3.1. Winter Nitrate Distribution
 Winter nitrate concentrations vary widely between the various methods used in this study (Figures 1a, 1b, 1e, 2a, and 2b). In the temperate northeast Atlantic (50°N–60°N), for example, minimum and maximum estimates differ by more than 10 μmole kg−1 or a factor of 3 to 4 (Figures 2a and 2b). Generally the data sets which are based on interpolations in space (WOA_winter, WOA_maximum, LN_winter, LN_maximum) show much lower concentrations in the temperate and subarctic North Atlantic compared with data sets which are based on summer to winter extrapolation techniques (GBexWOA, KexWOA). At 50°N, for example, the LN_winter data set shows winter nitrate concentrations between about 5 μmole kg−1 at 20°W and 10 μmole kg−1 at 50°W (Figure 1a). At the same latitude the winter nitrate distribution based on the extrapolation method of Glover and Brewer  (GBexWOAvd) shows concentrations around 15 μmole kg−1 over most of the basin (Figure 1b). One major caveat of all attempts to describe the winter nitrate distribution from regional interpolation of winter time observations is the scarcity of data during that season. In Figure 1a, diamonds indicate grid cells for which data are available for the time period January to March. (WOA 2001 was checked for winter nitrate data, too, and showed a small number of additional grid cells with winter data around Ireland, but no additional data in the open ocean.) It becomes clear from this that the N-S distribution of winter nitrate in WOA_winter and hence the predicted location of the nutrient front between oligotrophic and nutrient replete waters builds on the interpolation of observations from close to Iceland in the north, stations at about 30°N in the south, and stations close to the continents in the east and west. The method of Louanchi and Najjar  to construct monthly nutrient data sets includes temporal data interpolation, and combines 1-monthly, 3-monthly and 5-monthly composites. If winter time observations are rare, like in the North Atlantic, this approach interpolates winter data from autumn and spring observations and hence, since a maximum can not be found from interpolation, will underestimate winter time concentrations in temperate and subpolar waters (data set LN_minimum in Figure 2a). Estimating the winter nitrate distribution in the Northern Hemisphere as the mean of January–February–March [Louanchi and Najjar, 2000] further smoothes the distribution (LN_winter in Figure 2a). The comparison with NABE data (shown as bars in Figure 2a), estimates from the outcrop analysis (squares in Figure 2a) and the winter nitrate estimates from 47°N/20°W [Koeve, 2001] indicates that, at least in the northeast Atlantic, winter estimates based on WOA98 data or the Louanchi and Najjar  data set underestimate the winter time concentrations in temperate and subarctic waters. It is noteworthy, though not in the focus of this study, that the Louanchi and Najjar  data appear to overestimate winter concentrations in the subtropics and hence most likely overestimate NCP there. Winter nitrate estimates based on the extrapolation technique (GBexWOA, Figure 2b) are subject to a number of systematic problems, like the influence of seasonal remineralization on winter nitrate estimates [Koeve, 2001] and the choice of the winter mixed layer criterion, which require attention.
Figure 1. Estimates of the winter nitrate distribution. (a) LN_winter: Winter surface nitrate distribution estimated as the seasonal mean for January, February, and March of the monthly resolving nitrate data set of Louanchi and Najjar . Diamonds indicate the grid points for which historical data are available for this time period (WOA unanalyzed data). (b) Winter nitrate data field estimated by applying the extrapolation method of Glover and Brewer  to the annual mean nitrate data set from the WOA98 (for details see the text). Diamonds indicate the grid points for which historical data are available in the unanalyzed annual nitrate composite of WOA98. (c) Oxygen saturation on the winter mixed layer depth plane. The extrapolation method is applied to the annual oxygen saturation data set of WOA98. (d) Oxygen saturation at the sea surface during Northern Hemisphere winter, extracted from the seasonal data set of the WOA98. The data distribution of the unanalyzed oxygen data set is not shown, however, unlike for nitrate winter data (see Figure 1a); oxygen data are available for about 50% of the grid points between 30°N and 65°N, with a very good representation of the open ocean. Data distribution plots are available from the NODC website http://www.nodc.noaa.gov/OC5/WOA98F/woaf_cd/search.html. (e) Winter nitrate data field estimated after correcting the estimates shown in Figure 1b by the nitrate equivalent of the seasonal apparent oxygen utilization (AOU) on the depth plane of the winter mixed layer depth. The AOU data set is from the annual WOA98 climatology. Like in Figure 1b, diamonds indicate the grid points for which historical data are available in the unanalyzed annual nitrate composite of WOA98.
Download figure to PowerPoint
Figure 2. A comparison of the distribution of different surface nitrate estimates along 20°W during winter. For acronyms, see Tables 2a, 2b, and 2c. (a) Estimates from published climatologies: LN_winter (thin dash-dotted line), LN_maximum (thin dashed line), WOA_winter (thick solid line), and WOA-maximum (thick dashed line). (b) Winter estimates from summer to winter extrapolation techniques. GBexWOAvd (thin dash-dotted line), GBexWOApd (thin dotted line), GBexWOApt (thin dash-dot-dotted line), KexWOAvd (thick solid line), KexWOA98pt (thick dashed line). In both plots, bars indicate ranges of observations from JGOFS process studies during early spring: 59°N–60°N, upper range (∼13 μmole kg−1) from R/V Atlantis cruise 119/4 during late April 1989 and lower range (10.5 μmole kg−1) from FS Meteor cruise 10/2 during late May 1989; 47°N, ∼ 6 μmole kg−1 from R/V Atlantis cruise 119/4 during late April 1989 and 5.0 to 6.5 μmole kg−1 from FS Meteor cruise 21/1 during late March 1992. All data are from U.S. and German JGOFS data archives. Squares indicate winter nitrate estimates from the isopycnal outcrop analysis [Körtzinger et al., 2001a]. The standard error of the latter ranged between 0.3 and 0.7 μmole kg−1. Star with vertical bar at 47°N refers to estimates from the “O2-step” method discussed by Koeve ; the vertical bar gives the range of interannual variation estimated for that site.
Download figure to PowerPoint
 The shallowing of the mixed layer in spring is a rapid process [Monterey and Levitus, 1997] which leaves the depth layer between the seasonal and the permanent thermoclines secluded from air-sea exchange at the surface of the ocean. Under these conditions remineralization of organic matter leads to a seasonal oxygen debt (apparent oxygen utilization, AOU) which can be used to estimate the size of the seasonal nitrate release between winter and summer [Koeve, 2001]. The magnitude of this problem is evident from the distribution of the oxygen saturation on the depth plane of the winter mixed layer (Figure 1c) which shows oxygen saturation below 90% in large areas north of about 40°N. This is much less compared with the “observed” winter time (March) oxygen saturation in the surface mixed layer of the WOA98 climatology, which is close to 100% saturation (Figure 1d). Owing to mixing with deep waters impoverished of oxygen and despite the effect of bubble injection [Craig and Hayward, 1987], a value of about 98% saturation appears to be more typical for winter condition in the deep mixing regimes of temperate and subarctic North Atlantic [Koeve, 2001] (Figure 1d). The area-weighted mean of the climatological winter oxygen saturation data at the oceans surface (WOA 98, January–March) between 40°N and 65°N is 97.9%; the standard error of this estimate is about 0.1%.
 The preformed nitrate concentration, NO3(pref), is estimated as described in section 2 (see equation (1)). I use two estimates of the seasonal AOU at the depth of the winter mixed layer plane. The first is extrapolated from the annual WOA98 AOU data set and assumes O2Sat (pref) = 100%. The second is recalculated from this AOU and the accompanying O2Sat data set and assumes a winter time oxygen saturation of O2Ssat (pref) = 98%.
 The estimated distribution of NO3(pref) (KexWOA, Figure 1e) shows consistently lower winter time nitrate values compared with the GBexWOA data set (Figure 1b), but still much higher values compared with LN_winter (Figure 1a). Estimated NO3(pref) concentrations at the NABE-47 station (47°N, 20°W) are 7.5 to 8.1 μmole kg−1 and compare well with results from a recent detailed local study (7.8 ± 0.8 μmole kg−1 [Koeve, 2001]) which was based on independent data and methods. For the section along 20°W (Figure 2b) the comparison with the estimates based on the isopycnal outcrop analysis, taken from Körtzinger et al. [2001a], shows that KexWOA tracks the position of the nitrate front the best among all data sets. The only exception is a moderate underestimate of winter nitrate concentrations in KexWOA in the very north.
 In subtropical regions, KexWOA usually predicts winter nitrate concentrations close to zero. Occasionally, however, negative preformed nitrate values are predicted at the northern and southern boundaries of the subtropical gyre (Figure 1e). This is a reminiscent of similar features observed in the subtropical North Pacific [Emerson and Hayward, 1995] and indicative that the remineralization of low-nitrogen dissolved organic matter [Kähler and Koeve, 2001] which can significantly contribute to oxygen consumption on shallow isopycnals below the subtropical gyre of the Pacific and Indian Ocean [Doval and Hansell, 2000; Abell et al., 2000] and thereby increases the AOU:NO3 ratio of remineralization (R in equation (1)). On deep isopycnals, which crop out in temperate and subarctic waters, however, the contribution of DOM remineralization to AOU was insignificant in the Pacific and Indian Ocean [Doval and Hansell, 2000]. Furthermore, regions of negative NO3(pref) reduce in size if an O2S(pref) = 98% is assumed. By analogy with the findings of Doval and Hansell  and backed up by the very good agreement between KexWOA and the winter estimates from the isopycnal outcrop analysis (Figure 2b), it is therefore concluded that the value of R = −9.1 is a reasonable choice for temperate and subarctic waters.
 Estimates of winter nitrate concentrations from the extrapolation method are sensitive to the choice of the mixed layer criterion. Along 20°W, for example, differences of up to a few micromoles between pt and pd based estimates are seen (GBexWOA estimates in Figure 2b). Fortunately, most of this sensitivity disappears after correction for seasonal remineralization [Koeve, 2001] (KexWOA estimates in Figure 2b). From this I estimate the uncertainty of the KexWOA winter nitrate estimate to be better than ±0.5 μmole kg−1.
3.2. Winter NCt Distribution
 Winter NCt along 20°W, calculated from T × NO3- and Alk × pCO2-algorithms, are shown in Figure 3. In high latitudes the range between the maximum and minimum estimates is about 50 μmole kg−1, which is the same magnitude as the seasonal NCt change in this region (Takahashi et al.  and this study; see below). Minimum and maximum estimates are from NCt computed with the T × NO3 approach from the LN_winter and GBexWOA nitrate estimates, respectively. The accompanying winter temperature data sets (not shown) differ by about 0.5°C since LN_winter (and WOA_winter) use SST while GBexWOA uses the winter temperature on the Zmax depth plane, which by definition of the mixed layer criterion is 0.5°C lower than the SST. For the section along 20°W (north of 30°N) this temperature difference corresponds to a difference in NCt of about 4–5 μmole kg−1. This is equivalent to 10–20% of the overall difference between LN_winter and GBexWOA based estimates. Hence the major source of the uncertainty in winter NCt estimates from T × NO3 algorithms is due to the uncertainty in winter nitrate estimates.
Figure 3. Salinity normalized total dissolved inorganic carbon (NCt) in the surface mixed layer during winter along 20°W in the North Atlantic (for acronyms, see text and Table 2a): KexLeeSST (thick solid line), KexLee (thick dashed line), Alk × pCO2 (thick dotted line), GBexLee (thin dash-dotted line), WOAwLee (thin long dashed line), and LNwLee (thin dash-dot-dotted line). For comparison, estimates from an isopycnal outcrop analysis (squares north of 40°N) and winter time observations (squares south of 40°N) from Körtzinger et al. [2001a] are shown.
Download figure to PowerPoint
 Moderate NCt values are found for the KexWOA based estimate (data set “KexLee”) and the Alk × pCO2 based estimate. Along 20°W the latter agree with the estimates from the isopycnal outcrop analysis [Körtzinger et al., 2001a] within a few μmole kg−1. In fact, most of the difference between KexLee and Alk × pCO2 is now due to the corresponding differences in temperature. KexLee utilizes the winter temperature on the Zmax depth plane while the surface mixed layer alkalinity, which is used to estimate the Alk × pCO2 based NCt data set, implies sea surface temperature (SST). The corresponding difference in NCt of 4–5 μmole kg−1 is about the size and direction of the difference between KexLee and Alk × pCO2. I conclude that, as for winter nitrate, reconciling estimates from KexLee, Alk × pCO2 and the isopycnal outcrop method allow an estimate of the most likely winter NCt distribution in the temperate and subarctic North Atlantic. If the winter nitrate estimates from KexWOA are combined with winter SST a NCt distribution (data set “KexLeeSST”) is computed which compares even better with the NCt from the Alk × pCO2 approach (Figure 4) and the isopycnal outcrop analysis. For both data sets, winter NCt increases from values around 1940 to 1960 μmole kg−1 in the subtropical gyre to values larger than 2100 μmole kg−1 north of 50°N. The mean difference (RMS) between both estimates is about 6 μmole kg−1.
Figure 4. Salinity normalized total dissolved inorganic carbon (NCt) in the surface mixed layer during winter in the North Atlantic Ocean. (a) Alk × pCO2. (b) KexLeeSST.
Download figure to PowerPoint
3.3. Summer Nitrate and NCt Distribution
 Owing to nitrate uptake during the phytoplankton spring bloom, most of the North Atlantic is nitrate depleted (NO3 < 0.5 μmole kg−1) in the surface mixed layer at the end of summer [Strass and Woods, 1991; Yentsch, 1990; Campbell and Aarup, 1992]. For waters south of Iceland, however, several authors [e.g., Veldhuis et al., 1993; Sambrotto et al., 1993b] report nitrate concentrations of several micromoles that persist over the growth season. All summer nitrate data sets (Table 2a) analyzed in this study reproduce this north-south distribution (Figure 5a). Differences between data sets are small (<2 μmole kg−1) compared with differences in winter nitrate estimates discussed above. Owing to the interpolation procedures applied [Louanchi and Najjar, 2000], gradients in the data set LN_summer are smooth. Along 20°W surface nitrate concentrations in this data set increase from slightly below 1 μmole kg−1 in subtropical waters toward about 2.5 μmole kg−1 south of Greenland. The data set WOA_minimum shows a stronger gradient between lower nitrate concentrations in the subtropics and concentrations around 4 μmole kg−1 around 60°N. A nutrient front is observed between 50°N and 55°N. An analysis of data compiled in the eWOCE archive (2nd edition [Schlitzer, 2000]; data not shown) supports the latter picture of a strong nitrate front at about 50°N–55°N. Surface summer data from south of this latitude are essentially zero, concentrations north of this front are around 5 μmole kg−1. Still, with the combined data from the WOCE program and the historical data from the WOA98, the subarctic North Atlantic is undersampled during the narrow time slot between the end of the summer phytoplankton bloom and the renewed cooling. The exact distribution of the minimum summer nitrate concentration in the subarctic North Atlantic remains vague. Whether, and why, nitrate uptake appears to stop at a concentration of about 4 to 5 μmole kg−1 needs further work.
Figure 5. Distribution of nitrate and salinity normalized total dissolved inorganic carbon in the surface mixed layer during summer along 20°W in the North Atlantic. (a) Summer nitrate distribution: WOA_minimum (solid line), WOA_summer (dotted line), and LN_summer (dashed line). (b) Summer NCt estimates: WOAsLee (solid line), LNsLee (dashed line), and Alk × pCO2 (dotted line).
Download figure to PowerPoint
 Differences in summer nitrate estimates propagate into very small differences when summer NCt is computed from the T × NO3 approach (Figure 5b). However, the Alk × pCO2 approach suggest annual minimum NCt along 20°W which are 13 ± 12 μmole kg−1 lower in the North Atlantic. From total-CO2 data compiled in the eWOCE archive, three east-west sections cross the 20°W longitude during summer (plus signs in Figure 5b). Two crossing-overs at 44°N and 59°N show NCt surface values of 2021 to 2024 μmole kg−1 and 2062 to 2066 μmole kg−1, respectively, which are very close to the T × NO3 based estimates. These stations where sampled in June (1997, WOCE section A24). The respective surface nitrate concentrations are 0 to 0.5 μmole kg−1 and 4.8 to 5 μmole kg−1. Another WOCE line (A05, sampled late July), crossed the 20°W line at 24.5°N and had a NCt value of 1969 μmole kg−1. A third section intersects at about 52°N and was sampled during September (1991, WOCE section A01E) and has a NCt concentration of 2023 μmole kg−1, which is very close to the annual minimum of the Alk × pCO2 based estimate of NCt. Unfortunately, only few stations from this late season section have total CO2 data from the surface mixed layer. They are always in between T × NO3 and Alk × pCO2 based NCt summer estimates from nearby grid points. This brief analysis suggests that ΔNCt estimates which use summer NCt data from the T × NO3 approach might be slight underestimates. However, since Alk × pCO2 based estimates are only available from the surface mixed layer and due to the uncertainty how to extrapolate this surface difference over depth, I regard the summer NCt estimates from the T × NO3 approach the most reasonable summer data set which can be used during this study.
3.4. Seasonal Nutrient and Carbon Budget
 From the larger number of winter and summer nitrate and NCt data sets a subset of three winter-summer pairs is chosen for further analysis (Tables 2a (seasonal differences) and 3). These represent high-end winter concentration estimates (GBexWOAmin), low-end winter estimates (LN_based) and one based on the AOU corrected winter nitrate data (KexWOAmin).
Table 3. Integrated Nitrate and Carbon Budgets (Σ0–100 m) of the Temperate and Subarctic North Atlantic (40°N to 65°N)a
|Estimate||NO3 Budget, mole N m−2||OC-NCt Budget, mole C m−2||ΔNCtseas, mole C m−2||ASE, mole C m−2||IC Prod., mole C m−2||C:N, mole:mole|
| EC-W92 (std)b||0.29||3.3||2.4||1.3||0.43||11.3|
| NCEP-41-W92c|| || || ||1.5|| ||12.2|
| NCEP-41-WM99d|| || || ||1.8|| ||13.2|
| NCEP-41-LM86e|| || || ||0.8|| ||9.7|
| NCEP-41-N00f|| || || ||1.0|| ||10.8|
|LN_basedg||0.22||2.8|| ||1.3|| ||12.6|
|GBexWOAminh||0.59||4.0|| ||1.3|| ||6.9|
 Both ΔNO3 and ΔNCt (Figures 6a and 6b) increase northward along 20°W and show maxima between 47°N (LN_based) and 55°N (GBexWOAmin). Large differences, however, exist for the magnitude of the three ΔNO3 and ΔNCt estimates shown, for example, the maxima along 20°W range from 0.2 mole NO3 m−2 (LN_based) to 1.15 mole NO3 m−2 (GBexWOAmin) and 2.2 mole NCt m−2 (LN_based) to 6.9 mole NCt m−2 (GBexWOAmin), respectively. Additional variation is added through the contribution of ASE and ΔIC to the carbon budget. For KexWOAmin, which is shown as an example (Figure 6c), CO2 air-sea gas exchange can add up to 50% to the observed ΔNCt. The relative importance of ASE for the carbon budget differs between data sets. It is largest for LN and lowest for GBexWOAmin. Production of CaCO3 on the other hand (Figure 6c) usually accounts for about 10% to 20% of the total carbon uptake. Estimates of this fraction and the ΔOC: ΔIC ‘rain ratio’ show only little variations between data sets but consistent regional patterns with an increase toward the north along 20°W (Figure 7).
Figure 6. Nitrate and total CO2 uptake along 20°W, integrated over the upper 100 m. (a) Seasonal nitrate uptake: KexWOAmin (solid line), GBexWOAmin (dashed line), and LN-based (dotted line). (b) Seasonal change of NCt (symbols like in Figure 6a). (c) NCt-budget of KexWOAmin. Seasonal NCt-change (ΔNCt; dash-dotted line), air-sea exchange (ASE; dashed line), CaCO3 production (given as negative values, −ΔIC, to indicate that CaCO3 production has to be subtracted from the seasonal NCt change; dotted line) and net production of organic carbon (ΔOC; solid line), computed according to equation (4), are shown. For acronyms, see Table 2a (seasonal differences).
Download figure to PowerPoint
Figure 7. Ratio of organic carbon to inorganic carbon uptake in the mixed layer along 20°W: KexWOAmin (solid line), GBexWOAmin (dashed line), and LN-based (dotted line). Dash-dotted line is an estimate of RRml [from Koeve et al., 2002] which assumes a constant C:N ratio (6.6) and winter and summer nitrate distribution similar to KexWOAmin.
Download figure to PowerPoint
 Not surprisingly, estimates of the C:N ratio from the three data sets are very different, too (Figure 8). For the LN data set an increase toward the north is predicted, C:N ratios along 20°W range from about 8 at 30°N to about 28 at 60°N. GBexWOAmin and KexWOAmin suggest increasing C:N ratios toward the oligotrophic subtropical North Atlantic. GBexWOAmin predicts a moderate increase from about 6.6 at 60°N to about 10 at 30°N. KexWOAmin varies between about 10 and 14 north of 37°N and rapidly increased south of 37°N. For the North Atlantic between 40°N and 65°N (260°E–380°E), mean annually integrated C:N ratios of 6.9 (GBexWOAmin), 11.3 (KexWOAmin) and 12.6 (LN_based) are estimated.
Figure 8. C:N uptake ratio along 20°W, estimated from integrals (upper 100 m) of the seasonal NCt budget (see Figure 6c) and nitrate uptake (see Figure 6a): KexWOAmin (solid line), GBexWOAmin (dashed line), and LN-based (dotted line).
Download figure to PowerPoint
3.6. Seasonal Development of the C:N Ratio of New Production in the (Temperate) North Atlantic
 Our current knowledge of the C:N ratio of new production in the temperate and subarctic North Atlantic is based on short-term process studies (days to weeks [Sambrotto et al., 1993a; Kähler and Koeve, 2001; Koeve, 2004]) and the analysis of data sets which cover variable parts of the growth season [Körtzinger et al., 2001a]. None of these studies attempted to estimate the C:N ratio of new production over the whole growth season and on the basin scale. In this section I briefly discuss evidence from the published studies and relate it to findings from this paper.
 Sambrotto et al. [1993a] analyzed the temporal development of salinity normalized total CO2 and nitrate data in the surface mixed layer observed during two drifting experiments in spring 1989 at the NABE-47°N site and found mean C:N ratios of new production of about 10 for the time period of a diatom bloom in late April to early May and larger than 9 (no ASE estimate available) during a flagellate bloom in late May [Sieracki et al., 1993]. Koeve  recently reanalyzed the same data set and found C:N estimates to depend on a proper correction of preformed values; the respectively corrected short term averaged (C:N)NCP ratio during the diatom spring bloom was 7.3 ± 0.7.
 Körtzinger et al. [2001a] suggested that the C:N uptake ratio was close to or even below 6.6 during the phytoplankton spring bloom, increased at the end of the bloom and was significantly elevated in post-bloom waters at the southern boundary of the temperate northeast Atlantic. In that study nitrate and carbon based estimates of NCP were computed from winter time estimates (isopycnal outcrop analysis; see the data used in Figures 2 and 3), a midsummer transect from 30°N to 60°N along 20°W and an estimate of ASE similar to the one described in section 2.4. CaCO3 formation was assumed to be an insignificant fraction of total CO2 drawdown. A similar analysis of the cumulative C:N ratio of net community production for the 1989-NABE data from 47°N, 20°W gave an average integral spring bloom C:N ratio of 7.2 ± 0.9 [Koeve, 2004]. The approach used by Körtzinger et al. [2001a] and this study necessarily integrates over larger time and space scales, and, in the absence of N2-fixation (see above), provides an estimate of the C:N ratio of NCP, (C:N)NCP.
 C:N assimilation ratios of primary production, (C:N)PP, can be computed from instantaneous isotope-based (15N, 14C) rate measurements of phytoplankton production. In systems where the f-ratio (NCP/PP) is high, the (C:N)PP may be taken as a first order proxy of (C:N)NCP simply since the majority of PP is NCP. During a process study at the NABE-47°N site (C:N)PP was close to Redfield, or lower, as long as nitrate was replete and the f-ratio was high [Bury et al., 2001]. In a similar study of a late bloom situation in the northeast Atlantic at 58°N, 20°W (surface nitrate 1–2 μmole L−1; surface chlorophyll a 1–1.2 μg L−1; f-ratio = 0.8 ± 0.1 [Boyd et al., 1997]), (C:N)PP ratios of 5.6 ± 2.3 (N = 8) were found. Both studies support the finding of Körtzinger et al. [2001a] and Koeve  that excess carbon production is not related to the spring bloom but a feature of the summer system. The analysis from this study (Figure 8 and Table 3) that the late winter to late summer integrated (C:N)NCP ratio is elevated throughout the temperate and subarctic North Atlantic suggests that the finding of a seasonal progression of (C:N)NCP from low values in spring to high values in summer [Körtzinger et al., 2001a] can be extended to the temperate and subarctic North Atlantic in general.
3.7. Source and Fate of Excess Carbon
 The presented evidence of elevated annual C:N ratios of net community production in large parts of the North Atlantic and the inferred finding that summertime carbon overconsumption is the most likely reason for the elevated spring + summer (C:N)NCP signal does not provide an a priori hint concerning the specific processes involved. Assuming that unrecognized N2-fixation is an unlikely explanation (see section 3.5.2), both carbon overflow during primary production and subsequent exudation of carbon rich DOM [Wood and Van Valen, 1990] or preferential recycling of nitrogen over carbon [Thomas et al., 1999] are possible source processes of carbon overconsumption. Whether the autotrophic pathway (carbon overflow) or the heterotrophic one (preferential recycling of N) dominates in the North Atlantic cannot be settled from the data presented in this study, but requires detailed in situ rate measurements. In either case, carbon rich (or nitrogen poor) material must be stored seasonally in the euphotic zone or become exported from the euphotic zone during summer in order to sustain a persistent (C:N)NCP until the end of summer. Seasonal storage of low-N dissolved organic carbon has been observed [Williams, 1995; Kähler and Koeve, 2001]; however, more data with an adequate seasonal coverage (including the wintertime start conditions) and from open ocean temperate or subarctic sites are needed to fully quantify the importance of this process in relation to the overall spring + summer (C:N)NCP ratios reported here. Seasonal storage of particulate organic material is unlikely to be important, since the C:N ratio of suspended POM is usually close to the Redfield value [Copin-Montégut and Copin-Montégut, 1983; Körtzinger et al., 2001a] and since no significant amounts of POM accumulate during summer in the open ocean North Atlantic. Rapid export of excess carbon during summer via sinking particles appears unlikely if suspended particles have C:N ratios close to Redfield. DOM export during summer is limited due to the strong vertical density stratification.
 While DOM accumulation could not fully explain the spring to summer gradient in mixed layer (C:N)NCP and since the computed residual cumulative C:N ratio of exported matter showed a seasonal progression from spring into summer systems in the northeast Atlantic, Körtzinger et al. [2001a] speculated that TEP (transparent exopolymer particles [Alldredge et al., 1993]) could be involved in export of excess carbon from the mixed layer. Extending this further, Engel and Passow  and Engel  have proposed that TEP, which has been shown to have a high C:N ratio [Engel and Passow, 2001; Mari et al., 2001], may provide a vehicle by which carbon-rich DOM from carbon overconsumption in the euphotic zone interacts with sinking POM and hence contribute to particle export, even into the deep ocean. TEP, which probably form from polysaccharides released by phytoplankton and bacteria, is very surface-active and easily coagulate among themselves and with other particles [Alldredge et al., 1993]. This reactivity gives rise to the mechanistic importance of TEP for particulate carbon fluxes in the ocean [Passow, 2002; Boyd and Stevens, 2002; Passow, 2004]. A non-Redfield “TEP-Pump” [Koeve, 2005] could provide a mechanism by which the total amount of carbon that is fixed in the ocean by phytoplankton and transported to the deep sea via the biological pump could well be more than calculated using the Redfield ratio [Engel, 2002].
 Evidence from particle flux studies provides constraints on the magnitude of the excess carbon export into the deep ocean. Studying a global data set of particle flux measurements, Schneider et al.  concluded that the global mean C:N export ratio (Z = 100 m) is 7.1 ± 0.1 and that the average C:N ratio of sinking particles increases by about 0.2 ± 0.1 per 1000 m. Elevated C:N ratios of about 10 (or even up to 20) as frequently observed in the deep ocean, for example in freshly deposited detritus on the seafloor [Smith et al., 1998; Lampitt et al., 2001; Beaulieu, 2002], do not hold a valid information concerning C:N ratios of material exported from the euphotic zone since they are primarily caused by preferential remineralization of N over C [Lee and Cronin, 1984] during the decent of particles into the deep ocean [Koeve, 2005].
 Interseasonal differences of the C:N ratio of sinking particles have been reported repeatedly from deep ocean sediment traps in the North Atlantic [von Bodungen et al., 1991; Honjo and Manganini, 1993]. At BATS, average C:N ratios of sinking particles in 150 m depth are around 7 during and after the spring bloom and increase to mean values around 8 during June to August (Figure 9a). In the temperate northeast Atlantic (47°N, 20°W) multiyear observations with moored sediment traps at 500 m depth show values around or below 7 during the spring bloom period (March to June) and values around 8 between August and October (Figure 9b). Both observations suggest that only a moderate fraction of the summertime carbon overconsumption appears to be exported from the euphotic zone (Figure 9a) or even sequestered below the winter mixed layer (Figure 9b).
Figure 9. Annual cycles of C:N ratios of sinking particles in the North Atlantic. (a) Export production C:N ratios at the Bermuda Time Series Station BATS. Data from surface tethered particle interceptor traps deployed at 150 m between 1988 and 2001 (http://www.bbsr.edu/cintoo/bats/bats.html) are binned into 2-weekly intervals (crosses, solid line). (b) Carbon sequestration C:N ratios at the Biotrans Site (47°N, 20°W). Data from bottom moored particle interceptor traps deployed at 500 m between 1992 and 1996 (W. Koeve and B. Zeitzschel, unpublished data, 1996) were binned into monthly intervals (crosses, solid line). Dashed and dotted lines in both plots show the cumulative C:N ratio and the Redfield value of 6.6, respectively.
Download figure to PowerPoint
 While direct export of dissolved organic matter is hindered during summer, it can be important during autumn when the stratification weakens and during winter when mixing reaches down to several hundred meters in the temperate and subarctic North Atlantic. Convective export of DOM out of the euphotic zone has been observed at the rim of the subtropical gyre [Carlson et al., 1994; Hansell and Carlson, 2001], where the deepest mixing and the phytoplankton spring bloom co-occur. For the temperate Northeast Atlantic, where deep mixing and the spring bloom are mutually exclusive, only sporadic observations of convective DOM export from the euphotic zone are available (P. Kähler, personal communication, 2002) and its relative importance as a sink of excess carbon from the summer system can not be quantified by now.
 The finding (Figure 2b) that the a priori chosen AOU/nitrate remineralization ratio of −9.1 (equation (1)) yields winter nitrate concentrations that are in very good agreement with independent estimates, which do not apply any Redfield ratio assumption, suggests that the seasonal AOU on the plane of the climatological winter mixed layer depth is due to remineralization of carbon and nitrogen in Redfield proportion. This indicates that the spring + summer remineralization at this depth layer is mainly driven by freshly deposited material from the spring bloom, which is characterized by Redfield (C:N)NCP [Körtzinger et al., 2001a; Koeve, 2004] as well as C:N ratios of export from the euphotic zone (Figure 9a) and C:N ratios of carbon sequestration (Figure 9b) close to the Redfield value. In turn this suggests that no significant (at least no detectable) amount of excess carbon reached the winter mixed layer plane, for example in the form of low-N DOM, by the end of summer. If, analogous to observations from the Pacific and Indian oceans [Doval and Hansell, 2000], no degradable DOC is subducted during the ventilation of mesopelagic waters at the end of winter, this may further point to limited deep export of excess carbon via low-N DOM during autumn and winter.
 If little excess carbon gets exported via particles (Figures 9a and 9b) [Schneider et al., 2003; Koeve, 2005] or in the form of low-N DOM (last paragraph) what is the final sink of the excess carbon from summertime carbon overconsumption? One possibility, though yet to be tested by measurements, could be remineralization during autumn and early winter within a continuously deepening mixed layer. If this would be the dominant sink, lasting, that is export on timescales longer than one year, export of excess carbon might be small.