Dataset: GP15 Water Column Dissolved 227Ac Leg 1
Deployment: RR1814

RaDeCC Data Reduction: The instrument opens a two-stage window when an alpha decay is detected. The first stage (219 window) spans 0.01 to 5.60 ms and the second (220 window) spans 5.61 to 600 ms. Any subsequent decays occurring in each window are recorded. Decays of 219Rn lead to decays of 215Po that primarily occur in the 219 window and decays of 220Rn lead to decays of 216Po that primarily occur in the 220 window. The total record for events observed in successive 10-minute intervals are stored in a text file. A Matlab program was written to read these files, correct the values in each window for random decays of other isotopes, and for decays of 219 that occur in the 220 window plus decays of 220 that decay in the 219 window. The program plots the corrected activity vs. time. Random events resulting in chance coincidence in each window build up over time, and when these become large enough to dominate the window signal, the count is terminated. Water evaporating from the wet sample can condense in the detector and reduce efficiency during the count. If air leaks into the helium counting matrix, efficiency can also drop. These decreases in efficiency can be evaluated from the window signals vs. time, and the count is terminated when a decrease in counting rate appears to pose a problem. For GP15 samples, each count was 120 to 350 minutes, depending on the behavior observed. Anomalous signals indicating counter noise can also be detected and removed (rare).

Computations have been described by Moore and Arnold (1996), although two corrections were made for this work. For each channel:

cc# = R²t_g/(1-Rt_g)

where cc# is chance counts for the window # and t_g, is the window open interval;

R = random single events = BC - AC

where BC = total count rate and AC = sum of 219 and 220 window count rates;

cc219 = (t1/tg) R²(tg)/(1-R*tg)

cc220 = (t2/tg) R²⁽tg)/(1-R*tg)

where t1 and t2 are the duration of the 219 and 220 windows and tg is t1+t2. This corrects the algorithm with the addition of the first term on the right of the cc219 and cc220 expressions above. It makes little difference for low activity samples, but becomes significant at high count rates.

The second change was to solve simultaneous equations to correct for 219 in the 220 window and 220 in the 219 window from the signals after correction for chance counts:

Final 219 cpm = corr219 - 0.0288*(corr220)

Final 220 cpm = corr220 - 0.1279*(corr219)

With the constants in the equations above determined from the window durations and the Po daughter decay constants.

The matlab routine calculates uncertainties as the standard deviation of the means of the 10 minute counting interval lines.

Geometry and Standard Calibration: Two types of sample geometry have been used extensively by the community. The traditional one was developed by Moore (1976) and uses acrylic fibers coated with MnO2, housed in an acrylic tube about 200 cc in volume (identified here as MF geometry). The second type (used in Geotraces sampling by several groups) uses a commercial water filter coated with MnO2, housed in a commercial filter holder (~450 cc, identified here as CC geometry). A standard solution of 227Ac obtained by our collaborators at WHOI was obtained from Eckert and Ziegler, diluted, and sent to USC. This solution was used to prepare standards in the Geotraces CC geometry by adding an aliquot of spike to about 1 L of Ra-free seawater (prepared by passing through Mn-fibers in a MF cartridge). The spiked seawater was then passed through a cartridge 5 times at about 1 L/minute to ensure all Ac was sorbed. The residual seawater was later checked for 223Ra activity to be sure no Ac had remained in the solution or on the walls of the bottle. No activity was found. Multiple standards were prepared, but standardization was primarily with Std K-21. In addition, standards were prepared on acrylic fibers (MF geometry). One historical standard prepared from a calibrated solution of dissolved urananite (std 27) has been in use in our lab and was run regularly during the analysis of GP15 samples to ascertain any deterioration of detectors during this time. The relative efficiency of our 6 detectors remained constant while GP15 samples were being analyzed.

Radon Release Factor: 227Ac standards in the MF geometry typically decline by 10-20% with time, over a few months, and then become stable. This is evidence that as the initially sorbed 223Ra decays and is replaced by atoms produced by 227Ac sorbed to the fiber, some 219Rn must be recoiled deeper in the fibers, hindering its escape during the 3.96 s half-life of these atoms. This decline with time is not seen in CC geometry standards, as we have found these to be stable for many years. However, we have found that samples in the CC geometry release 219Rn about 30% more slowly than the initial release rate of standards in the MF geometry, resulting in lowered counting efficiency, when counters are run at our normal He flow rate.

Flow Rate Effects for Radecc: The CC geometry has been found to be much more sensitive to flow rate than the MF geometry (Fig. 2). It is critical for the CC geometry to keep flow rate in the normal range to obtain consistent results.

Channel Crosstalk: Signals in the 219Rn channel are created by high activity in the 220Rn channel (Fig. 3). The effect remains after chance count correction and also correcting the 219 counts for contributions from decay of the 216Po that is expected in the 219 window. Scholten et al. (2010) had previously noted this feature. We have confirmed it, and found it seems to be independent of channel identity. There does not seem to be an effect of high 219Rn on the 220Rn channel. The cause of the signal is likely from the fraction of decays of 216Po that occur after the 220 window is closed. The result is that the gate windows re-open, and some additional events are captured. The effect depends on the activity of 220Rn and can be removed by an empirical algorithm:

220 cpm: Crosstalk in 219 cpm

0-2: no correction

2-6.6: 0.0358*(220cpm) - 0.012

>6.6: 0.06433*(220cpm-6) + 0.0358

For samples from GP15, this effect was generally negligible.

Cartridge Sorption Efficiency: When cartridges are in series and cartridge quality is uniform, the sorption efficiency can be determined from the ratio of activities on the two cartridges, A_B/A_A:

Eff = 1 - A_B/A_A

However, most GP15 samples were run with A cartridges prepared at WHOI and B cartridges prepared at USC. The USC cartridges were used in both positions at station 37, and found to be 65% efficient. This is similar to our previous observations in the South Pacific (65%) and Arctic (62%), as well as the 65% observed by Geibert (2002) at similar flow rates and cartridge size. However, A cartridges were usually less efficient than B cartridges and somewhat more variable, averaging 23%. Consequently, total sample activities were calculated as (A activity) + (B activity)/(0.65), when two cartridges were used for sampling. In the upper water column, most sampling used only 1 WHOI cartridge, and there the activity was calculated as (A activity)/(0.23). The uncertainty in efficiency is difficult to evaluate, but appears to be about 0.05, based on data previously collected in the South Pacific. This introduces an uncertainty of 11% in the calculation of sample activity when A and B cartridges were used, and 20% if only one cartridge was used. Interestingly, the sorption efficiency for Ra on USC cartridges was 80%, higher than for Ac. Data based on a single cartridge has been flagged with 2, probably ok, but with greater uncertainty than samples collected with paired cartridges.

227Ac Concentration, Blanks, and Uncertainty Calculations: Typically, each cartridge was counted 2-6 times using different counters. Results were averaged and the standard deviation of the mean (sdom) was calculated. The uncertainty was also calculated from the uncertainties of each measurement. These were usually quite close to the sdom, and the larger of the two counting uncertainties was assumed. Contributions of uncertainties from crosstalk corrections and the uncertainty from cartridge efficiency were included by error propagation expressions. Total Ac and its uncertainty were divided by the water volume pumped to find concentration. Blanks were negligible, based on measuring unused cartridges or those deployed in pumps that did not pump.

Excess 227Ac: Analyses of the 227Ac parent 231Pa were done on Niskin water samples that were usually a few meters vertically from the pumps. These were performed by collaborating groups (C. Hayes of U. So. Miss., R. Anderson of LDEO and L. Edwards of U. Minnesota). Their results were subtracted from total Ac to calculate excess 227Ac. In a few cases, it was necessary to interpolate over larger distances, but the uncertainty from this is expected to be small. No effect on Ac uncertainties is expected, as the 231Pa measurements have a precision of only a few percent. Most samples that were not near the bottom have negligible excess 227Ac in comparison to the uncertainties, indicating the 227Ac is accurately calibrated. This is illustrated in Fig. 5, as a plot of total dissolved 227Ac vs. total 231Pa for samples from the upper 2 km, where paired cartridges were deployed. Measured excess activities were corrected for radioactive decay between collection and analysis using the 227Ac half life of 21.77 yrs.

Quality Flags: Quality flags have been assigned to data based on the following descriptions:

0 = no quality control.
1 = good value; A & B cartridges deployed.
2 = probably good; only A cartridge deployed; larger uncertainty; total Ac less than zero but within uncertainty of zero; adjusted to zero.
3 = probably bad; inconsistent with other data.
4 = bad value.
5 = changed value.
6 = below detection; total Ac less than zero; adjust to zero.
7 = value in excess of detection.
8 = interpolated value.
9 = missing value; insufficient pump volume, pump failure, or no cartridge delivered.