3.3 The Multiple-Parameter Linear Regression Method for Checking Overlapping Cruises
Several of the cruise lines were completely or partially reoccupied during the 1990s, and the MLR technique was used to compare the overlapping cruise data as well. This method is preferable to direct comparison because biogeochemical variables often co-vary in a systematic fashion in water masses. Water mass movement between the times of investigation is thus (partially) accounted for in the MLR technique. The following cruises were compared: A05 with A05R, A16N with A16S, A01W with A24, A01W with A01WR, and A12B with A12BR. The suffix "R" stands for repeat occupation.1 The coefficients for the MLR were determined by combining data from both cruises in question. Multiple-parameter linear fits of the form
DIC = α1 + α2 S + α3 Θ + α4 AOU + α5NO3 + α6 SiO2,
TAlk = β1 + β2 S + β3 Θ + β4 NO3 + β5 SiO2
were created for the combined data set, and the average differences between the observed points and calculated values (observed minus predicted from the fit) were determined (Table 8).
The MLR was applied over different depth ranges, depending on number of samples and the density structure. In general, samples at depths greater than 1500 m were used to avoid biases from variations in the upper water column. For the A16N/A16S comparison between 4° N and 4° S along 25° W, the criteria were relaxed to include depths up to 200 m to get a sufficient number of samples. Because both cruises were run in the summer, seasonal variability should not have unduly affected the values. For the high-northern-latitude A24 and A01W cruises, only samples deeper than 3000 m were analyzed because the high-latitude regions experience deep ventilation and the two cruises did not overlap exactly in space. No significant differences were observed for DIC between overlapping cruises. Significant differences were observed for TAlk for A24, A01W, and A01WR, but these differences were generally within one standard deviation. As described below, these results were used as corroborative evidence of systematic offsets between cruises.
1Throughout this report the standard WOCE/WHP cruise identifiers were used except for the repeat occupations, which are listed by the WOCE/WHP primary cruise identifier with an "R" attached rather than the official WOCE/WHP repeat cruise designation, "ARXX."