Accelerated oxygen consumption in eastern Mediterranean deep waters following the recent changes in thermohaline circulation

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. C9, 8107, doi: /2002jc001454, 2003 Accelerated oxygen consumption in eastern Mediterranean deep waters following the recent changes in thermohaline circulation Birgit Klein, 1 Wolfgang Roether, 1 Nurit Kress, 2 Beniamino Bruno Manca, 3 Maurizio Ribera d Alcala, 4 Ekaterini Souvermezoglou, 5 Alexander Theocharis, 5 Guiseppe Civitarese, 6 and Anna Luchetta 6 Received 23 April 2002; revised 4 October 2002; accepted 9 October 2002; published 18 September [1] Using an intercalibrated set of oxygen data for the eastern Mediterranean, , we study the evolution of oxygen concentrations that accompanied the recent changes in the thermohaline circulation of this sea (the so-called Eastern Mediterranean Transient (EMT)). We find that, by way of massively transferring oxygen-rich nearsurface waters into the deep layers, the EMT by 1995 had raised oxygen concentrations considerably relative to the pre-emt situation in Between 1995 and 1999, however, the oxygen concentrations decreased noticeably, we identify oxygen decreases of about 5 mmol/kg for the layers below 2200 m and between about 500 and 1000 m depth, for which layers previous work found vanishing or little replenishment during Supporting evidence for absence of replenishment is obtained from the temporal evolution of tritium- 3 He ages. The oxygen decreases convert into utilization rates of approximately 1.3 mmol/(kg yr). An even higher rate, 2.3 mmol/(kg yr), is obtained for the deep waters of the Cretan Sea below 1000 m. The utilization rate found for the deepest waters significantly exceeds the pre-emt value of 0.53 mmol/(kg yr). We propose that the massive invasion of near surface waters affected by the EMT made available large amounts of dissolved organic carbon with an unusually high fraction of labile material, which in turn enhanced oxygen consumption. Supporting evidence is obtained from data of dissolved organic carbon, and from mesozooplankton ecology data. The enhanced oxygen utilization represents a further example of EMT-related disturbances in the biogeochemistry of the eastern Mediterranean. INDEX TERMS: 4215 Oceanography: General: Climate and interannual variability (3309); 4243 Oceanography: General: Marginal and semienclosed seas; 4805 Oceanography: Biological and Chemical: Biogeochemical cycles (1615); 4283 Oceanography: General: Water masses; KEYWORDS: oxygen consumption, Mediterranean, thermohaline circulation Citation: Klein, B., W. Roether, N. Kress, B. B. Manca, M. Ribera d Alcala, E. Souvermezoglou, A. Theocharis, G. Civitarese, and A. Luchetta, Accelerated oxygen consumption in eastern Mediterranean deep waters following the recent changes in thermohaline circulation, J. Geophys. Res., 108(C9), 8107, doi: /2002jc001454, Introduction [2] In the late 1980s to early 1990s, a combination of meteorological and hydrological factors led to the establishment of the Aegean Sea as a new source of deep water in the eastern Mediterranean, in addition or in place of the classical Adriatic source, which previously dominated the waters below about 1200 m depth [Roether et al., 1996; 1 Department of Oceanography, Institute of Environmental Physics, University of Bremen, Bremen, Germany. 2 Israel Oceanographic & Limnological Research Ltd., Haifa, Israel. 3 Istituto Nazionale di Oceanografia e di Geofisica Sperimentale (OGS), Sgonico/Trieste, Italy. 4 Stazione Zoologica Anton Dohrn, Napoli, Italy. 5 National Center for Marine Research, Athens, Greece. 6 Istituto Talassografico di Trieste, Trieste, Italy. Copyright 2003 by the American Geophysical Union /03/2002JC Klein et al., 1999; Lascaratos et al., 1999]; these waters were commonly termed Eastern Mediterranean Deep Water, EMDW. The resulting profound disturbances in the thermohaline circulation, which are still ongoing, are now commonly referred to as the Eastern Mediterranean Transient (EMT). The new Aegean source not only provided deep waters at a far larger rate (7-year average 1 Sv= 10 6 m 3 /s) than the Adriatic source previously (0.3 Sv [Roether and Schlitzer, 1991; Gacic et al., 1996; Astraldi et al., 1999]), but the waters produced also had different properties (higher temperature and salinity than for the Adriatic source). During the early years of the transient, the Aegean-derived waters were of particularly high density and thus affected primarily the near-bottom layers. Their massive addition enforced an uplifting of the residing EMDW. By early 1995, about 20% of the EMDW below about 1500 m depth had been replaced by Aegean-derived waters. Although presence of the latter was less evident upward of 1500 m, the uplifting apparently affected the PBE 8-1

2 PBE 8-2 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN entire water column up to at least the Levantine Intermediate Water [Roether et al., 1998a]. The influence of the Aegean waters was most pronounced in the vicinity of the Cretan Arc and decreased toward the east and west, becoming very small at the western margin of the Ionian Sea. In the subsequent years, gradually, the Aegean-derived contribution became more equally distributed horizontally by mixing. [3] Hydrographic data and current meter measurements have shown that the Aegean source peaked in 1992 with an output of more than 2 Sv [Tsimplis et al., 1997]. After 1993 the dense water production in the southern Aegean (the Cretan Sea) slowed down considerably [Theocharis et al., 1999], and at the same time, the waters became warmer and less saline, and thus also less dense. During the period newly formed waters only penetrated to m depth in the Cretan Sea, and the overflow from the Cretan Sea was, outside the Cretan Arc, limited to between 1500 and 2000 m; replenishment was also low farther up in the water column ( m) [Manca and Scarazzato, 2001; Theocharis et al., 2002]. The EMT in its later stages, thus implied a more moderate addition of Aegean dense waters into a restricted midwater column depth range only. As for the classical Adriatic deep-water source, the density of the waters produced after 1992 was insufficient to replenish bottom waters in the Ionian Sea, and presently, waters exiting the Adriatic through Otranto Strait are settling upward of 1500 m depth [Manca et al., 1998; Klein et al., 2000]. Little is known about the performance of the Adriatic source close to the onset of the EMT. A much reduced ADW supply was however indicated in observations in 1995 at a location close to the western slope of the Ionian basin where in 1987 the core of the most recently ventilated EMDW was prominent. Concentrations of the transient tracer CFC-12 at this location had not increased since 1987 [Klein et al., 1999] indicative of reduced deepwater discharge from the Adriatic. [4] There is only limited information concerning ecological effects of the EMT. One obvious feature was an uplifting of the nutricline [Klein et al., 1999], which brought nutrients much closer to the euphotic zone than was the case previously. It has been argued that the shoaling of the nutricline should result in higher productivity but, to date, observational evidence for this is limited. Mesozooplankton standing crop below 1000 m in the Ierapetra basin has been reported to be up to two orders of magnitude higher than observed prior to the EMT [Weikert et al., 2001]. The increase was mostly due to the copepod species C. helgolandicus, which had previously been indigenous to the northern marginal basins of the Mediterranean (Adriatic, Aegean) only. As the appearance of this species coincided with the changes in deepwater properties [Weikert et al., 2001], it is highly plausible that the event was related to the deposition of the dense waters overflowing from the Aegean into the adjacent Ierapetra basin. A natural consequence of the invasion of near-surface waters into the deep layers was a decrease in nutrient levels and an increase in oxygen concentrations in the deep waters between 1987 and 1995 [Klein et al., 1999]. Unexpectedly, however, a cruise in 1999 revealed a distinct decrease in oxygen over most of the deep waters of the eastern Mediterranean relative to 1995, and in part even relative to the oxygen concentrations observed prior to the EMT [Kress et al., 2003]. In the deep layers of the Cretan Sea, Souvermezoglou et al. [1999] reported decreasing oxygen already between 1992 and 1994, except for a near-bottom layer at a station in the eastern corner of the sea where the density during the period still increased. Given the mentioned lack of near-bottom water replenishment in recent years, a certain decrease in oxygen near to the bottom should be natural. But, as is outlined below, the decrease in was stronger than current knowledge on deepwater oxygen utilizations rates (OURs) would predict. [5] Oxygen consumption below the euphotic zone arises mostly from bacterial utilization of dissolved and particulate organic matter (DOM, POM), of which the former, which represents the far larger fraction [Doval et al., 1999], follows the water circulation and mixing, while the concentration of the latter is also influenced by the flux from the surface. Consumption is believed to decrease strongly with depth, as more easily digestible matter is attacked preferentially so that, continually, the remainder becomes more refractory. The classical view has been that much of the DOM found in the water column is old and too refractory to undergo much oxidation, but in recent years this view has been challenged and DOM has been assigned a more active role in the microbial food web [Lefèvre et al., 1996, Fasham et al., 2001]. A comparatively limited role of POM even in the upper waters ( m) has been claimed from sediment trap observations in the northwestern Mediterranean Sea [Lefèvre et al., 1996]. The traps were however placed rather close to the continental slope so that a disturbance by lateral POM input from the slope remains a possibility. The limited information available on OURs in deep ocean waters points to rather low rates. Craig and Weiss [1970] estimated an OUR of about 0.17 mmol/(kg yr) at 2000 m depth in the northeast Pacific from a 14 C-calibrated vertical advection-diffusion balance, and Broecker et al. [1991] deduced a rather similar rate of 0.12 mmol/(kg yr) for the Atlantic below 2000 m based on the correlation between oxygen deficiency and 14 C concentration. Jenkins [1987] reported a higher value for the North Atlantic (1.5 mmol/ (kg yr)), but this value refers to a shallower depth (1000 m) and, above all, represents an upper limit only. A comparatively higher OUR has been deduced for the deep waters of the eastern Mediterranean prior to the EMT (average below 1000 m 0.53 mmol/(kg yr) ± 35%), on the basis of a two-dimensional kinematic model calibrated using hydrographic and tracer data [Roether and Well, 2001]. Such a rather high OUR is unexpected in view of the oligotrophy of the eastern Mediterranean, but the comparatively high temperatures of the deep waters (>12 C) and the resulting higher activity of bacteria may offer an explanation. [6] The present work assesses the temporal evolution of oxygen within the eastern Mediterranean subsurface waters that accompanied the EMT. It furthermore attempts to convert decreasing oxygen concentrations into oxygen utilization rates and to relate these to changes in biogeochemical patterns. However, the oxygen concentration changes, absolutely speaking, have been quite moderate, so that absence of data biases is a prerequisite for such an analysis. This item is addressed in extent in the next section. Tritium and helium data are used to identify the layers in the eastern Mediterranean that show signs of stagnation. The temporal

3 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN PBE 8-3 Table 1. Measurement Techniques and Precision for Oxygen Data Sets Considered in This Work a Data Set Date Area Institution End-Point Detection Mesurement Precision Mean Surface Saturation M5/6 Aug. Sept entire EMED IfM Kiel Manual 0.5 mm 103 ± 1.6% M31/1 Jan. Feb entire EMED ITT Potentiometric 0.51 mm 99 ± 4.0% LIWEX/Shikmona April 1995 Levantine IOLR Potentiometric 0.4% 101 ± 1.0% LIWEX/Aegaeo April 1995 Levantine NCMR Colorimetric 0.05 ml/l 101 ± 1.8% LIWEX/Bilim b Feb Levantine METU 101 ± 3.0% LIWEX/Urania March April 1995 Levantine SZN Potentiometric % 101 ± 3.9% Aegaeo Oct. Nov around Crete NCMR Colorimetric 2.2 um 101 ± 0.6% M44/4 April May 1999 entire EMED IOLR Potentiometric 0.3% 102 ± 1.7% EMTEC April 1999 Ionian SZN Potentiometric % 101 ± 1.1% Sinapsi Jan Ionian SZN Potentiometric % 100 ± 0.8% a Adjusted data in column 7. The following abbreviations have been used for the involved institutions (Ifm Kiel, Institute of Marine Science in Kiel; Germany), (ITT, Istituto Talassografico di Trieste, Italy), (IOLR Israel Oceanographic and Limnological Res., Haifa, Israel), (NCMR, National Center for Marine Research, Athens, Greece), (SZN, Stazione Zoologica Anton Dohrn, Naples, Italy) and (METU, Middle East Technical University, Erdemli, Turkey). b Bilim data are only used in Figure 2. evolution of oxygen concentrations in the stagnant layers is corrected for the effects of mixing and advection and is converted to utilization rates. 2. Data Sets, Methods, and Intercalibration [7] The oxygen data sets that we use in this study cover the period 1987 to 1999, and are contributed by five institutions (Table 1). They originate from a basin-wide survey of R/V Meteor in the summer of 1987 (POEM V- 1987; Figure 1a) and from a multiship effort in early 1995 (LIWEX-95; Figure 1b), from a fall cruise of R/V Aegaeo in 1998 (MATER-Spreading 1998), and from observations by three ships in January to May 1999 (SINAPSI, EMTEC, Meteor M44, Figure 1c). The basin-wide surveys were organized by the cooperative programs POEM (Physical Oceanography of the eastern Mediterranean) and POEM-BC (POEM-Biology and Chemistry), and LIWEX stands for Levantine Intermediate Water Experiment. SINAPSI is the acronym of the Italian program (Seasonal, Interannual and Decadal Variability of Atmosphere, Oceans and Related Marine Ecosystem) and EMTEC stands for Eastern Mediterranean Transient and Ecosystem, a cooperative investigation, with R/V Meteor and R/V Urania concurrently at sea. [8] For oxygen measurement, seawater samples were drawn into volume-calibrated Erlenmeyer flasks (volumes between 60 ml (IOLR), 80 ml (SZN), 110 ml (NCMR)) and pickled. Dissolved oxygen was measured at sea using the Carpenter-Winkler titration procedure [Carpenter, 1965a, 1965b; Grasshoff, 1983]. Different automatic titrators were used (Radiometer TTT80 (IOLR), Metrohm Titroprocessor 636 (ITT), Metrohm Dosimat 665 (NCMR) and DMS 716 Titrino (SZN)), equipped with a dual platinum electrode, in the dead-stop end point mode. The samples were titrated in the original Erlenmeyer flask without any transfer. The reported precisions varied from % (IOLR) to % (SZN), 0.5% (ITT) and 1% (NCMR) (see Table 1). The thiosulfate solution was calibrated with standard potassium iodate solution either bought from Merck (IOLR, SZN, NCMR) or prepared at the lab (ITT). Concentrations are reported in mmol/kg. [9] To detect any biases between the oxygen data sets, we first compared concentrations between selected pairs of hydrographic stations within confined distance and with an adequately small time lag in sampling. The LIWEX-95 situation, when five institutions were operating in a coordinated manner within a rather short period (3 months), is adressed in Figure 2, presenting comparisons of ITT (Meteor), IOLR (Shikmona), NCMR (Aegaeo) and METU (Bilim; METU data are not used beyond Figure 2) with SZN (Urania). The SZN data, chosen as the reference, evidently agree satisfactorily with those of NCMR and of METU (Figures 2c and 2d). On the other hand, ITT is significantly high relative to SZN (Figure 2a), and IOLR consistently low (Figure 2b), the offsets being of a similar magnitude as the temporal changes in oxygen concentration that we address below. To quantify the deviations, deepwater oxygen data for all available stations pairs were linearly interpolated to set density levels. The combined ITT - SZN differences at these density levels are shown in Figure 3a, and mean deviations for the various station pairs are listed in Table 2 (s q > kg/m 3, depth > 1200 m). The average of these differences is 3.3 ± 0.3 mmol/kg (error is 1-sigma error of the mean). The IOLR data set had only very few data below 1000 m, and the comparison to SZN (Figure 3b, Table 2) therefore includes data from somewhat shallower depths (s q > kg/m 3, depth > 800 m). The resulting average IOLR - SZN difference is 5.5 ± 0.7 mmol/kg (Table 2). The same method was applied to the 1999 data that required a comparison between the oxygen scales of IOLR (M44) and SZN (S398 and EMTEC). Like for the LIWEX data set, IOLR is low compared to SZN (Figures 4 and 5, Table 2; s q > kg/m 3, depth > 1200 m), The mean IOLR offset is obtained as 4.9 ± 0.5 mmol/kg. [10] In consequence, we use in the following 1995 ITT (Meteor) oxygen data lowered by 3.4 mmol/kg, 1995 IOLR (Shikmona) data raised by 5.5 mmol/kg, and 1999 IOLR (Meteor) data raised by 4.9 mmol/kg. Unfortunately, we could not determine the cause(s) of the offsets, but we suspect standardization problems, so that the offsets should be rather uniform. In the comparisons of Figures 3 and 5, the apparent scatter of the oxygen concentration differences admittedly is somewhat larger than would correspond to the data uncertainties of Table 1, so that the true uncertainties are probably a little larger than given in Table 2. Moreover, there may be a certain contribution of true differences between the station pairs, like it is suggested by the apparently decreasing trend of the difference with density in Figure 5. However, in view of the large number of data points on which the averages

4 PBE 8-4 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN Figure 1. Station positions (a) for Meteor cruise M5/6, August/September 1987, (b) for the multiship LIWEX in 1995 (Aegaeo, Bilim, Meteor, Shikmona, and Urania stations are denoted by solid rhombes, open triangles, solid triangles, crosses, and open rhombes, respectively), and (c) for three data sets in 1999 (Meteor cruise M44/4, Sinapsi cruise S398, and the EMTEC cruise (Urania) are denoted by solid triangles, crosses, and open triangles) and the Aegaeo cruise in 1998 (denoted by solid circles). are based (Tables 2), the adjustment is judged reliable. Support is obtained by the fact that the IOLR-SZN comparisons for 1995 and 1999 agree within the errors, indicating internal consistency and coherence in time of the calibrations. A further point of support is that the adjusted data yield average surface water oxygen saturations that agree among the various cruises within the errors (Table 1). Our conclusion is that the adjusted data should be internally consistent within ±1.0 mmol/kg or better. The adjusted oxygen data in effect are normalized to the oxygen scales of SZN, NCMR, and METU. The accurate absolute concentrations cannot be ascertained, but this aspect is not dealt with further because the principal concern in addressing the questions posed in this study is the internal consistency of the observations. [11] A similar adjustment of the 1987 Meteor data was not attempted, because it was considered difficult to ascertain continuity in calibration over the 8-year gap until However, Figure 6 provides evidence that a deviation of the 1987 data should be small. Shown is an oxygen versus salinity plot of the 1995 deep waters in the vicinity of the Cretan Passage subject to an EMT-induced oxygen enhancement (see below), and superimposed the properties of (1) the corresponding deep waters in 1987 and (2) those of the dense outflow in Kasos Strait in Evidently, the 1995 deep waters are close to being apparent twocomponent mixtures of the pre-emt deep waters in 1987 and the new outflow. The point is that a very similar mixing situation was implied in a corresponding T-S plot reported previously [Roether et al., 1996]. From the consistency between the previous T-S relationship and Figure 6, and also from the fact that the average surface water saturation for the 1987 cruise is similar to those of the other cruises (Table 1; The 1987 Meteor cruise was in early fall so that saturation may in fact have been a little higher, and an opposite argument applies for the 1995 Meteor cruise), we conclude that there is no obvious deviation of the 1987 oxygen scale from the one adopted here. [12] We furthermore use tritium and 3 He data that were obtained by mass spectrometric measurement at the Bremen helium isotope lab (tritium measurement by the 3 He ingrowth method except for part of the 1987 data), using procedures reported previously [Roether et al., 1998b, 1999]. The data are combined to obtain tritium- 3 He ages according to equation (1) (see below), which requires the separation of the portion of 3 He due to the in situ decay of tritium. The uncertainty of this component [Roether et al., 1998b] is the main contributor to the error of the ages. These errors turn out to be between about ±0.5 year and ±1 year (the high value applies for rather low tritium concentrations in combination with a moderate age). For differences between ages obtained in different years of observation, adressed below (Figure 11), similar uncertainties apply, as much of the uncertainty of the tritiugenic 3 He is of a systematic nature and therefore tends to cancel in the differences. Note that the 1998 ages have a higher error due to a 3 He measurement problem. 3. Temporal Evolution of Oxygen in the Eastern Mediterranean [13] Having adjusted the oxygen data to a common scale, we now address the temporal evolution of oxygen. Figure 7 shows oxygen sections running all along the eastern Mediterranean in 1987, 1995, and We note that the structure of these sections is fully consistent with that of concurrent sections for other properties [Klein et al., 1999; Roether and Well, 2001; Kress et al., 2003], so that we only repeat the very basic features here. [14] In the pre-emt situation (1987, Figure 7a), there was an extended deepwater oxygen minimum centered in

5 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN PBE 8-5 Figure 2. Intercomparison of oxygen profiles at common stations, LIWEX experiment, (a) SZN (Urania, solid triangles), station 59, and ITT (Meteor, open squares), station 46, at 34.5 N, 26.0 E. (b) SZN (Urania, solid triangles), station 99, and IOLR (Shikmona, open circles), station 11, 33.5 N, 27.5 E. (c) SZN (Urania, solid triangles), station 63, and NCMR (Aegaeo, open rhombes), station 59, 34.5 N, 27.0 E. (d) SZN (Urania, solid triangles), station 28, and METU (Bilim, inverted triangles), station 3900, 35.5 N, 29.0 E m depth, deepening and intensifying eastward. Newly added ADW higher in oxygen was apparent toward the base of the western continental slope, in keeping with rather high CFC-12 concentrations reported in these waters previously [Schlitzer et al., 1991]. The upper ocean highoxygen layer was less deep in the Levantine basin (200 mmol/kg isoline at appr. 400 m) than in the Cretan Passage and in the Ionian Sea ( m) due to absence of an intermediate water mass in the east that intruded from the Cretan Sea (Cretan Intermediate Water) [Roether et al., 1999] and also reflecting the different dynamics in the two basins, with cyclonic circulation dominating the northern parts of the Levantine compared to anticyclonic motion in the central Ionian in By 1995, the intense Aegean outflow of newly formed dense waters had raised the oxygen concentrations through most of the deep waters (Figure 7b), primarily in and around the Cretan Passage below about 1600 m depth. An exception are the oxygen-rich upper waters, which were compressed due to the upwelling of older, comparatively oxygendepleted waters below them, so that oxygen concentrations in a few hundred m depth decreased, particularly so in the Ionian Sea. Oxygen also decreased below about 3000 m at the western slope in the Ionian Sea where previously the rather oxygen-rich admixtures of ADW were present. This decrease suggests a strongly reduced ADW supply, also indicated in CFC concentrations at this location in 1995 that are about the same as in 1987 [Klein et al., 1999]. The

6 PBE 8-6 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN Figure 3. Oxygen concentration differences versus density (s 0 ), (a) ITT-SZN below 1200 m and (b) IOLR-SZN below 800 m, for all common stations available. Differences are based on concentrations linearly interpolated to set densities for each station. Relative to the SZN data, ITT data are consistently higher, and IOLR data consistently lower section (Figure 7c) still shows a relative oxygen maximum in the Cretan Passage below about 1800 m, but spatial differences in deep-water oxygen are far smaller than in The reduced oxygen gradients are a result of mixing, in connection with a greatly reduced supply of dense Aegean waters that furthermore only reached intermediate-depth layers due to a reduced density (see section 1). [15] Most striking, however, is the fact that oxygen concentrations in 1999 (Figure 7c) are distinctly lower than in 1995 (Figure 7b) essentially all through the deep waters. The lowest concentrations are found at middepth, like in 1987, but the depths of the minimum are distinctly less ( m), presumably in response to the upward displacement of the pre-emt oxygen-minimum layer noted for the 1995 section. In the Ionian, the high-oxygen upper layer appears to be even shallower in 1999 than in 1995, suggesting that upward displacement continued aided further by the change in circulation in this basin from anticyclonic to cyclonic. Furthermore, minimum values are now even lower than prior to the EMT (Figure 7a). A comparison of the 190 and 185 mmol/kg isolines in the 1995 and 1999 sections, respectively, indicates oxygen decreases of roughly 5 mmol/kg, between about 500 and 1000 m depth in the Ionian, and reaching somewhat deeper in the Levantine. We note in passing that the general oxygen decrease is not an artifact of the data adjustments (section 2), as these act to reduce, rather than to increase, the 1995 to 1999 differences. [16] For a more detailed view, we next examine the temporal evolution of oxygen at selected stations that have been sampled repeatedly between 1987 and 1999 (Figures 8a 10a). The oxygen profiles are accompanied by salinity profiles (Figures 8b 10b) to reveal effects of advection and mixing. Figure 8a shows the collection for the middle Cretan Sea. Below about 900 m, the oxygen concentrations in 1987 and 1995 were rather similar, while farther up ( m) the 1995 values are appreciably lower. The 1998 values are the lowest below about 400 m and only between 250 and 400 m exceed the ones for The similarity of the deep-water oxygen concentrations in 1987 and 1995 reflects the fact that the Cretan Sea deep waters have always been rather well ventilated [Schlitzer et al., 1991], so that the accelerated deepwater formation in the 1990s made only a moderate difference; at the same time, oxygen concentrations by 1995 had probably already passed their maximum [Souvermezoglou et al., 1999]. The Table 2. Intercalibration Results for Data Sets Used in the Manuscript a Year 1995 LIWEX Institutions (Ships) ITT (Meteor) versus SZN (Urania) 1995 IOLR (Shikmona) versus LIWEX SZN (Urania) 1999 IOLR (Meteor) SZN (Urania) Density Interval Mean Oxygen Difference for Individual Profiles , , , , , , , , , , Average Oxygen Difference for Cruise No. of Individual Data Points Error of the Mean, mmol/kg a For the LIWEX experiment in 1995 oxygen data sets from ITT, IOLR and SZN have been compared on isopycnals. A similar comparison was made for the 1999 data sets from IOLR and SZN. Individual differences for station pairs are listed as well as average offsets for entire cruises.

7 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN PBE 8-7 Figure 4. Intercomparison of oxygen profiles at nearby stations, SZN (Urania, Sinapsi cruise, solid triangles), station C01, 35.5 N, 21.7 E, and IOLR (Meteor cruise M44/4, open circles), station 294, 35.2 N, 21.5 E oxygen minimum at 400 m is due to rather old water from the upper layers of the EMDW that intruded into the Aegean, as has been observed similarly for other properties [Klein et al., 1999]. The decrease in the deep-water oxygen values (>1000 m) amounts to as much as mmol/kg, which is in fact the highest decrease that we observed anywhere in the eastern Mediterranean. The salinity evolution follows a different pattern (Figure 8b). Salinity increased greatly during the early period of the EMT [Roether et al., 1996] and evidently came back only slowly thereafter, such that the decrease in salinity is much less pronounced than the one for oxygen. [17] The profile in the southern Rhodes Gyre area (Figure 9a) is representative of changes in a broad area around the Cretan Arc and the western Levantine basin. Oxygen values below 1000 m increased from their pre-emt values in 1987 near 180 mmol/kg to between 200 and 210 mmol/kg in 1995, while the opposite is true farther up in the water column. After 1995, the oxygen concentrations over the entire water column below the surface layer decreased up to 1998 and farther up to There are two depth ranges in which the oxygen decrease after 1995 is rather pronounced. The shallower of these layers is situated at middepth ( m) while the deeper one covers the water column below 2200 m. Around 2000 m depth, the changes in the profiles in the Rhodes Gyre are apparently smaller. The inflection toward smaller oxygen decrease around 2000 m is found in various profiles around the Cretan Arc, although the depth range varied and is related to discharge of less dense Aegean deep waters after Salinities also decreased after the initial increase in 1995 (Figure 9b) below about 2200 m depth, while farther up there is a certain variability. [18] Interestingly, in the western Ionian oxygen was already decreasing in the lower 500 m from 1987 to 1995 (Figure 10a). This part of the water column was in 1987 occupied by the ADW flowing along the Italian continental slope as a near-bottom boundary current. The decrease in oxygen of about 8 mmol/kg below 3000 m can thus be ascribed to a near-stagnant situation, but mixing with loweroxygen waters off the slope (Figure 7a) presumably contributed. During , oxygen decreased rather uniformly below 800 m by approximately 4 mmol/kg, and even more between there and 300 m depth. The shallow oxygen decline is accompanied by uplifting judging from the upward shift of the related salinity profiles in this depth range (Figure 10b), as noted already in connection with Figure 7. The 1999 station is appreciable farther off the slope than the previous ones, but due to the nearhomogeneity in oxygen already in 1995 this should not matter much. Salinities below 1500 m showed a slight increase between 1987 and 1995, and a more distinct one thereafter (Figure 10b) due to the westward propagation of the CDW in the Ionian. [19] The extent to which the deep waters in which we note decreases in oxygen have been isolated, in particular concerning the addition of younger, comparatively oxygenrich waters from upper layers, can be learned from transient tracer ages. Such isolation will be manifested as ages increasing with time. Water ages have been computed on the basis of 3 He grown in from tritium decay. The age t (years) is given by 3 ½ HeŠ t ¼ 17:69 log e 1 þ ½ 3 HŠ where is the radioactive mean life of tritium (years), and [ 3 H] and [ 3 He] are the concentrations of tritium and its decay product 3 He (section 2). Tritium- 3 He ages refer to the young portion of a water parcel, as a contribution of old, tritium-free water does not affect the 3 He/ 3 H ratio. For mixtures of (young) waters of different age, a component Figure 5. Oxygen concentration differences versus density (s 0 ), IOLR-SZN, below 1200 m, for all nearby stations available. Procedure as for Figure 3. IOLR data are consistently lower. ð1þ

8 PBE 8-8 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN Figure 6. Oxygen/salinity diagram for deepwater data ( m depth) of cruise M31/1 in the vicinity of the Cretan Passage (crosses), in comparison with old EMDW in the same region in 1987 (with error bar representing data variance) and values in the outflow from the Cretan Sea in Kasos Strait. The solid line is an apparent mixing line between the presumed parent water masses (circles). with a higher tritium content carries a higher weight [Doney et al., 1997], but in recent years, tritium concentrations according to unpublished Bremen data were sufficiently uniform to make this a secondary effect. Thus a true isolation will raise the tritium/ 3 He age by very nearly the time elapsed, while in a true steady state of subsurface water renewal and recirculation, water ages will be invariant in time. [20] In the Cretan Sea (Figure 11a), tritium/ 3 He ages in 1987 were maximal at about 7 years near 900 m depth (about the depth of the deepest outflow) and decreased to 3 years at the bottom and even more up to the surface. Evidently, newly formed water was added near to the bottom and increased in age by about 4 years prior to outflow. The finite age of the near-bottom waters indicates contributions of older waters being entrained in the densewater formation process. In 1995, the near-bottom ages are nearly the same. The faster renewal presumably is compensated by a higher age of the entrained subsurface waters due to mentioned addition of uplifted, old EMDW, which shows up as ages of up to 10 years between about m. The age difference between bottom and outflow-depth is now far less, reflecting the fast turnover of the waters that is implied by the EMT. By 1998, the age of the near-bottom waters increased relative to 1995 by about 5 years, even somewhat more than the actual time elapsed since the 1995 sampling (3.7 years, January 1995 to September 1998). The discrepancy may in part be due to vertical mixing prior to 1998 when the upper waters had rather higher age (see 1995 profile), but the higher analytical error of the He data (section 2) may also contribute. The data thus do support an isolation of the deep waters between 1995 and 1998 below about 1200 m. [21] Ages for repeated stations in the Levantine Sea are shown in Figure 11b. The depth resolution is somewhat limited and 1987 data are only available down to 300 m. For , one notes age increases below 2100 m depth and between about 500 and 1000 m. The age change in between is small, in keeping with the mentioned ongoing renewal in this depth range (see section 1). The large age increases somewhat exceed the period of time between the related surveys, like found in Figure 11a. Near to the western slope of the Ionian Sea (Figure 11c), the near-bottom age in 1987 was 6 years, and it rose to 14 years by The age difference closely matches the time difference between the two surveys. Part of the apparent ageing may be due to a replacement of young ADW by older waters from off the slope, but the principal effect presumably is isolation by a greatly diminished or ceasing supply of new ADW, as noted above. Between 1995 and 1999 there is further ageing below about 2200 m, which also closely matches the time difference between the observations. Because, as mentioned above, lateral property gradients had become small by 1995, the ageing can be ascribed essentially to stagnation of the water alone. Ages steadily increasing in time are also found farther up in the water column between about 500 and 1000 m. We note in passing that the lowest ages in Figure 11b and 11c, marking the most recently ventilated parts of the water column, distinctly exceed the lowest ages in the Aegean (Figure 11a). The difference is due to admixture of older waters during outflow from the formation areas (Cretan Sea,

9 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN PBE 8-9 Figure 7. Zonal sections (see inset maps) of oxygen concentrations (mmol/kg; adjusted data; see section 2) obtained by Meteor: (a) 1987 (M5/6), (b) 1995 (M31/1), and (c) 1999 (M44/4), station numbers are shown on top and data points are indicated by dots. Please note that because of the uncertainty in absolute concentrations in other contributions, this oxygen section is shown on the original scales used by the respective institutions.

10 PBE 8-10 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN Figure 8. Temporal evolution of (adjusted) (top left) oxygen concentrations and (top right) salinity in the central Cretan Sea (southern Aegean). Profiles are shown samples at nearby stations in 1987 (M5, station 753), 1995 (M31, station 41), and 1998 (Aegaeo, station 13). For station positions, see map. Adriatic) or during transit to the location of observation. We conclude from Figures 11b and 11c, that the deepest waters (>2200 m in the Levantine and >2000 m in the Ionian), and also an intermediate depth range ( m), have been largely isolated between 1995 and 1999, whereas there has been replenishment of the layers in between ( m). 4. Oxygen Utilization Rates [22] To convert the observed decreases in oxygen during the period 1995 to 1999 to oxygen utilization rates, one must account for effects of advection and mixing. Wherever advection and mixing during the period in question raised the contribution of dense Aegean overflow, salinity and oxygen would increase. This holds, no matter whether such overflow dates from the early EMT or occurred more recently, because the overflow was always comparatively high in both oxygen and salinity. The opposite situation is a net admixture of relatively more original old EMDW, which is lower in both oxygen and salinity. Both situations imply that salinity and oxygen should be positively correlated, and this is in fact borne out by the data of Figures In the first case, oxygen may rise despite utilization, while in the second one oxygen may decrease even in the absence of utilization. [23] For an assessment we consider the temporal evolution of the oxygen-salinity correlations in the form of oxygen-salinity diagrams, assisted by T-S diagrams, for certain subregions of the eastern Mediterranean at depths exceeding 2200 m (Figure 12) (1000 m for the Cretan Sea).

11 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN PBE 8-11 Figure 9. Same as Figure 8, but Rhodes Gyre area: 1987 (M5, station 723), 1995 (M31, station 53), 1998 (Aegaeo, station 43), and 1999 (M44, station 239). For station positions, see map. In constructing Figure 12, all available data also beyond those in Figures 8 10 were used. An exception are some stations within the Rhodes Gyre that in 1995 displayed unusually low salinity and oxygen, indicating a dynamically restricted addition of Aegean overflow waters or deep convection in the winter of 1992 [Sur et al., 1992]. As the affected volume is rather restricted, the effect on the situation at large should be small. We use two criteria to exclude effects other than consumption, such as mixing and advection, in producing differences in oxygen concentration between repeated observations. The first criterion is that the T-S diagrams in Figure 12 should indicate that the waters at the later date are compatible with being mixtures of the waters covered at the earlier date only. The second criterion is that, furthermore, the property ranges at the later date are less extended, reflecting the expectation that any mixing between the properties as observed at the earlier date would wipe out extreme values. [24] For the Levantine basin the T-S relationship is linear and identical for 1995 and 1999, and the 1999 data range is less extended than in 1995 (Figure 12b). As the criteria named above are thus met, we conclude that the offset for oxygen between the 2 years (Figure 12a) can be ascribed to consumption. The offset is taken to be the distance, parallel to the oxygen axis, between the straight lines fitted to the data points that are shown in the figure. The resulting oxygen loss is 5.4 mmol/kg. For the Ionian basin, the 1995 T-S relationship (Figure 12d) has a moderate kink close to 13.3 C. The 1999 relationship points to an average value of the 1995 data below the kink, so that there is no evidence for a contribution outside the waters covered by the 1995 T-S data. The 1999 data range is also more restricted than that for The oxygen-s relationship (Figure 12c) again indicates an oxygen deficit, which, proceeding as before, is obtained as 5.6 mmol/kg.

12 PBE 8-12 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN Figure 10. Same as Figure 8, but western Ionian Sea: 1987 (M5, station 779), 1995 (M31, station 8), and 1999 (M44, station 303). For station positions, see map. [25] A different situation is met in the Cretan Passage (Figures 12e and 12f ). Because in 1995 no observations were made near to center of the Passage (Figure 1b), the graphs make a distinction between more westerly and more easterly stations in the Passage. The first point is that temperature and salinity decreased between 1995 and 1999, such that the T-S data hardly overlap (Figure 12f ), and that an oxygen decrease due to consumption cannot be read from the oxygen-salinity correlations (Figure 12e) in the fashion above. Mean correlations for the Levantine and Ionian (Figures 12a 12d) have been added in the figures (dashed lines), showing that, as expected, mixing with such waters is compatible with the 1995 to 1999 T-S changes. Moreover, if we assume the waters near to the eastern end to have mixed essentially with Levantine waters, and those in the western part with Ionian waters, losses of oxygen become evident (Figure 12e). Rough guesses are 6 mmol/kg in the western part and distinctly less in the east. The losses are thus of similar magnitude as the losses above, but definite values cannot be obtained because the mixing in detail remains uncertain. [26] For the deepest waters in the Cretan Sea, the 1995 T-S relationship for T <14 C (S > , depth > 1500 m) shows some scatter (Figure 12h), but the 1998 data points fall within a subregion of those in 1995, so that we can assume that the criteria are met. Any addition of shallower 1995 waters by vertical mixing must be small to be compatible with the T-S changes, and even if such mixing occurred, it would make little effect in the oxygen distributions (Figure 12g). Because newly formed dense waters as mentioned did not penetrate beyond 1000 m between 1995 and 1998 [Theocharis et al., 2002], addition of upper waters to the depth range in question can be ruled out. We therefore take the oxygen decline due to consumption, to be the distance between the oxygen-salinity relationships at S = 39.05, yielding a value of 8.5 mmol/kg. Farther up (T > 14 C), the mixing situation is unclear and an oxygen consumption cannot be determined. [27] Taking into account the time intervals between the related observations, the oxygen deficiencies obtained for the deep layers of the Levantine, Ionian, and Cretan Sea

13 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN PBE 8-13 Figure 11. Vertical profiles of tritium- 3 He water ages (a) in the Aegean for the profiles shown in Figure 8, (b) for stations in the western Levantine for the profiles shown in Figure 9, and (c) for the western Ionian profiles shown in Figure 10. The age uncertainties arising from the tracer measurement errors are year (1998 errors are presumably larger). The same symbols as in Figures 8 10 have been used to indicate the year of observation. convert to utilization rates of 1.25 ± 0.35, 1.3 ± 0.35, and 2.3 ± 0.5 mmol/(kg yr), respectively. The errors are those due to the intercalibration uncertainties, taken to be ±1.5 mmol/kg for the concentration differences. Because quite a few data enter into the determination of the oxygen concentration differences, data precision as such is ignored. The noted oxygen decreases farther up in the water column (approximately m depth) are similar in magnitude (5 mmol/kg), yielding utilization rates of approximately 1.2 mmol/(kg yr). 5. Discussion [28] Oxygen utilization rates of mmol/(kg yr) for the period have been deduced both for the deepest waters of the Ionian and Levantine Seas and for

14 PBE 8-14 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN Figure 12. Oxygen-salinity and temperature-salinity correlation for 1995 (open circles), 1998 (solid rhombes), and 1999 (solid triangles) below 2000 m depth. (a) and (b) Levantine Ionian Sea, (c) and (d) Ionian Sea, (e) and (f ) Cretan Passage area, and (g) and (h) Cretan Sea. Data from all available stations in the respective areas are shown. For the Cretan Passage a distinction is made between more easterly stations, shown in 1995 as open circles (Levantine side) and crossed circles (Ionian side) and in 1999 as solid triangles (Levantine side) and inverted solid triangles (Ionian side); furthermore, straight dashed lines representing the mean correlations for the Ionian and Levantine are shown for reference. waters at intermediate depths (about m). The basis is that, as we have shown, these waters have been essentially stagnant during this period. Bacterial consumption therefore remained unbalanced, and the observed oxygen concentration decreases can be converted into utilization rates. Loss of oxygen was also found for the layers in between, but these waters have been subject to partial renewal. The renewal reduced the water ages (Figure 11), and therefore presumably also raised the oxygen concentrations, counteracting consumption. For these waters therefore, at most lower limits of oxygen utilization are obtainable. The applied adjustment of the oxygen data (section 2) is a source of uncertainty, but we maintain that we have taken into account the related

15 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN PBE 8-15 Figure 12. (continued) uncertainty realistically. Moreover, we mix in Figure 12 the data of different laboratories, and the evolving structure of the oxygen distribution (Figure 7) is entirely reasonable, supporting a view that biases are small. [29] The utilization rates obtained for the deepest waters not only greatly exceed the values reported for the major oceans (see section 1), but also the higher rate for the eastern Mediterranean deep waters prior to the EMT mentioned above (0.53 ± 0.19 mmol/(kg yr)) [Roether and Well, 2001]. The increase is more than twofold and clearly outside the errors. We cannot exclude that the oxygen decrease to a moderate degree is due to benthic consumption, but the mean distance to the seafloor of the layers in question should be sufficient to make this a minor effect. An even higher value is deduced for the deep Cretan Sea, which exceeds any literature value by so much that enhancement by a special ecological situation is indicated. However, the degree of enhancement remains uncertain as no related pre-emt reference value exists. Uncertain enhancement also holds for the waters of the intermediate depth range. The rates for this range moderately exceed the pre-emt values of Roether and Well [2001] ( m average 0.85 mmol/(kg yr), but these values have a considerable uncertainty in particular upward of 700 m, and the new rates are not really at variance with the literature.

16 PBE 8-16 KLEIN ET AL.: OXYGEN CONSUMPTION IN EASTERN MEDITERRANEAN [30] The oxygen decreases, on the order of 5 mmol/kg, should be accompanied by increased nutrient concentrations. Although classical Redfield ratios (molar changes related by O 2 :C:NO 3 :PO 4 = 138:106:16:1) have a limited relevance for the eastern Mediterranean [Kress and Herut, 2001], they should be adequate for a first guess. The resulting changes are C =3.8mmol/kg, NO 3 = 0.58 mmol/kg, and PO 4 = 0.04 mmol/kg. Nutrient data are available for all cruises, and at least the changes in NO 3 are judged large enough to be detectable in principle, but data were not intercalibrated among the institutions and at present it is impossible to verify the small NO 3 and PO 4 changes expected. Data for carbon are unsufficient for an assessment. [31] The natural explanation of an enhanced oxygen utilization in the deepest waters, in which such enhancement is judged definite, is correlated to the EMT. We propose that the massive invasion of Aegean waters into the deep layers outside the Cretan Arc provided an unusually massive and fast transfer of DOC from near-surface layers to great depths. This resulted in a higher survival of the more labile DOC fractions than in the pre-emt situation, the availability of which naturally led to enhanced oxygen utilization. Such an explanation would preferentially apply to the waters below about 1600 m only, because the Aegean dense water addition was distinctly less at shallower depths, except possibly for the vicinity of deep Cretan Sea outlets. The suspected increased productivity due to the EMT (see section 1) might also have contributed, by additionally enhancing DOC supply in the euphotic zone. An increased productivity could also have raised the supply of POM to the water column, which could have added to oxygen consumption also at shallower depths, such as in the mentioned intermediate depth range ( m). However, it seems unlikely that increased productivity could support the observed doubling of the deep consumption rates. Furthermore, the effect should be rather short-lived as POC transported to depth with the Aegean dense water is soon buried in the sediments. [32] There are two lines of evidence supporting such a DOC scenario. First, observations in the Ionian Sea point to DOC concentrations in 1999 of about 70 mmol/kg for surface water, 35 mmol/kg for old EMDW, and 45 ± 2 mmol/kg in near-bottom waters of the Ionian basin associated with the spreading of the Aegean-derived deep waters; POC was found to be far lower, only 1 2 mmol/kg [Seritti et al., 2003]. The Aegean derived waters thus have an elevated DOC content relative to older deep waters, which is consistent with the addition of surface waters affected by the EMT. Our scenario requests that by 1995 the value in the Aegean-influenced waters was only approximately 10% higher (i.e., mmol/kg), which is judged modest enough to be realistic. The DOC data, although limited in number, thus are consistent with the deduced oxygen consumption. A contribution of POC to the enhanced consumption rates cannot be ruled out. [33] The other line of evidence is related to the large increase of mesozooplankton biomass in the Ierapetra basin mentioned above, which first of all indicates a profound shift in the functioning of the deep food web. Integrated vertically between 2250 and 4200 m, the biomass dropped from 12.2 g/m 2 in 1993 to 0.3 g/m 2 in 1999 (wet weight), when only a slight excess over the levels of 1987 had survived [Weikert et al., 2001; Koppelmann et al., 2003]. The missing 11.9 g/m 2 of biomass correspond to an oxygen consumption of 0.04 mmol/kg (R. Koppelmann, personal communication, 2002). The mesozooplankton biomass at least partly depended on feeding at depth. It can be anticipated that the surplus in fresh DOC enhanced bacterial production, but as the mesozooplankton does not feed on bacteria, a sustained population of bacterivores was additionally required. Taking into account the additional trophic levels of bacteria and bacterivores, and assuming a factor of 10 increase in biomass per trophic level, the implied total oxygen consumption rises to 4 mmol/kg, about matching the estimated enhancement in oxygen utilization (1.2 mmol/ (kg yr) in place of regularly 0.5 mmol/(kg yr) for 4 years). This specific event probably was a localized effect, but similar additions of labile DOC elsewhere should have supported a standing crop of bacteria of a comparable magnitude, resulting in a similar oxygen demand. [34] Our scenario implies that oxygen utilization should have been decreasing since the onset of stagnation, as no more DOC was supplied so that the concentrations not only decreases but became more refractory. A similar situation should hold in the Aegean, where, indeed, Danovaro et al. [2001] report decreasing benthic meiofauna and bacterial abundance during the period for which we determined the oxygen decrease ( ). We would thus predict that future surveys will find lower rates of oxygen consumption in the deepest waters of both the eastern Mediterranean and the Cretan Sea. 6. Summary [35] Assessing the temporal evolution of oxygen concentrations during the Eastern Mediterranean Transient based on data during , we obtained oxygen utilization rates during , a period when the output of dense waters from the Cretan Sea had subsided and renewal of the eastern Mediterranean waters became restricted to a depth range between about 1500 and 2200 m. For the waters deeper than 2200 m, the rate obtained, approximately 1.3 mmol/(kg yr) is more than twice the value obtained for the pre-emt situation previously [Roether and Well, 2001]. We argue that dissolved organic material which under normal circumstances is transferred rather slowly from the surface to the deep layers, being constantly degraded, reached the deep waters in a comparatively short time by way of the EMT, increasing oxygen consumption over the pre-emt values. In a way, the new utilization rates support the pre-emt value which might look rather high in comparison to reported values in the deep waters of the major oceans. [36] If this scenario is correct, it follows that the oxygen utilization rates were the highest just when the replenishment stopped, and have since decreased, until replenishment may start again. Such prediction can be checked in future observations. The present study adds to investigating the biogeochemical changes induced by the EMT. [37] Acknowledgments. We are grateful to Allan Robinson and Paola Rizzoli, both Cambridge Massachusetts, USA, chair persons of POEM and POEM BC for organizing the 1987 and 1995 surveys. We are also grateful to Gerd Fraas and Wilfried Plep, Bremen, for carrying out most of the tritium and 3 He measurements. The Meteor cruises M5/6, M31/1, and M44/4 and the data evaluation by B. Klein were funded by the Deutsche