Statistical analysis of the CO2 and CH4 annual cycle on the northern plateau of the Iberian peninsula

Abstract

Outliers are frequent in CO2 and CH4 observations at rural sites. The aim of this paper is to establish a procedure based on the lag-1 autocorrelation to form measurement groups, some of which include outliers, and the rest include regular measurements. Once observations are classified, a second objective is to determine the number of harmonics in order to suitably describe the annual evolution of both gases. Monthly CO2 and CH4 percentiles were calculated over a six-year period. Linear trends for most of the percentiles were around 2.24 and 0.0097 ppm year−1, and the interquartile ranges of residuals calculated from detrended concentrations were 6 and 0.02 ppm for CO2 and CH4, respectively. Five concentration groups were proposed for CO2 and six were proposed for CH4 from the lag-1 autocorrelation applied to detrended observations. Monthly medians were calculated in each group, and combinations of harmonics were applied in an effort to fit the annual cycle. Finally, adding annual and semi-annual harmonics successfully described the cycle where one step was observed in the concentration decrease in spring, not only for high CO2 percentiles but also for low CH4 percentiles.