has traditionally (e.g. Follows and Williams, 2011) been attributed to the temperature dependence of CO2 solubility (cold water holds more CO2 gas than warm water; if both are in equilibrium with the same atmospheric CO2; in addition, seawater with 60 higher CO2 also has higher DIC, all else being equal). A recent analysis (Wu et al., 2019) has shown, however, that the latitudinal gradient is driven in more or less equal part by upwelling/entrainment as it is by temperature-induced solubility variation.
Most recently, higher surface nickel concentrations at either end of an Atlantic-long transect have been reported (Middag et al., 2020).
where bioutilised elements are defined here to be those which are taken up from surface water by organisms and which, as a result, 75 increase in concentration with depth in the ocean. We distinguish here between bioutilised and biounutilised elements. These 2 categories map onto the 3 categories of an earlier scheme (Broecker and Peng, 1982): our bioutilised category encompasses both biolimiting and biointermediate categories of the earlier scheme; our biounutilised corresponds to their biounlimiting. In this study we do not care whether elements are limiting nutrients for biological production, we care only whether (1) an element is transported to depth by the particle flux, and (2) this gives rise to a vertical gradient in the concentration of that element. Following the earlier 80 scheme, some essential elements for plankton growth, such as manganese (Raven, 1990), are in this way classified as biounutilised, because their uptake and vertical transport does not result in a vertical gradient of increasing concentration with depth ( Fig. 1; see also Sarmiento and Gruber, 2006). Conversely, some elements are classified here as bioutilised even though they are not required by plankton but rather are taken up "by accident" because of their chemical similarity to useful elements (for instance Ge, which is taken up inadvertently by diatoms in place of Si (Azam and Volcani, 1981), and increases with depth as a result (Sarmiento and 85 Gruber, 2006)). Here we use data from the international GEOTRACES programme to explore the extent to which this hypothesis holds true (for 90 instance, as just described, we know it to be true for nTA but not so much for TA). GEOTRACES is the first programme to make https://doi.org/10.5194/bg-2020-371 Preprint. Discussion started: 19 October 2020 c Author(s) 2020. CC BY 4.0 License. a global survey of the distributions of a large number of different elements, including trace elements and micronutrients as well as the more commonly measured variables. Here we look to see if the rule of high latitude surface enrichment applies also to the wider suite of GEOTRACES variables.

95
An earlier global survey campaign (WOCE, the World Ocean Circulation Experiment (Chapman, 1998]) measured macronutrients (NO3, PO4, SiO4) and carbonate chemistry (e.g. DIC & TA) on a global scale. Large data syntheses now exist for macronutrients (WOCE, WOA) and carbonate chemistry (GLODAP) (Olsen et al., 2016(Olsen et al., , 2019Key et al., 2015). GEOTRACES is the first programme to measure a larger number of different biogeochemical variables in a comprehensive manner on a global scale. The GEOTRACES second Intermediate Data Product (IDP2017v2) contains bottle data from 39 cruises, collected during the period 100 2007 to 2014. Only dissolved concentration data is used here; particulate concentrations are not included in our analyses. The recommended methods and protocols used to measure the different dissolved element concentrations are included in the GEOTRACES cookbook (available at: https://www.geotraces.org/cookbook). Subsequent to collection, the data was subjected to quality control procedures and, where such procedures indicated systematic offsets, adjustment . Table 1 shows which elements were included in our study and on which GEOTRACES cruises they were measured. In this study we only 105 included those GEOTRACES cruises for which final data has been released (the programme is part-way through). At the time of writing, sufficient GEOTRACES data to test the hypothesis of this paper was only available for the Atlantic; for this reason, data from other ocean basins is not considered further here. Figure 2 shows the geographical distribution of the Atlantic data for the various elements.   GLODAPv2_2019 was used here. Low latitudes are defined in this study as between 30° S and 30° N and high latitudes as between 50° and 70° in the relevant hemisphere. The only elements used in our analysis were those with more than 15 surface layer measurements in both the low and high latitude categories. As shown later (Sect. 4.1), although some of the datasets are relatively small, there is enough data to obtain statistically significant results when testing the hypothesis. However, it is important to note that there are differences between elements in terms of data availability; for instance, there is no high latitude North Atlantic data 175 for Ba and there is no high latitude South Atlantic data for La, Ni, Pb or Y (Fig. 2).

110
The data processing followed a similar method to that of Fry et al. (2015) and Wu et al. (2019). The method was applied to both GEOTRACES and GLODAPv2 data. Only bottle data accompanied by a 'good' flag value was used. The surface ocean was defined as 0-20 m between 30° S and 30° N and 0-30 m elsewhere. Data from locations with a seafloor depth shallower than 200 180 m or salinities less than 33 were excluded so that only open ocean data relatively unaffected by the seafloor and/or rivers is included. Data from the Mediterranean Sea (enclosed basin) was not included. Data from > 65° N was excluded so that the heavily freshwater-influenced Arctic Ocean was omitted from the main analysis, although considered separately later (Sect. 4.5).
Salinity normalisation was carried out according to: where is the salinity, the measured value, and is the salinity-normalised value.
To distinguish between bioutilised and other elements, vertical profiles were plotted from a typical GEOTRACES station at 22° S, 33° W ( Fig. 1). On this basis, Ba, Cd, Fe, Ni, NO3, PO4, SiO4 and Zn are all categorised as bioutilised because their vertical profiles all show surface depletion relative to deep values. Of these elements, all were categorised previously by Broecker and Peng (1982) 190 as bio-limiting or bio-intermediate, except iron and nickel which could not at that time be categorised due to lack of reliable data.
Mn, La, Y and Pb, in contrast, do not show surface depletion ( Fig. 1) and are categorised here as biounutilised.
Statistical hypothesis testing was carried out using a null hypothesis (H0) of: "High latitude concentrations are lower than or equal to low latitude concentrations." Most of the element datasets are not normally distributed and so a Mann-Whitney U test was used 195 (one-tailed, assuming equal variances), although similar conclusions (not shown) were obtained when applying a t-test (one-tailed, unequal variances) to the same data.
Spearman's rank order correlation coefficients were calculated between absolute latitude (equivalent, for correlation purposes, to distance from the equator) and concentration of the element. The routine used (the 'corr' function in Matlab) also tested the 200 hypothesis that the two variables are correlated (null hypothesis (H0) of "no correlation"). All data were included in the correlation calculations, not just the low and high latitude data.

Results
All plots and analyses were carried out on salinity-normalised data. In this section, NO3 therefore refers to nNO3, and so on for other elements.

Relationship with latitude
The geographical distributions of the GEOTRACES data are shown for each element in Fig. 2, with points colour-coded to indicate concentration. The hypothesis can be visually evaluated by inspection of Fig. 2. Figure 3 shows more clearly the presence or absence of a connection with latitude by plotting of data values (concentrations) against latitude.

Statistical tests
Results of the Mann Whitney U test are shown in

240
Moving on to consider biounutilised elements, Al has a low latitude mean that is greater than the high latitude means. Pb does not have any Southern Ocean data but does have a greater mean in the high latitude North Atlantic compared to low latitudes. Both Y and La have no Southern Ocean data and both have greater mean concentrations in the high latitude North Atlantic than in low latitudes (Table 2). Applying the Mann-Whitney U-test to Al, Pb, Y and La data results in a failure to reject H0 for all cases (Table   245 2).

Correlation with proximity to Poles
Those elements that have significantly higher mean concentrations at high than at low latitudes also tend to have the strongest Spearman rank correlations with latitude (Table 3)   The GEOTRACES IDP2017v2 dataset consists of a smaller number of cruises in the Atlantic (13) than does GLODAPv2_2019 275 (247). In addition, for some variables (Ba, Ni, La, Y, Pb), there is high-latitude data from only one hemisphere rather than from both. We now consider whether there is sufficient data in GEOTRACES to carry out a meaningful examination of the hypothesis.
This question can first be examined by comparing results from GEOTRACES to those from the larger GLODAPv2 dataset. As can be seen in Fig. 4, distributions of the same elements within the two datasets are similar and the major relationships with latitude 280 are apparent in both. From this it can be concluded that there is enough data in the smaller GEOTRACES dataset to pick out the https://doi.org/10.5194/bg-2020-371 Preprint. Discussion started: 19 October 2020 c Author(s) 2020. CC BY 4.0 License. main trends in the data, even though one or two differences are observed between the two datasets. Most noticeably, PO4 values are elevated (as high as 2 μmol kg -1 ) at 25° S in GEOTRACES but not to the same degree in GLODAPv2 (Fig. 4). The elevated values are from samples taken in the Benguela upwelling region (Fig. 1). Although GLODAPv2 cruises also sampled there, GEOTRACES sampling may have coincided with a time of intense upwelling whereas GLODAPv2 sampling did not. It seems, 285 therefore, that details in the data patterns are sensitive to the particular times and places of sampling but that the main trends in the data are not. Most results are significant not only at the 5 % but also at the 1 % and the 0.1 % levels (Tables 2 and 3). In other words, it is surpassingly unlikely in most cases that the differences between low latitude and high latitude means are just down to chance 295 variations in small datasets. In addition, if there is no real trend towards higher values at higher latitudes then we would not expect https://doi.org/10.5194/bg-2020-371 Preprint. Discussion started: 19 October 2020 c Author(s) 2020. CC BY 4.0 License.
correlations between absolute latitude and element concentration. Again, the results obtained are very clear: most elements are highly correlated (ρ > 0.6) with absolute latitude and, in addition, a hypothesis of no correlation is strongly rejected.

300
All of the macronutrients as well as DIC and TA have higher concentrations at high latitudes. Other bioutilised elements measured by GEOTRACES (specifically barium, cadmium, nickel and zinc) also share the same pattern (Table 2).

Outcome 2 of tests: Not all bioutilised elements increase with latitude
However, one bioutilised element does not share the behaviour: (i) a plot of Fe against latitude does not show increasing values towards high latitudes; (ii) there is no strong correlation with absolute latitude; (iii) statistical hypothesis testing is unable to reject 305 either the hypothesis of "high latitude concentrations ≤ low latitude concentrations" or the hypothesis of "no correlation between concentration and absolute latitude". We do not attempt to find a definitive answer here to the question of why Fe and Mn do not share the same pattern. However, it seems likely that Fe does not share the pattern because it is the proximate limiting nutrient in regions of upwelling / strong winter mixing (Moore et al., 2016) (including most high-latitude regions) and as a result is depleted to near-exhaustion in those surface waters (particularly in summer, when most ship-based observations are made at high latitudes). Al may be used in the siliceous frustules of diatoms (Gehlen et al., 2002). These elements were also tested to see if they exhibited a general pattern of higher values at high latitudes, but none did. Plots of values against latitude did not show elevation at high latitudes. Although yttrium exhibits a northwards increase along the Atlantic from lowest values at ~50° S to highest values at 315 ~60° N, yttrium values do not increase away from the equator in both directions. For La, Al and Pb, hypothesis testing was unable to reject the null hypotheses above and correlation analyses did not find any strong positive correlations with absolute latitude.

Probable physical cause of the enrichment of bioutilised elements at high latitudes
The phenomenon of high latitude enrichment has been shown here to hold true for all bioutilised elements examined except Fe.
We now speculate as to the most likely cause of this phenomenon.

320
Deep waters have higher concentrations of most bioutilised elements through the actions of the biological pump (Fig. 1). Therefore, when deep water is introduced into the surface layer it usually produces an increase in the concentrations there. However, deep waters are not often brought to the surface at low latitudes, because of the strong temperature differential between warm surface waters and cold deep waters. Deep waters are uniformly cold: (4-15 °C below 500 m, according to GLODAPv2 data) whereas 325 surface waters between 30° S and 30° N are uniformly warm (15-27 °C above 100 m). The density of seawater is strongly affected by its temperature, with warm water less dense than cold water. Although wind driven upwelling does occur at low latitudes (Kämpf and Chapman, 2016), such upwelling comes from relatively shallow depths (e.g. < 300 m) rather than from very deep in the ocean. Water from these intermediate depths is not so cold (e.g. 10-25 °C between 100 and 300 m) and is therefore able to be brought to the surface. In short, the overall thermal structure of the ocean (Fig. 5) inhibits water rising from considerable depth to the surface at low latitudes but, in contrast, permits it to do so at high latitudes. However, the water that is entrained in autumn and winter into the deepening surface layer is water that has only relatively recently left the surface (because the high-latitude North Atlantic is a deep water formation region, at the beginning of the deep ocean conveyor). The deep water that is stirred up into the surface of the high latitude North Atlantic is not greatly enriched in elements because it is 'young' deep water that has not had time to accumulate the products of particle flux remineralisation to any great 345 extent; for this reason, the degree of enrichment of surface waters in the high-latitude North Atlantic is not as great as in the Southern Ocean. This can be seen in the results of our analyses. Of those elements which exhibit high latitude enrichment, the enrichment in the high latitude South Atlantic (Southern Ocean) is always greater than that in the high latitude North Atlantic, without exception (Table 2). There is some enrichment in the high latitude North Atlantic, but it is less intense than in the Southern Ocean.

Input of deep water is not the only process
While it is argued here that there is a general tendency towards increased concentrations of bioutilised elements at high latitudes because of deep water inputs, we do not suggest that this is the only process affecting latitudinal patterns in elemental concentrations. For instance, an excess of evaporation over precipitation in the subtropical gyres tends to raise concentrations of elements there (i.e. at lower latitudes), opposing the pattern described here. This is seen for instance in TA before salinity 355 normalisation. As described above, biological uptake of the most limiting nutrient until it is exhausted overrides the tendency for iron to have higher concentrations at high latitudes. Higher concentrations of DIC at high latitudes are known to be in large part due to solubility variations with temperature, as well as the effects of upwelling (Wu et al., 2019).

Arctic counter-example
Data from the Arctic Ocean (> 65° N) has not been included in this study. However, previous studies have reported generally low 360 levels of macronutrients, DIC and total alkalinity in the surface waters of the western basin of the Arctic Ocean (e.g. Codispoti et al., 2005;Woosley and Millero, 2020), contrary to the general pattern shown here. It is clear that the western Arctic Ocean does not follow the rule of high-latitude enrichment.
We presume that the reason why the Arctic does not follow the general pattern is that large volumes of river water enter the Arctic 365 Ocean each year. Because of this, surface waters in the Arctic are considerably less saline (30-33.5; (Swift et al. 1997)) than deep waters (34.93-34.95; (Aagaard, 1981;Jones et al., 1995)). There is a strong halocline across the whole of the Arctic basin . So, although there is not a strong vertical temperature gradient in the Arctic, there is a strong vertical salinity gradient.
Salinity also influences the density of seawater (more saline waters are more dense). For this reason the exchange of surface and deep waters is impeded in the Arctic by salinity difference rather than temperature difference, and it is this that prevents supply of 370 elements from below, allowing concentrations to stay low at the surface (Codispoti et al., 2005;Reigstad et al., 2002;Jensen et al., 2019). Because of buoyancy of surface waters due to low salinity, the physical mechanism that increases concentrations in most high-latitude oceans cannot operate in the Arctic Ocean.

Relationship to previous work
Much previous work on trace element distributions has focussed on understanding vertical patterns (Bruland et al., 2014) and 375 similarities between vertical distributions of different elements, for instance Cd and P (de Baar et al., 1994;Middag et al., 2018) and Zn and Si (Vance et al., 2017;Middag et al., 2019). This paper investigates instead horizontal distributions.
Other work has, like us, investigated controls on horizontal surface distributions of different elements. Many studies have identified a pattern of high dissolved Fe concentrations above shelves and close to shore, because of Fe release from sediments (e.g. Johnson 380 et al., 1997). Continental shelves and margins are an important source of other elements as well, including Mn (Laës et al., 2007) and Co (Tagliabue et al., 2018). Inputs of Saharan dust are responsible for the higher dissolved Fe (and Al and Mn) concentrations in the subtropical North Atlantic, leading, indirectly, to lower PO4 concentrations in the subtropical North Atlantic than in the subtropical South Atlantic (Mather et al., 2008). Rare earth elements (REE) have been found to increase towards coastal and high latitude regions in the North Pacific (Hongo et al., 2006). Other known drivers of surface distributions include Fe from volcanic 385 ash (Duggen et al., 2010), temperature for DIC and river inputs near the mouths of large rivers (Cooley and Yager, 2006). Although there is clearly a great variety of controls, the results presented here point to the existence of one common driver that is important for all bioutilised elements except iron: deep water exchange at high latitudes exerts a strong influence over the horizontal distributions of them all.

390
Based on our results from the Atlantic and for a limited number of elements, we make the following testable predictions: (1) That other bioutilised elements (ones not measured on GEOTRACES cruisesfor instance Ca and Ge) will be found to exhibit the same pattern.
(2) That the patterns found in the Atlantic will also be found in the Pacific.
(3) That the degree of enrichment in the high-latitude North Pacific will be greater than that in the high-latitude North Atlantic 395 (because the deep waters there, at the end of the deep ocean conveyor, have higher concentrations).
(4) That bioutilised elements not yet measured in the western Arctic will be found to occur at low concentrations.

Conclusions
The main finding of this study is a general principle that elements taken up by biology tend to be more abundant in high-latitude surface waters than in low-latitude surface waters. The reason is that element-rich deep water is able to rise to the surface at high  410 Author Contributions. TT was responsible for the initial idea that formed the basis of this research. DP carried out data acquisition, analysis, and investigation with supervision from TT. The manuscript was drafted by both DP and TT with review and editing contributions made equally.
Competing Interests. The authors declare that they have no conflicts of interest.