3.1 Crossover Analyses
Before the detailed crossover analyses were performed, all relevant parameters at crossover locations were inspected for overall data integrity. Data were imported into Ocean Data View (ODV), a data-visualization program available as freeware at http://www.awi.de/de/. For each crossover, all stations within a 1º (≈100 km) radius of each other were included for analysis. Individual ODV profiles were created of the carbon parameters, temperature, salinity, oxygen, nitrate, phosphate, and silicate vs full water depth. Each crossover plot was visually inspected for offsets among the cruises that might be consistent across the hydrographic, nutrient, and carbon parameters. Such consistent offsets might indicate that different water masses were sampled on the two cruises. This first look also allowed us to detect errors such as improper recording of the measurement temperature for fCO2 or an incorrectly labeled reference scale or measurement temperature for pH. After this visual inspection was completed, more rigorous approaches were performed to quantify offsets in carbon system parameters.
The crossover analyses involved the comparison of one to four inorganic carbon-system parameters (DIC, fCO2, TAlk, and pH) at each crossover against the density referenced to 4000 dB, σ4. σ4 was calculated from conductivity-temperature-depth sensor (CTD) salinity, CTD temperature, and pressure with the algorithms of Millero and Poisson (1981). When pressure was not included in the provided data set, the algorithms of Saunders and Fofonoff (1976) were used to calculate it from depth. A commercial plotting program, KaleidaGraph, was then used to create plots. Within each plot, an expanded area representing deep-water values was then selected (typically σ4 ≈45.5-45.9) and was plotted separately (Fig. 5). A second-order polynomial fit of the relevant carbon system parameter vs σ4 was determined for this limited σ4 range for each of the stations included in the crossover. For many of the crossovers, a strong curvature of DIC or TAlk vs density in the bottom waters complicated the curve-fitting analysis. Thus, the deepest water (Antarctic bottom water) often was not included in the regression. For high southern latitude crossovers, the inorganic carbon parameters were plotted against depth because the water masses south of 40º S have a very narrow density range.
Table 4 provides a summary of the derived constants M0, M1, M2, and r2 from the polynomial X = M0 + M1 (σ4) + M2 (σ4)2 curve-fitting routine for each station profile, where X is a value for DIC, TAlk, fCO2, or pH. The correlation coefficient r2 indicates the goodness of fit and often is a good indicator of precision of the data for a particular station. The calculated curve for each station in each crossover analysis was divided into ten evenly spaced σ4 intervals, on average, 0.04 units apart, over the data range ( σ4≈ 45.5-45.9). The value of X was calculated for each point in the interval from the derived M0, M1, and M2 for each of the stations. In all cases, the data bracketed the σ4 interval. The average difference between station profiles was then determined and was used as the calculated offset in Table 4. The standard deviation of the difference of the ten points indicated whether the offset values were systematic or scattered around the mean. As a convention, the calculated difference is always the value for a crossover station being subtracted from the first cruise/station of that crossover listed in Table 4. A search radius of 1º (≈100 km) sometimes included more than one station from a given cruise. This situation offered the opportunity to assess offsets of data for adjacent stations for a particular cruise as well as differences between cruises. Table 5 contains information similar to Table 4, but it is ordered according to the cruise. This table contains an average of the means of the crossovers listed for each cruise and is useful for determining whether parameters on a particular cruise show consistent offsets. Plots of the crossovers can be found on the following website: http://www.aoml.noaa.gov/ocd/gcc/atlantic_synthesis.html.