The provided datasets contain the original (non-normalized) high-frequency soil and meteorological observations that were fed to the Self-Organizing Map (SOM) in order to identify ranges of values associated with low and high soil O2 conditions. For the Champlain Valley (CV) site we used the natural breaks algorithm to subset the data into high and low O2 datasets. O2 values were consistently low at the Green Mountains (GM) site, so we ran a single SOM for all O2 values at this site. The original values were then range-normalized before they were fed to the SOM.