Influence of dataset density on CO2 and CH4 trend calculation

Trend analysis requires long-time series of observations, the continuity of which is not usually obtained since gaps emerge linked to problems with the device or external reasons. This paper explores the influence of dataset density on the parameters involved in linear trends of CO2 and CH4 half-hourly observations and daily ranges measured at a semi-rural site over a 6-year period. Half-hourly observation trends were 2.40 and 0.0085 ppm year−1 for CO2 and CH4 respectively, and a noticeable value of 1.98 ppm year−1 was obtained for the CO2 daily range, whereas the CH4 daily range remained steady. Random samples of variable numbers of observations were extracted from these time series. Robust statistics of location, spread, symmetry and concentration of observations were calculated and fitted with a third-degree polynomial expression for half-hourly observations and a linear equation for the daily range. In general, medians and interquartile ranges provided the best fits. Confidence intervals were also obtained. Around 350 and 760 half-hourly observations for CO2 and CH4 provided 95% of statistically significant correlations at the 0.1% level. Finally, daily evolution revealed the contrast between the two trace gases where CO2 increased with a ramp during the night that ended with a cliff whereas a sinusoidal evolution was associated with CH4. Moreover, the interquartile range presented a daily cycle for CO2 but not for CH4.