IN SITU Carbon Dioxide Discrete Flask Sample Measurements - Mixing Ratio

This data set represents the most comprehensive carbon dioxide flask data set available in the world. The data set contains carbon dioxide mixing ratio in parts per million (ppm) based on measurements from the NOAA/CMDL Cooperative Flask Sampling Network from 1967-1991. There are 34 worldwide sites of which 32 were still active in 1991. These data are updated periodically as part of a cooperative effort between CMDL and NCDC which is supported as a core project by NOAA's Climate and Global Change Program. These data are also available on-line via internet on a CMDL workstation. Surface flask samples were made to ensure that the collected sample was not contaminated from local sources of carbon dioxide. The observer was instructed to take an air sample only when the wind was in excess of 2 meters per second blowing from the clean air sector near the site. All flask samples were shipped form the station to the NOAA/CMDL laboratory located in Boulder, CO and analyzed for carbon dioxide. The air samples were collected by two general methods: by flushing and then pressurizing the glass flasks with a pump or by opening a stopcock on an evacuated glass flask. Normally, the collection involved the connecting of two flasks in series. The time between sample collection and analysis usually ranged from several days to about 1 month. However, for some remote sites such as Amsterdam Island and all Antarctica sites, the time period between collection analysis averages 6 months. Some flask samples from these remote sites were stored for more than one year. Extensive quality checks are performed at the CMDL laboratory on all flask samples. Samples are occasionally found to be contaminated by local sources of carbon dioxide which do not represent the background concentration. In other cases, flask samples were found not to be adequately pressurized or not adequately flushed before pressurization. Since samples are collected in pairs, the pair difference is calculated, and samples with a pair difference greater than 0.5 ppm are flagged. Through 1988, one member of a bad pair was sometimes retained if it fell within +/- 3 sigma from a fitted curve. From 1989 to the present, both members of bad pairs have been automatically rejected. All original observations have been retained in this data set but have been designated with various quality flags if problems were found.