Noble gas concentrations were sampled using copper tube sampling (~13g samples) with a handheld pump, and extracted and analyzed at the WHOI Isotope Geochemistry Facility using peak height manometry and isotope dilution mass spectrometry techniques. Uncertainties are calculated from replicate precision and are similar to expected analytical uncertainties reported by WHOI IGF. Noble gas saturation concentrations are calculated using the unpublished solubility functions of Dempsey Lott III and William Jenkins (WHOI Isotope Geochemistry Facility (http://www.whoi.edu/IGF)), derived for all five gases from the same instrument over conditions near one atmosphere and atmospheric partial pressures of the noble gases. Saturation concentrations were calculated using a Monte Carlo approach in which temperature and salinity (the dominant sources of uncertainty in the saturation concentration) were allowed to randomly vary within the standard deviation bounds for the environmental variability (1 degree Celsius, 0.5 on PSS-78) with 3000 iterations, and the mean variance of all iterations converted to a one standard deviation uncertainty for the saturation concentration based on the measured temperature and salinity at the sonde location. The same approach was used to combine measurement uncertainties with saturation concentration uncertainties and determine the one standard deviation uncertainty of the gas saturation states. These uncertainties were simultaneously determined for all gases at each timepoint, so that the full suite of noble gases vary with respect to each other. Variations of temperature and salinity around the pond were not random with respect to time (i.e. a location with high temperature relative to the sonde at 10am generally had higher temperature at noon and 2pm as well). Therefore the saturation state of any given noble gas varies systematically across each time series; the difference between concentrations at consecutive timepoints is conserved because the random uncertainty associated with the individual measurements is an order of magnitude lower than the systematic uncertainty associated with shifts in temperature and salinity compared to that expected from the sonde.

Because of the sampling method (smaller than usual copper tube samples, relatively slow and uneven flow from hand pumping compared to Niskin sampling), it is possible that atmospheric He was not adequately flushed from the copper tube and contaminated the samples for that gas only—we urge caution in the interpretation of the He results, and do not use these measurements in the associated manuscript (Howard et al., submitted to Limnol. Oceanog. Letters). There was no evidence that this issue would affect the other gases in a detectable manner, because of their higher solubility. One noble gas sample was flagged as having a pattern of saturation states consistent with contamination by a bubble during sampling. He, Ne, and Ar concentrations and saturation states from this sample are likely too large, but Kr and Xe are insensitive to bubble processes and any effect of sample contamination in minor on these more soluble gases.