Detecting Seasonal and Decadal Oscillations

Atmospheric carbon dioxide (CO2) levels are rising globally, yet their multiscale variability in remote oceanic regions remains poorly characterized. This study examines a 45-year monthly CO2 record (1980–2024) from the Azores, a subtropical North Atlantic site, using a spectral and statistical framework. The series was decomposed into high- and low-frequency components via Butterworth filtering and analyzed with the Correlogram-Based Periodogram (CBP) and Monte Carlo significance testing. The residual component robustly recovered the expected seasonal cycle (~12 months), validating the methodology. The trend component revealed an apparent enhancement in low-frequency spectral power, largely explained by the accelerating long-term increase. Control tests with a synthetic quadratic trend and polynomial detrending indicate a weak ~11-year enhancement in low-frequency power that is not robust under a red-noise (AR(1)) null. Segmented regressions showed a sustained and accelerating increase in CO2 accumulation over the past four decades, consistent with Mauna Loa. These results demonstrate the importance of long-term monitoring in remote regions while highlighting both the potential and limitations of spectral methods for detecting weak low-frequency signals in greenhouse gas records.