Trend and Multi‐Frequency Analysis Through Empirical Mode Decomposition: An Application to a 20‐Year Record of Atmospheric Carbonyl Sulfide Measurements

Abstract

AbstractThe Empirical Mode Decomposition (EMD) is a fully non‐parametric analysis of frequency modes and trends in a given series that is based on the data alone. We have devised an improved strategy based on a series of best practices to use EMD successfully in the analysis of the monthly time series of carbonyl sulfide (OCS) atmospheric mole fractions measured at NOAA network stations (2000–2020). Long‐term trends and intra‐ and inter‐annual variability has been assessed. After a phase of generally increasing mole fractions up to 2015, with a temporary decline around 2009, we found that the OCS atmospheric mole fraction subsequently decreased at all stations, reflecting a recent imbalance in its total sources and losses. Our analysis has revealed a characteristic time scale for variation of 8–10 years. The variance associated with this long‐term behavior ranges from 15% to 40% of the total strength of the signal, depending on location. Apart from this complex long‐term behavior, the OCS time series show a strong annual cycle, which primarily results from the well‐known OCS uptake by vegetation. In addition, we have also found one more frequency of minor variance intensity in the measured mole fraction time‐history, which corresponds to periods in the range of 2–3 years. This inter‐annual variability of OCS may be linked to the Quasi‐Biennial Oscillation.