On long-term SABER CO2 trends and effects due to nonuniform space and time sampling

The Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) instrument on board the TIMED satellite has been continuously operating for more than 16 years, since 2002, monitoring the CO2 concentration on nearly a global scale in the middle and upper atmosphere (from 65 km up to 110 km). A recent reanalysis (Qian et al., 2017, ) concluded that different deseasonalizing methodologies may have a strong impact on long-term trend analysis, ultimately yielding different altitude profiles of the global mean CO2 trend. In this work, we aim to understand how the nonuniform spatial and temporal sampling inherent in the SABER CO2 data set affects the determination of the long-term trends. In addition, our goal is to disentangle reported differences in SABER CO2 trends due to different time averaging windows and methodologies used for trend estimation. The Whole Atmosphere Community Climate Model is used for synthetic studies of the time series. We demonstrate that, due to the time varying data gaps and nonuniform sampling of local times, different time binning of the SABER CO2 data may indeed bias the long-term trend estimation. We show and discuss how the 60-day averaging reduces the bias in relative trends. We also conclude that different deseasonalizing methodologies (averaged over the same temporal bins) yield negligible differences on the trend determination. Taking this into account the global mean CO2 relative trend does not deviate statistically from the tropospheric value below 1 x 10(-3) mb (90 km). Above about 90 km, there is a positive slope in the global CO2 trend profile, but with substantially reduced magnitude for 60-day binned data.