Data recovery of A06 and A07 WOCE cruises

Abstract. The WOCE cruises were carried out during the 1990s and were included in GLODAP, which is an easily usable, available and fully calibrated global database. AT and CT data, together with the rest of carbon variables, were subjected to rigorous quality control and some adjustments were done assuming biases, in case of AT and CT, not greater than ±6 μmol kg−1 and ±4 μmol kg−1, respectively. The A06 and A07 cruises were deleted from GLODAP database owing to AT and CT data were not suitable for analysis. However, these data are still available in CLIVAR and Carbon Hydrographic Data Office web site, demonstrated the unreliable quality of AT and CT, but contrarily, the more realistic profiles of pH data. The main goal of the present work is to recover AT and CT data of A06 and A07 using GLODAP database combining with CARINA database and the most contemporary cruise MOC2Equatorial 2010. Thus, AT data of A06 and A07 will be renewed using directly these data in a particular application of Multiple Linear Regression: the 3-D moving window MLR estimation method. Moreover, CT data will be recalculated using the CT AT ratio together with the obtained results from the crossovers analysis method. In order to demonstrate the quality of the recovered AT and CT, the new pH has been calculated, showing the good agreement in terms of pH obtained between A06 and A07 related to MOC2. To sum up, the entire carbon databases of A06 and A07 were checked and recovered.


Introduction
The World Ocean Circulation Experiment (WOCE) together with the Joint Global Ocean Flux Study (JGOFS), and the Ocean-Atmosphere Carbon Exchange Study (OACES) were carried out during the 1990s.The three programs covered different ocean regions, improving the combined global coverage.The collaborative efforts between scientists concluded in the known Global Ocean Data Analysis Project (GLODAP) (Sabine et al., 2005) as an easily usable and fully calibrated global database.Between the general goals of GLODAP, it is worth highlighting the distribution of the variables of Figures carbon system in order to estimate their changes and the possibility of evaluate the inventory of natural and anthropogenic carbon (Key et al., 2004).Thus, the data were merged into a common format database divided by ocean, which is available in the GLODAP website hosted by the Carbon Dioxide Information Analysis Center (CDIAC) (http://cdiac.ornl.gov/oceans/glodap/index.html).
In the Atlantic Ocean, not only there are several cruises merged in GLODAP (Key et al., 2004) database but also in CARbon IN the Atlantic Ocean (CARINA) (Key et al., 2010).CARINA database was generated in order to create a merged calibrated database from open ocean measurements of biogeochemical investigations, in particular, studies involving the carbon system.The wide set of historical and recent hydrographic cruises include the entire Atlantic, the Artic and Southern Ocean giving one CARINA data product for each region.The CARINA products and a significant volume of supporting information are available in the CARINA web site also hosted by CDIAC (http://cdiac.ornl.gov/oceans/CARINA/).Despite the fact that the CARINA data products are compatible with the GLODAP data products, they are not identical, differing in column order, included parameters and number of data.Thus, in order to obtain a complete database, historical and recent cruises have been taken from both GLODAP and CARINA databases in the present work.
One key point in both GLODAP and CARINA databases is that an extensive calibration and quality control procedures have been designed to remove measurement bias and bad data in the carbonate system measurements sampled during the cruises involved in CARINA and GLODAP.The carbonate system in seawater describes the equilibrium between dissolved CO 2 and carbonate and is defined by four parameters, i.e. total inorganic carbon (C T ), total alkalinity (A T ), pH and fugacity of CO 2 (f CO 2 ).Knowing at least two of these parameters allows calculating the unknowns.Only A T and C T data, focused on the Equatorial Atlantic Ocean, will receive the whole attention hereinafter, being their biases, assumed by both GLODAP and CARINA databases, not greater than ±6 and ±4 µmol kg −1 for A T and C T , respectively (Key et al., 2004(Key et al., , 2010)).Figures

Back Close
Full The combined GLODAP + CARINA database contains bottle and carbon data from surface to deep ocean waters in the Atlantic Ocean from historical and recent cruises.In order to complete this combined database and to contribute with more modern data, the most contemporary cruise in the equatorial Atlantic Ocean is included in that study: MOC 2 Equatorial (Meridional Overturning Circulation -Memoria Oce ánica del Clima, http://www.icm.csic.es/oce/ca/content/moc2/),hereafter MOC 2 .MOC 2 cruise was spanned along 7.5 • N during April-May in 2010 on board of BIO Hesp érides.Carbon system water samples, i.e.C T , A T and pH, have been sampled and analysed in the whole water column.The A06 (Oudot, 1993a) and A07 WOCE (Oudot, 1993b) cruises were carried out on board of N/O l'Atalante as part of the French CITHER project.A07 took place in January 1993 along 4.5 • S while A06 took place in February-March 1993 along 7.5 • N.
The A T and C T data of both cruises were deleted from GLODAP database because these data were not suitable for analysis (Wanninkhof et al., 2003).However, these data are still available in CLIVAR and Carbon Hydrographic Data Office (CCHDO) web site (http://cchdo.ucsd.edu/)inside the Atlantic Ocean data section.Thus, taking into account the combined GLODAP + CARINA + MOC 2 database, the complete database of A06 and A07 cruises can be objectively revised and corrected.Despite the fact that the A T and C T values of A06 and A07 cruises are out of quality controls of CARINA and GLODAP, the calculated pH data, involving these A T and C T , seem to show realistic and reliable profiles, as it can be seen in mind that A T is the parameter of the carbonic system with less variability, A T data A06 and A07 will be used directly in a 3-DwMLR method (Velo et al., 2011).Moreover, C T data will be checked and complete using the C T A T ratio and reassert the obtained A T and C T results using crossovers analysis (Tanhua, 2010).To sum up, the entire carbon databases of A06 and A07 will be tested and recovered.

Data
Data of GLODAP and CARINA databases were widely revised and involved in different quality control processes in order to assure the highest veracity and consistency to their included measurements.Thus, carbon data of cruises of the Atlantic Ocean merged in both databases were downloaded from their respective web sites (http://cdiac.ornl.gov/oceans/).Table 1 shows all the cruises selected in this work including A06 and A07 cruises, specifying cruise number, expocode, alias, database to which belong each cruise, year and, in the last column, if the defined cruise was carried out near A06, A07 or both of them.The geographical distribution of these cruises is showed in Fig. 2, where the entire Atlantic Ocean basin is represented.
Databases of A06 and A07 cruises are available in CCHDO web site (http://cchdo.ucsd.edu/).A06 and A07 cruises were carried out in the winter season along 7.5 • N and 4.5 • S, respectively, sampling and analysing some variables that defined the carbon system, i.e. pH and C T .Measurements of C T were made by gas chromatography according to the modified method as described in Oudot et al. (1995).pH measurements were based on the total hydrogen ion concentration scale (pH SWS ) using a Ross combination electrode calibrated in Tris buffer.Because of the unknown temperature and calibration data of pH data of A06 and A07 cruises (Wanninkhof, 2003), in the present work the pH data were calculated from A T and C T using the Excel CO 2 sys program (Pierrot et al., 2006) (http://cdiac.ornl.gov/ftp/co2sys/CO2SYScalc XLS/) and the acid constants from Mehrbach et al. (1973) fitted by Dickson and Millero (1987) together with the boron to chlorinity ratio from Lee et al. (2010).The A T data of cruises A06 Figures

Back Close
Full and A07 collected in the CCHDO were determined from C T and pH measurements using the equation describing the terms of alkalinity contributions defined by UNESCO (Oudot et al., 1995).MOC 2 cruise was carried out at spring season along 7.5 • N in 2010 on board of BIO Hesp érides.This 2010 cruise is very close repetition of the leg spanned along 7.5 • N by the cruise A06 in 1993 (see geographical location of A06 in Fig. 2).A T and pH samples have been analysed in the progression of the cruise, and the C T samples data were analysed in the laboratory of IIM-CSIC (Vigo).Measurements of A T were done by a one endpoint method using an automatic potentiometric titrator with a combined glass electrode (Mintrop et al., 2000), in which the potentiometric titration was carried out using the two pH endpoints method according to P érez and Fraga (1987).Seawater pH measurements were made using the spectrophotometric method described in Clayton and Byrne (1993) adding m-cresol purple as indicator and controlling the temperature at 25 • C by a thermostatic bath.The pH scale fixed in MOC 2 was the total scale (pH T ), but, with the aim of normalized pH data, MOC 2 pH T were rescaled to pH SWS using the CO 2 sys program.C T data were calculated from A T and pH using the CO 2 sys program because there were not C T measured in the whole levels of the water column.The coulometric C T measured (Johnson et al., 1993(Johnson et al., , 1998) ) is relevant to check if the calculated C T fits well with measured C T , observing a good consistence between them.In order to estimate the accuracy of the A T and C T methods, calibrations were performed with certified reference material (CRM) of CO 2 provided by Andrew Dickson.

Methodology
As previously said, the A T and C T of A06 and A07 cruises were deleted from GLODAP because of being out of quality controls.However, the combination of A T and C T produces suitable pH results, being the Multiple Linear Regression method (3-DwMLR) developed by Velo et al. (2011) was applied to try to obtain the consistent A T data of these cruises, defining successfully new A T data (A T MLR ).In order to check if the adjustments of A T MLR are well done, the widely applied in CARINA secondary quality control (2nd QC) crossovers analysis (Tanhua, 2010;Tanhua et al., 2010a, b;Velo et al., 2010a) method was used.
(2) Taking into account the ratio, the C T data were calculated using the Eq. ( 1): Where C T orig and A T orig are the C T and A T original data for the A06 and A07 cruises and A T MLR is the modified data set of A T for the A06 and A07 cruises.Nonetheless, a little bias in pH profiles is slightly appreciated (Fig. 1).Therefore, the same little bias should be transfer to the C T A T ratio.In order to solve this deviation, the crossover analysis method was applied to the cruises listed in Table 1, calculating C T data by Eq. ( 2): Where the factor, one for each A06 and A07 cruises, is calculated as function of the value of crossover offset obtained from crossovers analysis and taking into account 2100 µmol kg −1 as the mean C T value for deep water in the Atlantic Ocean.

The 3-D moving window MLR estimation (3-DwMLR)
A T is the parameter that can be computed most accurately with less uncertainty due to its low variability in the ocean.Moreover, A T can be well correlated with salinity, silicate, and even temperature (Wallace, 1995;Millero, 1995;Lee et al., 2006).Due to that, A T data can be directly used for MLR computations (Velo et al., 2011).The objective is to improve the results of a MLR by using a 3-D moving window around the node where A T is being calculated.An algorithm extracts a pool of data from a box around each A T data to perform the calculation.That box is the central node of the moving window, and each A06 and A07 A T data have assigned their specific data window.Each data window has a particular width and depth, which have been selected according to the experience taking in Velo et al. (2010b and2011).Thus, the unreliable quality A T data of A06 and A07 has been replaced by the 3-DwMLR method, giving back the successfully fitted A T MLR data for A06 and A07.

2nd QC: crossovers analysis
Crossovers analysis is an objective comparison of deep water data from one cruise with data from other cruises in the same area, available as a MATLAB tool-box in CDIAC web site (http://cdiac.ornl.gov/oceans/2ndQC Tool/) (Tanhua, 2010).The result of a crossover analysis is an offset ± standard deviation, which is defined as the difference between two cruises, A (cruise will be analysed) and B (cruise as reference), derived from the analysis.This tool interpolates vertical profiles of stations from A and B cruises in the nearest area and calculates the "difference profile".This procedure is repeated in each pair of stations and the result is a crossover offset media for each crossover pair.For C T , A T and salinity, these offsets are quantified as an additive adjustment, while, in case of O 2 and nutrient data are quantified as a lineal correction factor.
Here, the crossovers analysis has two main goals, (1) to confirm that the A T MLR data obtained using 3-DwMLR agree with GLODAP/CARINA quality criteria, i.e. biases not greater than ±6 µmol kg −1 and (2) to recover C T data (Eq.2), to obtain a valid offset to minimize the deviation of C T orig A T orig ratio giving a new coherent ratio.

Recovery of A T data
In order to check and complete the A T data of both A06 and A07 cruises along the Equatorial Atlantic Ocean, the 3-DwMLR method was used in this dataset.The evident improvements reached by using this method is observed in Fig. 3 (Fig. 3a and b) and A T MLR profiles (Fig. 3c and d) of A06 and A07 cruises are shown.The Fig. 3a and b show the A T original data calculated from pH and C T , while the A T MLR profiles (Fig. 3c and d) show more reliable profiles for both 7.5 • N and 4.5 • S legs, respectively from A06 and A07 cruises.Even though the A T orig profiles show a columnar distribution proceeding from the poor quality of C T orig measurements, there are common patterns in both A T orig and A T MLR profiles.For instance, the distribution of the upper waters is roughly similar reaching in both profiles the highest A T values and the influence of the bottom waters are also shared in these profiles.Moreover, the intermediate waters are also lightly correlated with the water masses present in the Equatorial Atlantic Ocean.However, more realistic and clearer profiles are shown by A T MLR profiles for A06 and A07 cruises (Fig. 3c and d), accordingly with the horizontal and vertical distribution of the water masses in this area.Importantly the minimum of A T associated to the Antarctic intermediate waters is clearly shown in A T MLR in both A06 and A07 profiles (Fig. 3c and d).
In order to confirm the good quality of A T MLR data recovered, the cruises listed in Table 1 have been used to the crossover analysis of A06 and A07.However, not all listed cruises have given an offset.The valid A T MLR offsets are summed up in the Table 2 together with the standard deviation and the mean offset value.The graphical representation of A06 and A07 A T crossovers is shown in upper panels of Fig. 4 (Fig. 4a  and b).Each point in these graphs is the offset ± standard deviation value of cruise A (cruise crossover ID listed in Table 2) taking as reference cruise B, A06 and A07, respectively, in Fig. 4a and b.In general terms, the offset value of both A06 and A07 crossover analysis are not greater than ±6 µmol kg −1 , which is the value of A T data bias accepted by CARINA and GLODAP databases.The mean crossovers offset values (black bold line in Fig. 4a and b and the last column of A T in Table 2) obtained for A T MLR A06 and A07 were −2.1 ± 3.7 µmol kg −1 and −1.0 ± 3.6 µmol kg −1 , respectively.These mean offset values are also in the quality control of GLODAP and CARINA.Therefore, A T MLR biases are not necessary to apply.From now on, A T MLR data can be used as A T data corrected and recovered from A06 and A07 cruises.

Recovery of C T data
The C T measurements of A06 and A07 cruises are not very reliable data, as it is shown in the Fig. 5a and b, where the distribution of C T lacks of continuity, even though the known patterns in surface, intermediate and bottom waters are lightly distinguished.As previously said, the pH orig , calculated with A T orig and C T orig , are realistic data (Fig. 1), which is an indicator of the validity of the ratio, C T was recalculated using Eq. ( 1), and therefore, the newest C T profiles would better fit than the C T orig profiles.However, the deviation of the C T A T ratio has more influence than the corrected A T MLR , giving the same deviated behaviour that the mean C T A T profiles.Therefore, in order to minimize that deviation a crossover analysis of C T orig is proposed.
The crossover analysis of C T orig from A06 and A07 cruises is applied including the cruises listed in Table 1.However, idem that happened in A T crossover analysis, not all the cruises have given an offset.The obtained C T offsets with their corresponding standard deviation for each cruises involved in the analysis are summed up in the Table 2, and their graphical representation is shown in the Fig. 4c and d  C T offset values for A06 crossover analysis vary from −8 to −14 µmol kg −1 , giving a mean offset of −11.5 ± 2.0 µmol kg −1 , while, in case of A07 crossover analysis, the offset values vary from −4 to −20 µmol kg −1 and their corresponding mean value is −9.2± 4.3 µmol kg −1 .Taking into account the mean offset value obtained for A06 and A07 crossover analysis, two factors are calculated assuming that the mean C T value in the Atlantic Ocean is 2100 µmol kg −1 .These factors are 1.0055 and 1.0045 for A06 and A07 C T data, respectively.Using Eq. ( 2), the new C T data (C T MLR ) for A06 and A07 cruises are calculated, as it is shown in lower panels of Fig. 5 (Fig. 5c and d).In these figures, the reliable profiles of C T MLR are observed, showing a distribution of the waters from surface to depth more homogenous than the C T orig .For example in A07 profile (Fig. 5d), the highest values of C T influenced by the Antarctic bottom waters are more outlined in C T MLR than C T orig , as well as the minimum values of intermediate water around 2000 m in A06 profile (Fig. 5c).
In order to check if these corrected C T MLR values are really reliable, the mean profiles of both A06 and A07 are calculated.In Fig. 6a, the suitability between A06 and A07 C T A T profiles related to MOC 2 profile is shown across the grey (light and dark) dashed lines, where is clearly identified the good agreement among cruises.

pH data
The pH orig data available of A06 and A07 data are not very reliable due to the unknown temperature and calibration data.Then, the pH orig was recalculated with A T orig and C T orig original using the CO 2 sys program showing reliable profiles (Fig. 1).In order to check if pH orig data A06 and A07 are in agreement with their mate cruise in location, MOC 2 , the mean pH profiles of these cruises have been made (Fig. 6b). the deviation of these A06 and A07 profiles is clearly shown related to the MOC 2 pH profile.
In order to assert the good quality of the recovered carbon variables, i.e.A T MLR and C T MLR , to A06 and A07 cruises, the pH MLR has been calculated.In the same way than mean pH orig profiles, the mean pH MLR profiles have been estimated, obtaining the expected coherence among A06, A07 and MOC 2 mean pH profiles, in the same way that it happens in mean profiles.In Fig. 6b, the clear concordance between A06 and A07 (light and dark grey dashed lines, respectively) and MOC 2 mean pH profiles (black continuous line) is demonstrated.Therefore, the whole carbon database was recovered from A06 and A07 cruises.

Conclusions
The recovery of A T and C T data of A06 and A07 cruises suggests new branches of study about the carbon distribution in the equatorial Atlantic Ocean.Taking into account the transport and the exchanges of mass, the carbon transport in that area could be estimated together with the storage of anthropogenic CO 2 in 1990s.Giving a regional estimate of the accumulative oceanic sink for anthropogenic CO 2 could also be done.Moreover, bear in mind the MOC 2 cruise, the possibility of study the changes in anthropogenic carbon storage, calcite/aragonite horizon and pH from 1990s to present times are planned as future work.
Taking into account this recovery of carbon data in the Atlantic Ocean, the increment in global carbon databases should be noted.The recovery and analysis of these A T and C T data from A06 and A07 cruises allows to future works having old references in the equatorial Atlantic Ocean because this area has never again been sampled since the 1990s WOCE cruises.The innovative 3-DwMLR method applied in this work to recover A T in the Equatorial of the Global Ocean being a useful tool to check and calculate data.On the other hand, the widely applied crossovers analysis is an easily and usable tool to check the quality of own data.Therefore, combining both 3-DwMLR and crossovers analysis methods could be used, as in the present work, to improve, for example, sampled and analysed data from a particular cruise and to recover data from rejected cruises.Full  Full  Full

ESSDD
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Fig. 1.In this figure, the pH profiles of A06, A07 and MOC 2 cruises are shown, observing the coherence between A06 and A07 profiles related to MOC 2 pH profile.The relevance of this good agreement among cruises is that the combination of A T and C T , in spite of being unreliable data, produces suitable pH results, hence, the ratio C T A T is valid.Taking into account not only that there are many stations or even cruises driven close in time and location to A06 and A07 cruises (Fig. 2) but also the suitability of calculated pH data together with the validity of the C T A T ratio, the possibility of recovery the A T and C T data of A06 and a07 cruises, is planned as main aim of the present manuscript.Bear in Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | valid.Therefore, the main phases of the present work were outlined.(1) Due to the low A T variability, the 3-D moving window ESSDD Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | , where the A T original ESSDD Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | A06 and A07 cruises have been made.The mean profile is the mean value of the defined layers of depth.The layers from surface to 1000 m are calculated each 200 m, while from 1000 m to bottom (∼5000 m) each 500 m.The resultant mean C T A T profiles of A06 (continuous grey light line), A07 (continuous grey dark line) and MOC 2 (black line) are shown in Fig. 6a.Bear in mind these both A06 and A07 profiles, a slight deviation of these profiles related to the MOC 2 C T A T profile is observed overall from 1000 m to bottom.Taking into account the recovered A T data (A T MLR ) and the reliability of the C T A T for A06 and A07 cruises, respectively.The whole offset values obtained for C T A06 and A07 are out of the quality controls assumed by GLODAP and CARINA (±4 µmol kg −1 ).The ESSDD Discussion Paper | Discussion Paper | Discussion Paper | The mean pH profiles of MOC 2 , A06 and A07 are shown in Fig. 6b.In this Fig.6b, the grey continuous lines (light and dark) represent A06 and A07 pH orig , respectively, while the black line represents MOC 2 pH.Following the same pattern than mean C Discussion Paper | Discussion Paper | Discussion Paper |

Atlantic
Ocean could be applied to other carbon variables, and of course, in other areas ESSDD Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Wallace, D. W. R.: Monitoring global ocean carbon inventories, Ocean Observing System Development Panel, Texas A&M University, College Station, TX, 54 pp., 1995.Wanninkhof, R., Peng, T.-H., Huss, B., Sabine, C. L., and Lee, K.: Comparison of inorganic carbon system parameters measured in the Atlantic Ocean from 1990 to 1998 and recommended adjustments, ORNL/CDIAC-140, 43 pp., 2003.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 2 .
Fig.2.Geographical distribution of selected cruises of the combined GLODAP and CARINA database carried out near A06 and A07 cruises.These both A06 and A07 cruise are also shown while MOC 2 cruise, which is not show, is a repetition of A06 stations.

Fig. 3 .
Fig. 3.A T data.The upper panels show the unreliable A T orig profiles of A06 and A07 cruises (a and b, respectively) while the lower panels, (c) and (d), show the recovered A T MLR profiles of A06 and A07, respectively.

Fig. 4 .
Fig. 4.Offsets obtained from crossover analysis method applied to A T orig (a and b for A06 and A07, respectively) and C T orig data (c and d for A06 and A07, respectively).For each individual graph, the "y-axis" represents the offset value in µmol kg −1 and the "x-axis" the crossover ID given to each cruises.Each point is the obtained offset with their associated uncertainty.

Fig. 5 .
Fig. 5. C T data.The (a) and (b) panels show the unsuitable C T orig profiles of A06 and A07 cruises, respectively, while the (c) and (d) panels show the recovered C T MLR profiles of A06 and A07, respectively.

Fig. 6 .
Fig. 6.Mean profiles of (a) C T A T ratio and (b) pH for A06 (light grey lines), A07 (dark grey lines) and MOC 2 (black lines).The A06 and A07 continuous lines display the original data, C T orig A T orig ratio and pH orig , while the dashed lines are the recovered data, C T MLR A T MLR ratio and pH MLR .

Table 1 .
Information about the cruises selected from combined GLODAP and CARINA database including A06 and A07 cruises.The cruise number is the number assigned by GLO-DAP or CARINA databases, Expocode is the international name for the cruise, Alias is a little name for each cruise, Year is the year when the cruise was carried out, Database is the open access database to which belong each cruise, and, in the last column, Crossovers means if the defined cruise was selected to use in crossover analysis of A06, A07 or both of them.

Table 2 .
Results from crossover analysis applied to A T MLR and C T orig of A06 and A07 cruises.Cruise number is the number assigned by GLODAP or CARINA databases.Cruise Crossover ID is the identity given to each cruise in the crossover analysis.Offset and Uncertainty are the offset ± standard deviation obtained from crossover analysis, while the mean offset is the mean value of all offsets including in each A T or C T analysis.