There are multiple levels of quality control for 12-pack flask samples taken at each aircraft site, however, there are also multiple sources of errors that need to be assessed. Sources of error to the 12-pack flask measurement fall into three different categories: flask induced bias, sampling errors and analysis errors.
Flask induced biases and reproducibility
Each flask is made with borosilicate glass that has been rinsed with hydrophosphoric acid and distilled water and then dried at ~500 C; this ensures that the inner surfaces of the flasks are clean. Once the flask package is assembled, it goes through both short-term (1-7 days) and long-term (~28 days) tests to identify leaks and possible contamination for gases measured on the MAGICC system or hydrocarbon and halocarbon system (Neff et al. in prep). For MAGICC gases a 12-pack will not pass the short term storage test unless the standard deviation of all 12 flasks is less than 0.055 ppm and offset from network flasks filled simultaneously is less than 0.155 ppm for CO2. Similarly, a 12-pack will not pass the long term test unless the standard deviation of all 12 flasks is less than 0.155 ppm and offset is less than 0.355 ppm. Although no limits are set for the other MAGICC gases, the table below illustrates typical offsets observed and the average standard deviation that has been observed from flask to flask within a 12-pack. The standard deviation of the individual flasks in a 12-pack package suggests that the 12-pack does add significant variability over and above the repeatability of the measurement for all the MAGICC gases except CH4, which shows relatively small offsets and variability between flask compared to the measurement precision (Error: Reference source not found). In considering the bias as represented by the offsets between measurements of network flasks filled and analyzed at the same time as the 12-pack flasks, both CO2 and N2O show a significant decrease with time.
Multiple storage tests over the last 5 years suggest that CO2 and probably N2O are diffusing through the Teflon O-ring seals at the end of each flask at the rate of about 0.007 ppm per day. This diffusion is driven, primarily, by the more than 175 kPa pressure gradient between the manifold and flask interior. Since both O2 and N2 diffuse more slowly across this pressure gradient, the net result is depletion in the CO2 mixing ratio over the time of storage (Neff et al. in prep). At this time no corrections have been made for the biases.
Sampling and Analysis Errors
During sampling the flow rate and time are monitored to determine how much air has been used to flush the manifold (>5 L) and each flask (>10 L) and to make sure that the target storage pressure (40 psia, 275 kPa) is reached. For most compressor packages these specifications are easily reached unless there is a leak in the system (i.e. a torn diaphragm or a break in the line between the compressor and flask packages). However, there are many high altitude samples that show no indication of contamination yet do not reach 275 kPa, which indicates that over time the PCP does not meet original specifications (Neff et al. in prep). For this reason, flask samples have been reviewed on a case-by-case basis to determine if there is obvious contamination due to sampling irregularities. In the sampling process, careful records of flows and instrument drift are examined to detect irregularities in the automated sample analysis. Because of the high cost and limited sampling resolution afforded by the 12-flask packages, duplicate sampling has not been performed on a regular basis, which means that we must rely on other indicators to detect contamination from the 12-flask system or aircraft.
For data presented here, there are two methods that have been used to identify potential contamination or leaks. The first approach is to bin the data by altitude and look for periods where there are at least three standard deviations from the mean annual trend and seasonal cycle (Masarie and Tans, 1995). The second approach for identifying potential contamination relies on a comparison of the multiple species that are measured in each sample. In particular, sample line leaks are typically identified by high CO from engine exhaust, C2H2F4 and CHClF2 from air conditioners, and CBrClF2 from fire extinguishers. Because the flow rates at higher altitudes are reduced and typical concentrations in the mid-troposphere are low for these gases, any enhancements from leak are generally more noticeable at the top of a profile. When enhanced mixing ratios of CO, C2H2F4, etc. are measured, the MAGICC species are flagged to reflect possible contamination.