Hot or not? An evaluation of methods for identifying hot moments of
nitrous oxide emissions from soils
Abstract
Effectively quantifying hot moments of nitrous oxide (N2O) emissions
from agricultural soils is critical for managing this potent greenhouse
gas. However, we are challenged by a lack of standard approaches for
identifying hot moments, including (1) determining thresholds above
which emissions are considered hot moments, and (2) considering seasonal
variation in the magnitude and frequency distribution of net N2O fluxes.
We used one year of hourly N2O flux measurements from 16 autochambers
that varied in flux magnitude and frequency distribution in a
conventionally tilled maize field in central Illinois, USA to compare
three approaches to identify hot moment thresholds: 4x the standard
deviation (SD) above the mean, 1.5x the interquartile range (IQR), and
isolation forest (IF) identification of anomalous values. We also
compared these approaches on seasonally subdivided data (early, late,
non-growing seasons) vs. the whole year. Our analyses of the datasets
revealed that 1.5x IQR method best identified N2O hot moments. In
contrast, the 4 SD method yielded hot moment threshold values too high,
and the IF method yielded threshold values too low, leading to missed
N2O hot moments or low net N2O fluxes mischaracterized as hot moments,
respectively. Furthermore, seasonally subdividing the dataset
facilitated identification of smaller hot moments in the late and
non-growing seasons when N2O hot moments were generally smaller, but it
also increased hot moment threshold values in the early growing season
when N2O hot moments were larger. Consequently, we recommend using the
1.5x IQR method on whole year datasets to identify N2O hot moments.